The Oxford Handbook of Land Economics

T H E OX F OR D HA N DB O OK OF
LAND
E C ON OM IC S
CONSULTING EDITORS
Michael Szenberg
Lubin School of Business, Pace University
Lall Ramrattan
University of California, Berkeley Extension
THE OXFORD HANDBOOK OF
LAND
ECONOMICS
Edited By
JOSHUA M. DUKE
and
JUNJIE WU
1
3
Oxford University Press is a department of the University of
Oxford. It furthers the University’s objective of excellence in research,
scholarship, and education by publishing worldwide.
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trademark of Oxford University Press

in the UK and certain other countries.
Published in the United States of America by

Oxford University Press
198 Madison Avenue, New York, NY 10016
© Oxford University Press 2014
All rights reserved. No part of this publication may be reproduced, stored in

a retrieval system, or transmitted, in any form or by any means, without the prior
permission in writing of Oxford University Press, or as expressly permitted by law,
by license, or under terms agreed with the appropriate reproduction rights organization.
Inquiries concerning reproduction outside the scope of the above should be sent to the
Rights Department, Oxford University Press, at the address above.
You must not circulate this work in any other form

and you must impose this same condition on any acquirer.
Library of Congress Cataloging-in-Publication Data

The Oxford handbook of land economics / edited by Joshua M. Duke and Junjie Wu.
pages cm
Includes bibliographical references and index.
ISBN 978–0–19–976374–0 (alk. paper)
1. Land use. 2. Economic development. I. Duke, Joshua M. II. Wu, JunJie.
HD111.O94 2013
333.73—dc23
2013024328
1 3 5 7 9 8 6 4 2
Printed in the United States of America
on acid-free paper
Contents
Foreword ix
Preface xiii
List of Contributors xv
I ntroduction: Land as an Integrating Theme in Economics 1

Joshua M. Duke and JunJie Wu
PA RT I â•‡ DE T E R M I NA N T S A N D DR I V E R S
OF L A N D U SE C HA N G E
1. Integrating Regional Economic Development Analysis and
Land Use Economics 23
Mark D. Partridge and Dan S. Rickman
2. Technology Adoption and Land Use 52
David Zilberman, Madhu Khanna, Scott Kaplan, and
Eunice Kim
3. Are Large Metropolitan Areas Still Viable? 74
Edwin S. Mills
4. Modeling the Land Use Change with Biofuels 85
Madhu Khanna, David Zilberman, and Christine L. Crago
5. Modeling the Determinants of Farmland Values in the United States 111
Cynthia J. Nickerson and Wendong Zhang
6. Land Use and Sustainable Economic Development:
Developing World 139
Edward B. Barbier
vi Contents
PA RT I I E N V I RON M E N TA L A N D
S O C IOE C ON OM IC C ON SE Q U E N C E S OF
L A N D U SE A N D L A N D U SE C HA N G E
7. The Economics of Wildlife Conservation 163
David J. Lewis and Erik Nelson
8. Connecting Ecosystem Services to Land Use: Implications
for Valuation and Policy 196
Robert J. Johnston, Stephen K. Swallow, Dana Marie Bauer,
Emi Uchida, and Christopher M. Anderson
9. Land Use and Climate Change 226
Bruce A. McCarl, Witsanu Attavanich, Mark Musumba,
Jianhong E. Mu, and Ruth Aisabokhae
10. Land Use, Climate Change, and Ecosystem Services 255
Witsanu Attavanich, Benjamin S. Rashford,
Richard M. Adams, and Bruce A. McCarl
11. Fire: An Agent and a Consequence of Land Use Change 281
Claire A. Montgomery
12. Land Use and Municipal Profiles 302
Edward Stone and JunJie Wu
PA RT I I I M E T HOD OL O G IC A L
DE V E L OP M E N T S
13. An Assessment of Empirical Methods for Modeling Land Use 327
Elena G. Irwin and Douglas H. Wrenn
14. Equilibrium Sorting Models of Land Use and Residential Choice 352
H. Allen Klaiber and Nicolai V. Kuminoff
15. Landscape Simulations with Econometric-Based Land Use Models 380
Andrew J. Plantinga and David J. Lewis
16. An Economic Perspective on Agent-Based Models of Land Use and
Land Cover Change 402
Dawn Cassandra Parker
Contents vii
17. Spatial Econometric Modeling of Land Use Change 430

Seong-Hoon Cho, Seung Gyu Kim, and Roland K. Roberts
18. Using Quasi-Experimental Methods to Evaluate Land

Policies: Application to Maryland’s Priority Funding Legislation 452
Charles Towe, Rebecca Lewis, and Lori Lynch
19. Applying Experiments to Land Economics: Public Information and

Auction Efficiency in Ecosystem Service Markets 481
Kent D. Messer, Joshua M. Duke, and Lori Lynch
PA RT I V T H E E C ON OM IC S OF
L A N D U SE L AW A N D P OL IC Y
20. Open Space Preservation: Direct Controls and Fiscal Incentives 513

Ekaterina Gnedenko and Dennis Heffley
21. Land Conservation in the United States 547

Jeffrey Ferris and Lori Lynch
22. European Agri-Environmental Policy: The Conservation and

Re-Creation of Cultural Landscapes 583
Ian Hodge
23. Agri-Environmental Policies: A Comparison of US and EU

Experiences 612
Roger Claassen, Joseph Cooper, Cristina Salvioni,
and Marcella Veronesi
24. Stigmatized Sites and Urban Brownfield Redevelopment 648

Joel B. Eisen
25. Regulatory Takings 668

Thomas J. Miceli and Kathleen Segerson
26. Eminent Domain and the Land Assembly Problem 698

Joshua M. Duke
27. Future Research Directions in Land Economics 723

Joshua M. Duke and JunJie Wu
Subject Index 737

Foreword
Few human sentiments are more urgent than place—the place we are born, the place we
become sentient, the place we engage others in a variety of pursuits, and finally the place
we will become dust. Some creatures have their territory. Humans are creatures of their
territory.
It cannot therefore be a surprise that place becomes conflated with land, and vice
versa. The primacy of land can be seen in a number of ways. In some societies, “prop-
erty rights”—shorthand for some presumptive imaginings about individual control
over land—often seem more important than “human rights” (whatever they might be).
History reveals the military and political importance of land. In agrarian societies the
connection between land and economic well-being is obvious. In that regard, it has
been claimed that economic development and attendant urbanization will diminish the
economic importance of land. This now seems improbable. Indeed, one could make a
plausible argument that land will become of increased importance in the future. The
contents of this marvelous volume would certainly support that hypothesis.
Those of us who are modern know well the Lockean Creation Myth—God gave land
to all in common and then admonished us to take control of it and make it flourish.
From this mischief all manner of tragedy has followed, whether we have in mind the
near-complete annihilation of indigenous peoples the world over, or the near-misses of
European wars of mutually assured destruction throughout recorded history. Land is
always worth a good (or bad) fight.
Happily for those of us who are economists, land is also—and always will be—worth a
good debate. And good debates lead to good science.
The chapters included here offer profound insights into many of those debates. Josh
Duke and JunJie Wu have arranged for an impressive lineup of experts to address, with
clarity and rigor, the important issues requiring good analysis and coherent solutions.
An important undercurrent here, and one that explains many of the difficulties in
crafting workable public policy to address problems in land use and land use change, is
the conceptual inconvenience that land is a fictitious commodity [Polanyi, 2001].1
. . . labor, land, and money are essential elements of industry; they also must be orga-
nized in markets; in fact, these markets form an absolutely vital part of the economic
system. But labor, land, and money are obviously not commodities; the postulate that
1 Karl Polanyi, The Great Transformation, Boston: Beacon Press, 2001.

x Foreword
anything that is bought and sold must have been produced for sale is emphatically
untrue in regard to them. . . . Labor is only another name for a human activity which
goes with life itself. . . .; land is only another name for nature, which is not produced
by man; actual money . . . is merely a token of purchasing power . . . which comes into
being through the mechanism of banking or state finance. None of them is produced
for sale. The commodity description of labor, land, and money is entirely fictitious
[Polanyi, 2001, pp. 75–6].
The inconvenience of land as a fictitious commodity arises because economic mod-

els can only do the necessary work when they are deployed in the service of answers to
questions that motivated their creation in the first place. All models are context spe-
cific, and they are only useful if their deployment in new settings is consistent with the
assumptions underlying their essential structure. The test of all models is whether or not
they are good to think with.
For most economic models, various quantities of a particular commodity can be
arrayed along one axis, and the various prices of that commodity can be arrayed along
another axis. Unfortunately for land and land use changes, the commodity fiction ren-
ders this problematic. It is, of course possible to plot acres/hectares of land along one
axis, and it is possible to refine that depiction by incorporating some index of “quality.”
But that may not satisfy some who refuse to see land in that light. Equally problematic,
the other axis in our models reflects yet another fictitious commodity—money.
Suddenly we see why there are so many profound debates about land. The very con-
cepts and models that allow us to analyze markets for “real” commodities—toothpaste,
bread, houses, cameras—offer up seriously contested concepts when we must deal with
land (nature). Two obvious problems arise. First, many people refuse to accept money
as a plausible measure of the value of land—the one fictitious commodity cannot be
mapped into the other fictitious commodity. Second, many people refuse to accept the
idea that land (nature) is a commodity. Note that for the concept of a commodity to
have any meaning in economics it must be capable of assignment (ownership). This
introduces the concept of “belonging to.” Native people say that land does not belong to
them—they belong to the land. The implication of this notion may warrant a brief elab-
oration. Recall that the essence of a normal commodity is that when it moves through
markets the only thing that really matters is that there is a change in its ownership. That
is what markets do—they mediate changes in ownership of those commodities that “pass
through” markets. And since ownership is itself yet another social construct, we see the
layering of contestation that will always attend economic analysis of land, land use, and
land use change.
We disregard these concerns at our peril. If we hope to produce policy relevant
insights concerning the contested realms of land, we must speak to a large audience of
sapient adults who refuse to accept quite fundamental presumptions in our models.
Science practiced in disregard for shared human meanings is impertinent.
The various chapters here admirably spell out the contested nature of figuring out how
to think about what is better to do with respect to land. I like to say that there is no such
thing as land, there is only land tenure—social rules that bestow on certain individuals
Foreword xi
a circumscribed suite of capacities concerning what can and cannot be done with that
thing we call land. And this reminds us that when we study land we are really poking
around at the outer limits of presumptions concerning who gets to define the rules by
which land use—and land use change—shall be determined. It seems we are back to the
matter of presumptive “rights” over land. And as we know, rights are not inherent but
worked out:
Only those economic advantages are rights which have the law back of
them . . . whether it is a property right is really the question to be answered [Justice
R. Jackson, Willow River Power Co. 324 US 499, 502 (1945)].
In other words, economic advantages are not protected because they are rights.
Rather, those settings and circumstances that a society chooses to consider valuable
are given protection under the cover of “rights.” We see that economic advantages are
bestowed by the political class. Suddenly we grasp the fount of contestation over the
manifold advantages of owning this thing called land.
Daniel W. Bromley
Madison
October, 2011
Preface
Land use change is arguably one of the most pervasive socioeconomic forces affecting
ecological systems, economic systems, and human wellbeing. Almost all major environ-
mental problems, including climate change, water pollution, and habitat destruction,
are rooted in land use change. Many socioeconomic phenomena, such as urban sprawl,
suburbanization, urban redevelopment, and economic segregation, are also deeply
ingrained in land use change. In response to the great need to study these environmental
and socioeconomic phenomena, many new developments have taken place in the field
of land economics during the past decade, justifying a new handbook in the field.
This volume draws on recent advances in several literatures that investigate land use
behavior and policy, including natural resource economics, environmental economics,
regional science, and urban economics. The contributors of this volume are the eminent
scholars in the field and the newer experts, who work at the frontier of the field. Starting
from inherited theories and analyses, this forward-looking handbook seeks to become a
“must” reading, not only for those who are new to the field, but also for those who want
to extend their knowledge to the frontier of land economics.
There are various ways to use this handbook. This comprehensive treatment of land
economics provides an excellent source of readings for a graduate course in land or
resource economics. Although the length and diversity of methods may make it difficult
to cover in a single semester course, instructors may seek to focus on a subset of chap-
ters. For instance, a course might be structured around the chapters on the ecosystem
services of land and a few related methods chapters. Or, the focus might be on cutting
edge methods in land economics, supplementing the methods chapters with seminal
articles in general economics on equilibrium modeling, auction theory, and specific
econometric techniques. Researchers and policy analysts will find that the book offers
the state-of-the-art in land economics research. The depth of coverage on the methods
chapters offers researchers a structure for setting up their own analyses. The applied
chapters can serve either as a starting point for learning about markets and incentive
problems associated with land topics, or as a source of citable research results and syn-
thetic conclusions from experts in the area. Those with less familiarity with economics
can also use this handbook to understand what is known and unknown on a given topic
area. This will help noneconomists, policy makers, and grant funders to articulate better
hypotheses, policy goals, and funding opportunities.
We are profoundly grateful to our chapter authors for their outstanding contributions
to this handbook. We also acknowledge the insights of our colleagues around the world,
who inspire us with their research and collegiality. Among a very long list, we would
xiv Preface
like to single out Daniel Bromley and Kathy Segerson as our mentors, who shaped our
lives—research and otherwise—at a deep level and to whom we owe a great debt. We
would also like to recognize the other leading lights in our professional lives, includ-
ing Emery N. Castle, Richard M. Adams, Bill Boggess, and David Zilberman. Finally,
we thank our friends and colleagues for their advice and encouragement during the
long process, especially Titus Awokuse, Kathleen Bell, Rob Johnston, Lori Lynch, Kent
Messer, and George Parsons. Finally, we are grateful for the support of our Universities,
whose combined land grant missions have promoted the advancement of integrated
land economics research.
Joshua M. Duke, Newark, Delaware

JunJie Wu, Corvallis, Oregon
List of Contributors
Richard M. Adams is Professor Emeritus in the Department of Applied Economics at

Oregon State University.
Ruth Aisabokhae is Economist, DuPont Pioneer, Lagos, Nigeria.
Christopher M. Anderson is Associate Professor in the School of Aquatic and Fishery
Sciences at University of Washington.
Witsanu Attavanich is Lecturer in the Department of Economics at Kasetsart University
in Thailand.
Edward B. Barbier is the John S. Bugas Professor of Economics in the Department of
Economics and Finance at University of Wyoming.
Dana Marie Bauer is Assistant Professor in the Department of Geography and Environment
at Boston University.
Daniel W. Bromley is Professor Emeritus in the Department of Agricultural and Applied
Economics at the University of Wisconsin-Madison.
Seong-Hoon Cho is Associate Professor in the Department of Agricultural and Resource
Economics at University of Tennessee.
Roger Claassen is a senior agricultural economist in the Resource, Environmental,
and Science Policy Branch of the US Department of Agriculture Economic Research
Service.
Joseph Cooper is Chief of the Agricultural Policy and Models Branch in the Market
and Trade Economics Division of the US Department of Agriculture Economic Research
Service.
Christine L. Crago is Assistant Professor in the Department of Resource Economics
and Commonwealth Honors College at University of Massachusetts Amherst.
Joshua M. Duke is Professor in the Department of Applied Economics and Statistics at
University of Delaware.
Joel B. Eisen is Professor of Law at the University of Richmond School of Law.
Jeffrey Ferris is a Graduate Student in the Department of Agricultural and Resource
Economics at University of Maryland.
xvi list of Contributors
Ekaterina Gnedenko is Lecturer in the Department of Economics at Tufts University.

Dennis Heffley is Professor in the Department of Economics at University of Connecticut.
Ian Hodge is Professor in the Department of Land Economy at University of Cambridge.
Elena G. Irwin is Professor in the Department of Agricultural, Environmental, and
Development Economics at the Ohio State University.
Robert J. Johnston is Director of the George Perkins Marsh Institute and Professor in the
Department of Economics at Clark University.
Scott Kaplan is an undergraduate student at the University of California, Berkeley.
Madhu Khanna is Professor in the Department of Agricultural and Consumer Economics
at University of Illinois.
Eunice Kim is Program Administrator in the Department of Agricultural and Resource
Economics at University of California, Berkeley.
Seung Gyu Kim in Assistant Professor of Agricultural Economics at Kyungpook
National University in South Korea.
H. Allen Klaiber is Assistant Professor in the Department of Agricultural, Environmental,
and Development Economics at the Ohio State University.
Nicolai V. Kuminoff is Assistant Professor in the Department of Economics at Arizona
State University.
David J. Lewis is Associate Professor of Applied Economics at Oregon State University.
Rebecca Lewis is Assistant Professor of Planning, Public Policy and Management at
University of Oregon.
Lori Lynch is Professor in the Department of Agricultural and Resource Economics at
University of Maryland.
Bruce A. McCarl is Regents and Distinguished Professor in the Department of Agricultural
Economics at Texas A&M University.
Kent D. Messer is Associate Professor and Unidel Howard Cosgrove Chair for the
Environment in the Department of Applied Economics and Statistics at University of
Delaware.
Thomas J. Miceli is Professor in the Department of Economics at University of
Connecticut.
Edwin S. Mills is Emeritus Professor of Real Estate and Finance in the Kellogg School of
Management at Northwestern University.
Claire A. Montgomery is Professor in the Department of Forest Engineering, Resources
and Management at Oregon State University.
list of Contributors xvii
Jianhong E. Mu is Postdoc Scholar in the Department of Applied Economics at Oregon

State University.
Mark Musumba is Postdoctoral Research Fellow at The Earth Institute of Columbia
University.
Erik Nelson is Assistant Professor of Economics at Bowdoin College.
Cynthia J. Nickerson is an agricultural economist in the Resource and Rural Economics
Divison of the US Department of Agriculture Economic Research Service.
Dawn Cassandra Parker is Associate Professor in the School of Planning at University
of Waterloo.
Mark D. Partridge is the C. William Swank Chair of Rural-Urban Policy at The Ohio
State University and a Professor in the Agricultural, Environment, and Development
Economics Department.
Andrew J. Plantinga is Professor of Natural Resource Economics and Policy in the Bren
School of Environmental Science and Management at University of California, Santa
Barbara.
Benjamin S. Rashford is Associate Professor in the Department of Agricultural and
Applied Economics at University of Wyoming.
Dan S. Rickman is Regents Professor of Economics and Oklahoma Gas and Electric
Services Chair in Regional Economic Analysis in the Department of Economics and
Legal Studies in Business at Oklahoma State University.
Roland K. Roberts is Professor and Director of Graduate Studies in the Department of
Agricultural and Resource Economics at University of Tennessee.
Cristina Salvioni is Associate Professor of Agricultural Economics in the Department of
Economics at University of Chieti-Pescara, Italy.
Kathleen Segerson is Philip E. Austin Professor in the Department of Economics at
University of Connecticut.
Edward Stone is a graduate research assistant in the Department of Applied Economics
at Oregon State University.
Stephen K. Swallow is Professor and DelFavero Faculty Fellow in the Department of
Agricultural and Resource Economics and the Center for Environmental Sciences and
Engineering at University of Connecticut.
Charles Towe is Assistant Professor in the Department of Agricultural and Resource
Economics at the University of Maryland.
Emi Uchida is Assistant Professor in the Department of Environmental and Natural
Resource Economics at University of Rhode Island.
xviii list of Contributors
Marcella Veronesi is Assistant Professor in the Department of Economics at the

University of Verona, Italy and the Institute for Environmental Decisions at ETH Zurich,
Switzerland.
Douglas H. Wrenn is Assistant Professor in the Department of Agricultural Economics,
Sociology, and Education at the Pennsylvania State University.
JunJie Wu is Emery N. Castle Chair in the Department of Applied Economics at Oregon
State University.
Wendong Zhang is a doctoral student in the Department of Agricultural, Environmental,
and Developmental Economics at the Ohio State University.
David Zilberman is Robinson Chair in the Department of Agricultural and Resource
Economics at University of California, Berkeley.
INTRODUCTION
Land as an Integrating Theme in Economics
JO SH UA M . DU K E A N D J U N J I E W U
1. Land Markets and Welfare
This handbook explains what economists know about land—and how they know. The
innumerable decisions about how to use land and how to change land uses over time per-
vade society, affecting human well-being both directly and indirectly through changes
in the performance of economic and ecological systems. Large shares of major environ-
mental problems (air pollution, water pollution, climate, habitat destruction, to name a
few) are rooted in land use change. Socioeconomic phenomena such as urban sprawl,
suburbanization, urban redevelopment, and jurisdictional fragmentation are essentially
land use changes by another name—and they affect opportunities for further land use
change. Fomenting all these forces are the inescapable land policies, which sanctify win-
ners, disappoint losers, and provide a setting for the baser forms of modern civil disputes.
The special status of land in history and culture serves to intensify these disputes.
With so much at stake and so many pressing environmental and socioeconomic chal-
lenges inextricably linked to land and land use change, future progress requires clear
economic insight about the functioning of land markets and the drivers of land use
behavior. Economists have long offered explanations about why land decision makers
behave suboptimally and how policy might redirect these decisions to enhance social
welfare. These insights often involve measuring utility impacts outside of markets, alter-
ing incentives with policy change, and creating markets to improve the allocation of
society’s resources. Economists seek to understand how land markets adjust in the face
of policy changes and changes in relative scarcity of resources, anticipating opportuni-
ties to enhance the effectiveness of policy. As research over the past few decades has
shown, explaining the processes of land use change poses great challenges because of
the simultaneous cause and effect of price changes in land markets and the oft-times
confounding role of local policies. Land economics covers more than explanation.
2 Joshua M. Duke and JunJie Wu
Economists use recent advances in theory and methods to predict the likely impacts of
novel and unimplemented policies. The large set of recent developments in land eco-
nomics warrant a new handbook for the field.
This handbook draws broadly from advances that investigate land use behavior,
markets, and policy, showing that land is a theme that integrates several fields of
economics. These fields include natural resource economics, environmental eco-
nomics, regional science, and urban economics. The emergence of the new eco-
nomic geography and the increasing recognition of the role of natural endowments
and amenities in determining urban development patterns and the spatial distribu-
tion of economic activities has led to a blurring of the lines among the traditional
fields, with land use and land use patterns as an integrating theme. The alignment of
interest and the development of spatial modeling approaches have made the poten-
tial gains from collaboration and cross-fertilization across fields much greater. One
goal of this handbook is to stimulate further collaboration and cross-fertilization
among the economics fields related to land use markets, behavior, patterns, and
policy.
This handbook presents studies of land use and land use changes from various eco-
nomic perspectives. Several other disciplines also take land use as their subject mat-
ter of study, including geography, anthropology, and sociology. What distinguishes
economics from those disciplines is that land economists largely focus on explain-
ing the economic incentives or institutions that drive land use behavior and policy.
Land economics investigates the benefits and costs of land use decisions and change.
These benefits and costs are broadly defined to include those associated with both
economic and ecological impacts, as well as the feedback effects from those systems.
Land economics emphasizes economic efficiency in land allocation. As the reader will
discover, many chapters in this handbook will be distinguished by whether the authors
assess efficiency in partial and general equilibrium settings. These approaches involve
tradeoffs in explaining the substantive and complex problems associated with land.
When benefits are difficult to measure, land economists often turn to cost effectiveness
in achieving a goal when evaluating a policy.
This handbook is organized into four sections. The first section investigates the major
drivers of land use behavior and land use change. The second section evaluates the envi-
ronmental and socioeconomic implications of these forces, including chapters focusing
on the impact of land use change on water, habitat, climate, and other ecosystem ser-
vices. The third section presents recent methodological advances in land market mod-
eling, involving spatial modeling techniques, agent-based approaches, econometric
methods, quasi-experiments, and experiments. The fourth section focuses on the per-
vasive set of institutions from law and policy that direct land use behavior and change.
The handbook concludes with a discussion of future research directions in land eco-
nomics. The remaining parts of this introduction establish a setting and offer a brief
overview of each section, in turn.
Introduction 3
2. Determinants and Drivers of Land Use

and Land Use Change
Numerous societal changes, including economic development, technological progress,

and urbanization, drive land use change. Land use change in turn influences societal
changes. The complexity of the relationships among land use, societal change, and the
spatial distribution of economic activity have been increasingly recognized with the
emergence of the new economic geography and new growth theory. Consequently,
much effort has been devoted to the development of spatially explicit models that iden-
tify the nature of forces that shape the spatial distribution of economic activities and
land use patterns.
The first section of this handbook reviews the economic literature for recent advances
in the analysis of the major drivers of land use change. It starts with Chapter 1 by
Partridge and Rickman, which examines the relationship between economic develop-
ment and land use change. The chapter contends that two largely distinct literatures
have emerged in regional economic development and land use economics, despite their
fundamental interrelationship. Partridge and Rickman argue that a lack of integration is
a shortcoming in both approaches and that a spatial equilibrium framework is especially
suited for a systematic understanding of the various feedback mechanisms that affect
development and land use.
Technological progress is another major force driving land use change. The astonish-
ing increase in agricultural productivity enables the consistently decreasing number of
farmers to feed the consistently increasing number of people in the world. Technological
advances in seed varieties, irrigation, and fertilizers, among many others, led to these
productivity gains. Although the negative impacts of intensive agriculture may garner
disproportionate attention in the developed world’s popular press, it is difficult to over-
state the importance of technology in land use—especially when one recognizes that a
more technologically efficient, intensive use of farmland directly impacts outcomes in
urban land markets (which will be encouraged to grow up, not out) and forests (which
will be less likely to be converted to agriculture). In Chapter 2, Zilberman, Khanna,
Kaplan, and Kim offer a comprehensive assessment of technology and land use. They
focus on adoption as an investment, and present results associated with the threshold
model, input use efficiency, diffusion, credit, learning, and risk.
Land use change is ultimately determined by the relative value derived from the alter-
native uses. Since Ricardo and von Thunen, economists have sought to understand the
determinants of land values and land use patterns. Modern manifestations of this work
include explaining urbanization and suburbanization processes and the external costs
from development at the fringe, which is the economic approximation of the wide-
spread notion of “sprawl.” In Chapter 3, Mills discusses the role of the metropolitan area
in explaining urbanization and suburbanization and assesses the impacts of proximity
and density on the costs of various land uses. Mills’s discussion of economies of scale
and scope and of congestion and pollution helps build a theory on the size and growth of
metropolitan areas, predicting that suburbs will continue to grow rapidly, at least in the
United States.
The relative value derived from alternative land uses is ultimately affected by the rela-
tive value of services or outputs from the land uses. Land can be used to produce bio-
energy crops. As returns to energy production increase either due to market forces or
policy changes, more land will be allocated to energy production. In Chapter 4, Khanna,
Zilberman, and Crago evaluate the phenomenon of biofuels—an emerging issue
that has propelled land allocation into public debates about “food or fuel.” The mod-
els reviewed directly link land markets with energy markets; interestingly, these mod-
els show that government policies other than zoning can trigger substantive land use
changes. Although the reviewed models lack agreement about the extent of land use
change needed to meet governmental biofuel targets, Khanna, Zilberman, and Crago
anticipate a moderate increase in crop prices. Technology will mitigate some of the
anticipated adverse impacts from large-scale biofuel production.
In Chapter 5, Nickerson and Zhang tackle perhaps the longest standing challenge in
land economics: explaining farmland value. They focus on hedonic estimations of mod-
els in which land rents are capitalized, considering both cross-sectional and dynamic
analyses. Nickerson and Zhang review the tools for addressing spatial dependence,
spatial heterogeneity, and sample selection bias, and then discuss recent innovations
in nonparametrics, quasi-experimental design, panel data, and structural econometric
models.
In the developing world, agriculture comprises a comparatively large part of the
economy and remains labor-intensive. The expansion of cultivation in the developing
world continues to decrease forests and other natural land uses. In the final chapter of
this section (Chapter 6), Barbier considers whether this farmland expansion will lead
to the same level of economic development experienced from similar patterns in the
past. Barbier models the processes of this “frontier economy,” in which a traditional
sector converts available land to produce a nontraded agricultural output, and a fully
developed, commercially oriented sector exploits available land and natural resources
for a variety of traded outputs. The model accounts for population increases, migra-
tion, and unskilled labor. The results suggest that although the frontier can help mit-
igate the adjustments of economic growth, it also induces considerable costs, such as
those associated with boom-and-bust cycles. The results provide a plausible explana-
tion for why land use expansion in developing economies may not be generating greater
economy-wide benefits.
Together, the chapters included in this section provide a critical assessment of the
recent analyses of the major drivers of land use change. They also lay a foundation for
understanding the environmental and socioeconomic impacts of land use change,
which are explored in the next part of this handbook.
Introduction 5
3. Environmental and Socioeconomic

Consequences of Land Use and
Land Use Change
Land use changes, such as deforestation, urbanization, intensification of agriculture,

and innumerable other human activities have substantially altered the Earth’s landscape.
Such disturbances affect important ecological processes and the provision of ecosys-
tem services, causing wide-ranging and long-term environmental and socioeconomic
consequences. The second section of this handbook presents economic research efforts
that examine the environmental and socioeconomic impacts of land use and land use
change. It shows the breadth of research in this area, and highlights the need for more
economic research that focuses on the socioeconomic impacts of land use and land use
change.
3.1 The Land Use and Environment Nexus

Land use is arguably the most pervasive socioeconomic force affecting the processes
and functions of ecosystems. Forests provide a clear example of the relationship
between land use and ecosystem services. Forests support biodiversity, provide critical
habitat for wildlife, hold carbon dioxide out of the atmosphere, intercept precipitation,
slow surface runoff, reduce soil erosion, and mitigate flooding. These ecosystem ser-
vices significantly affect human well-being, although they are rarely priced by markets.
Substantive welfare losses may result when the ecosystem services of forests are ignored
in decisions to convert forest to agriculture or urban development. In one current and
leading example of these phenomena, deforestation substantially alters and fragments
the Earth’s vegetative cover. Such disturbance can change the global atmospheric con-
centration of carbon dioxide, the principal heat-trapping gas, and is suspected to affect
local, regional, and global climate by changing the energy balance on the Earth’s surface
(Marland et al. 2003).
Agriculture is a dominant form of land use, with 38% of land in agricultural uses
globally (Food and Agricultural Organization [FAO] 2004). Agro-ecosystems gener-
ate beneficial ecosystem services, including food and materials for human consump-
tion, but intensive agriculture can have a wide range of negative ecological impacts. For
example, it has long been recognized that agricultural land use and practices can affect
water quality and quantity, and the effect is influenced by government policies. Soil ero-
sion and nutrient runoff from agricultural lands are a leading source of water pollution
both in inland and coastal waters.
Urban development has also been linked to many environmental problems, including
air pollution, water pollution, and loss of wildlife habitat. Urban runoff often contains
nutrients, sediment, and toxic contaminants, and it can cause large variations in stream
flow and temperatures. Habitat destruction, fragmentation, and alteration associ-
ated with urban development have been identified as the leading causes of biodiversity
decline and species extinctions (Czech et al. 2000; Soulé 1991). Urban development and
intensive agriculture in inland and coastal areas damage the health, productivity, and
biodiversity of the marine environment throughout the world.
The first five chapters of the second section present economic research efforts to
understand the linkages between land use decisions and environmental outcomes.
Chapters 7, 8, and 10 all address habitat conservation, but with three different foci. In
Chapter 7, Lewis and Nelson evaluate the effectiveness of three leading approaches to
securing the public goods associated with wildlife conservation: regulation, direct pur-
chase, and incentive-based policy. They evaluate the challenges to these policies, includ-
ing the perverse result of preemptive habitat destruction, or “shoot, shovel, and shut up,”
and issues arising from spatially dependent benefits.
In Chapter 8, Johnston et al. offer a review of the most current methods for deter-
mining the nonmarket values of ecosystem services. The authors show that economists’
focus on valuation methods—that is, how ecosystem services are evaluated—is only
one part of a very complex process. Researchers must also determine what ecosystem
services are to be evaluated and at what scope and scale. These latter two challenges
involved many uncertainties, including unknown scientific information on the pro-
cesses and the linkages among various services.
In Chapter 10, Attavanich et al. develop an integrated model to predict the joint
effect of climate change and resulting land use responses on a specific ecosystem
service (waterfowl productivity). Land use change in a specific location (the Prairie
Pothole Region of North America) is explicitly modeled as a function of climate
change. One important finding from this analysis is that land use response to cli-
mate change exacerbates the direct negative effects of climate change on waterfowl
populations.
In Chapter 9, McCarl et al. present a thorough assessment of what is known about
land use and climate change. Although climate change is an area with massive uncer-
tainties, economists offer a great deal of recent research on how climate change and land
use interact; for instance, this research predicts how agricultural growing regions will
alter with a warming planet. McCarl et al. divide the research results into three types of
studies: vulnerability, adaptation, and mitigation research.
In Chapter 11, Montgomery provides a comprehensive treatment of the topic of
fire as an agent and a consequence of land use change. The chapter presents the lit-
erature on the economics of fire management, institutions, and policy and examines
emerging challenges for fire policy. Montgomery also discusses the three core themes
in the economics of wildfire: spatial externalities, incentives, and risk-based decision
analysis.
These chapters highlight two challenges for the evaluation of environmental impacts.
The first is related to the challenges of linking specific ecosystem functions to land use deci-
sions. Two of the chapters offer examples of how economists are beginning to overcome
Introduction 7
this challenge. Johnston et al. offer a bioeconomics model linking specific land manage-
ment decisions with bird habitat outcomes for a specific species (bobolink) in a specific
location. Similarly, Attavanich et al. develop an integrated model to predict how a specific
ecosystem service (waterfowl productivity) will change as a result of land use changes.
The second challenge is related to the problem of asymmetric information, a theme
arising throughout the handbook. Chapter 7 addresses this challenge in the context of the
design of incentive-based policies for habitat conservation. Several other chapters study
various manifestations of this problem (see for instance Chapter 15 on empirically mod-
eling landowner returns, Chapter 19 on private information in conservation auctions,
and Chapter 26 on landowner information on reservation value and land assembly).
This information asymmetry problem refers to the policy maker’s inability to design
first-best or cost-effective mechanisms when landowners have private information
about their willingness to accept compensation to provide ecosystem services. With
constrained budgets, policy makers would prefer to secure the greatest ecosystem ser-
vices supply possible, which in theory means paying each landowner their minimum
willingness to accept compensation. However, information asymmetry prevents policy
makers from targeting the least-cost suppliers, manifesting as at least three overlapping
problems examined in recent literature. First, fiscal inefficiency occurs when a land-
owner is paid more than his or her minimum willingness to accept because the policy
maker cannot sort by types. For instance, Kirwan et al. (2005) find evidence that 10–40%
of the US Conservation Reserve Program expenditures were rent premiums. Second,
adverse selection occurs when a landowner is paid for supplying an ecosystem service
even though he or she would supply that service in the absence of the policy. Third, addi-
tionality is not achieved when a current supplier of ecosystem services is credited for
future supply, even though that supply currently exists and would likely continue had
no policy been implemented. Much research has explored solutions to these problems
for the design of conservation and environmental policies. For example, economists
have developed contracts to achieve second-best outcomes in the face of these infor-
mation problems when targeting land for conservation (Smith 1995; Wu and Babcock
1996). Several methodological chapters in this handbook address economic techniques
to predict and/or sort by types, given the underlying censoring from this information
asymmetry.
3.2 Socioeconomic Impacts of Land Use Change

Perhaps the most visible social impacts of land use change are those associated with
urbanization and suburbanization, which affects both urban and rural communi-
ties. As more people leave rural areas to live in cities, demand for housing increases,
congestion intensifies, and urban air quality declines. All these changes will drive up
housing prices in desirable locations, which some see pushing “the new labouring poor
into great morasses of misery outside the centres of government and business and the
newly specialised residential areas of the bourgeoisie” (Hobsbawm 1962, Chapter 11).
The locational pattern of different income groups and community characteristics, such
as economic segregation and jurisdictional fragmentation, are strongly influenced by
the spatial distribution of environmental amenities (Wu 2006). For example, it has been
suggested that “the almost universal European division into a ‘good’ west end and a
‘poor’ east end of large cities” documented by Hobsbawm (1962, chapter 11) is “likely
due to the prevailing south-west wind which carries coal smoke and other airborne pol-
lutants downwind, making the western edges of towns preferable to the eastern ones.”
As congestion associated with urbanization surpasses an acceptable cost level, a
reverse pattern of migration, known as suburbanization, may occur. In fact, during the
past 50 years, the proportion of the US population living in suburban areas increased
from about one-third in 1960 to 63% in 1998 (US Department of Housing and Urban
Development [USDHUD], 2000). With suburbanization, cities tend to gain lower
income residents and lose upper income population, causing income segregation and
economic disparities between urban and suburban communities to manifest and inten-
sify. From 1969 to 1998, the share of low-income families in central cities grew from
21.9% to 25.5% compared with a decline from 18.3% to 16.6% for high-income house-
holds (USDHUD, 2000). The change in income mix led to a smaller tax base and more
need to finance social services in urban communities. Wu (2010) developed a spatially
explicit model to investigate how urban and suburban communities evolve differently
with changes in local economic fundamentals such as rising income or falling commut-
ing costs. The model highlights the importance of environmental amenities and the
economy of scale in the provision of public services as determinants of urban spatial
structure.
Urbanization has also changed rural communities. In some areas, migration to cit-
ies has turned once-viable rural communities into ghost towns. In other rural areas,
urban sprawl has encroached to such an extent that the community itself has been lost
(Wu et al. 2008, vii). Urbanization also presents challenges for farmers on the urban
fringe, especially those who lease and therefore do not benefit from land appreciation.
As neighboring farms are converted to development, farmers will no longer be able to
take advantage of economies of scale from information sharing and business relation-
ships with neighboring farmers. Urbanization may also cause the “impermanence syn-
drome,” leading to a reduction in investment in new technology or machinery or idling
of farmland (Lopez et al. 1988).
As urbanization intensifies, agricultural and nonagricultural land use conflicts
become more severe (Lisansky 1986). This may lead to an increase in local ordinances
designed to force farmers to internalize some of the negative externalities normally
generated by agriculture. As the nearest input suppliers close because of insufficient
demand for farm inputs, a farmer may have to pay more for inputs or spend more time
to obtain equipment repairs (Lynch and Carpenter 2003; Wu et al. 2011). Competition
for labor from nonagricultural sectors may raise farmers’ labor costs. When the total
amount of farmland falls below a critical mass, the local agricultural economy may col-
lapse (Wu et al. 2011).
Introduction 9
Urbanization also presents opportunities to farmers. The emergence of a new cus-

tomer base provides farmers with new opportunities for higher value crops. For exam-
ple, vegetable producers receive higher prices in urbanized areas (Lopez et al. 1988).
Many farmers have shown remarkable adaptability in adjusting their enterprises to take
advantage of new economic opportunities at the urban fringe. They farm more inten-
sively in areas with high population density (Lockeretz 1988). More than half the value
of total US farm production is derived from counties facing urbanization pressure
(Larson et al. 2001).
Although there is strong evidence that land use change affects social structures in
both rural and urban communities, relatively few studies have focused on the socioeco-
nomic impacts of land use change. Nechyba and Walsh (2004) recognized this gap. After
an extensive review of the literature, Nechyba and Walsh point out that, although many
previous studies have investigated the drivers of urban development, relatively few
have examined how city landscapes evolve within expanding boundaries. Chapter 12
by Stone and Wu, focuses on a fundamental problem in analyzing the socioeconomic
impacts of urbanizing land use: household location choice. Stone and Wu first survey
the most significant developments in theory and analyses that examine the interactions
among household location decisions, land use patterns, and municipal profiles and then
explore strategies to model these interactions using a case study from Portland, Oregon.
We hope this chapter will help stimulate more research on the socioeconomic impacts
of land use change.
4. Methodological Developments
The increasing importance of the environmental and socioeconomic issues associ-

ated with land use and the increasing complexity of land policy has led researchers to
develop ever more sophisticated methods. Section III of this handbook addresses six
cutting-edge approaches in three general categories. The section also includes a syn-
thetic chapter critically reviewing methodological advances.
Spatial econometric analysis is one approach presented. The most recent spa-
tial methods seek to maximize the information potential of explicitly spatial data.
Economists have developed conceptual frameworks that attempt to address the spa-
tial characteristics of benefits and costs of land use changes in tandem with the natu-
ral scientists, who develop spatial models of resource systems but also a variety of new
methodologies to analyze spatially explicit data. A second approach is simulation,
including spatial-equilibrium and agent-based methods. A third approach is inferen-
tial and experimental, using reduced-form and structural econometric models, to better
understand the drivers of land behavior, tease out causation, and predict hidden policy
impacts.
The initial, synthetic chapter of this section, by Irwin and Wrenn, pro-
vides an overview and assessment of the main methods used to model spatially
explicit data on land use and land use change. The chapter offers a valuable com-
parison of reduced-form and structural econometric models. It also compares
spatial-equilibrium and agent-based simulation. Irwin and Wrenn provide a critical
assessment of three important questions: what are the advantages and disadvantages
of these various empirical approaches to modeling land use and land use change?
Which questions are best suited to be answered using one versus the other approach?
And, where are the gaps in the current literature? Chapters about specific modeling
approaches follow this overview.
4.1 Spatial Econometric Methods in Land Use

During the past two decades, spatial econometric methods have matured into a
formal, insightful, and widely used method for assessing land use and land policy.
Cho, Kim, and Roberts, in Chapter 17, provide an overview of the state-of-art spa-
tial econometric methods and a comparison of different approaches. The chapter
offers an application for predicting development rates in Nashville-Davidson County,
Tennessee.
Chapter 14, by Klaiber and Kuminoff, provides a comprehensive treatment of the
equilibrium sorting methodology. It clarifies the relationship between an equilibrium
sorting model, in which households make location decisions as well as being spatially
explicit and characterizing household preferences, and a reduced-form hedonic model
that does not involve sorting. The chapter presents a detailed description of the econo-
metric procedures for estimating equilibrium sorting models and the simulation proce-
dures for policy evaluation.
4.2 Simulation Methods in Land Use

Two chapters describe simulation methods. Plantinga and Lewis, in Chapter 15,
describe econometric-based landscape simulation models for policy evaluation in a
spatially heterogeneous landscape. They identify modeling challenges such as captur-
ing the variation in private returns to land use at the right analytical level and the private
information landowners possess about these returns. Plantinga and Lewis also present
an application that connects land use (shoreline development in Wisconsin) with habi-
tat provision (green frog population).
In Chapter 16, Parker offers a very different simulation method: agent-based mod-
eling. In a spatially explicit virtual landscape, these models capture the decisions and
interactions of economic agents. Parker reviews this approach in terms of model attri-
butes, computational issues, and the questions the models can answer. The models are
well positioned to address, jointly, spatial and agent heterogeneity.
Introduction 11
4.3 Experimental Methods in Land Use
Two chapters describe experimental methods, both of which can inform the likely effec-
tiveness of land policies without necessarily having to evaluate an existing policy in a
specific location. In Chapter 18, Towe, Lewis, and Lynch, discuss the methods and chal-
lenges of quasi-experimental econometric estimation for evaluating land policies. It
has long been recognized that selection issues confound inference of policy impacts;
simply, land outcomes cannot be modeled as a result of a policy because these policy
treatments were not randomly assigned. But methods for addressing selection prob-
lems have continually advanced. Towe et al. describe these inferential problems and how
one approach, the propensity score matching method, can solve them. Advantages of,
challenges with, and steps to be taken when employing the propensity score matching
method are explained. The chapter also provides a detailed application that analyzes
how a smart growth policy affects land development outcomes in Maryland.
In Chapter 19, Messer, Duke, and Lynch present recent uses of laboratory and field
experiments to inform land use and policy problems. The authors develop a framework
to understand the tradeoffs in experimental control, problem context, and the represen-
tativeness of the participants to actual land decision makers. An application investigates
the impact of different types of information on the performance of reverse auctions for
ecosystem services. The results suggest that different levels of public information affect
sellers’ bidding behavior as well as auction competitiveness. Overbidding and too little
market competition leads to significant auction efficiency loss.
5. The Economics of Land Use

Law and Policy
Land use provides many economic benefits and costs that are not figured into the pri-
vate landowner’s decisions. These externalities lead to an inefficient allocation of land
uses. Land market inefficiencies take diverse forms on the ground. For instance, devel-
opers may not bear all the environmental and infrastructural costs generated by their
projects. Natural land owners do not enjoy all the social benefits they supply. Owners of
small urban parcels hold out, thereby preventing optimally sized redevelopments and
driving economic activity to the suburbs. Other market failures also characterize land
markets—for instance, given that location makes many land uses perfectly heteroge-
neous, imperfect competition may arise.
Politicians, legal scholars, planners, policy makers, and the general public have long
understood the problems associated with these land market failures—even though
efficient resource allocation is probably not driving their thinking—because these fail-
ures often mirror readily understood notions of appropriate neighborly behavior and
the interdependencies of modern life. In other words, people often believe that neigh-
bors act inappropriately when they make decisions with negative externalities, foisting
unwanted costs on them. Of course, Coase (1960) reframed the way economists think
about causation in externalities, providing a recognition that two parties are competing
for use of the same resource (see Duke 2004). But part of the innovation of Coase (1960)
was to help economists recognize that the conventional explanation of external costs
was incomplete.
The conventional story of land use conflict persists in the public imagination, despite
Coase’s (1960) efforts and those of the economists who followed. Residential voters and
their representatives will support local zoning, recognizing that external costs of mix-
ing commercial and residential land uses hurts their property values. Similarly, many
will support revitalization of an urban brownfield, not to prevent “sprawl” and conver-
sion of greenfields, but instead because they envision a future in which the local urban
economy is revitalized and land is “clean.” Policies to promote provision of positive land
use externalities may be more recent, but preservation and conservation seem to “make
sense” to many members of society. The alignment of economic rationale and the way
the general public thinks about its well-being has likely led to the developed world’s long
history of land use policy.
Duke and Lynch (2006) derive a framework to explain the different forms land con-
trols may take in the context of land retention. Some controls are regulatory, such as
zoning. Other controls are incentive-based, such as an impact fee (tax) on new resi-
dential development to fund sewers. Techniques such as conservation easements are
best framed as participatory, rather than incentive-based, because the public or pri-
vate demander secures positive externalities by triggering demand in a market for a
less-than-fee right in land (say, a negative easement). In other words, the easement mar-
ket always existed, but it was the newly created demand from government or a private
group that created the viability of this “new” market for conservation. A final set of con-
trols is a hybrid of two of the preceding; for instance, a transferable development right
program is part regulatory (the cap) and part incentive-based (the trading).
The aforementioned types of land use control, in effect, establish the specific markets
for land. Framed differently, any land unit may be sold into various overlapping and/or
mutually exclusive land use markets. For instance, one parcel of farmland may be sup-
plied in agricultural land use markets and conservation markets, but that same parcel
may not be supplied in preservation and developed use markets. Institutions establish
the property rights that define markets (Schmid 1999). Resource allocation efficiency
is a function of the prevailing institutional arrangements (Bromley 1989, c hapter 5).
Once rights are established, each policy can be assessed for its potential efficiency impli-
cations. Note also that all rights associated with land are not assigned, and remaining
externality conflicts arise from the absence of rights (Duke 2004).
Legislatures and quasi-judicial bodies create these land markets, and courts sanction
and refine the allocation of rights. As this section clarifies, land economists have partici-
pated in the land use policy debate in several ways. The unifying theme, however, is that
work in this area tends to be applied. Economists’ applications focus on specific policies
Introduction 13
(conservation easements, zoning, etc.) or specific areas of law (regulatory takings, emi-
nent domain, etc.). A common result is that there are unintended consequences; seeking
to solve one failure can trigger substantial welfare losses in the form of higher housing
prices, smaller houses, and inefficient land use patterns (Cheshire and Sheppard 2002;
Hascic and Wu 2012).
Most economic attention falls on incentive-based and participatory policies because
these policies are new and match most economists’ underlying aversion to inflexible
regulations of any type. The incentive-based and participatory approaches seemingly
have many advantages over direct regulatory land use control. A development impact
fee can be used to achieve both the optimal pace and pattern of land development, a
shortcoming of zoning regulations (Wu and Irwin 2008). However, zoning may be pre-
ferred from a practical viewpoint, as well as in cases in which the environmental costs
of land conversion are highly uncertain. Zoning may also be preferable when one also
considers the costs of implementing policies because regulations are less expensive than
many participatory policies (Johnston and Duke 2007). In situations in which the natu-
ral and human systems interact in complex ways, thresholds and nonlinear dynamics
are likely to exist, and the environmental costs could be very high and sensitive to land
development. In such cases, zoning may be preferred. The policy challenge, however, is
to know when the system is in the neighborhood of such thresholds.
Although federal spending on land-related conservation programs, such as the
Conservation Reserve Program (CRP) and the Wetland Reserve Program (WRP), has
increased substantially over the past 25 years, the federal government has yet to articu-
late a clear vision of how land use should be managed. Most land use controls are in
the hands of local governments, and the level of government involvement in land use
planning and regulation varies considerably across counties and municipalities in the
United States. Some local governments have few land use controls, whereas others are
actively involved in land use planning and regulation.
The forces of urbanization have motivated many local governments to impose strict
land use control. Economists research whether these policies achieve their goals and
how they impact associated markets. Evidence suggests that some of the efforts have
successfully slowed development. For example, Wu and Cho (2007) found that local
land use regulations reduced the total supply of developed land by 10% in the five west-
ern states between 1982 and 1997, with the largest percent reduction in Washington
(13.0%), followed by Oregon (12.6%), California (9.5%), Idaho (4.7%), and Nevada
(2.8%). Yet a predictable but unintended consequence of land use regulation is higher
housing prices, which make housing less affordable to middle- and low-income house-
holds (Glaeser and Gyourko 2002; Cho et al. 2003; Glaeser and Ward 2006).
Private trusts and nonprofit organizations play an increasingly important, albeit
uncoordinated, role in the mix of local and federal land use through their efforts
to promote land conservation. For example, the Nature Conservancy (2013) has
helped to protect approximately 15 million acres of ecologically important lands
in the United States. However, some have questioned whether private conservation
efforts crowd out or complement public efforts for land conservation (Albers
et al. 2008).
Most land use controls prove contentious, especially in areas facing rapid urbaniza-
tion. The simplistic view on land use control does not capture the complex motivations
driving the decisions of the public and government. In this view, proponents envision
protection of farmland, forests, water quality, open space, and wildlife habitat. They
anticipate increases in property values and human health. Opponents argue that urban
development is an orderly market process that allocates land from agriculture to urban
use and that governments tend to overregulate because they rarely bear the costs of reg-
ulation (Hascic and Wu 2012).
A more complex perspective recognizes that land use controls generate both benefits
and costs, and, in most cases, create both winners and losers, at least, in the short term.
Each side attempts to marshall the forces of “good” (clean environment, job creation,
good schools, health, etc.) against the other side’s “evil” (pollution, job destruction,
crime, etc.). Both sides recognize that any policy measures that aim at curbing urban
development will ultimately affect a key element of the traditional “good” life—such as
the ability to consume a large amount of living space at affordable prices.
Economists have much to contribute to the debate. Social welfare accrues from many
sources, so the grip of advocacy need not necessarily determine the outcomes of the
analysis. Conflict arises from poorly designed incentives and an absence of markets.
Information asymmetry is rampant, driving many conflicts, and a poorly designed pol-
icy does not overcome this challenge. Policy makers ought to resist the temptation to
attribute all “irregular” land use patterns to market failures and impose stringent land
use regulations that may hinder the function of market forces. They should try to iden-
tify and understand the sources of market failures—such as those that cause “excessive
development”—and address problems at their roots.
Part IV of this handbook presents seven chapters that analyze the economics of land
use law and policy. The first four chapters disentangle the economics of land conser-
vation and preservation, which has emerged as an increasingly important tool in the
past several decades as land use regulations have waned. Chapter 20, by Gnedenko and
Heffley, presents a rich open-city model of the tax, spending, and land use zoning pol-
icies of local government and applies the model to analyze the impact of open space
preservation on local land use and community characteristics. The chapter provides an
applied analysis, suggesting that policies that seemingly promote open space may in fact
work against that goal.
In Chapters 21–23, land conservation policies in the United States and the
European Union are explained and compared. Ferris and Lynch, in Chapter 21,
categorize US conservation policies into the four-part scheme of Duke and Lynch
(2006)—the categorization described earlier. The authors use this categorization to
organize the economics literature on US conservation and derive synthetic results on
Introduction 15
the effectiveness of each approach. Hodge, in Chapter 22, conducts an assessment of

European agri-environmental policies, focusing on the voluntary contracts in place
since the mid-1980s under the Common Agricultural Policy. Hodge’s analysis con-
siders policy challenges with asymmetric information and transaction costs, and he
concludes that future policy will need to enhance payment targeting, competition
among suppliers, coordination, and security of the environmental benefits obtained.
Chapter 23, by Claassen et al. complements Chapters 21 and 22 with a systematic
comparison of US and EU agri-environmental policies. In both locations, these poli-
cies began in earnest in the 1980s, and both employ a mix of three types of poli-
cies: incentive-based, regulatory, and cross-compliance. Although the policies have
similarities, most economists tend to focus their research on a policy in one loca-
tion or the other. The chapter focuses on a comparison of the economic research
results on the effectiveness of these policies and the data used by economists in these
studies.
Chapters 24–26 turn to the economic analysis of the legal institutions of land use.
These chapters focus on law and economic problems about the limits of permissi-
ble government control of land in the US context. In Chapter 24, Eisen reviews the
broader context of “brownfields” redevelopment in urban areas, analyzes developers’
brownfields development decisions, and assesses how state and federal laws affect the
decisions. Contamination imposes a stigma on these sites, and stigma is affected by the
cleanup policies employed. Information problems are highlighted, as are novel incen-
tives to overcome stigma, including voluntary programs, public input, and the lifecycle
impacts of remediation.
Chapters 25 and 26 investigate takings, both regulatory and eminent domain. Miceli
and Segerson in Chapter 25 explain the legal and economic theories determining when
a regulation crosses the compensation threshold, thus becoming a compensable tak-
ing. They explain the seminal Blume, Rubinfeld, and Shapiro (1984) model of takings,
the connection to regulatory takings, and the surprising result of efficient zero com-
pensation. Then Miceli and Segerson interpret and expand this model with a review of
decades of subsequent work. Collectively, their chapter clarifies the efficiency implica-
tions of various compensation rules.
In Chapter 26, Duke assesses the economic literature on eminent domain, focusing
solely on physical appropriations as opposed to the nonpossessory actions described
in Chapter 25. Rather than assessing efficient compensation—which is well covered in
Chapter 25—Duke examines two other themes in the economic literature on eminent
domain. First, eminent domain solves inefficiencies in land assembly: holdouts and the
provision of inefficiently low levels of public goods from urban redevelopment. The
chapter builds a model of information asymmetry in assembly markets, then compares
the conditions under which eminent domain does and does not enhance efficiency rela-
tive to market assembly.
6. Future Research Directions
This handbook concludes with a synthetic chapter (Chapter 27) on future research
directions in land economics. All the chapters offer assessments of where research is
going—or should be going—in the areas covered. Duke and Wu assimilate these sugges-
tions and predictions into five general directions for future work.
The first direction involves spatially explicit structural modeling. The interde-
pendence of land use patterns and economic growth highlights the need for spatially
explicit structural modeling. Such a modeling approach better explains economic per-
formance, the distribution of economic activity in a region, the impact of shocks, and
poverty. Although economists are accustomed to structural models, a great deal of the
current empirical work relies on reduced-form models. Irwin et al. (2009) call for struc-
tural modeling to better identify the potential causal linkages among the many interde-
pendent processes that affect urban-rural growth.
The second direction is toward greater integrated economic and ecological modeling.
Integrated modeling gives economists a way to increase explanatory power by linking
economic models with quantitative modeling efforts in different economic fields and in
noneconomic disciplines. Integration may include linking land use and development or
linking land use and ecosystem services. Increasingly, economists’ research interests are
aligned with questions from other scientists. For instance, the linkage of economic and
ecological systems offers great promise, with many economists interested in research
questions that build on or are nested with models traditionally addressed by ecologists,
hydrologists, and other natural scientists.
Advancing methods to understand and uncover agents’ behavior offers a third direc-
tion. The methods chapters cover spatial, econometric, simulation, and experimental
approaches. Most of these methods have been developed to better understand selection
issues. In other words, the methods allow economists to understand agents’ interactions
and decisions without necessarily observing, in a given location, a real-world policy or
a real-world market. Inferences can be made without bias. Collectively, economists are
becoming better able to understand land use behavior and phenomena.
A fourth direction involves land economists’ efforts to build models that best
employ newly abundant spatial and other land data. Many chapters highlight new
sources of data on land prices, uses, and services. Despite their increasing abundance,
these data are often incomplete and inconsistent. There are many suggestions for how
to use these data, but the chapter authors also identify areas in which some measures
need to be developed, where data need to be collected, and where datasets ought to be
linked.
The fifth and final direction concerns economists’ efforts to overcome information
challenges in policy design. Not surprisingly, economists see spatial analysis playing a
key role in future analyses. Many chapters suggest moving toward more structural mod-
eling or various experimental methods to draw more broadly applicable results on land
Introduction 17
use policy. Several of the chapters also suggest increasingly sophisticated approaches to
the underlying information problems that prevent the creation of first-best policies. In
large part, information problems in land manifest as incentive problems in voluntary
policies—especially in agri-environmental policies. Land economists see great oppor-
tunities to improve policy performance with respect to additionality, leakage/slippage,
and other incentive problems. Some chapters recommend that economists take a step
back from the focus on solving these problems and, instead, encourage research about
how multiple policies interact and how policy/market complexities affect performance.
7. Conclusion
This handbook is framed with the idea that an integrated approach to land use econom-
ics is needed. Why is this approach needed? First, partial equilibrium analysis is not
always adequate to examine the questions society needs answered. Second, land eco-
nomic problem settings are often too fluid to warrant the simplification economists
seek to derive tight and tractable results, ready lab experiments, and empirically testable
theoretically derived results. Third, integrated work may help prevent unexpected sub-
optimal recommendations.
Integration can occur within economics, but across fields. Or, it can occur between
economic and noneconomic models. Even within the discipline, greater recognition and
integration stimulates cross-fertilization between the fields of land economics research.
By providing a comprehensive survey of land-related work in several economics fields,
we hope this handbook will provide the basic tools needed for new and established land
economists to redefine the scope and focus of their work, to better incorporate the con-
temporary thinking from other fields, and to push out the frontiers of land economics in
the areas identified.
References
Albers, H. J., A. W. Ando, and M. Batz. 2008. Equilibrium patterns of land conserva
tion:Crowding in/out, agglomeration, and policy. Resources and Energy Economics 30(4):
492–508.
Attavanich, W., B. S. Rashford, R. M. Adams, and B. A. McCarl. 2014. Land use, climate change,
and ecosystem services. In The Oxford handbook of land economics, eds. J. M. Duke and J.
Wu, 255–280. New York: Oxford University Press.
Barbier, E. B. 2014. Land use and sustainable economic development: developing world. In The
Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 139–159. New York: Oxford
University Press.
Blume, L., D. Rubinfeld, and P. Shapiro. 1984. The taking of land: When should compensation
be paid? Quarterly Journal of Economics 99: 71–92.
Bromley, D. W. 1989. Economic interests and institutions: The conceptual foundations of public

policy. Oxford: Basil Blackwell.
Cheshire, P., and S. Sheppard. 2002. The welfare economics of land use planning. Journal of
Urban Economics 52: 242–269.
Cho, S.-H., S. G. Kim, and R. K. Roberts. 2014. Spatial econometric modeling of land use
change. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 430–451. New
York: Oxford University Press.
Cho, S.-H., J. Wu, and W. G. Boggess. 2003. Measuring interactions among urbanization,
land use regulations, and public finance. American Journal of Agricultural Economics
85: 988–999.
Claassen, R., J. Cooper, C. Salvioni, and M. Veronesi. 2014. Agri-environmental policies: A
comparison of US and EU experiences. In The Oxford handbook of land economics, eds. J. M.
Duke and J. Wu, 648–667. New York: Oxford University Press.
Coase, R. H. 1960. The problem of social cost. Journal of Law and Economics 3:1–44.
Czech, B., P. R. Krausman, and P. K. Devers. 2000. Economic associations among causes of spe-
cies endangerment in the United States. BioScience 50: 593–601.
Duke, J. M. 2004. Institutions and land-use conflicts: Harm, dispute processing, and transac-
tions. Journal of Economic Issues 38(1): 227–252.
Duke, J. M. 2014. Eminent domain and the land assembly problem. In The Oxford handbook
of land economics, eds. J. M. Duke and J. Wu, 723–735. New York: Oxford University Press.
Duke, J. M., and L. Lynch. 2006. Four classes of farmland retention techniques: Comparative
evaluation and property rights implications. Land Economics 82(2): 189–213.
Duke, J. M., and J. Wu. 2014. Future research directions in land economics. In The Oxford hand-
book of land economics, eds. J. M. Duke and J. Wu. New York: Oxford University Press.
Eisen, J. B. 2014. Stigmatized sites and urban brownfield redevelopment. In The Oxford handbook
Ferris, J., and L. Lynch. 2014. Land conservation in the United States. In The Oxford hand-
book of land economics, eds. J. M. Duke and J. Wu, 583–611. New York: Oxford University
Press.
Food and Agriculture Organization (FAO). 2004. Statistics from www.faostat.fao.org, updated
February 2004.
Glaeser, E. L., and J. Gyourko. 2002. The impact of zoning on housing affordability. Harvard
Institute of Economic Research, Discussion Paper Number 1948. Accessed November 19,
2007. http://post.economics.harvard.edu/hier/2002papers/HIER1948.pdf
Glaeser, E. L., and B. A. Ward. 2006. The causes and consequences of land use regula-
tion: Evidence from greater Boston. Harvard Institute of Economic Research, Discussion
Paper Number 2140. Accessed November 19, 2007. http://www.economics.harvard.edu/
hier/2006papers/HIER2124.pdf
Gnedenko, E., and D. Heffley. 2014. Open space preservation: Direct controls and fiscal incen-
tives. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 547–582. New
Hascic, I., and J. Wu. 2012. The cost of land use regulation versus the value of individual exemp-
tion: Oregon’s measures 37 and 49. Contemporary Economic Policy 30: 159–214.
Hobsbawm, E. 1962. The age of revolution: 1789–1848. London: Weidenfeld & Nicolson.
Hodge, I. 2014. European agri-environmental policy: The conservation and re-creation of cul-
tural landscapes. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu,
612–647. New York: Oxford University Press.
Introduction 19
Irwin, E. G., and D. Wrenn. 2014. An assessment of empirical methods for modeling land use.
In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 327–351. New York:
Oxford University Press.
Irwin, E., K. P. Bell, N. E. Bockstael, D. Newburn, M. D. Partridge, and J. Wu. 2009. The econom-
ics of urban-rural space.” Annual Review of Resource Economics 1(October): 1–26.
Johnston, R. J., and J. M. Duke. 2007. Willingness to pay for agricultural land preservation
and policy process attributes: Does the method matter? American Journal of Agricultural
Economics 89(4): 1098–1115.
Johnston, R. J., S. K. Swallow, D. M. Bauer, E. Uchida, and C. M. Anderson. 2014. Connecting
ecosystem services to land use: Implications for valuation and policy. In The Oxford handbook
Khanna, M., D. Zilberman, and C. L. Crago. 2014. Modeling the land use change with biofuels.
Kirwan, B., R. N. Lubowski, and M. J. Roberts. 2005. How cost-effective are land retirement
auctions? Estimating the difference between payments and willingness to accept in the
Conservation Reserve Program. American Journal of Agricultural Economics 87(5): 1239–1247.
Klaiber, H. A., and N. V. Kuminoff. 2014. Equilibrium sorting models of land use and residen-
tial choice. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 352–379.
New York: Oxford University Press.
Lewis, D. J., and E. Nelson. 2014. The economics of wildlife conservation. In The Oxford handbook
Larson, J., J. Findeis, and S. Smith. 2001. Agricultural adaptation to urbanization in southeast-
ern Pennsylvania. Agricultural and Resource Economics Review 30: 32–43.
Lisansky, J. 1986. Farming in an urbanizing environment: Agricultural land use conflicts and
rights to farm. Human Organization 45: 363–371.
Lockeretz, W. 1988. Urban influences on the amount and structure of agriculture in the
north-eastern United States. Landscape and Urban Planning 16: 229–244.
Lopez, R. A., A. O. Adelaja, and M. S. Andrews. 1988. The effects of suburbanization on agricul-
ture. American Journal of Agricultural Economics 70: 346–358.
Lynch, L., and J. Carpenter. 2003. Is there evidence of a critical mass in the mid-Atlantic agricul-
tural sector between 1949 and 1997? Agricultural and Resource Economics Review 32: 116–128.
Marland G., R. A. Pielke, M. Apps, R. Avissar, R. A. Betts, K. J. Davis, P. C. Frumhoff, S.
T. Jackson, L. A. Joyce, P. Kauppi, J. Katzenberger, K. G. MacDicken, R. P. Neilson, J. O. Niles,
D. S. Niyogi, R. J. Norby, N. Pena, N. Sampson, and Y. Xue. 2003. The climatic impacts of land
surface change and carbon management, and the implications for climate-change mitiga-
tion policy. Climate Policy 3:149–157.
McCarl, B. A., W. Attavanich, M. Musumba, J. E. Mu, and R. Aisabokhae. 2014. Land use and
climate change. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 226–
254. New York: Oxford University Press.
Messer, K. D., J. M. Duke, and L. Lynch. 2014. Applying experiments to land economics: Public
information and auction efficiency in ecosystem service markets. In The Oxford handbook
Miceli, T. J., and K. Segerson. 2014. Regulatory takings. In The Oxford handbook of land econom-
ics, eds. J. M. Duke and J. Wu, 698–722. New York: Oxford University Press.
Mills, E. S. 2014. Are large metropolitan areas still viable? In The Oxford handbook of land eco-
nomics, eds. J. M. Duke and J. Wu, 74–84. New York: Oxford University Press.
Montgomery, C. A. 2014. Fire: An agent and a consequence of land use change. In The Oxford
handbook of land economics, eds. J. M. Duke and J. Wu, 281–301. New York: Oxford University
Press.
The Nature Conservancy. 2013. Private lands conservation. Accessed January 31, 2013. http://
www.nature.org/about-us/private-lands-conservation/index.htm
Nickerson, C. J., and W. Zhang. 2014. Modeling the determinants of farmland values in the
United States. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 111–
Parker, D. C. 2014. An economic perspective on agent-based models of land use and land cover
Partridge, M. D., and D. S. Rickman. 2014. Integrating economic development analysis and
land use economics. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu,
Plantinga, A. J., and D. J. Lewis. 2014. Landscape simulations with econometric-based land use
models. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 380–401. New
Nechyba, T. J., and R. P. Walsh. 2004. Urban sprawl. Journal of Economic Perspectives 18: 177–200.
Schmid, A. A. 1999. Government, property, markets . . . In that order . . . Not government ver-
sus markets. In The fundamental interrelationships between government and property, ed. N.
Mercuro and W. J. Samuels. Greenwich, CT: JAI Press, 237–242.
Smith, R. B. W. 1995. The conservation reserve program as a least-cost land retirement mecha-
nism. American Journal of Agricultural Economics 77 (1995): 93–105.
Soulé, M. E. 1991. Conservation: Tactics for a constant crisis. Science 253: 744–750.
Stone, E., and J. Wu. 2014. Land use and municipal profiles. In The Oxford handbook of land
economics, eds. J. M. Duke and J. Wu, 302–324. New York: Oxford University Press.
Towe, C., R. Lewis, and L. Lynch. 2014. Using quasi-experimental methods to evaluate land
policies: Application to Maryland’s priority funding legislation. In The Oxford handbook
US Department of Housing and Urban Development (USDHUD). 2000. The state of the cities
2000. Washington, DC: US Department of Housing and Urban Development.
Wu, J. 2006. Environmental amenities, urban sprawl, and community characteristics. Journal of
Environmental Economics and Management 52: 527–547.
Wu, J. 2010. Economic fundamentals and urban-suburban disparities. Journal of Regional
Science 50: 570–591.
Wu, J., and B. A. Babcock. 1996. Contract design for the purchase of environmental goods from
agriculture. American Journal of Agricultural Economics 78: 935–945.
Wu, J., P. W. Barkley, and B. A. Weber, eds. 2008. Frontiers in resource and rural economics.
Washington DC: Resources for the Future.
Wu, J., and S. Cho. 2007. The effect of local land use regulations on urban development in the
western United States. Regional Science and Urban Economics 37: 69–86.
Wu, J., M. Fisher, and U. Pascual. 2011. Urbanization and the viability of local agricultural
economies. Land Economics 87: 109–125.
Wu, J., and E. Irwin. 2008. Optimal land development with endogenous environmental ameni-
ties. American Journal of Agricultural Economics 90: 232–248.
Zilberman, D., M. Khanna, S. Kaplan, and E. Kim. 2014. Technology adoption and land use. In
The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 52–73. New York: Oxford
University Press.
PA R T I
DE T E R M I NA N T S A N D
DR I V E R S OF L A N D
U SE C HA N G E
C HA P T E R 1
I N T E G R AT I N G R E G I O NA L
ECONOMIC DEVELOPMENT
A NA LYS I S A N D L A N D U S E
ECONOMICS
M A R K D. PA RT R I D G E A N D DA N S . R IC K M A N
Academic economists historically separated issues related to land use from those
related to regional economic development. One reason is that land use studies typically
do not consider the connectedness of firm and household location decisions, whereas
regional economic development studies rarely account for land (McDonald, 2001).
Moreover, it appears that land use researchers think more at the microscale of neighbor-
hoods (or intraregional), whereas economic-development researchers think more at the
macroscale (or interregional).
The division between the two fields does not reflect how local economic develop-
ment policy is undertaken. Economic development is inherently about land because it is
about activity in a place or on a specific land area. Local governments compete with one
another in trying to attract households and firms to their place.
Land use and economic development, then, are inherently linked through zon-
ing, transportation, infrastructure, sprawl, and environmental attributes that jointly
affect firm productivity and household utility. Because local policy is about place, land
economics is linked to economic development policy through competition for new
development. This raises further questions about governance and local government
effectiveness in delivering public services that underlie development through Tiebout
(1956) sorting and spatial equilibrium processes generally.
In this chapter, we attempt to tie together the two separate literatures. We stress the
economic development literature in regional and urban economics that most closely
relates to land economics. An implicit theme is that land economic studies should pay
closer attention to joint firm/household location decisions, whereas the regional eco-
nomic development literature should pay closer attention to land as it defines the place
24 Mark D. Partridge and Dan S. Rickman
Amenities and
Household migration
quality of life
Public services
taxes
expenditures
Land use Local and regional
job creation and firm
zoning
productivity
climate
topography
soil quality
Urban system/hierarchy
Central place theory
new economic geography
Italics are examples
FIGURE 1.1 Model of regional economic development with land use.
that the activity occurs. Likewise, another theme is that research should focus more on
the regional interaction of activity across space. Models and empirical approaches are
needed that recognize regions as complex systems, fully understanding and model-
ing the interplay between land use and economic development, including the linkages
between the intraregion distribution of economic activity and overall regional eco-
nomic performance.
Figure 1.1 shows the interdependence of local and regional economic development
(depicted as job creation and firm productivity) with several key factors including land
use, amenities and quality of life, household migration, public services, and the urban
system. Italics indicate some examples of these factors. The figure reflects the key role of
land use in directly affecting economic development and, in turn, being directly affected
by economic development. Land use also indirectly influences economic development
through its interactions with the other factors. These interactions also illustrate the
difficulties of identification of causality in empirical analysis. The chapter will outline
these direct and indirect effects that land use and economic development have with each
other, illustrating the central connections between land economics and regional and
urban economics.
Before describing the contents of the chapter, we note that some important topics are
given brief treatment or omitted because of space limitations. Examples include public
infrastructure, tax competition, urban amenities, and spatial econometrics. Section 1
outlines the basic spatial equilibrium approach used in modern regional economic
development studies and outlines ways to include land. Section 2 describes the natural
link between land economics and economic development through proximity to urban
Integrating Regional Economic Development Analysis 25
centers. Economic activity across space is strongly affected by access to agglomeration

economies that influences economic location both within and across regions, with the
latter being our focus. Section 3 describes how land use affects the provision of natural
and urban amenities that influence whether households and businesses want to locate in
a particular place. We focus on the common features in the two literatures in which the
land use literature focuses on microscale amenities such as open space, whereas the eco-
nomic development literature focuses more on amenities at the regional scale that affect
regional economic growth.
Section 4 provides a brief introduction to government policy aimed to improve
land use and increase economic activity. This literature is extensive and we can only
provide a cursory treatment. Section 5 describes some of the empirical approaches
used in the economic development literature, focusing on the quasi-experimental and
structural approaches that currently predominate. We note that both have advantages
for empirical assessment but they suffer from shortcomings. The unifying theme is
that studies using either approach need to more rigorously assess the legitimacy of
their identifying assumptions and check robustness. Section 6 briefly highlights areas
ripe for future research while Section 7 presents our conclusions.
1. Land Use in Regional Economic

Development Analysis
Despite its central role in firm and household location decisions and regional economic
activity generally, land routinely is omitted in regional economic development analysis.
In part, this results from the traditional tools used in economic development analysis,
which often are chosen for convenience rather than demonstrated accuracy (Partridge
and Rickman, 1998, 2010). In studies where land use is the focus, regional economic
development considerations often are ignored or are of secondary importance.
Nevertheless, there is growing recognition of the central role of land use in regional eco-
nomic development.
Land is completely removed from consideration in economic impact analysis that
involves application of an input–output model because of its implicit assumption of per-
fectly elastic supply. Factors of production implicitly are assumed in excess supply in
short-run analysis or perfectly mobile in long-run analysis. As a fixed factor, often in
limited supply, the implicit omission of land from consideration questions the routine
use of input–output models in regional economic development analysis. This omission
likely leads to highly inaccurate impact assessments when land prices are highly respon-
sive to economic development or when there is intraregional heterogeneity in how land
prices respond.
Computable general equilibrium (CGE) models incorporate factor supply con-

straints, making them more general than input–output models (Partridge and Rickman,
1998). Although CGE models potentially are more accurate in a wide range of appli-
cations, this depends critically on the formulation of the CGE models. For example,
McGregor, Swales, and Yin (1996) formulate a CGE model with short-run labor supply
and capital adjustment constraints. They relax the constraints in the long run in demon-
strating how the CGE model then functions as an input–output model. Partridge and
Rickman (2010) argue that the traditional method of formulating regional CGE models
limits their applicability for regional economic development analysis; rather than pat-
terning regional CGE models after their national counterparts, they should be based on
spatial equilibrium theory, including explicitly incorporating land.
Rickman (1992) incorporates fixed land and imperfectly mobile capital and labor in a
regional CGE model, demonstrating how this produces dramatically smaller economic
multipliers than what is obtained by assuming factors of production are elastically supplied.
Fixity of land drives up its price when exports increase, crowding out other production (the
model did not separately consider residential land though). The CGE multiplier effects
then greatly depend on the elasticity of substitution between land and the mobile factors.
Despite von Thünen’s (1966) model of land use and the general importance of land
in location theory, land has largely been ignored in the increasing-returns literature
(Combes, Duranton, and Overman 2005). Helpman (1998) added a nontradeable hous-
ing sector to the New Economic Geography (NEG) model to introduce congestion
costs, though land use is not explicitly modeled. Pflüger and Tabuchi (2010) incorpo-
rate land used in housing and in production by an increasing returns sector in a general
equilibrium model, which produces a differing pattern of economic development than
if land is only used in housing.
McDonald (2001) effectively argues for connecting regional economic development
policies to both labor and land markets. Land markets not only affect predicted out-
comes, but also may be a source of economic development gains. Consistent with Bartik
(1991), benefits of regional economic development policies that allocate land to indus-
trial uses include employment of previously unemployed or underemployed members
of the labor force and higher land values. Welfare gains to original residents of the area
from economic development are enhanced to the extent land is owned by residents
(Morgan, Mutti, and Rickman 1996).
Burnett, Cutler, and Thresher (2007) incorporate land in a CGE model of Fort
Collins, Colorado to examine potential crowding out effects on other industries from
increased tourist activity and to assess whether tourism is an optimal land use. The sup-
ply of land is specified as price elastic for both commercial and residential uses. They
found land used in tourism as having the largest per-acre effect on gross city product
and real household income. A notable feature of the model is the connection between sec-
toral land use, direct job creation, in-migration, and residential land use. Tourism reduced
in-migration and hence less residential demand for land. Using the same framework,
Cutler and Davies (2007) report that sectors primarily employing low-skilled labor gen-
erally reduce in-migration and demand for residential land use compared to high-skilled
sectors, producing a larger per acre contribution of gross product and income. Kim and Ju
(2003) integrate an urban land supply module with a CGE model for Seoul in examining
the impacts on gross regional product, welfare, and income distribution from converting
industrial land and green space into residential use.
Another long-standing omission in the regional economic development literature is
the positive role land plays as a natural amenity. Land used for public parks, or left as open
space, for example, create recreational opportunities and provide attractive vistas, increas-
ing the local quality of life. Higher quality of life increases retiree and labor force migration,
stimulating regional growth.
Land’s contribution to the local quality of life then provides another feedback loop in a
regional economy. Changes in land use that enhance quality of life increase in-migration
and growth (Rickman and Rickman 2011). Regional economic development analyses then
must not only consider the relative direct benefits of alternative commercial or residential
uses, they also should consider the effects on local quality of life.
Thus, we advocate that regional economic development analysis be conducted using a
modeling framework broadly capable of capturing important feedback loops within a
regional economy. One such framework is the widely used spatial equilibrium approach
(Roback 1982; Beeson and Eberts 1989). The spatial equilibrium approach is sufficiently
flexible to reflect an array of quality of life and firm agglomeration considerations (Tabuchi
and Thisse 2006).
In the spatial equilibrium approach, households geographically locate so as to maximize
utility, whereas firms maximize profits in their location. Central to both decisions are nom-
inal wage rates and land costs, as well as perfect mobility. Higher wages, adjusted for land
costs, attract households. Lower wage rates and land costs attract firms. In addition, the
framework incorporates site specific characteristics, reflecting the quality of life and qual-
ity of the business environment. Quality of life includes benefits households derive from
land use beyond those obtained from residential housing. In equilibrium, the values of
site-specific characteristics are capitalized into wages and land costs. The approach can be
formulated in growth terms by assuming that economies transition across spatial equilib-
ria as exogenous conditions change (Dumais et al. 2002). Besides predictive equations for
wages and land costs, equations can be derived from a spatial equilibrium model for growth
in employment, gross regional product, investment, and population (Brown, Hayes, and
Taylor 2003; Partridge and Rickman 2003; Brown and Taylor 2006).
Both traded and nontraded goods can be included in the model, in which the traded
good can be specified with varying elasticity of demand.1 Alternative theories of agglom-
eration economies can be captured in the approach, ranging from NEG (Ottaviano and
Pinelli 2006) to urbanization economies, and those related to Central Place Theory
1
The traditional approach assumes that firms producing a traded good are price takers. Alternatively,
traded goods can be modeled using the Armington assumption, in which there is imperfect substitution
between traded goods of differing origins (Partridge and Rickman 2010). McDonald (2001) examines the
significance of alternative assumptions on the elasticity of demand for export goods in assessing regional
economic development policies.
(Partridge, Ali, and Olfert 2010). Quality of life includes exogenous attributes such as
weather, proximity to oceans or freshwater, or mountains. Other natural amenity attri-
butes may be endogenous, being affected by local economic activity, including, air and
water quality, forests, open space, attractive vistas. Endogenous quality of life attributes
also include manmade amenities such as public infrastructure.
In the traditional spatial equilibrium framework, regions are assumed to have uni-
form land use policies. However, within a growth context, Glaeser and Tobio (2008)
extend the model to allow for the effects of differential changes in land use and hous-
ing policies. They find that in former Confederate states, policies favorable to housing
development were more likely responsible for strong population growth near the end of
the 20th century than favorable weather.
Along these lines, Rappaport (2009) numerically simulates a structural spatial equi-
librium model to produce a series of equilibriums in examining US metropolitan popu-
lation growth. The model’s sole congestion force is land, which is used to produce both
a traded good and residential housing. Simulated feedback effects include population
growth effects on area amenity attractiveness and the effects of increased population
density on productivity.
2. Economic Development: Distance

and Proximity
Land economics and economic development are linked through the location of house-
holds and firms. Although urban economists often emphasize the location of house-
holds and businesses within a given urban or metropolitan area, regional economists
tend to focus on the relative differences across space, that is, comparing outcomes across
economic regions that could be metropolitan, nonmetropolitan, or some combina-
tion. Because intrametropolitan area location patterns are discussed elsewhere, we only
briefly highlight them, instead emphasizing broader regional patterns.
2.1 Distance and Regional Economic Development

Both land use and economic development are tied to a given place with its economic
activity closely tied to proximity within the urban system. A first effort was Von Thünen’s
(1966) classic model of land use surrounding a single urban center on a featureless plane
(Hite 1997). He shows that high-value-added products with high transportation costs
locate closest to the urban center. The missing feature is it does not reflect the interaction
of cities and regions across an urban system.
Central Place Theory (CPT) represented the first formal effort to model the urban
system (Christaller 1933; Lösch 1940). Under assumptions including a featureless plane,
CPT shows how a multitiered urban system could develop in which the type of services
determine the size and location of urban center—for example, the top of the urban
system has all higher-ordered services such as patent attorneys, whereas the very bot-
tom has basic services such as convenience stores.2 CPT is adept at predicting the loca-
tion of cities within urban systems, particularly in areas such as the North American
Great Plains with traditionally high farm intensities (Fox and Kumar 1965; Wensley and
Stabler 1998; Olfert and Stabler 1999). CPT is useful in predicting the location of actual
business and consumer services and their population thresholds. A primary critique of
CPT is its static nature. It is usually necessary to impose ad hoc assumptions regarding
changes in technology and transport costs to describe an evolving urban system.
Nevertheless, CPT is still quite useful in understanding the organic process of how
urban-centered regions have expanded since the 1950s (Irwin et al. 2010). This pro-
cess is driven by many factors such as labor saving productivity gains in the primary
sector that released labor for urban employment, further facilitated by the rising use
of automobiles that aid long-distance rural–urban commuting. Increasing population
thresholds for public and private services also led more services to be provided from a
central location. The inherent spillovers as economies began to regionalize have long led
to calls for government consolidation and regional collaboration around the functional
economic regions delineated from CPT (Fox and Kumar 1965; Tweeten and Brinkman
1976). Increasing agglomeration economies imply that growth prospects are better in
regions with critical mass (Portnov and Schwartz 2009). Conversely, promoting growth
in small communities in isolation would be ineffective because they lack the agglom-
eration economies necessary to generate endogenous growth (Fox and Kumar 1965;
Berry 1970).
The question whether urban-centered growth helps the surrounding hinterlands
spawned a regional version of the spread and backwash literature that originated in
international development,3 namely, does prosperity in urban growth centers “spread”
into the countryside and create economic opportunities, primarily through commut-
ing, or does it create a “backwash” where urban growth pulls rural workers and capi-
tal into cities? United States results suggest urban growth spreads into the countryside
(Hughes and Holland 1994; Barkley, Henry, and Bao 1996; Henry et al. 1997), while
spreading up to 200 kilometers in Canada (Partridge et al. 2007). Yet, urban spread is
more likely when rural communities have sufficient quality of life and services to sup-
port a commuting residential population (Henry et al. 1997; Kahn, Orazem, and Otto
2001; Partridge, Ali, and Olfert 2010). Likewise, Ke and Feser (2010) found that spread
effects predominate in China, though Chen and Partridge (2011) find that growth in the
three Chinese mega cities (Beijing, Shanghai, Guangzhou) creates widescale backwash.
2 See Mulligan (1984) for a review of the CPT literature.

3 See Myrdal (1957) for early applications and reviews by Richardson (1976) and Gaile (1980).
A key economic development question then is whether urban-led growth can reduce
rural unemployment. There are reasons for pessimism. Renkow (2003) found that
about 60% of the adjustment to local nonmetropolitan employment growth is accom-
modated through changes in commuting flows and another 30% is through changes in
migration—that is, employment growth is only partially met through increases in local
labor-force participation.
Although CPT inspired a large economic development literature, CPT theoretical
research waned after the 1980s. One reason is that CPT was rather mature, and enthu-
siasm shifted to NEG. Another is that Geographical Information System (GIS) technol-
ogy was not sufficiently developed to produce reliable empirical measures. Not until
Partridge et al. (2008a, 2008b) was there a full test of CPT across a broad landscape. They
used US county data to consider hundreds of metropolitan areas that are typically sepa-
rated by rural space, forming a perfect setting for assessing the urban hierarchy’s inter-
vening effects on job and population growth. They employed detailed measures of access
to the five nearest higher-ordered tiers in the urban hierarchy. Their results show that
urban proximity has strong intervening effects that act through access to all the nearest
higher-tiered urban areas.4 Partridge et al. (2008b) also investigated the so-called “dis-
tance is dead” hypothesis that enhanced information technology and transportation had
slayed the “tyranny of distance.” They found that not only is distance not dead, but its
effects are actually becoming stronger over time, most likely due to spatial transactions
costs (e.g., face-to-face contact) in the expanding service sector. If distance is more prob-
lematic for rural areas and small cities, there are policy implications for the provision of
broadband, transportation, business development, and regional governance.
Hedonic studies further support the notion that distance is a key factor behind spa-
tial variation in wages and housing costs—which ultimately reflects how remoteness
affects productivity and quality of life. Defining remoteness as being nonadjacent to a
metropolitan area, Wu and Gopinath (2008) find that remoteness accounts for 76% of
the expected differences in average wages between the highest and lowest US county
quintiles, exceeding the importance of other factors such as amenities and human capi-
tal. Partridge et al. (2009, 2010) further confirm that remoteness is a key factor behind
wages and housing prices. Partridge et al. (2010) find that most of the distance effects
relate to productivity disadvantages (not household effects) and that these disadvan-
tages are rising over time even with new technologies.
NEG models generated significant enthusiasm after Krugman’s (1991) seminal work.
They capture agglomeration economies and product variety (both as inputs to firms and
4
Partridge et al. (2008b) find that distance from the nearest metropolitan area of at least 50,000
population leads to an economic penalty. If the nearest metropolitan area is not at least 1.5 million
people, there are added penalties for the distance to reach metropolitan areas of at least 250,000 people,
to reach metropolitan areas of at least 500,000 people, and to reach metropolitan areas of at least
1.5 million. For a clever application of the attenuation of agglomeration economies within metropolitan
areas, see Rosenthal and Strange (2008). For applications of how the CPT urban hierarchy affects locale
industry composition, see Wensley and Stabler (1998) and Polèse and Shearmur (2004).
to consumers) that can lead to core-periphery patterns (Brakman, Garretsen, and van
Marrewijk 2009a). Economists are attracted to NEG models because they have explicit
microfoundations, are analytically tractable, and they can explain uneven regional
development (World Bank 2009). For example, Fujita, Krugman, and Mori (1999) show
how a CPT urban hierarchy could initially form and Tabuchi and Thisse (2011) show
how shocks affect the hierarchy. There are relative few empirical NEG applications, but
examples include Brülhart and Koenig (2006) (transition economies), Volpe-Martincus
(2010) (Brazil); Redding and Sturm (2008) (Postwar Germany); Brakman et al. (2009b)
(European Union); and Hering and Poncet (2010) (China).
NEG has been used to inform regional development policy, often suggesting that tra-
ditional place-based policy to support lagging regions is misguided. The World Bank
(2009) uses NEG to support its contention that regional policy should be spatially
neutral because excessive support of peripheral regions shifts resources from cen-
tral regions, leading to lower aggregate growth due to lost agglomeration economies.
Likewise, providing infrastructure to peripheral regions could actually hurt them
because it lowers transportation costs from central regions, allowing central firms to
supply peripheral regions, further taking advantage of their agglomeration economies
(Puga 1999). NEG frameworks have also been used to argue that large cities can have
higher tax rates, allowing them to capture some of the “agglomeration rents” they pro-
vide businesses (Baldwin and Krugman 2004).
Despite their mathematical elegance, NEG models are criticized for lacking relevance
for economic development policymaking. Several strict assumptions are typically
employed to make these models solvable including a simplistic production function,
iceberg transportation costs, little consideration of institutional factors, and house-
hold location preferences that are crude (Partridge 2010). NEG models often produce
knife-edge results in which small parameter changes generate unstable outcomes.
Partridge (2010) argues that the patterns uncovered in NEG models have limited appli-
cability in North America, especially when compared to factors such as amenities and
human capital. Partridge et al. (2008b, 2009, 2010) find that standard CPT significantly
outperforms NEG in explaining US population movement, wages, and land costs.
Krugman even notes that NEG models better described American development at the
dawn of the 20th century, not the dawn of the 21st century, though he argues that con-
temporary China is a better setting.
2.2 Land Economics and Intrametropolitan-Area Economic

Development
There are two workhorse models that economists use to describe urban location theory.
First is the Alonso, Mills, Muth Monocentric City Model (MCM) (Alonso 1964; Mills
1967; Muth 1969). The MCM postulates an inverse relationship between land prices and
distance to the central business district to compensate for longer commutes, though the
rise of polycentric cities has reduced some of its applicability (McDonald and McMillen
2000). Yet, in an MCM framework, lower transport costs and higher incomes imply
an expanding city footprint—or sprawl (Glaeser and Kahn 2004; Nechyba and Walsh,
2004; Wu 2010). Although sprawl has ambiguous impacts on social welfare (Glaeser
and Kahn 2004), Fallah, Partridge, and Olfert (2011) find that sprawl is associated with
decreased firm productivity, presumably due to diminished agglomeration economies,
suggesting businesses are less competitive in sprawling cities.
The Tiebout (1956) model is the second major model describing intra-urban loca-
tion. People “vote with their feet” by sorting to places that offer higher utility on the basis
of economic and noneconomic factors. Quality of life and environmental services could
be one factor that induces self-sorting within metropolitan areas (Banzhaf and Walsh,
2008). Public finance applications stress intrametropolitan differences in public services
and their tax price.
Self-sorting in the Tiebout model gives communities incentives to use exclusionary
zoning to attract the type of residents who will positively contribute toward public ser-
vice provision. This could lead to equity and efficiency concerns if there is spatial mis-
match between the location of workers and jobs (Kain 1968; Ihlanfeldt and Sjoquist 1998;
Houston 2005). For example, zoning (and segregation) may limit affordable housing for
lower skilled workers to the central cities, but firms that employ low-skilled workers relo-
cate in the suburbs (Martin 2004; Stoll 2006). Blumenberg and Shiki (2004) argue that spa-
tial mismatch may even be more severe in remote rural areas because thin labor markets
and longer distances could further reduce employment access for specific skill groups.
Raphael and Stoll (2002) provide evidence that job accessibility for minority work-
ers remains problematic, though it improved during the 1990s. Partridge and Rickman
(2008) report indirect evidence that job accessibility is one reason for high poverty in
central cities by showing that job growth has a stronger inverse association with lower
poverty in central counties. Conversely, sorting of residents with weak labor market
attachment into central cities would have suggested a smaller job growth-poverty link-
age. Providing low-skilled households better employment access through providing
cars or public transit and finding ways to relocate households closer to employment
seems to be sensible as this benefits the workers and the employers. Yet, the notion of
Tiebout sorting and exclusionary practices by local governments may limit the effec-
tiveness of such policies.
3. Land Use, Quality of Life, and

Regional Economic Development
The quality of life afforded by natural amenities has long been recognized as a critical
factor in regional growth. An area with high quality of life attracts both working-age
adults and retirees (Vias 1999; Deller et al. 2001; Gunderson, Pinto, and Williams 2008;
Whisler et al. 2008). In-migration of working-age adults shifts labor supply and the
demand for land outward, reducing the real-wage rate through lower nominal wages
and/or higher land prices. Firms also may consider the amenity attractiveness of an area
in their location decision in order to attract skilled workers (Gottlieb 1995), and because
of preferences of managers or owners for amenity consumption. Retiree in-migration
and new firms shift labor demand outward, particularly for workers employed in local
service sectors, and increase land prices. Natural amenities especially may attract those
with greater human capital, further boosting employment (Shapiro 2006; McGranahan
and Wojan, 2007), wages, and land prices. Whether nominal wages are lower in areas
with a high quality of life depends on the balance of these forces in addition to a number
of structural characteristics of the local economy (Rappaport 2008).5
As a normal good, the demand for amenities in the United States increased in the 20th
century with rising income (Costa and Kahn 2003; Rappaport 2007). In fact, argued
to be fueled by rising income, increased wealth, and an aging population, Partridge
(2010) reports natural amenities as dominating other theories, such as NEG, as the pri-
mary reason for US regional growth differentials in the latter half of the 20th century.6
However, although increased demand for amenities increases household willingness to
pay higher land prices, the extent that it leads to in-migration depends on amenity con-
sumption’s elasticity of substitution with nonamenity goods and services; a lower elas-
ticity leads to greater in-migration (Rappaport 2009).
There are limits to the growth that can be attained in areas with high levels of natu-
ral amenities. For one, as amenities become capitalized into wages and land prices,
household utility advantages in the region are reduced, causing growth to become
more spatially equalized (Partridge et al. 2008a). Even with continued rising income,
forward-looking households can lead to capitalization of amenities in the near term,
shutting off growth.
Inelastic land supply is one reason for many cities having faster housing price growth
and an increasingly right-skewed distribution of income (Gyourko, Mayer, and Sinai
2006). These cities often have limited land supply because of geographical barriers
such as coastlines and mountains, and often enact policies that limit the development
of new housing. Many also are places with perceived high levels of natural amenities
(Rappaport 2009).
Yet, if regional policies allow growth to diminish quality of life (Gabriel, Mattey,
and Wascher 2003), negative feedback effects on growth will occur (Chen, Irwin, and
Jayaprakash 2009). Rickman and Rickman (2011) find evidence of within-Census
region deterioration of quality of life in nonmetropolitan areas possessing high levels
of natural amenities during the 1990s. They conclude that localities should manage
5
Rappaport’s (2008) model predicts that high quality of life is capitalized much more into land prices
than wages. Empirically, Wu and Gopinath (2008) and Rickman and Rickman (2011) find that natural
amenities are capitalized much more into housing prices than wages.
6 See Partridge (2010) for discussion of amenity migration studies for other countries.
growth in ways to reduce negative amenity effects lest both the quality of life and growth
be diminished.
Land use affects an area’s quality of life through several channels, which is a con-
sideration particularly critical for areas primarily dependent on quality of life for eco-
nomic growth. Unmanaged growth in high-amenity areas can lead to sprawl, and the
associated traffic congestion and pollution (Hansen et al. 2002). There also may be
development-related losses in valued attributes such as open space (Vias and Carruthers
2005; Cho, Poudyal, and Roberts 2008), wildlife and its diversity (Hansen et al. 2002),
the quantity or quality of vegetation and forests (Cho et al. 2009), and scenic views
(Benson et al. 1998).
Proximate public lands, land owned by nonprofit organizations, and restrictive zon-
ing may contribute to an area’s amenity attractiveness and its economy in some ways,
but also may inhibit the economy in other ways (McGranahan 2008). Henderson and
McDaniel (2005) suggest that restrictive zoning in high-amenity areas may be one rea-
son why they found manufacturing growth lagging that of other sectors. Yet, Rickman
and Rickman (2011) did not find evidence of changes in land use regulations or reduced
productivity affecting population growth in high amenity nonmetropolitan areas dur-
ing the 1990s.
Lewis, Hunt, and Plantinga (2002) find slightly higher net migration rates for coun-
ties with more conservation land in the US Northern Forest region but no differences
in employment growth. In an evaluation of the Northwest Forest Plan by the US Forest
Service, Eichman et al. (2010) found negative employment effects from reduced tim-
ber use that are only slightly offset by positive effects of increased in-migration, which
contrasts with findings reported in other studies. They attributed the difference in find-
ings in part to the productiveness of the timberland withdrawn from production in the
northwest. Rosenberger, Sperow, and English (2008) concluded that official wilderness
designation did not greatly affect the transition of Appalachian Region counties from
being primarily dependent on natural-resource and manufacturing activity to primary
dependence on nonlabor sources of income and services. In a review of studies on wil-
derness designation and local growth, Rosenberger and English (2005) concluded that
the link depends on the structure of the local economy and its longer-term trend.
Land use in cities also may adversely affect their environmental quality and feed-
back negatively on growth. City size can be associated with increases in various con-
gestion forces such as crowded roads and increased pollution. Not only city size but
also the degree of urban sprawl has often been identified as having a number of adverse
environmental impacts (Johnson 2001; Hasse and Lathrop 2003; Nechyba and Walsh
2004). Stone (2008) found sprawl to be associated with the number of times monitored
ozone concentrations exceeded the National Ambient Air Quality Standards across 45
US cities. Other impacts include loss of open space, reduced diversity of wildlife spe-
cies, increased water pollution, and emission of particulates, significant losses of native
vegetation and forests, loss of natural wetlands, blocking of mountain views, and eco-
system fragmentation.
Some studies question the perceived negative relationship between sprawl and envi-
ronmental quality. Despite growing numbers of higher-income households living in
suburbs and commuting to work, Kahn and Schwartz (2008) found reduced air pol-
lution in California cities, which they attributed to technological improvements in
auto emissions. Although Kahn (2001) found evidence of reduced quality of life in
fast-growth California cities, he did not attribute this to air pollution because it had
decreased, which suggested other causes such as increased traffic congestion. In survey-
ing research on the dynamics of urban growth and ecological systems in the western
world, Czamanski et al. (2008) concluded that “peri-urban” areas associated with sprawl
provide ecosystem benefits because of their position between developed urban areas
and agricultural lands. In an analysis of the impact on ecosystem services from urban
sprawl in San Antonio Texas from 1976 to 1991, Kreuter et al. (2001) found that despite
a dramatic increase in the area of urban land use and reduction in the size of rangelands,
the shift of rangelands to woodlands greatly helped limit the loss of ecosystem services.
Wu (2006) demonstrates how spatial variation in environmental amenities themselves
can contribute to what is perceived as sprawl.
Therefore, an assessment of what constitutes sprawl and how it affects the quality of
life is critical for sustainable regional economic development. More research is needed
to assess the channels through which land use, growth, and environmental impacts
interrelate. How these environmental changes affect perceived quality of life also require
further investigation along the lines of hedonic studies of regional differences in quality
of life.
4. Fiscal Federalism, Land Use, and

Regional Economic Development
The spatial location of economic activity and land use also are affected by regional fis-
cal and land use policies. Both fiscal and land use policies can affect sprawl and regional
economic development. The complexity of regional economies also makes the policies
interdependent, both within and across jurisdictions.
Within the spatial equilibrium framework, variation in state and local fiscal policies
has been found to be as important as individual characteristics in explaining wage dif-
ferentials across US metropolitan areas and to matter as much for metropolitan quality
of life as natural amenities (Gyourko and Tracy 1989, 1991). They also have been found
to be important in explaining US nonmetropolitan county wage and land rent differ-
entials, in which some policies primarily affect quality of life, whereas others affect the
business climate (Yu and Rickman 2011). State and local fiscal policies directly affect
quality of life through the taxes that households pay and government services they
receive. Likewise, firm profits are affected by taxes and government services. Indirectly,
however, local fiscal policies may have spillover effects, affecting economic activity and
land use in neighboring jurisdictions.
High taxes and inadequate services in central cities can push economic activity out-
ward into the suburbs and beyond, creating sprawl. Although there are potential social
welfare gains from Tiebout-sorting of individuals according to their preferences for
government services, the deconcentration of local government can affect the relative
efficiency of the provision of government services, and hence the quality of life and
productivity (Mattoon 1995; Innes and Booher, 1999). Public infrastructure exhibiting
economies of scale or network effects (Dalenberg, Partridge, and Rickman, 1998) may
be underprovided in a deconcentrated environment.
In reviewing the literature, Mattoon (1995) lists water and sewerage disposal as most
efficiently provided by centralized metropolitan governments, whereas services such as
education are reported as better provided with decentralized government. As discussed
earlier in the chapter, increased sprawl can affect the amenity attractiveness in the
broader metropolitan area such as through increased air and water pollution. Increased
traffic congestion associated with sprawl can affect firm productivity. Therefore, the
relative centralization and coordination of local fiscal policies can affect land use and
economic development of the broader region.
Using state level data, Akai and Sakata (2002) find measures of local government
expenditures and revenue relative to those for the state to be positively related to growth.
In an examination of all US metropolitan areas, Stansel (2005) reports that decentraliza-
tion increased growth (though state fixed effects are not accounted for and state laws
and constitutions set the framework for local governments). In a related study, he found
that the negative effect was weaker in the largest 100 metropolitan areas (Stansel 2002).
Hammond and Tosun (2011) examined all US counties and found that decentraliza-
tion in metropolitan areas, as measured by increased fragmentation of single-purpose
governments, increased employment growth, whereas reduced revenue centralization
increased income growth. In contrast, they found that general purpose government
fragmentation reduced population and employment growth in nonmetropolitan coun-
ties. They concluded that their varied results suggest that general claims could not be
made regarding fiscal deconcentration and regional growth.
Deconcentration also may occur in land use regulations. Jurisdictions in metropoli-
tan areas with tighter controls push building activity into neighboring jurisdictions pos-
sessing fewer controls, which often are positioned at the periphery of the metropolitan
area and beyond, creating sprawl (Carruthers 2003). Mills (2006) argued that Tiebout
competition increases jurisdictional competition and reduces inefficient low-density
development, a point disputed by Vigdor (2006). Brueckner (2000) argued that urban
expansion reflects consumer demands for larger houses and yards, as well as proxim-
ity to consumer amenities. If these suburban options are unavailable, this could reduce
a metropolitan area’s attractiveness to households. Lax land use regulations and an
absence of charging for social costs of development such as damage to ecosystem ser-
vices also can lead to rural sprawl (Weiler 2003), which may feed back negatively on
growth.
So-called “smart-growth” policies have been widely enacted to promote sustainable

development (Wu 2006; Braun and Scott 2007) through increased efficiency of govern-
ment services and added environmental protection. Yin and Sun (2007) report that
metropolitan smart-growth policies increased the population share living in dense areas
in the 1990s, whereas state-level smart-growth policies did not. Wu and Cho (2007)
found that local and state land use restrictions reduced land development in five western
states. Boyle and Mohamed (2007) concluded that state, regional, and local attempts
to limit urban sprawl in Michigan largely failed. Kline and Alig (1999) concluded that
Oregon’s land use planning program concentrated development within urban growth
boundaries, but the effect on land use in forest and farm land use zones was uncertain.
In comparing Portland, Oregon; Orange County, Florida; and Montgomery County,
Maryland, Song (2005) reported a long-term increase in population density in resi-
dential neighborhoods, which is partly attributed to growth management policies. Yet,
reduced external connectivity and a lack of mixed land use are bemoaned, including low
access to mass transit.
Glaeser and Kahn (2010) considered the effect on national carbon emissions associ-
ated with transportation, home heating and cooling, and electricity use, from spatial
variation in local land use restrictions. They suggest that strict land use restrictions in
lower emissions areas might cause their economic activity to shift to areas with high
emissions. A potential policy recommendation would be to impose federal fees on
development in high emission areas.
Overall, a complex relationship exists between fiscal federalism, land use decisions,
and economic development. Sustainable economic development at all levels of spatial
aggregation requires conceptualizing local and regional economies as complex systems,
including land use and economic development policy interactions (Innes and Booher
1999). The extent of externalities across jurisdictions in a region, state, or nation sug-
gests a need for some government coordination and more government involvement.
Considerably more research on the complexity of interactions is needed to inform poli-
cymakers regarding government’s proper role.
5. Empirically Assessing Economic

Development
When assessing economic development, one needs to consider several issues such as
(a) firm and household self-sorting; (b) the endogeneity of public policy (e.g., roads
are built where policymakers expect future growth or maybe where they do not expect
future growth); (c) unobservable factors that are correlated with both the dependent
and independent variables that cause endogeneity and omitted variable bias; and
(d) sample heterogeneity. The four main approaches in assessing economic devel-
opment are CGE models; simulations of theoretical models; instrumental variable
(IV)/quasi-experimental approaches; and structural models.7 CGE models have already

been discussed. Brakman, Garretsen, and Marrewijk (2009a) describe NEG simula-
tions, whereas other simulation approaches are covered elsewhere. Thus, we outline the
latter two econometric approaches.8
Ordinary least squares (OLS) consists of regressing the dependent variable y (e.g.,
population growth) on several explanatory variables X (e.g., job growth, taxes).
y = βX + e (1)
in which e is the residual. A key assumption is Cov(e, X)= 0, or there is no endogeneity

bias. Endogeneity bias can arise from direct reverse causality—for example, regressing
population growth on average wages—which is less of a problem in contemporary work
because of improved research design. The more likely cause is omitted variables (unob-
servables) that are correlated with some of the X—for example, persistent factors such as
a good harbor that is correlated with job growth in the population model.
A Hausman test can be used to determine the existence of statistical endogeneity
in Equation (1), which requires instrument(s) Z that predicts the potentially endoge-
nous explanatory variable(s) X1, but Z cannot have a causal relationship with y (Cov(e,
Z) = 0)–that is, the exclusion restriction. In other words, Z only influences y indirectly
through how it affects X1. It is essential that Z be “strong” (Stock and Watson 2007), or
does a good job of predicting X1 in the first stage. Strong economic rationale and insti-
tutional features often are used to find Z. For example, a good instrument for interstate
highway mileage is how many miles were in the original World War II era military plan
for the interstate system (Duranton and Turner 2011). In a population growth model, a
good instrument for job growth is the predicted job growth if all of the local area’s indus-
tries grew at the national rate—that is, from shift-share analysis (Bartik 1991). A related
question is deciding which variables should be tested for endogeneity. Local economies
are general equilibrium systems in which feedback loops are endemic. Good judgment
needs to distinguish between statistical endogeneity that biases the coefficients in an
economically meaningful way from trivial “endogeneity” that can arise from almost all
any variable.
The primary solution for endogeneity is the IV approach (Stock and Watson 2007). In
a careful study of how roads influence driving, Duranton and Turner (2011) use the IV
7
Holmes (2010) also labels reduced-form and descriptive exercises as another approach, noting its
limitations for establishing causality. However, Angrist and Pischke (2009, 213) describe the inherent
value of reduced-form models for careful empirical analysis. We do not separately consider descriptive
approaches because the dividing line between IV and reduced-form approaches has greatly blurred.
8
See Holmes (2010), Angrist and Pischke (2009), and Stock and Watson (2007) for more econometric
details. We do not describe spatial econometric methods because they are well known. In addition, their
value has recently been questioned due to specification issues including a lack of theoretical motivation
for their use and identification problems. See Overman and Gibbons (2010), McMillen (2010), and
Pinske and Slade (2010) for recent critiques.
approach. Building a good economic case for their use, “clever” instruments are devel-
oped for contemporary interstate highway mileage: military road plans, early explorer
routes, and late 1890s railroad mileage. Further, they test for the strength of these instru-
ments and illustrate that an instrument can be conditionally valid after accounting for
other control variables.9 Alternative models such as limited information maximum like-
lihood estimators are used as robustness checks for weak instruments (see Angrist and
Pischke 2009 for related discussion).
Random experiments are the gold standard of empirical assessment, but rarely
exist in economic development practice (Holmes 2010). Quasi-experimental (QE)
approaches are used to approximate this setting (Card 1990). Holmes’ (1998) study of
business climate is one example. He examined the influence of state business climate on
manufacturing employment growth in the border counties between US states with and
without right-to-work union laws. The key identifying assumption is that productivity
would be the same at the border, in which state policy would be the main factor that
causes employment growth to vary. Of course, there could be many other factors that
could influence productivity such as historic location of cities. Holmes spent consider-
able effort in controlling for these persistent factors to strengthen identification.
Another QE approach is the difference-in-difference approach (DID) (Stock and
Watson 2007, Chapter 13). One example is Funderberg et al.’s (2010) examination of
1990s-era highway expansions in California. They examined population and employ-
ment growth in the immediate surrounding census tracts around selected highway
projects, comparing this to growth in nearby control tracts. Essentially, in the treatment
tracts, they differenced growth in the years after the completion of the road from growth
in the years immediately preceding completion. They did the same for the control tracts
that did not receive a new project. If the treatment experienced significantly higher
growth after the project, then the DID would be positive.10 The identifying assumption
is that the main factor affecting trend differences between the two groups is the road
construction, a strong assumption. Funderberg et al. (2010) control for other factors
that could account for different growth rates between the groups to strengthen their
identification. A possible research design weakness is that the control tracts were very
close to the treatment tracts. The new roads could shift growth from the treatment to the
control tracts, positively biasing the impact of the road construction, which needs to be
considered in research design.
9
Duranton and Turner (2011) argue that 19th-century railroads were built for short-term profits
and indirectly affect population today by affecting historic population. Thus, controlling for historic
population from the early-20th century would remove any correlation of the instrument with the
residual—that is, Cov(Z, ε │X) = 0.
10 Suppose that the DID window was five years before (period 0) and after (period 1) for employment
growth. Then the difference across the two periods for the treated region: ∆T = EmpGrowth1 –
EmpGrowth0. The analogous can be written for the control region ∆C. The DID estimator is ∆T – ∆C.
Greenstone, Hornbeck, and Moretti (2010) (GHM) is an example of the advantages

and potential pitfalls of QE design. They examined how large plant openings affect
total factor productivity (TFP) of other manufacturers in the county with the opening.
Greenstone, Hornbeck, and Moretti argued that comparing winning county TFP to that
in all other counties would produce biased results due to unobservables.11 To develop a
counterfactual, they compare “winner” county TFP to the runner up or “loser” coun-
ty’s TFP. Loser counties are identified in a monthly article in the trade publication
Site Selection Magazine, which reported location announcements of large plants. The
article lists the “loser” counties that GHM contend “narrowly” lost the competition.
Greenstone, Hornbeck, and Moretti’s identifying assumption is that the loser county is
like the winner county in every economically consequential way, forming a good coun-
terfactual. They employed best-practice DID methodology augmented by time trends,
industry dummies, and other plant-specific inputs to account for other factors associ-
ated with that plant’s TFP.
Greenstone, Hornbeck, and Moretti found that the winner’s TFP growth averaged 5%
to 12% more than in losing counties. Such strong agglomeration economies far exceed
the typical estimates from the agglomeration literature (Rosenthal and Strange 2004).
Greenstone, Hornbeck, and Moretti concluded that these spillovers justify the generous
tax incentives offered by local governments to new firms. Yet, in an odd result, when
they compared winner TFP to all US counties, they found that all counties had TFP
growth that was about 5% greater than the winners, suggesting that either their com-
plete set of DID controls were ineffective (which seems unlikely) or their identifying
assumption is suspect.
Greenstone, Hornbeck, and Moretti’s identifying assumption does not square with the
institutional features of local governments bidding for firms. Profit-maximizing firms
would not engage in a publically announced bidding war to establish counterfactuals
for researchers, but to strategically affect the bidding, thereby possibly creating endoge-
neity. Take GHM’s example of Greenville, SC beating Omaha, NE for a large Mercedes
plant in the 1990s. Is Omaha a true counterfactual? It is located far from ports and far
from markets and auto suppliers. Indeed, despite “narrowly” losing to Greensville for
Mercedes, Omaha has never landed an auto assembly plant. Was Mercedes simply using
Omaha to sweeten their deal from Greensville—that is, “losers” may be more willing
to offer large tax incentives to help their economy. Wouldn’t a better true counterfac-
tual have been in the Southeast with similar market attributes and low union densities
as Greensville? The point is QE studies should engage in robustness checks to assess
their experiment. For GHM, a good robustness check would use matching or propensity
score approaches.
11 Although GHM did not predict the sign of this bias, it seems reasonable that comparing the
winning county’s TFP to all counties would overstate the TFP effects of a large plant opening because the
firm would likely locate in counties with underlying factors that would raise TFP for all firms.
Another econometric problem is unobservable variables. This is especially problem-

atic when there are unobserved location-specific factors that are correlated with the X
variables, producing omitted variable bias. If a researcher has pooled-time-series data,
they can control for location fixed effects that account for persistent factors associated
with the place. When including fixed effects, all cross-sectional differences are in the
location fixed effects, meaning that only within-location time-series variation in the
variables is identifying the coefficients.
Including fixed effects, however, does not account for unobserved time-varying
effects for the location. Also, if there is measurement error in the X variables, then the
time-series variation will be increasingly dominated by noise, substantially biasing
the coefficients toward zero. Finally, fixed effect models incorporate the very strong
assumption that the Xt variables and the residuals are not only contemporaneously
uncorrelated, but Xt has to be uncorrelated with the residuals across all time periods
(Wooldridge 2002). Conversely, first-difference models that net out location fixed
effects do not need this strong assumption.
Heterogeneous responses can greatly alter the interpretation of the results. In such
cases, locally weighted regression (LWR) approaches (or geographically weighted
regression) can estimate different regression coefficients βi that vary across locations.12
Locally weighted regression typically requires a separate regression for each observa-
tion on a sample of neighboring observations that is usually determined by proximity.
Locally weighted regressions have gained prominence and have been used to exam-
ine factors such as employment density (McMillen 2004), housing prices (McMillen
and Redfearn 2010; Redfearn 2009), population growth (Ali, Partridge, and Olfert
2007), and environmental hazards (Carruthers and Clark 2010). Ali et al. (2007) and
Carruthers and Clark (2010) show how to decompose the variance of the predicted
effects into that due to variation in the X variables and that due to spatial variation in the
regression coefficients.
Structural models use theory to derive identifying restrictions to help establish cau-
sality when there are heterogeneous agents (Keane 2010). Yet, they have only been used
at the edge of the economic development literature with most applications occurring in
the fields of environmental economics or public finance (see Holmes 2010; Kuminoff,
Smith, and Timmins 2010; “An Assessment of Empirical Methods for Modeling Land
Use” by Irwin and Wrenn and “Equilibrium Sorting Models of Land Use and Residential
Choice” by Klaiber and Kuminoff for reviews). If the correct theoretical model is
employed, then structural models better inform policy because the underlying causal
mechanisms are uncovered. Moreover, they are useful for evaluating nonmarginal
changes in policies or amenities.
The disadvantage of structural approaches is that the results can be sensitive to the
underlying model or functional form of, say, the utility function (Kuminoff and Jarrah
2010). Others criticize them for imposing too much structure and not “letting the data
12 See McMillen (1996) and Fotheringham, Brunsdon, and Charlton (2002) for details.
speak” (Angrist and Pischke 2010), though structural proponents argue they are more
upfront about explicitly stating the model’s assumptions (Keane 2010). Angrist and
Pischke (2010) convincingly argue that another shortcoming is that authors do not sub-
ject structural models to sufficient robustness tests of their assumptions.
Structural models require further advances to capture the multiple dimensions
of modeling economic development and land use. Modeling of forward-looking
household behavior and place of work/place of residence behavior are in its infancy
(Kuminoff, Smith, and Timmins 2010) and both of these are key features of economic
development and land use processes. Likewise, modeling firm behavior is still emerg-
ing; thus, the joint firm/household decision making that characterizes the special equi-
librium approach is another area needing further research for developing structural
models.
6. Conclusion and Future Research
The primary theme of this chapter is the need to fully integrate land use in economic
development analysis. The complexity of regional economies combined with data and
methodological limitations have too often led to piecemeal analysis of regional eco-
nomic development and land use. Unfortunately, this has resulted in widely varying
findings and an incomplete understanding of key issues. Too little is known about the
interconnectedness of regional economic development and land use.
We outlined some areas for future research in the chapter, but there are other pos-
sibilities that warrant mention. We have already noted that sprawl studies typically do
not assess the interrelation between land use, regional economic growth, and envi-
ronmental quality. Likewise, firm location and workplace decisions are understudied
within this context. Modeling metropolitan areas or functional economic regions in
isolation of the interaction of cities across the entire hierarchy may produce mislead-
ing findings as shown by Polèse and Shearmur (2004) and by Partridge et al. (2008a,
2008b, 2009). Likewise, we know little about how structural shocks such as energy
shocks, housing bubbles/busts, and economic recessions such as the Great Recession
alter the course of land use and local economic development trajectories. The CGE
model is one tool that can be further utilized to structurally assess these complex inter-
actions with studies by Burnett et al. (2007) and Cutler and Davies (2007) representing
a good first step.
With income inequality reaching very high levels in the United States and elsewhere
(Atkinson, Piketty, and Saez 2011), another topic warranting more attention is how land
use and its interrelation with economic development affect poverty rates and income
inequality. The spatial mismatch literature shows that housing availability and employ-
ment access can affect employment outcomes for low-income households. Likewise,
land use decisions affect housing costs, which further affect income inequality.
Examining these issues requires better data. More work has been done with micro
geo-coded housing data using GIS than with geo-coded firm-level data, although
Greenstone, Hornbeck, and Moretti (2010) demonstrate the possibilities. Very little
research brings both geo-coded firm and household data together, although the plan-
ning literature is one exception (e.g., Funderberg et al. 2010).
Combined with the increased availability of GIS and microdata, and improved meth-
ods of empirical estimation and modeling, the spatial equilibrium approach offers sig-
nificant promise for increasing our understanding of the relationship between regional
economic development and land use issues. Without a greater understanding of the
connection between the two, regional economic development and land use policies may
prove to be ineffective or harmful.
References
Akai, N., and M. Sakata. 2002. Fiscal decentralization contributes to economic growth: Evidence
from state-level cross section data for the United States. Journal of Urban Economics
52: 93–108.
Ali, K., M. D. Partridge, and M. R. Olfert. 2007. Can geographically weighted regressions
improve regional analysis and policymaking? International Regional Science Review
30: 300–329.
Alonso, W. 1964. Location and land use. Cambridge, MA: Harvard University Press.
Angrist, J. D., and J. Pischke. 2009. Mostly harmless economics: An empiricist’s companion.
Princeton, NJ: Princeton University Press.
Angrist, J. D., and J. Pischke. 2010. The credibility revolution in empirical economics: How bet-
ter research design is taking the con out of econometrics. Journal of Economic Perspectives
24: 3–30.
Atkinson, A. B., T. Piketty, and E. Saez. 2011. Top incomes in the long run of history. Journal of
Economic Literature 49: 3–71.
Baldwin, R. E., and P. Krugman. 2004. Agglomeration, integration and tax harmonisation.
European Economic Review 48: 1–23.
Banzhaf, H. S., and R. P. Walsh. 2008. Do people vote with their feet? An empirical test of
Tiebout’s mechanism. American Economic Review 98: 843–863.
Barkley, D. L., M. S. Henry, and S. Shuming Bao. 1996. Identifying “spread” versus “back-
wash” effects in regional economic areas: A density functions approach. Land Economics
72(3): 336–357.
Bartik, T. J. 1991. Who benefits from state and local economic development policies? Kalamazoo,
MI: W. E. Upjohn Institute.
Beeson, P. E., and R. W. Eberts. 1989. Identifying productivity and amenity effects in interurban
wage differentials. Review of Economics and Statistics 71(3): 443–452.
Benson, E. D., J. L. Hansen, A. L. Schwartz, and G. T. Smersh. 1998. Pricing residential ameni-
ties: The value of a view. Journal of Real Estate Finance and Economics 16(1): 55–73.
Berry, B. J. L. 1970. Labor market participation and regional potential. Growth and Change
1: 3–10.
Blumenberg, E., and K. Shiki. 2004. Spatial mismatch outside of large urban areas: An analysis of
welfare recipients in Fresno County, California. Environment and Planning C: Government
and Policy 22: 401–421.
Boyle, R., and R. Mohamed. 2007. State growth management, smart growth and urban contain-
ment: A review of the US and a study of the heartland. Journal of Environmental Planning
and Management 50(5): 677–697.
Brakman, S., H. Garretsen, and C. van Marrewijk. 2009a. The introduction to geographical eco-
nomics. Cambridge, UK: Cambridge University Press.
Brakman, S., H. Garretsen, and C. van Marrewijk. 2009b. Economic geography within and
between European nations: The role of market potential and density across space and time.
Journal of Regional Science 49: 777–800.
Braun, G. O., and J. W. Scott. 2007. Smart growth as new metropolitan governance: Observations
on U.S. experience. In The international handbook of urban policy: Contentious global issues,
Vol. 1, ed. H. S. Geyer, 213–223. Cheltenham, UK: Edward Elgar.
Brown, S. P. A., and L. L. Taylor. 2006. The private sector impact of state and local government.
Contemporary Economic Policy 24(4): 548–562.
Brown, S. P. A., K. J. Hayes, and L. L. Taylor. 2003. State and local policy, factor markets and
regional growth. Review of Regional Studies 33(1): 40–60.
Brueckner, J. K. 2000. Urban sprawl: Diagnosis and remedies. International Regional Science
Review 23(2): 160–171.
Brülhart, M., and P. Koenig. 2006. New economic geography meets Comecon. Economics of
Transition 14: 245–267.
Burnett, P., H. Cutler, and R. Thresher. 2007. The impact of tourism for a small city: A CGE
approach. Journal of Regional Analysis and Policy 37(3): 233–242.
Card, D. 1990. The impact of the Mariel boatlift on the Miami labor market. Industrial and
Labor Relations Review 43: 245–257.
Carruthers, J. I. 2003. Growth at the fringe: The influence of political fragmentation in United
States metropolitan areas. Papers in Regional Science 82: 475–499.
Carruthers, J. I., and D. E. Clark. 2010. Valuing environmental quality: A space-based strategy.
Chen, A., and M. D. Partridge. 2011. When are cities engines of growth? Spread and backwash
effects across the Chinese urban hierarchy. Regional Studies doi:10.1080/00343404.2011.58
9831.
Chen, Y., E. G. Irwin, and C. Jayaprakash. 2009. Dynamic modeling of environmen-
tal amenity-driven migration with ecological feedbacks. Ecological Economics
68: 2498–2510.
Cho, S-H., N. C. Poudyal, and R. K. Roberts. 2008. Spatial analysis of the amenity value of green
open space. Ecological Economics 66(2–3): 403–416.
Cho, S-H., Kim, R. K. Roberts, and S. Jung. 2009. Amenity values of spatial configurations of
forest landscapes over space and time in the southern Appalachian highlands. Ecological
Economics 68(10): 2646–2657.
Christaller, W. 1933. Die zentralen Orte in Süddeutschland. Jena: Gustav Fischer. (Partial English
translation: 1966. Central places in southern Germany. Prentice Hall.)
Combes, P-P., G. Gilles Duranton, and H. G. Overman. 2005. Agglomeration and the adjust-
ment of the spatial economy. Papers in Regional Science 84: 311–349.
Costa, D. L., and M. E. Kahn. 2003. The rising price of nonmarket goods. American Economic
Review 93(2): 227–232.
Cutler, H., and S. Davies. 2007. The impact of specific-sector changes in employment on
economic growth, labor market performance and migration. Journal of Regional Science
47(5): 935–963.
Czamanski, D., I. Benenson, D. Malkinson, M. Marinov, R. Roth, and L. Wittenberg. 2008.
Urban sprawl and ecosystems—can nature survive? International Review of Environmental
and Resource Economics 2(4): 321–366.
Dalenberg, D. R., M. D. Partridge, and D. S. Rickman. 1998. Public infrastructure: Pork or jobs
creator? Public Finance Review 26(1): 24–52.
Deller, S., T-H. Tsai, D. Marcouiller, and D. English. 2001. The role of amenities and quality of
life in rural economic growth. American Journal of Agricultural Economics 83(2): 352–365.
Dumais, G., G. Ellison, and E. Glaeser. 2002. Geographic concentration as a dynamic process.
Review of Economics and Statistics 84: 193–204.
Duranton, G., and M. A. Turner. 2011. The fundamental law of road congestion: Evidence from
US cities. American Economic Review 101(6): 2616–2652.
Eichman, H., G. L. Hunt, J. Kerkvliet, and A. J. Plantinga. 2010. Local employment growth,
migration, and public land policy: Evidence from the northwest forest plan. Journal of
Agricultural and Resource Economics 35(2): 316–333.
Fallah, B., M. D. Partridge, and M. R. Olfert. 2011. Urban sprawl and productivity: Evidence
from U.S. metropolitan areas. Papers in Regional Science 90(3): 451–472.
Fotheringham, S., C. Brunsdon, and M. Charlton. 2002. Geographically weighted regression: The
analysis of spatially varying relationships. Chichester, UK: John Wiley & Sons.
Fox, K. A., and T. K. Kumar. 1965. The functional economic area: Delineation and implications
for economic analysis and policy. Papers of the Regional Science Association 15: 57–85.
Fujita, M., P. Krugman, and T. Mori. 1999. On the evolution of hierarchical urban systems.
European Economic Review 43: 209–251.
Funderberg, R. G., H. Nixon, M. G. Boarnet, and G. Ferguson. 2010. New highways and land
use change: Results from a quasi-experimental research design. Transportation Research
Part A 44: 76–98.
Gabriel, S. A., J. P. Mattey, and W. L. Wascher. 2003. Compensating differentials and evo-
lution in the quality of life among U.S. states. Regional Science and Urban Economics
33: 619–649.
Gaile, G. L. 1980. The spread-backwash concept. Regional Studies 14(1): 15–25.
Glaeser, E. L., and M. E. Kahn. 2004. Sprawl and urban growth. In Handbook of regional and urban
economics, Vol. 4, eds. V. Henderson and J. F. Thisse, 2481–2527. Amsterdam: North-Holland.
Glaeser, E. L., and M. E. Kahn. 2010. The greenness of cities: Carbon dioxide emissions and
urban development. Journal of Urban Economics 67: 404–418.
Glaeser, E. L., and K. Tobio. 2008. The rise of the sunbelt. Southern Economic Journal
74(3): 610–643.
Gottlieb, P. D. 1995. Residential amenities, firm location and economic development. Urban
Studies 32(9): 1413–1436.
Greenstone, M., R. Hornbeck, and E. Enrico Moretti. 2010. Identifying agglomeration spill-
overs: Evidence from million dollar plants. Journal of Political Economy 118: 536–598.
Gunderson, R. J., J. V. Pinto, and R. H. Williams. 2008. Economic or amenity driven migration?
A cluster-based analysis of county migration in the four corners states. Journal of Regional
Analysis and Policy 38(3): 243–254.
Gyourko, J., C. Mayer, and T. Sinai. 2006. Superstar cities. NBER Working Paper 12355.
Gyourko, J., and J. Tracy. 1989. The importance of local fiscal conditions in analyzing local labor
markets. Journal of Political Economy 97(5): 1208–1231.
Gyourko, J., and J. Tracy. 1991. The structure of local public finance and the quality of life.
Journal of Political Economy 99(4): 774–806.
Hammond, G., and T. Mehmet. 2011. The impact of decentralization on economic
growth: Evidence from U.S. counties. Journal of Regional Science 51(1): 47–64.
Hansen, A. J., R. Rasker, B. Maxwell, J. J. Rotella, J. D. Johnson, A. Wright Parmenter, U. Ute
Langner, W. B. Cohen, R. L. Lawrence, and M. P. V. Kraska. 2002. Ecological causes and con-
sequences of demographic change in the new west. BioScience 52(2): 151–162.
Hasse, J. E., and R. G. Lathrop. 2003. Land resource impact indicators of urban sprawl. Applied
Geography 23: 159–175.
Helpman, E. 1998. The size of regions. In Topics in public economics, eds. D. Pines, E. Sadka, and
I. Zilcha. Cambridge, MA: Belknap Press.
Henderson, J. R., and K. McDaniel. 2005. Natural amenities and rural employment
growth: A sector analysis. The Review of Regional Studies 35(1): 80–96.
Henry, M. S., D. L. Barkley, and S. Bao. 1997. The hinterland’s stake in metropolitan area growth.
Hering, L., and S. Poncet. 2010. Market access and individual wages: Evidence from China.
Hite, J. 1997. The Thunen model and the new economic geography as a paradigm for rural
development policy. Review of Agricultural Economics 19: 230–240.
Holmes, T. J. 1998. The effect of state policies on the location of manufacturing: Evidence from
state borders. Journal of Political Economy 106: 667–705.
Holmes, T. J. 2010. Structural, experimentalist, and descriptive approaches to empirical work in
regional economics. Journal of Regional Science 50: 5–22.
Houston, D. 2005. Methods to test the spatial mismatch hypothesis. Economic Geography
81(4): 407–434.
Hughes, D. W., and D. W. Holland. 1994. Core-periphery economic linkage: A measure
of spread and possible backwash effects for the Washington economy. Land Economics
70: 364–377.
Ihlanfeldt, K. R., and D. L. Sjoquist. 1998. The spatial mismatch hypothesis: A review of recent
studies and their implications for welfare reform. Housing Policy Debate 9: 849–892.
Innes, J. E., and D. E. Booher. 1999. Metropolitan development as a complex system: A new
approach to sustainability. Economic Development Quarterly 13: 141–156.
Irwin, E. G., A. M. Isserman, M. Kilkenny, and M. D. Partridge. 2010. A century of research
on rural development and regional issues. American Journal of Agricultural Economics
92(2): 522–553.
Johnson, M. P. 2001. Environmental impacts of urban sprawl: A survey of the literature and
proposed research agenda. Environment and Planning A 33: 717–735.
Kahn, M. E. 2001. City quality-of-life dynamics: Measuring the costs of growth. Journal of Real
Estate Finance and Economics 22(2/3): 339–352.
Kahn, M. E., and J. Schwartz. 2008. Urban air pollution progress despite sprawl: The ‘greening’
of the vehicle fleet. Journal of Urban Economics 63: 775–787.
Kahn, R., P. F. Orazem, and D. M. Otto. 2001. Deriving empirical definitions of spatial labor
markets: The roles of competing versus complementary growth. Journal of Regional Science
41: 735–756.
Kain, J. 1968. Housing segregation, negro employment, and metropolitan decentralization.

Quarterly Journal of Economics 82: 175–183.
Ke, S., and E. Feser, 2010. Count on the growth pole strategy for regional economic growth?
Spread–backwash effects in Greater Central China. Regional Studies 44: 1131–1147.
Keane, M. P. 2010. Structural vs. atheoretic approaches to econometrics. Journal of Econometrics
156: 3–20.
Kim, E., and J. Ju. 2003. Growth and distributional impacts of urban housing supply: An
application of urban land use and a CGE model for Seoul. Review of Urban and Regional
Development Studies 15(1): 66–81.
Kline, J. D., and R. J. Alig. 1999. Does land use planning slow the conversion of forest and agri-
cultural land? Growth and Change 30(1): 3–22.
Kreuter, U. P., H. G. Harris, M. D. Matlock, and R. E. Lacey. 2001. Change in ecosystem service
values in the San Antonio area, Texas. Ecological Economics 39: 333–346.
Krugman, P. 1991. Increasing returns and economic geography. Journal of Political Economy
99(3): 483–499.
Krugman, P. 2010. The new economic geography, now middle aged. Paper prepared for presen-
tation to the Association of American Geographers Meetings, Washington, DC, April 16,
2010. http://www.princeton.edu/~pkrugman/aag.pdf
Kuminoff, N. V., and A. S. Jarrah. 2010. A new approach to computing hedonic equilibria
and investigating the properties of locational sorting models. Journal of Urban Economics
67: 322–335.
Kuminoff, N. V., V. K. Smith, and C. Timmins. 2010. The new economics of equilibrium sorting
and its transformational role for policy evaluation. NBER Working Paper 16349.
Lewis, D., G. Hunt, and A. Plantinga. 2002. Public conservation land and employment growth
in the Northern Forest Region. Land Economics 78(2): 245–259.
Lösch, A. 1940. Die räumliche Ordnung der Wirtschaft. Jena: G. Fischer. English translation (of
the 2nd rev. ed.): The Economics of Location. New Haven, CT: Yale University Press, 1954.
Martin, R. W. 2004. Spatial mismatch and the structure of American metropolitan areas, 1970–
2000. Journal of Regional Science 44: 467–488.
Mattoon, R. H. 1995. Can alternative forms of governance help metropolitan areas? Economic
Perspectives (Federal Reserve Bank of Chicago) 19(6): 20–32.
McDonald, J. F. 2001. Cost-benefit analysis of local land use allocation decisions. Journal of
Regional Science 41(2): 277–299.
McDonald, J. F., and D. P. McMillen. 2000. Employment subcenters and subsequent real estate
development in suburban Chicago. Journal of Urban Economics 48: 135–157.
McGranahan, D. A. 2008. Landscape influence on recent rural migration in the U.S. Landscape
and Urban Planning 85: 228–240.
McGranahan, D. A., and T. R. Wojan. 2007. Recasting the creative class to examine growth
processes in rural and urban counties. Regional Studies 41(2): 197–216.
McGregor, P. G., J. K. Swales, and Y. P. Yin. 1996. A long-run interpretation of regional
input-output analysis. Journal of Regional Science 36(3): 479–501.
McMillen, D. P. 1996. One-hundred fifty years of land values in Chicago: A nonparametric
approach. Journal of Urban Economics 40: 100–124.
McMillen, D. P. 2004. Employment densities, spatial autocorrelation, and subcenters in large
metropolitan areas. Journal of Regional Science 44: 224–243.
McMillen, D. P. 2010. Issues in spatial data analysis. Journal of Regional Science 50: 119–141.
McMillen, D. P., and C. L. Redfearn. 2010. Estimation and hypothesis testing for nonparametric
hedonic house price functions. Journal of Regional Science 50: 712–733.
Mills, E. S. 1967. An aggregative model of resource allocation in a metropolitan area. American
Economic Review 57(2): 197–210.
Mills, E. S. 2006. Sprawl and jurisdictional fragmentation. Brookings-Wharton Papers on Urban
Affairs, 231–256.
Morgan, W., J. Mutti, and D. Rickman. 1996. Tax exporting, regional economic growth, and
welfare. Journal of Urban Economics 39: 131–159.
Mulligan, G. F. 1984. Agglomeration and central place theory: A review of the literature.
International Regional Science Review 9: 1–42.
Muth, R. 1969. Cities and housing. Chicago: University of Chicago Press.
Myrdal, G. 1957. Economic theory and underdeveloped regions. London: G. Duckworth.
Nechyba, T. J., and R. P. Walsh. 2004. Urban sprawl. Journal of Economic Perspectives
18(4): 177–200.
Olfert, M. R., and J. C. Stabler. 1999. Multipliers in a central place hierarchy. Growth and Change
30: 288–302.
Ottaviano, G., and D. Pinelli. 2006. Market potential and productivity: Evidence from Finnish
regions. Regional Science and Urban Economics 36: 636–657.
Overman, H., and S. Gibbons. 2010. Mostly pointless spatial econometrics? Unpublished man-
uscript, London School of Economics.
Partridge, M. D. 2010. The duelling models: NEG vs amenity migration in explaining US
engines of growth. Papers in Regional Science 89(3): 513–536.
Partridge, M. D., K. Ali, and M. R. Olfert. 2010. Rural-to-urban commuting: Three degrees of
integration. Growth and Change 41: 303–335.
Partridge, M. D., R. Bollman, M. R. Olfert, and A. Alasia. 2007. Riding the wave of urban growth
in the countryside: Spread, backwash, or stagnation. Land Economics 83: 128–152.
Partridge, M. D., and D. S. Rickman. 1998. Regional computable general equilibrium
modeling: A survey and critical appraisal. International Regional Science Review
21: 205–248.
Partridge, M. D., and D. S. Rickman. 2003. Do we know economic development when we see it?
Review of Regional Studies 33(1): 17–39.
Partridge, M. D., and D. S. Rickman. 2008. Does a rising tide lift all boats? Assessing
employment-poverty dynamics by metropolitan size and county type. Growth and Change
39: 283–312.
Partridge, M. D., and D. S. Rickman. 2010. CGE modelling for regional economic development
analysis. Regional Studies 44(10): 1311–1328.
Partridge, M. D., D. S. Rickman, K. Ali, and M. R. Olfert. 2008a. Employment growth in the
American urban hierarchy: Long live distance. B.E. Journal of Macroeconomics: Contributions
8(1): 1–36.
Partridge, M. D., D. S. Rickman, K. Ali, and M. R. Olfert. 2008b. Lost in space: Population
dynamics in the American hinterlands and small cities. Journal of Economic Geography
8: 727–757.
Partridge, M. D., D. S. Rickman, K. Ali, and M. R. Olfert. 2009. Agglomeration spillovers
and wage and housing cost gradients across the urban hierarchy. Journal of International
Economics 78: 126–140.
Partridge, M. D., D. S. Rickman, K. Ali, and M. R. Olfert. 2010. Recent spatial growth dynam-
ics in wages and housing costs: Proximity to urban production externalities and consumer
amenities. Regional Science and Urban Economics 40(6): 440–452.
Pflüger, M., and T. Tabuchi. 2010. The size of regions with land use for production. Regional
Science and Urban Economics 40: 481–489.
Pinske, J., and M. E. Slade. 2010. The future of spatial econometrics. Journal of Regional Science
50: 103–117.
Polèse, M., and R. Shearmur. 2004. Is distance really dead? Comparing industrial location over
time in Canada. International Regional Science Review 27: 431–457.
Portnov, B. A., and M. Schwartz. 2009. Urban clusters as growth foci. Journal of Regional Science
49: 287–310.
Puga, D. 1999. The rise and fall of regional inequalities. European Economic Review 43: 303–334.
Raphael, S., and M. A. Stoll. 2002. Modest progress: The narrowing spatial mismatch between
blacks and jobs in the 1990s. Washington, DC: The Brookings Institution.
Rappaport, J. 2007. Moving to nice weather. Regional Science and Urban Economics
37(3): 375–398.
Rappaport, J. 2008. Consumption amenities and city population density. Regional Science and
Rappaport, J. 2009. The increasing importance of quality of life. Journal of Economic Geography
9: 779–804.
Redding, S. J., and D. M. Sturm. 2008. The costs of remoteness: Evidence from German division
and reunification. American Economic Review 98: 1766–1797.
Redfearn, C. L. 2009. How informative are average effects? Hedonic regression and amenity
capitalization in complex urban housing markets. Regional Science and Urban Economics
39: 297–306.
Renkow, M. 2003. Employment growth, worker mobility, and rural economic development.
American Journal of Agricultural Economics 85: 503–513.
Richardson, H. W. 1976. Growth pole spillover: The dynamics of backwash and spread. Regional
Studies 5: 1–9.
Rickman, D. S. 1992. Estimating the impacts of regional business assistance pro-
grams: Alternative closures in a computable general equilibrium model. Papers in Regional
Science 71(4): 421–435.
Rickman, D. S., and S. D. Rickman. 2011. Population growth in high-amenity nonmetropolitan
areas: What’s the prognosis? Journal of Regional Science 51(5): 863–879.
Roback, J. 1982. Wages, rents, and the quality of life. Journal of Political Economy 90:
1257–1278.
Rosenberger, R. S., and D. B. K. English. 2005. Impacts of wilderness on local economic devel-
opment. In The multiple values of wilderness, eds. H. K. Cordell, J. C. Bergstrom, and J.
M. Bowker, 181–204. State College, PA: Venture Publishing.
Rosenberger, R. S., M. Sperow, and D. B. K. English. 2008. Economies in transition and pub-
lic land use policy: Discrete duration models of eastern wilderness designation. Land
Economics 84(2): 267–281.
Rosenthal, S. S., and W. C. Strange. 2004. Evidence on the nature and sources of agglomeration
economies. In Handbook of regional and urban economics, Vol. 4, eds. V. Henderson and J.
F. Thisse, 2119–2171. Amsterdam: North-Holland.
Rosenthal, S. S., and W. C. Strange. 2008. The attenuation of human capital spillovers. Journal
of Urban Economics 41: 373–389.
Shapiro, J. M. 2006. Smart cities: Quality of life, productivity, and the growth effects of human
capital. Review of Economics and Statistics 88(2): 324–335.
Song, Y. 2005. Smart growth and urban development pattern: A comparative study. International
Regional Science Review 28(2): 239–265.
Stansel, D. 2002. Interjurisdictional competition and local economic perfor-
mance: A cross-sectional examination of US metropolitan areas. In George Mason
University Working Papers in Economics, 02.16. (www.gmu.edu/departments/economics/
working/directory.html).
Stansel, D. 2005. Local decentralization and local economic growth: A cross-sectional exami-
nation of US metropolitan areas. Journal of Urban Economics 57: 55–72.
Stock, J. H., and M. W. Watson. 2007. Introduction to econometrics, 2nd ed. Boston: Pearson.
Stoll, M. A. 2006. Job sprawl, spatial mismatch and black employment disadvantage. Journal
Policy Analysis and Management 25: 827–854.
Stone, B, Jr. 2008. Urban sprawl and air quality in large US cities. Journal of Environmental
Management 86: 688–698.
Tabuchi, T., and J. F. Thisse. 2006. Regional specialization, urban hierarchy, and commuting
costs. International Economic Review 47: 1295–1317.
Tabuchi, T., and J. F. Thisse. 2011. A new economic geography model of central places. Journal
Tiebout, C. M. 1956. A pure theory of local expenditures. Journal of Political Economy
64: 416–424.
Tweeten, L. G., and G. L. Brinkman. 1976. Micropolitan development: Theory and practice of
greater rural development. Ames: Iowa State University Press.
Vias, A C. 1999. Jobs follow people in the rural Rocky Mountain west. Rural Development
Perspectives 14(2): 14–23.
Vias, A. C., and J. I. Carruthers. 2005. Regional development and land use change in the Rocky
Mountain west, 1982–1997. Growth and Change 36(2): 244–272.
Vigdor, J. L. 2006. Comment on Mills, sprawl and jurisdictional fragmentation.
Brookings-Wharton Papers on Urban Affairs, 231–256.
Volpe-Martincus, C. 2010. Spatial effects of trade policy: Evidence from Brazil. Journal of
Regional Science 50: 541–569.
Von Thünen, J. H. 1966. Von Thünen’s isolated state: An English edition of “Der Isolierte Staat.”
In C. M. Wartenberg (trans.), P. Hall (ed.). New York: Pergamon (Published in German in
1826.)
Weiler, S. 2003. Pioneers of rural sprawl in the Rocky Mountain west. The Review of Regional
Studies 33(3): 264–283.
Wensley, R. D. M., and J. C. Stabler. 1998. Demand-threshold estimation for business activities
in rural Saskatchewan. Journal of Regional Science 38: 155–177.
Whisler, R. L., B. S. Waldorf, G. F. Mulligan, and D. A. Plane. 2008. Quality of life and the migra-
tion of the college educated. Growth and Change 39(1): 58–94.
Wooldridge, J. M. 2002. Econometric analysis of cross section and panel data. Cambridge,
MA: MIT Press.
World Bank. 2009. World development report reshaping economic geography. Washington,
DC: The World Bank.
Wu, J., and S-H. Cho. 2007. The effect of local land use regulations on urban development in the
western United States. Regional Science and Urban Economics 37: 69–86.
Wu, J., and M. Gopinath. 2008. What causes spatial variations in economic development in the
United States? American Journal of Agricultural Economics 90: 392–408.
Wu, J. 2010. Economic fundamentals and urban-suburban disparities. Journal of Regional
Science 50: 570–591.
Yin, M., and J. Sun. 2007. The impact of state growth management programs on urban sprawl
in the 1990s. Journal of Urban Affairs 29(2): 149–179.
Yu, Y., and D. S. Rickman. 2011. U.S. state and local fiscal policies and nonmetropolitan
area economic performance: A spatial equilibrium analysis. Papers in Regional Science
doi:10.1111/j.1435-5957.2012.00423.x.
C HA P T E R 2
TECHNOLOGY ADOPTION
A N D L A N D U S E
DAV I D Z I L BE R M A N , M A DH U K HA N NA ,
S C OT T KA PL A N , A N D E U N IC E K I M
The adoption of new technologies in agriculture (such as new irrigation technologies,

crops for new biofuels, high-yielding seed varieties, etc.) have been crucial contributors
to technological change. Innovations have enabled support for a growing global popula-
tion, which has increased sevenfold from 1 billion in 1800, while increasing acreage by
only 150%. Adoption of new technologies has transformed agriculture from a labor- to
a capital-intensive industry in the developed world, and it is crucial for the progress
of agriculture in the developing world. It has also expanded the range of goods pro-
duced in agriculture, going beyond food and fiber to include biofuel and fine chemicals.
Much of the economic literature on adoption and diffusion originated from research
on adoption of technologies in agriculture, be it hybrid corn in the United States or
Green Revolution varieties (varieties discovered and implemented during the Green
Revolution) throughout the world.
The adoption of these innovations has drastically affected land use and land values.
Much of the adoption of innovations has been embodied in changes in land use. For
example, adoption may result in the growth of new varieties and crops, as well as in
the installation of new irrigation equipment. In turn, much of the literature on adop-
tion (Feder et al. 1985), to a large extent, models adoption decisions as land use choices.
Adoption of new innovations in certain locations will affect outputs and costs and, thus,
spatial patterns of land prices. Thus, understanding the economics of adoption of new
technologies and innovation in agriculture is important in studying agriculture land use
and its value.
This chapter first presents the basic theories of adoption. Then, it identifies how pat-
terns of adoption vary depending on the characteristics of these technologies. Next, it
assesses how various economic and noneconomic factors affect patterns of adoption.
And, finally, it considers how technology adoption and its economic implications are
evolving in a modern world with an integrated supply chain and contracting.
Technology Adoption and Land Use 53
1. Adoption and Diffusion
Adoption and diffusion are two processes that affect the introduction of technological
innovations. Adoption represents individual decisions regarding the technology and is
measured as a discrete choice; that is, whether or not a technology was adopted by a
farmer or used on a piece of land. But it may also be accompanied by measures of inten-
sity; that is, the extent to which adoption occurs (degree of land share devoted to new
varieties). When new technologies have multiple components, they may be adopted
jointly or sequentially (Khanna 2001). For example, adoption of Green Revolution vari-
eties is also associated with choice of complementary inputs, such as fertilizer. In this
case, both land share of the new technology and the extent of fertilizer use are measures
of adoption.
Diffusion represents aggregate adoption. One measure of diffusion is the fraction of
farmers who adopt a new technology, whereas another measure is the fraction of the
land that is switched to using the new technology. When new technologies are embod-
ied in capital goods, they are often rented, and purchase decisions are only made after
sufficient experiences are accumulated. In these cases, diffusion is measured by the use
of new technology rather than the ownership of equipment.
Early empirical studies of diffusion were conducted by sociologists, such as Rogers
(2003), who collected data on aggregate adoption of different technologies and found
that diffusion was an S-shaped function of time, reflecting slow initial diffusion, then
a period of takeoff, and then an eventual tapering off. Rogers established the imitation
model, and, assuming homogeneity among farmers, he was able to model the spread
of a technological innovation as a process of imitation, which is similar to the spread
of an epidemic. In particular, if p(t) is the land share of the new technology over time,
K
then P (t ) = − ( α + βt )
, where K is the maximum diffusion rate, α is a measure of the
1+ e
initial rate of adoption, and β is the measure of the speed of adoption. Griliches (1957)
expanded the Rogers model by suggesting that the relative profitability of new technolo-
gies affects the speed of imitation. The more profitable the new technology, the faster the
imitation, the steeper the slope of the S-shaped curve (higher β), and the larger the value
of the maximum adoption, K.
David (1975) and Feder, Just, and Zilberman (1985) argue that the imitation model
does not include an explicit economic decision-making model, and so David introduced
the threshold model. The threshold model incorporates three major components. First,
farmers consider multiple factors in making economic decisions, including profit, util-
ity, risk, and other criteria. Second, it takes into consideration heterogeneity of farm size,
human capital, and/or land quality. Third, it is a dynamic model. Frequently, studies have
assumed static profit maximization or expected utility maximization by the decision
maker. Recent studies have assumed dynamic optimization, with the timing of adoption
54 David Zilberman Et al.
being determined by considering the tradeoff of benefits from use in the present with
reduced prices as production expands in the future (McWilliams and Zilberman 1996).
Sometimes, the dynamic processes that affect returns or costs are stochastic, such as
additive and multiplicative random walk. In these cases, decision makers are taking a
real option approach; thus, timing of adoption is selected so that marginal benefit over-
comes marginal cost plus the hurdle rates that increase with uncertainty (Khanna et al.
2000; Seo et al. 2008). The threshold model emphasizes the importance of the effective
rollout of a technology, as well as its introduction in locations with the highest returns
and willingness to experiment with the product. People who adopt early are those who
have the most favorable conditions. But, over time, a new technology may become more
attractive because of learning by doing (i.e., knowledge acquisition from experience in
production of a product), learning by using (i.e., learning through use of a technology),
or network externalities, causing more adopters to join in. When there is partial adop-
tion, increase in adoption over time may be both at the intensive and extensive margins.
For example, over time, the adoption of Green Revolution varieties may expand because
adopters increase the relative land share used by the technology (intensive margin) and
because nonadopters enter and allocate land to the technology (extensive margin). In
the case of mechanical innovation, larger scale farmers will adopt first, but as a technol-
ogy becomes cheaper and custom services are developed, smaller farmers will adopt the
technology (Sunding and Zilberman 2001). In the case of drip irrigation, the sources of
heterogeneity are represented by the differences in water-holding capacity of the soil,
and adoption occurs on land that previously utilized traditional technologies, as well as
on land with low water holding capacity that was unable to be used previously. Adoption
of technologies such as drip irrigation and pesticides tends to increase the acreage of
usable agricultural land by adding land that could not be utilized previously because of
water or pest constraints. With adoption, the value of this land increases as a result of its
new use.
The threshold model emphasizes the importance of heterogeneity among farmers
and has been applied using data on technology, as well as on land use choices at the
plot or the farm level. Discrete-choice econometric approaches (using a probit or logit
model) are used to explain factors that affect the selection of specific divisible technol-
ogies by farmers (e.g., whether a farmer uses a tractor), whereas the use of the Tobit
model allows for the investigation of situations in which adoption is partial. This can
be seen when farmers allocate some of their land to Green Revolution technologies as
opposed to traditional technologies. Panel data on changes in technology choice and
land use over time, at the plot or farm level, identify sources of heterogeneity, as well as
the patterns of the evolution of diffusion and adoption (Sunding and Zilberman 2001).
Studies have also used treatment effect models to analyze the effects of adoption on land
use or input use (Khanna 2001).
2. Technology Characteristics
The threshold model emphasizes the importance of heterogeneity in the adoption

process, and the characteristics of the technology determine the source of heterogene-
ity and its impacts on land use and other key factors. Several characteristics of special
importance are described in the following sections.
2.1 Divisibility
Some technologies are embodied in indivisible equipment—like tractors or combines—
whereas others, like new seed varieties, are divisible. When an indivisible technology
has to be purchased, scale becomes the dominant source of heterogeneity. If a technol-
ogy that requires a fixed annual cost of Ft dollars and increases profit per acre in period
t by Δπt, then profit-maximizing firms of farm size Lt = Ft / ∆π t greater than the critical
farm size will adopt the technology at time t, showing farm size as a source of heteroge-
neity. The dynamics of diffusion will be affected by the distribution of both farm size and
critical size. Learning by doing, which acts to reduce Ft, and learning by using, which
acts to increase Δπt, will reduce Lt and drive diffusion. We plausibly assume that the
farm size distribution is unimodal and the diffusion curve is S-shaped (Sunding and
Zilberman 2001). Farm size distributions can be altered, and when technologies are
indivisible, owners of small farms have to expand the size of their operation to make
adoption profitable.
Thus, the distribution of land among farms may affect the timing and dynamics of
diffusion, and the introduction of new technologies may alter farm size distribution.
The introduction of indivisible technologies may have contributed to increases in aver-
age farm size in the United States and other Organisation for Economic Co-operation
and Development (OECD) countries. One mechanism that enables smaller farmers to
benefit from large, nondivisible machinery has been the introduction of custom service
provision. In locations where there are fewer barriers to the establishment of such ser-
vices, farmers can benefit from the technology without buying it, and the diffusion rates
measured by percentage of land used with machinery are much higher. Furthermore,
when farmers are uncertain about benefits of a technology, the introduction of rental
services allows them to gain experience with the technology prior to purchasing it.
In the case of divisible technologies, adoption may be partial. Farmers may adopt
new crop varieties or a pest control treatment on part of their land first, and then vary
the land share over time. Even in cases of technologies that are seemingly divisible, like
new seed varieties, farmers have fixed costs of learning and adjustment. Thus, a certain
amount of scale is needed to adopt some of these technologies, especially early in the
innovation process, in order to cover these fixed costs.
2.2 Impact on Input Use Efficiency

Technologies vary in the efficiency of the use of variable inputs, such as water or fertil-
izer. Caswell and Zilberman (1986) distinguished between applied and effective input
and mention that the ratio of applied to effective input is input use efficiency. Input use
efficiency varies across locations and among technologies. For example, traditional irri-
gation, such as furrow irrigation, may have an input use efficiency of 0.6 in relatively
heavy soil and 0.1 in sandy soil, whereas the efficiency of drip irrigation may be 0.95 in
heavy soil and 0.85 in sandy soil. The residual input (irrigation water runoff or pesticide
residue) is frequently a pollutant, and adoption of technologies that increase input use
efficiency leads to improved pollution control.
Input use efficiency augmenting technologies include improved pesticides and
chemical application and fuel efficiency, and their adoption may be induced by higher
input prices or environmental regulations (Khanna and Zilberman 1997). They tend to
require higher fixed cost per unit of land but often increase operational profits (revenue
minus variable costs per acre). Related technology types include damage control agents
(e.g., various forms of disease and pest controls; Lichtenberg and Zilberman 1986), soil
erosion control technologies (Ervin and Ervin 1982), and input augmenting technolo-
gies, notably the introduction of irrigation to augment rainfall or the use of synthetic
fertilizers.
Technologies that augment input use efficiency enable farmers to overcome the limi-
tations imposed by low land qualities. These technologies tend to be adopted first on
locations of lower land quality, which is measured by the input use efficiency while using
the traditional technology. They also tend to affect land prices significantly. In particular,
they may lead to reductions in the premium for land of higher quality, as is demonstrated
in Figure 2.1. The traditional technology has higher profits on high-quality lands because
it does not require the extra fixed costs associated with the new technology, so the gain
from the modern technology is relatively small. Before the introduction of the modern
technology, all of the land with qualities in the range B–D utilized traditional technol-
ogy, whereas the lower quality lands did not. After the introduction of the modern tech-
nology, lands in the range B–C were switched from using the traditional technology
to using the modern technology, and land in the range A–B was added to production.
Before the introduction of the modern technology, profit per acre, which represents land
rent, was denoted by the curve BF. After the introduction, the rent per acre is denoted
by the curve AEF, which reflects higher premiums for lower quality land. If adoption of
new technologies increases supply and reduces output price, then low-quality land that
enters production tends to gain from the technology and high-quality lands tend to lose.
Thus, a decline in the land quality premium is evident. The survey of Schoengold and
Zilberman (2007) confirms that adoption of water conserving technologies (e.g., drip
irrigation) increases input use efficiency, tends to increase yield, leads to reduced drain-
age, and, in some cases, water use per acre decreases. Ward and Pulido-Velazquez (2008)
showed that the expansion in acreage may increase water use after adoption of modern
irrigation technologies despite a reduction in per-acre use.
Profit per
Acre, $
Profit with traditional technology

F
E
Profit of modern technology
A B C D Land quality
FIGURE 2.1 The effect of technology adoption on land use.
Introduction of new technologies alters the relative value of land and may lead to
expansion of farmland into areas that weren’t previously utilized; this expansion is called
the extensive margin effect of adoption. An example of increasing returns to scale due
to the extensive margin effect of adoption is seen with the invention of pumps. Before
the invention of pumps, areas located below rivers were considered superior to areas
located above rivers because canals could irrigate them, but pumps raised the value of
land located above rivers and expanded farming to these areas. Drip irrigation increased
the relative value of sandy soil that has low water-holding capacity and led to farming in
areas that were previously deserted (Caswell and Zilberman 1986).
Each of these technology categories must overcome constraints but may lead to the
expansion of agricultural land, change the relative prices of land, and may actually
turn inferior land into superior land (as is the case in the introduction of irrigation to
California’s central valley).
2.3 Impact on Risk

Agriculture is subject to a high degree of variability. When farmers are risk averse, they
will pay high premiums to avoid it and, in turn, evaluate technologies by not only their
impact on average profit but by their riskiness as well. Just and Pope introduced the
Just-Pope production function, y = f (x ) + g (x )ε , where y is output and x is input, ε is
a random variable with a mean of 0, f(x) denotes impact of input on average output,
and g(x) is impact on risk. If g(x) is positive, then the technology is risk-enhancing. The
adoption of risk-reducing innovations or crop insurance programs was modeled using
a portfolio in which land and other inputs are allocated among risky alternatives. Some
technologies, including Green Revolution varieties that both enhance average profit
and risk, were selected by risk-averse farmers to diversify their land portfolio among
varieties in order to balance overall risk with expected gain. One of the advantages of
irrigation is that it both increases yield and reduces risk, so if the adoption cost is suf-
ficiently low, then irrigation technologies stochastically dominate dry farming. One
advantage of genetically modified (GM) cotton is that it both increases yield and reduces
risk, and the fixed cost associated with adoption is more than recovered in regions with
sufficient pest damage (NRC 2010).
2.4 Transport Cost Intensity

The von Thunen model established the importance of transportation cost consideration
in the allocation of agricultural activities. When farmers adopt technologies at a given
location, they consider the price of transportation costs for both outputs and inputs.
Thus, the adoption of technologies that reduce transportation costs may contribute to
changes in land use patterns and the introduction of new practices where they didn’t
exist before. Both the railroad and steamships allowed expansion of grain production
in the Midwest. The introduction of refrigerated trucks helped shift the production of
fruit and vegetables to California. The introduction of refrigerated air-freight facilitated
the adoption of intensive cash crops in various parts of Africa. Adoption of improved
transportation technologies outside of the farm affects land use and land value within
farming regions.
2.5 Complementary Technologies

Production may consist of several complementary processes, which include pest con-
trol, irrigation, and fertilization. Namely, a reduction in pest damage will increase the
value of fertilization or irrigation. Precision farming involves a bundle of technologies,
such as soil testing, variable rate fertilizer application, and yield monitoring. In some
cases, adoption of technologies that affect one process of production may not be profit-
able, but adoption of a package of complementary components may be profitable. The
Green Revolution consisted of many technology packages that combined new varieties
with modern inputs like fertilizer and irrigation. However, the various components may
be sold individually or as a combined package. Not all farmers will adopt all components
of the package at once (Byerlee and de Polanco 1986), and, in fact, farmers often prefer
to adopt technologies sequentially based on risk considerations, supply constraints, and
due to a lack of knowledge. Khanna (2001) found that although adoption of soil test-
ing for fertilizer requirements of the land was scale neutral, the subsequent adoption of
variable-rate fertilizer applicators was more likely to take place with larger, more experi-
enced and innovative farmers with greater human capital skills. Some technology pack-
ages combine improved modes of transportation with higher value crops, and together
they increase land values in remote regions. For example, the adoption of air-freight for
high value cash crops was essential to the adoption of such crops in Africa.
2.6 Economies of Scale and Scope

Industries characterized by decreasing returns to scale lead to competitive market out-
comes, whereas those characterized by increasing returns to scale favor oligopolistic
structures (Arthur 1994). Notably, technologies with increasing return to scale have
completely different patterns of evolution and adoption than do traditional technolo-
gies with increasing marginal cost. Most crop production technologies have decreasing
returns to scale, which has led to a primarily competitive structure. But the minimum
cost associated with crop production has increased over time, leading to increased farm
size and a decline in the number of farms. In some sectors of animal agriculture, the
least-cost scale grew immensely, leading to concentration and emergence of an oligopo-
listic industrial structure. The high cost of investment in livestock resulted in institu-
tional innovations in management and finance, as well as in the emergence of industrial
agriculture (Boehlje 1999).
3. Economic Considerations Affecting

Adoption
Technology adoption decisions are basically investment decisions. Assume that a new
technology requires an investment of I dollars and has a life horizon of T years. At the
beginning of each year, the farmer has to allocate his or her land (L between the tra-
ditional L0t and the modern technology L1t ). The modern technology is also likely to
change output (∆ yt) as well as input use (∆ xt ) and pollution (∆ zt ), but these impacts are
uncertain. The prices of output, input, and pollution at time t are pt, wt, and vt, respec-
tively. The change in profit at period t is ∆π t = ( pt ∆yt − wt ∆xt − vt ∆zt )L1t. Basic model-
ing suggests that a risk-neutral farmer will adopt the technology if its net present value
 T ∆π t 
∑
 t =0 (1 + r )
t − I  is positive. This model suggests that the likelihood of adoption increases;

as the discount rate and initial investment for the farmer become lower, the planning
horizon, the price of output (if the technology increases yield), the price of input (if the
technology saves input), and the pollution penalty (if the technology reduces pollution)
become higher. The analysis suggests that larger farms are more likely to adopt the tech-
nology and that larger initial investments, as well as higher discount rates, increase the
critical farm size for adoption.
The net present value approach emphasizes the role of financial incentives in induc-
ing adoption. Linn (2008) showed that financial incentives have a positive effect on
adoption of energy-conserving technology, but the elasticity of adoption in response
to financial incentives is low. Thus, financial incentives alone are not significant in
determining the feasibility of adoption, which points to the need to incorporate other
considerations that may affect adoption choices. These considerations may include
imperfect capital markets, risk aversion, and government policies, which are discussed
in detail in the next sections.
3.1 Credit Constraints

Potential adopters may need to finance the high up-front costs of new technology, as
well be willing to burden negative income streams associated with the establishment
phase of adoption of these technologies. The ability to finance investment in new tech-
nologies is constrained and is highly correlated to both a borrower’s wealth and the
capacity to pledge assets as collateral (Foster and Rosenzweig 2010b). Both transac-
tion costs and asymmetric information have been major causes of credit constraints in
which farmers are unable to get loans that can repay themselves, including for adop-
tion of innovations (Stiglitz and Weiss 1981). Credit provision and even subsidies
have been crucial for small farmers to adopt new technologies, particularly during
the initial stage of the Green Revolution in the late 1960s and 1970s (Fan et al. 2008).
Credit can be attained by the use of collateral, and, frequently, land is used for this
purpose. Thus, higher prices of land expand credit availability and may facilitate adop-
tion of new technologies that, in turn, may raise land value even further (Hochman
et al. 1977).
3.2 Risk Consideration

An extensive literature shows that risk considerations affect the technology adoption
decision, and differences in risk preferences and attitudes across individuals also lead
to heterogeneity in the adoption decision. Early studies of adoption under risk apply
safety rules, including the safety-first rule, which suggests that farmers and other land
users will select technologies that minimize the probability of a disaster—defined as a
situation in which their income falls below a subsistence level. The safety first rule was
used by Roumasset (1976) to explain why farmers in certain parts of Asia did not adopt
Green Revolution rice varieties.
The second, more widely used framework is based on the expected utility model,
which assumes that farmers are aware of the riskiness of different technologies and
account for it in calculating expected benefits. Frequently, adoption choices are analyzed
as optimal land portfolio management. Applications are based on the assumption that
farmers value higher profits but associate negative benefits to the riskiness (frequently
measured by variance) of those profits. Risk has been a major cause for diversification of
land among divisible technologies. Let L1 be the land area allocated to the modern tech-
nology of farm size L and μ0, σ02, μ1, σ12 the mean and variance of profits per acre of the
traditional and modern technologies, respectively, and σ12 the correlation between the
profits per acre for the two technologies. Just and Zilberman (1983) found that the area
allocated to the modern technology is
µ1 − µ0 σ12 − σ 02
L1 = − L, (1)
φ(σ12 − σ 02 + 2σ12 ) (σ12 − σ 02 + 2σ12 )
where ϕ is the measure of risk aversion, assuming that the modern technology has
higher mean and variance of profits. Equation (1) suggests that more land will be allo-
cated to the new technology the higher the yield gain is from this technology, the smaller
the risk aversion of this new technology, and the riskier the traditional technology is
relative to the new technology. The equation emphasizes the role of correlation in land
allocation; adoption of the new technology will increase as the correlation between the
traditional and new technology becomes smaller.
There is heterogeneity in the degree of risk aversion and loss aversion across farm
sizes and farmer wealth. Studies have found that larger and richer farmers tend to allo-
cate more acreage to a modern technology, but, in some cases, the land share of modern
technology is higher on smaller farms. For example, Marra and Carlson (1990) show
that the pattern of adoption of double cropping soybeans with wheat in the United States
is consistent with risk aversion and the covariance of returns between the old and new
technologies. In making decisions about allocating land for crops or other products,
farmers have to consider a number of risks, such as variability in yields due to weather,
difficulties in establishing the crop, and volatility in prices because of variable demand
and supply. High returns from other possible uses of the land may also play a primary
role in the farmers’ willingness to adopt a given technology.
A third approach is based on prospect theory (Kahneman and Tversky 1979) and has
more predictive power than expected utility theory in explaining decisions to adopt new
technologies, under certain conditions (Zellner and Zilberman 2011). The three key
features of prospect theory are (1) loss aversion, which implies that farmers are more
sensitive to losses below a reference; (2) framing of alternatives, namely specific simpli-
fication of risky alternatives considered in adoption choices; and (3) the difference in
perceived risk used for decision making and the actual associated risks. An empirical
study conducted by Malawi, Smale et al. (1994) shows that land use allocations between
new and old crops are explained by risk management strategies that combine portfolio
diversification, safety-first rules, and experimentation. Similarly, Huang and Liu (2013)
showed that both risk and loss aversion may delay the adoption of GM cotton in China.
A fourth approach to include risk in technology adoption is the real option approach
developed by Dixit, Pindyk, and Davis (1994). Whereas the other three approaches are
based on static analysis, Dixit, Pindyk, and Davis view adoption as a dynamic invest-
ment decision but also suggest that, instead of using net present value to decide whether
or not to adopt a technology at a given time, decision makers have another degree of
freedom—they can also determine the timing of the adoption. For example, if there is
uncertainty in the properties of the technologies or the behavior of prices in the future,
there may be gains (option-value) from taking advantage of waiting until uncertainties
are clarified. McWilliams and Zilberman (1996) showed that when prices of new tech-
nologies tend to decline over time, seen in the case of adoption of computers, optimal
timing balanced the gains from the decline of prices versus the loss of the services of
the new technology. Carey and Zilberman (2002) show that when considering adop-
tion of new irrigation technologies when water pricing is fluctuating, the critical price of
water that triggers adoption is the critical price under certainty plus a “hurdle rate” that
takes into account the fluctuation of water prices. Greiner, Patterson, and Miller (2009)
combined option-value consideration and risk aversion to explain barriers of adoption
of conservation technologies in Australia. Khanna et al. (2000) and Isik et al. (2001)
show that uncertainty about output prices and expectations of declining fixed costs of
adoption can create incentives to delay investment in precision technologies, particu-
larly on components that have relatively high fixed costs. This is particularly the case on
land parcels with low soil quality and low variability in soil quality, where the benefits
of adopting these technologies are relatively small. Thus, one venue through which risk
and uncertainty affect land use and land values is through their impact on technology
adoption. Risk consideration will affect adoption, and, at the same time, adoption of
new technologies will affect the magnitude of risk.
3.3 Information and Learning

The uncertainty about new technologies declines over time as farmers acquire knowl-
edge on their own and from others. A recent study by Conley and Udry (2010) suggests
that individuals tend to learn from the experience of members of the community and
adopt the practices used by successful individuals. As agriculture modernizes, farmers
become more specialized and information increases immensely, causing farmers to rely
on various sources for their decision making. For example, when it comes to informa-
tion on new technologies, farmers reported that the most important sources of infor-
mation were the agricultural media, informal sources (other farmers), extension, and
commercial vendors (Wolf et al. 2001). Although much of the information on new tech-
nologies is provided by formal networks, Just et al. (2002) estimated that roughly 50% of
the information farmers used for economic decision making comes through informal
networks that are often perceived to be less reliable than formal networks. Expansion
of formal networks associated with the increased information provided through the
Internet is likely to increase adoption. Internet-based information and e-commerce are
also improving the quality of information available to places farther from the urban cen-
ter, thus reducing the distance barrier to adoption.
Adoption of information technologies themselves are motivated by the benefits of
the network externalities they create (Shapiro and Varian 1999). These technologies
enable farmers to develop virtual networks of buyers and sellers that allow them to
negotiate volume discounts and obtain better prices for their output, sometimes using
e-marketing (Schmitz et al. 2005).
4. Sectoral Policies and Institutions
The agricultural sector has historically been characterized by institutional arrangements

to reduce risk, as well as by government policies to stabilize farm income. Several differ-
ent policies play a big role in the adoption of new land use technologies.
4.1 Crop Insurance

Crop insurance tends to increase adoption and even intensify adoption of high-yield,
high-risk varieties because it reduces risk while at the same time increases mean
yield (Just and Zilberman 1988). Farmers are more willing to incorporate additional
land with insurance programs because it provides a safety net for the risk involved
in marginal land use. Empirically, there is some evidence linking crop insurance to
adoption. The study by Christiaensen and Dercon (2010) found that lack of insurance
and ability to smooth consumption discouraged adoption of fertilizer by farmers in
Ethiopia.
4.2 Price Insurance Scheme

The low elasticity of agricultural demand combined with variability in supplies leads to
significant price fluctuations (Gardner 1987). Several institutional mechanisms reduce
price risks and future markets, which are markets where farmers can sell a contract of
a given level of output at a particular price. Theoretically, reduced risk will enhance
adoption, but more empirical evidence is needed. Price support policies provide a
floor on the price received by farmers. Price supports both increase the average price
expected by farmers and reduce risk of adoption of technologies by risk-averse farm-
ers. Availability of price insurance schemes tends to increase land values, and that may
increase adoption through availability of credit. Several studies have suggested that
expansion of agricultural supply through adoption and land use intensification over
the years was related to price support programs that reduce risk and increase average
profit (Gardner 1987).
4.3 Decoupled Income Support

Farmers place increasing reliance on “decoupled” payments; namely, payments that
assure a certain income regardless of actual production choices and yields. These pay-
ments assure a certain income based on historical planting decisions or regional income
average and are independent of actual choices. Under risk neutrality, decoupled support
is neutral in its impact on crop choices, so the way land is allocated to different crops is
unaffected. However, it can affect a farmer’s decision to remain in agriculture or use the
land for nonagricultural activities. Serra et al. (2005, 2006) show that decoupled policies
reduce farmers’ aversion to risk through the wealth effect and contribute to the intensi-
fication of farming.
4.4 Credit Subsidies

Lack of credit has been documented as a major constraint on adoption, especially by
small farmers. Governments established policies to overcome credit constraints. Giné
and Yang (2009) show that credit subsidies enhance the adoption of modern corn
technologies in Malawi. Hochman et al. (1977) suggest that credit supports enable
further adoption of waste management technologies in the context of animal waste in
California.
4.5 International Trade Policies

Tariffs and export subsidies, foreign exchange insurance, and exchange rates have sig-
nificantly affected the evolution of global agriculture (Schuh 1974). Government policy
may enhance adoption that expands supply of exporting industries by instituting export
subsidies, as well as by policies and regulation to reduce transportation costs. Trade bar-
riers on imports may lead to expansion of domestic industries to enhance input sub-
stitution. Foreign direct investment has been associated with the introduction of new
technologies, especially in developing countries, but there has been concern about for-
eign ownership of land, especially in Africa, and the tradeoff between increased devel-
opment and “neo-colonialism” (Cotula et al. 2009).
4.6 Regulations
Firms and farms are subject to various regulations, including worker safety regulations,
environmental regulations, and the like that can affect the costs and returns to alterna-
tive technologies. Regulations can increase the net gains from adopting environmen-
tally friendly technologies, rewarding farmers for reducing externalities associated with
land use. Some regulations are performance based, a criterion that may constrain the
outcomes of economic activity (e.g., upper bounds on concentration of chemicals in
water disposed by farms), whereas other regulations are practice based, which limit or
even ban specific practices. Lichtenberg (2002) documents how environmental regu-
lations (water quality regulations, runoff controls, pesticide residue regulations) led to
adoption of conservation and precision technologies. Casey et al. (1999) describe the
role that regulations have played in inducing adoption or disadoption of technologies.
Khanna et al. (2002) show that cost share subsidies and input reduction subsidies can
induce greater adoption of modern irrigation technologies than pollution taxes that
achieve the same level of pollution abatement.
4.7 Supply Chains and Contracting

The work by James Jr. et al. (2011) suggests that farmers in the United States and globally
are relying less on cash transactions and more on contractual relationships and vertical
integration for managing exchange. According to MacDonald and Korb (2008), 40%
of the value of US agricultural production is sold through contractual arrangements.
The deployment of contracting and vertical integration varies by crops and activities,
but it affects technology adoption, land allocation, and land value. It is also useful to
distinguish between marketing and production contracts. Marketing contracts specify
the price and/or quantity of the product sold by the farmer, as well as the condition of
delivery. In production contracts, the contractor owns the commodity when it is being
produced by the farmer and pays the farmer for services provided.
There are divisions of responsibilities between the farmer and contractor. For exam-
ple, the contractor may provide genetic materials and specialized inputs (feed for live-
stock), whereas the farmer may own specialized capital (farm, structures, land) and
conduct production activities subject to specified conditions from the contractors. The
farmer is paid a fee for services provided, rather than the market value of the output pro-
duced, although this fee may depend on the output’s market value.
MacDonald and Korb (2011) suggest that the nature of the product and the technol-
ogy used determine the use of contracting or vertical integration. For example, in 2008,
around a quarter of the total corn crop, 90% of sugar beets, and 68% of hogs were pro-
duced under production contracts. MacDonald and Korb (2011) show that contracting
enabled producers to assume more debts and that the use of contracting varies among
regions. James Jr. et al. (2011) suggest that establishment of contractual relationships, as
well as vertical integration, is associated with multiprocess production systems, where
each stage requires specialized capital. Thus, the introduction of contractual relation-
ships affects patterns of technology choice and land use. There is not much research
explicitly addressing the design of contracts as part of a technology diffusion process.
However, the literature on contracts (Alexander et al. 2011) emphasizes that they have
to be flexible enough to accommodate heterogeneity among farmers, which will lead
to variations in the responses to contracts by farmers. Barry et al. (1992) suggest that
the institutional designs for agricultural production systems have to take into account
financial considerations. Contractual arrangements are likely to increase the ability to
borrow, as well as affect the scale of operation of a system. James Jr. et al. (2011) sug-
gest that marketing contracts are associated with modification of existing production
systems, whereas production contracts and vertical integration are associated with the
introduction of new production systems. For example, the introduction and adoption of
new crops (e.g., kiwi fruit) or new products (e.g., broilers) occurred under vertical inte-
gration or production contracts.
The deployment of contract farming has been increasing globally, and the terms of
these contracts are shaped by the conditions of the country and the product (Rehber
1998). In many developing countries, the introduction of contracts was part of the
introduction of new products or new production systems that aimed at improving prod-
uct quality. This improved quality allowed market expansion of various products such as
fresh fruit, vegetables, and meats and the introduction of new technologies, such as the
enclosed industrial systems for producing poultry. In the developing world, contracts
are used as coordination mechanisms in terms of quality, quantity, and time, and they
provide incentives for performance, as well as provide protection against financial risks.
The growing importance of an integrated supply chain in agriculture suggests that more
emphasis should be placed on studying technology adoption and land use choices under
contracts and vertical integration.
5. Noneconomic Factors
There is growing evidence that adoption choices depend not only on monetary benefits,
but also on nonpecuniary benefits as well. The household production function litera-
ture spawned by Becker (1965) and Lancaster (1966) has shown that households make
choices that consider both market and nonmarket goods and consider factors such as lei-
sure, health, aesthetic beauty, and lifestyle in allocating resources, including technology
adoption. Marra and Piggot (2006) document that one major reason that farmers have
adopted GMO varieties in the United States, sometimes in spite of low-yield gains, is that
they entail less health risk, environmental damages, and effort than traditional varieties.
The theory of planned behavior (Ajzen 1991) has been used to understand decision
making by agricultural producers. This theory considers attitudes, subjective norms,
and perceived behavioral controls to be the primary determinants of behavioral inten-
tions, and it seeks to understand the factors that determine these behaviors.
5.1 Attitudes
There is not much quantitative assessment on the impact of attitude on adoption.
Positive attitudes among farmers toward environmentally friendly practices have led to
the adoption of crop rotations, sustainable agriculture, soil conservation practices, and
best management practices in dairy farms (Villamil et al. 2008).
5.2 Social Perception

Positive attitudes toward the adoption of a practice may not always be sufficient to
induce adoption. Perceived inability to successfully adopt the practice and social
pressures from important reference groups were major factors contributing to the
inability of farmers to convert positive attitudes into adoption of a technology. The effect
of social norms regarding visual appearance of a crop is an indicator of the success of a
farmer and may affect technology choices (Villamil et al. 2008). Social perception and
attitudes may trigger adoption of technologies that will benefit the community as a
whole, which may result in an increase in the value of land in the community.
6. Demographic and Socioeconomic

Characteristics
Farmer demographic and socioeconomic characteristics are expected to influence

technology adoption for a number of reasons. First, they are an indicator of hetero-
geneity among adopters that may affect the economic gains and costs of adoption.
Following the threshold model of adoption, these characteristics could influence the
dynamics of the diffusion process. Second, these characteristics can be correlated with
farmer attitudes, and knowing the extent to which those attitudes influence inten-
tions (following the theory of planned behavior), they could affect adoption decisions.
Some of the key demographic and socioeconomic factors that affect adoption are
described here.
6.1 Human Capital

Nobel laureate Theodore Schultz distinguished between two types of human capi-
tal: “worker ability,” which is the capacity to perform hard manual tasks, and “alloc-
ative ability,” or the “ability to deal with disequilibrium,” namely, the ability to assess
problems, make rational choices, and adjust to change (Schultz 2003). The adop-
tion of more advanced technologies and their effects on land use is clearly related
to allocative ability; however, this ability is not easily observable. One good proxy
is education. The literature on adoption of various technologies—computers, new
seed varieties, machinery, and better management systems—shows that more
educated farmers are early adopters (Sunding and Zilberman 2001). On the other
hand, some innovations, such as pesticides and management consulting, are human
capital augmenting technologies and are more likely to be adopted by human
capital-challenged or less educated farmers. New seed varieties, like GM varieties
that reduce the complexity of pest control, may also have special appeal for less edu-
cated individuals.
6.2 Wealth
This factor contributes to adoption in a variety of ways. Wealthier individuals face less
credit and other financial constraints that may hinder adoption, and they are often less
averse to risk; thus, they are more likely to invest in high-risk, high-return technologies.
Higher wealth may also lead to riskier decisions in terms of how to use land. Finally,
early adoption of some new and advanced technologies (tractors, computers, etc.) is
often prestigious, and a wealthier individual can more easily afford it. To the extent that
wealth contributes to adoption of technologies or crop varieties that increase the aver-
age profitability of farmers, it also serves as a mechanism to further increase the value of
land in wealthier regions.
6.3 Scale
A large body of evidence suggests that several dimensions of operational scale con-
tribute to adoption. Larger farm size is likely to enhance the adoption of technologies
that are either indivisible or have economics of scale. Size is likely to reduce risk aver-
sion and thus enhance adoption of high-risk, high-return technologies. Scale allows
farmers to buy inputs, including both physical (farm machinery, e.g., combines)
and human (expert advice) capital, which reduces the per-unit costs of these inputs
and enhances the utilization of these assets. A stronger capital asset base reduces
the cost of adoption of technologies that use these assets as complementary inputs.
Foster and Rosenzweig (2010a) argue that, in addition to being a catalyst for adop-
tion of technologies in developing countries, size is a major contributor to increased
productivity. The notion of “small farms” varies by crop and activity. For example, a
five-acre wheat farm will be minuscule whereas a nursery of the same size will be a
viable business.
6.4 Health
The intellectual capacity that is crucial for allocative ability also depends on health sta-
tus and good nutrition. The study by Croppenstedt and Muller (2000) documents that
improved nutritional status improves productivity, presumably through improved allo-
cation and technology choices.
6.5 Age
Several studies have found that age affects adoption choices. Younger farmers are
more knowledgeable about new practices and may be more willing to bear risk and
invest in new technologies (Gould et al. 1989; Polson and Spencer 1991; Adesina and
Zinnah 1993). The major reason that younger agents are more likely to adopt new
technologies is that they have a longer planning horizon and are thus likely to get
more use from the technology over their lifetime. Some authors find positive corre-
lation between age and adoption (Hussain et al. 1994). Older farmers may be more
likely to adopt technologies with shorter repayment periods that may reduce effort
and allow improved lifestyle.
6.6 Location
The early studies of adoption emphasized the role of location in explaining diffusion.
Distance and access to markets and experts can have significant impacts on adoption.
Villages farther away from centers of commerce were less likely to adopt technologies
such as hybrid corn (Rogers 2003). Key factors like climate and soil quality also influ-
ence the profitability of adoption and depend on location of the farmer.
7. Conclusion
Allocation of land for different activities is affected by adoption of new technologies.

The adoption process is gradual and depends on economic incentives, technological
incentives, policies, and regulation. The threshold model of adoption suggests that the
rate of adoption is affected by the heterogeneity of potential adopters and by dynamic
processes, including technological improvements and knowledge acquisition, that
increase the relative advantage of new technologies over time. Adoption of new tech-
nologies may expand utilized land and introduce agriculture to regions that were pre-
viously unable to be farmed. It also may change the relative value of various types of
land. Adoption behavior is frequently an investment and is subject to uncertainties,
thus crop-sharing institutions and various insurance mechanisms can affect the rate
of adoption and the spread of the technology. Government may encourage adoption
by providing financial incentives, but also by enhancing research and extension activi-
ties, as well as by establishing mechanisms for expanding knowledge among farmers.
Enhanced profitability is a major motivation behind adoption, but adoption may be
motivated by nonpecuniary factors, such improved convenience and increased safety.
Environmental regulation may also serve as a major mechanism for inducing adop-
tion and introducing technological change. Investment in research and activities that
enhance productivity of farming systems may contribute in reducing pressure on land
resources and slowing processes of deforestation by accelerating the adoption of tech-
nologies that intensify agricultural production, thus leading to more output per unit
of land.
Acknowledgments
The research leading to this chapter was supported by the Energy Biosciences Institute,
U.C. Berkeley and BP Environmental Science Challenge Project.
References
Adesina, A., and M. Zinnah. 1993. Technology characteristics, farmers’ perceptions and
adoption decisions: A Tobit model application in Sierra Leone. Agricultural Economics
9(4): 297–311.
Ajzen, I. 1991. The theory of planned behavior. Organizational Behavior and Human Decision
Processes 50:179–211.
Alexander, C., R. Ivanic, S. Rosch, W. Tyner, S. Y. Wu, and J. R. Yoder. 2012. Contract theory
and implications for perennial energy crop contracting. Energy Economics 34(4): 970–979.
Arthur, W. B. 1994. Increasing returns and path dependence in the economy. Ann Arbor: University
of Michigan Press.
Barry, P. J., S. T. Sonka, and K. Lajili. 1992. Vertical coordination, financial structure, and the
changing theory of the firm. American Journal of Agricultural Economics 74(5): 1219–1225.
Becker, G. S. 1965. A theory of the allocation of time. Economic Journal 75(299): 493–517.
Boehlje, M. A. 1999. Structural changes in the agricultural industries: How do we measure, ana-
lyze and understand them? American Journal of Agricultural Economics 81(5): 1028–1041.
Byerlee, D., and E. H. de Polanco. 1986. Farmers’ stepwise adoption of technological pack-
ages: Evidence from the Mexican Altiplano. American Journal of Agricultural Economics
68(3): 519–527.
Carey, J., and D. Zilberman. 2002. A model of investment under uncertainty: Modern irriga-
tion technology and emerging markets in water American Journal of Agricultural Economics
84(1): 171–183.
Casey, F., A. Schmitz, S. Swinton, and D. Zilberman. 1999. Flexible incentives for the adoption
of environmental technologies in agriculture. Norwell, MA: Kluwer Academic Publishers.
Caswell, M. F., and D. Zilberman. 1986. The effects of well depth and land quality on the choice
of irrigation technology. American Journal of Agricultural Economics 68(4): 798–811.
Conley, T. G., and C. R. Udry. 2010. Learning about a new technology: Pineapple in Ghana.
American Economic Review 100(1): 35–69.
Cotula, L., S. Vermeulen, L. Leonard, and J. Keeley. 2009. Land grab or development oppor-
tunity? Agricultural investment and international land deals in Africa. Report to the
International Institute for Environment and Development, London.
Croppenstedt, A., and C. Muller. 2000. The impact of farmers’ health and nutritional status
on their productivity and efficiency: Evidence from Ethiopia. Economic Development and
Cultural Change 48(3): 475–502.
David, P. 1975. Technical choice, innovation and economic growth. Cambridge, UK: Cambridge
University Press.
Dercon, S., and L. Christiaensen. 2010. Consumption risk, technology adoption and poverty
traps: Evidence from Ethiopia. Journal of Development Economics 96(2): 159–173
Dixit, A. K., R. S. Pindyck, and G. A. Davis. 1994. Investment under uncertainty. Princeton,
NJ: Princeton University Press.
Ervin, C. A., and D. E. Ervin. 1982. Factors affecting the use of soil conservation prac-
tices: Hypotheses, evidence, and policy implications. Land Economics 58(3): 277–292.
Fan, S., Gulati, A. and S. Thorat, S. 2008. Investment, subsidies, and pro-poor growth in rural
India Agricultural Economics 39(2): 163–170.
Feder, G., R. E. Just, and D. Zilberman. 1985. Adoption of agricultural innovations in develop-
ing countries: A survey. Economic Development and Cultural Change 33(2): 255–298.
Foster, A., and M. Rosenzweig. 2010a. Barriers to farm profitability in India: Mechanization,
scale, and credit markets. Paper presented at World Bank-UC Berkeley Conference on
Agriculture for Development, Berkeley, CA.
Foster, A. D., and M. R. Rosenzweig. 2010b. Microeconomics of technology adoption. Annual
Review of Economics 2(1): 395–424.
Gardner, B. L. 1987. The economics of agricultural policies. New York: Macmillan.
Giné, X., and D. Yang. 2009. Insurance, credit, and technology adoption: Field experimental
evidence from Malawi. Journal of Development Economics 89(1): 1–11.
Gould, B. W., W. E. Saupe, and R. M. Klemme. 1989. Conservation tillage: The role of farm and
operator characteristics and the perception of soil erosion. Land Economics 65(2): 167–182.
Greiner, R., Patterson, L., and Miller, O. 2009. Motivations, risk perceptions and adoption of
conservation practices by farmers. Agricultural Systems 99: 86–104.
Griliches, Z. 1957. Hybrid corn: An exploration in the economics of technological change.
Econometrica, Journal of the Econometric Society 25(4): 501–522.
Hochman, E., D. Zilberman, and R. E. Just. 1977. Two-goal regional environmental policy: The
case of the Santa Ana river basin. Journal of Environmental Economics and Management
4(1): 25–39.
Huang, J., and E. M. Liu. 2013. Risk preferences and pesticide use by cotton farmers in China.
Journal of Development Economics 103: 202–215.
Hussain, S. S., D. Byerlee, and P. W. Heisey. 1994. Impacts of the training and visit exten-
sion system on farmers’ knowledge and adoption of technology: Evidence from Pakistan.
Agricultural Economics 10(1): 39–47.
Isik, M., M. Khanna, and A. Winter-Nelson. 2001. Sequential investment in site-specific crop
management under output price uncertainty. Journal of Agricultural and Resource Economics
26(1): 212–229.
James Jr., H. S., P. G. Klein, and M. E. Sykuta. 2011. The adoption, diffusion, and evolution of
organizational form: Insights from the agrifood sector. Managerial and Decision Economics
32(4): 243–259.
Just, D. R., S. A. Wolf, S. Wu, and D. Zilberman. 2002. Consumption of economic information
in agriculture. American Journal of Agricultural Economics 84(1): 39–52.
Just, R. E., and D. Zilberman. 1983. Stochastic structure, farm size and technology adoption in
developing agriculture. Oxford Economic Papers 35(2): 307–328.
Just, R. E., and D. Zilberman. 1988. The effects of agricultural development policies on income
distribution and technological change in agriculture. Journal of Development Economics
28(2): 193–216.
Kahneman, D., and A. Tversky. 1979. Prospect theory: An analysis of decision under risk.
Econometrica: Journal of the Econometric Society 47(2): 263–291.
Khanna, M. 2001. Sequential adoption of site-specific technologies and its implications for
nitrogen productivity: A double selectivity model. American Journal of Agricultural
Khanna, M., M. Isik, and A. Winter-Nelson. 2000. Investment in site-specific crop management
under uncertainty: Implications for nitrate pollution control and environmental policy.
Khanna, M., M. Isik, and D. Zilberman. 2002. Cost-Effectiveness of alternative green payment
policies for conservation technology adoption with heterogeneous land quality. Agricultural
Economics 21(2): 157–174.
Khanna, M., and D. Zilberman. 1997. Incentives, precision technology, and environmental
quality. Ecological Economics 23: 25–43.
Lancaster, K. J. 1966. A new approach to consumer theory. Journal of Political Economy
74(2): 132–157.
Lichtenberg, E., and D. Zilberman. 1986. The econometrics of damage control: Why specifica-
tion matters. American Journal of Agricultural Economics 68(2): 261–273.
Lichtenberg, E. 2002. Agriculture and the environment. In Handbook of agricultural economics,
Vol. 21, eds. B. Gardner and G. Rausser, 249–1313. Amsterdam: Elsevier.
Linn, J. 2008. Energy prices and the adoption of energy-saving technology. Economic Journal
118(553): 1986–2012.
MacDonald, J. M., and P. Korb. 2008. Agricultural contracting update. Washington, DC: USDA
Economic Research Service.
MacDonald, J. M., and P. Korb. 2011. Agricultural contracting update: Contracts in 2008.
Economic Information Bulletin 72. Washington, DC: U.S. Department of Agriculture.
Marra, M. C., and G. A. Carlson. 1990. The decision to double crop: An application of
expected utility theory using Stein’s theorem. American Journal of Agricultural Economics
72(2): 337–345.
Marra, M., and N. Piggott. 2006. The value of non-pecuniary characteristics of crop biotech-
nologies: A new look at the evidence. Regulating Agricultural Biotechnology: Economics and
Policy. Natural Resource Management and Policy 30(1): 145–177.
McWilliams, B., and D. Zilberman. 1996. Time of technology adoption and learning by using.
Economics of Innovation and New Technology 4(2): 139–154.
NRC. 2010. Impact of genetically engineered crops on farm sustainability in the U.S. Washington,
DC: National Academies Press.
Polson, R. A., and D. S. C. Spencer. 1991. The technology adoption process in subsistence agri-
culture: The case of cassava in southwestern Nigeria. Agricultural Systems 36(1): 65–78.
Rehber, E. 1998. Vertical integration in agriculture and contract farming. Food Marketing Policy
Center, University of Connecticut.
Rogers, E. M. 2003. Diffusion of innovations, 5th ed. New York: Free Press.
Roumasset, J. A. 1976. Rice and risk: Decision making among low-income farmers. Amsterdam
and New York: Elsevier.
Schmitz, T. G., C. B. Moss, A. Schmitz, A. Kagan, and B. Babcock. 2005. E-Commerce in
Agribusiness. Longboat Key, FL: Florida Science Source.
Schuh, E. 1974. The exchange rate and US agriculture American Journal of Agricultural
Schoengold, K., and D. Zilberman. 2007. The economics of water, irrigation, and develop-
ment. In Agricultural development: Farmers, farm production and farm markets, 2940–2966,
Handbook of agricultural economics, Vol. 3, eds. Robert E. Evenson, Prabhu Pingali, and T.
Paul Schultz. Amsterdam: Elsevier.
Schultz, T. P. 2003. Human capital, schooling and health. Economics & Human Biology
1(2): 207–221.
Seo, S., Segarrab, E., Mitchell, P., and Leathamd, D. 2008. Irrigation technology adoption and
its implication for water conservation in the Texas High Plains: A real options approach.
Agricultural Economics 38: 47–55.
Serra, T., D. Zilberman, B. K. Goodwin, and A. Featherstone. 2006. Effects of decoupling on
the mean and variability of output. European Review of Agricultural Economics 33(3): 269.
Serra, T., D. Zilberman, B. K. Goodwin, and K. Hyvonen. 2005. Replacement of agricultural
price supports by area payments in the European Union and the effects on pesticide use.
American Journal of Agricultural Economics 87(4): 870.
Shapiro, C., and H. R. Varian. 1999. Information rules: A strategic guide to the network economy.
Cambridge, MA: Harvard Business School Press.
Smale, M., R. E. Just, and H. D. Leathers. 1994. Land allocation in HYV adoption models: An
investigation of alternative explanations. American Journal of Agricultural Economics
76(3): 535–546.
Stiglitz, J. E., and A. Weiss. 1981. Credit rationing in markets with imperfect information.
American Economic Review 71(3): 393–410.
Sunding, D., and D. Zilberman. 2001. The agricultural innovation process: Research and tech-
nology adoption in a changing agricultural sector. In Handbook of agricultural economics,
Vol. 1, eds. B. L. Gardner and G. C. Rausser, 207–261. Amsterdam: Elsevier.
Villamil, M. B., A. H. Silvis, and G. A. Bollero. 2008. Potential for miscanthus’ adoption in
Illinois: Information needs and preferred information channels. Biomass and Bioenergy
32: 1338–1348.
Ward, F. A., and M. Pulido-Velazquez. 2008. Water conservation in irrigation can increase
water use. Proceedings of the National Academy of Sciences of the USA 105(47): 18215–18220.
Wolf, S., D. Just, and D. Zilberman. 2001. Between data and decisions: The organization of agri-
cultural economic information systems. Research Policy 30(1): 121–141.
Zellner, A. and D. Zilberman. 2011. The economics and econometrics of risk: An introduction
to the special issue. Journal of Econometrics 162(1): 1–5.
C HA P T E R 3
A R E L A R G E M E T R O P O L I TA N
A R E A S S T I L L V IA B L E ?
E DW I N S . M I L L S
This chapter1 concerns the functions of and prospects for large metropolitan areas. In
the United States, the federal government recently revised and expanded its metropoli-
tan concepts2. In 2004, there were 375 generic metropolitan areas (MAs) that contained
80% of the US population. Since MAs consist of entire counties, they contain much
rural land, perhaps 25–40% of total MA land areas. One result is that 2–5% of metro-
politan residents are rural. Since only 2–3% of US workers are employed on farms, the
vast majority of rural workers are engaged in the same work that urban residents do. In
fact, a substantial number of US rural residents work in urban or metro areas, assisted
by our superb interstate highway system that enables rural residents to commute long
distances to urban jobs.
International comparisons of MAs are approximate. Nearly all governments define
a metropolitan concept, but not in quite the same way. By any reasonable definition,
Tokyo is the world’s largest metropolitan area, with about 25 million people, or 20% of
the Japanese population. Mexico City may be the world’s second largest MA, although it
is difficult to decide where the MA ends. In the United States, New York (18 million peo-
ple), Los Angeles (13 million people), and Chicago (9 millions people) have for decades
been ranked in that order as the three largest MAs.
Most high-income countries are 60–85% urban. Middle-income countries are mostly
40–60% urban. Low-income countries are mostly in the 20–40% range. Good cocktail
party conversation can be made of the fact that two countries that are popularly thought
1
This chapter is a substantial revision and updating of Mills (1992b).
2 See Gacquin and DeBrandt (eds.) (2006, 774, 775) for concise definitions. This annual 1,200-page
volume contains by far the best summary of US data for states, metropolitan areas, counties, and cities.
The statistics in this section are from it and recent issues of the World Development Report.
Are Large Metropolitan Areas Still Viable? 75
of as agricultural are among the most highly urbanized countries in the world: Israel and
New Zealand are 91% and 86% urban, respectively.
Although most MAs in the world have been suburbanizing during the post–World
War II period, the process has gone much further in the United States than in most
countries. Beyond 1–5 miles from the city center, population and employment densities
do not vary systematically with distance from US MA centers. The result is that US sub-
urbs are extremely low density by comparison with those in almost any other country.
More cocktail party conversation: one MA, Cheyenne, has a lower MA population den-
sity than the entire 48 contiguous states.
Why do MAs exist? They exist because they perform functions that cannot be per-
formed as well by any other form of spatial organization. In the United States, an acre
of prime downtown land in a large MA might sell for upward of $50 million, whereas
an acre of prime agricultural land 50 miles away might sell for $5,000–10,000, making
downtown land 5,000–10,000 times as valuable as nearby farmland. In large European
and Asian MAs, comparisons are equally dramatic. There is hardly any comparably
dramatic social comparison. The comparison suggests that MA land is extremely pro-
ductive. People pay so much more for downtown land only because it provides com-
mensurate benefits.
1. Functions of Metropolitan Areas
The literature on the functions of large MAs is confused and emotional, but the truth is
prosaic. MAs provide no technology and no form of social or business organization that
is not available elsewhere. The only characteristic that is unique to large MAs is proxim-
ity among tens of thousands of businesses and households within a few miles.
Why is proximity so valuable that it may drive up the price of land that provides the
best access by a factor of 5,000–10,000? The reason, of course, is that proximity econo-
mizes on transportation and communication costs. Transportation and communication
are expensive. A downtown location is worth more than a suburban location to a highly
paid professional who must meet frequently with other similar professionals. The travel
times to such meetings are much shorter downtown even if travel speeds are faster at
suburban locations but travel distances are greater. Also, the cost of moving people is
much greater than the costs of moving commodities, and, as people costs have risen rel-
ative to commodity costs, commodity production has almost completely moved away
from central locations in large MAs.
The high cost of transporting people and goods is a necessary but not sufficient condi-
tion for MAs. If all commodities and services could be produced as cheaply at small vol-
ume as in large volume, most transportation costs could be avoided if small businesses
located very close to each other, to their customers and employees, and to their material
suppliers. But it is uneconomical to produce cars, education, or almost anything else in
facilities that supply only a few consumers. Economies of scale require that production
76 Edwin S. Mills
be on a substantial scale if it is to be at low cost. Economies of scope make it advanta-

geous to produce a variety of related commodities and/or services that are related in
production and/or marketing in a single facility. It is thus economical to produce com-
modities and services in large volume and for consumers and producers to locate close
to each other if transportation and communication among them are necessary.
The final factor, which finishes the story and permits high-density MAs, is the techni-
cal ability to substitute structures for land where land is expensive. Offices and dwell-
ings permit such substitution most easily. For given costs of land and construction, it is
hardly more expensive per square foot of usable space to produce office or residential
space in a 100-story building than in a 10-story building. Substitution of structures for
land is much more difficult for manufacturing plants and warehouses and somewhat
more difficult for retail establishments. An important reason is the high cost of moving
commodities among floors. Vertical transportation of people is also expensive, and that
requires a balancing of costs of horizontal and vertical transportation in choosing office
and residential heights.3
This analysis applies to MAs of all sizes; indeed, to the smallest agricultural market
town. Small towns exist because of scale and scope economies in processing agricultural
products and in providing commodities and services to the townspeople and to the sur-
rounding rural population. Nevertheless, Peoria is different from Chicago. The number
and variety of commodities and services produced is much greater in large than in small
MAs or in small towns. Most of the world’s large MAs are on navigable waterways that
provide access to the oceans. The exceptions are a few national capitals, such as Brasilia,
Delhi, Mexico City, Paris, Seoul, and Washington. They are large because they produce
government services rather than commodities or services for export. Their locations
were chosen for political, not economic, reasons. (Most such capitals are locations of
centralized and intrusive governments.) These days, road and air transportation are at
least as important as water and rail transportation. Of course, large MAs are well served
by roads and airports, but that is both cause and effect. Roads and airports are built
where large MAs are, but they also promote MA growth. Sorting out cause and effect is
difficult.
In recent decades some large MAs, but not the largest, have become centers for scien-
tific research, development, and innovation. Route 128 near Boston was an early post-
war example. Silicon Valley near San Francisco; Research Triangle in North Carolina;
Austin, Texas; and, more recently, Bangalore in India are other examples. Undoubtedly,
proximity among such activities facilitates exchange of people and ideas. (See Jaffe,
Trajtenberg, and Henderson in Henderson 2005.) All are on the fringe of large MAs and
are close to one or more research universities.
3 Substantial opinion among real estate professionals holds that the tallest recently constructed
office towers are excessively tall. Even in the early days of New York’s World Trade Center, the top floors
were more difficult to rent than lower floors. It is perhaps indicative that most of the world’s recently
built towers, including the World Trade Center, were built by governments or with large government
subsidies. The top floors of such structures are sometimes referred to as “vanity floors.”
2. Growth and Sizes of Large

Metropolitan Areas
There is enormous stability in the relative population sizes of MAs within a country,
although the MAs that occupy particular ranks change from time to time (see Gabaix in
Henderson 2005). New York has been the country’s largest MA since the first census in
1790. The same MAs have occupied each of the top 5 size ranks since 1970. Over a lon-
ger period, Los Angeles has risen and Baltimore has fallen in rank.
Throughout the post–World War II period, the largest MAs have grown relatively
slowly. Of the 10 largest MAs in 2004, only Atlanta and Dallas were among the 10 fast-
est growing MAs from 2000 to 2004. Among the 10 fastest growing MAs from 2000 to
2004, all were in the Sunbelt except Sacramento. The fastest growing, Las Vegas, grew 3.4
times as fast as the average US MA. (That the largest MAs grow more slowly than smaller
MAs does not mean that MA sizes are converging. Just because tall parents tend to have
children who are shorter than their parents, and short parents tend to have children who
are taller than their parents does not mean people are converging to a uniform height.)
Five US MAs had annual growth rates in excess of 10% from 2000 to 2004. Such growth
rates rival those of the most rapidly growing third-world MAs. (The data in this para-
graph are from DeBrandt and Gaquin 2006, 774.)
What limits the sizes of the largest MAs? First is the size and geography of the country.
Only countries with large populations have large MAs. No MA with more than about
8 million population is in a country with fewer than 50 million people. Large MAs tend
to be distant from each other. Bombay and Calcutta are on the opposite coasts of India,
as are New York and Los Angeles in the United States. In many countries, the best natu-
ral harbor is also the site of the largest metropolitan area: New York, Tokyo, Mumbai,
Manila, and London (World Bank, various issues).
Second, and most fundamentally, are limits to the demand for commodities and
services produced in the MA. Every MA “exports” some commodities and services to
buyers outside the MA; nearly all commodities manufactured in an MA are sold out-
side the MA. (Similar comments apply to material inputs purchased from outside the
MA.) Many services also are sold outside the MA where they are produced, perhaps as
many as one-third. Patients come from great distances to the Johns Hopkins Hospital
in Baltimore, as do students to the major universities. Many of the sales on financial
exchanges in MAs are among buyers and sellers located outside the MA. As interna-
tional trade has increased in recent decades, foreign demand has added to the growth of
large MAs in some countries, including New York, London, Los Angeles, and Mumbai.
Because of lower transportation and communication costs, the cost of an MA’s export
tends to increase as a function of distance. Far away customers not only will be served
at greater cost, but the competition from other MAs will increase as well. Many studies
have shown that foreign demand cannot be explained by distance; presumably, the same
holds for MA exports. Large MAs often export commodities and services, sometimes at
78 Edwin S. Mills
great distance, that are not produced by smaller MAs. New York has the highest quality
maritime attorneys in the country, and Chicago has the most sophisticated commodi-
ties exchanges.
The final factor related to MA exports relates to costs. As noted earlier, land values
increase with MA size. Rents, wages, and other input prices are raised accordingly, mak-
ing large MAs expensive places to produce. In the final analysis this is the signal that the
MA is as big as it should be. In recent years, both workers and businesses have discov-
ered that Southern California is an expensive place to work or to locate a business. The
same appears to be true in many Asian MAs, such as Mumbai.
To this point, no mention has been made in this chapter of congestion and pol-
lution, factors many believe are limits to the size of large MAs. Absent remedial
measures, both problems tend to become worse as the size of an MA increases.
Nevertheless, both can be alleviated by government and private expenditures.
Transportation facilities can be built and improved. Sewage treatment facilities can
be built and upgraded, and emission standards can be upgraded. The additional
costs are a logical cost of large MAs. In the United States, the federal government
intervenes extensively, for example by financing MA public transit construction
with nationally raised taxes. The result is to understate the true costs of large MAs
to people and businesses in the MA. Such monies could be raised by MA govern-
ments, with oversight by state governments. Then, the costs of the MA would be
reflected in prices that would be charged for commodities and services produced
in the MA. Thus, congestion and pollution can be alleviated by appropriate expen-
ditures, and such expenditures are a logical part of the cost of living and doing
business in the large MAs. If this cost is reflected in prices of commodities and ser-
vices produced in large MAs, markets will get the right signals about appropriate
MA sizes.
The final issue discussed here pertains to crime, homelessness, poverty, illegitimacy,
racial tensions, and other forms of alienation that increase with MA size and tend to
limit size. With poverty, the claim is demonstrably false. The incidence of poverty is
lower in MAs than elsewhere and does not increase with MA size. There is some evi-
dence that welfare-prone people are attracted to MAs with unusually generous welfare
programs and that large MAs have more generous welfare programs than small MAs.
The appropriate measure is welfare payments relative to living costs, and “real” welfare
payments hardly rise with MA size. In any case, such effects are small, and the claims are
often thinly disguised forms of racism.
Street crime rates also rise with MA size, but, again, the correlation is not strong.
One key observation is that large MAs are more impersonal and consequently less
civil than towns or small MAs. No one who has lived in both a small town and a large
city can doubt this, but it is difficult to imagine that impersonal relations increase
significantly in places with more than 1 or 2 million people. Such MAs are already
impersonal.
3. Suburbanization
As previously noted, US MAs have suburbanized more than those in almost any other
country. Carefully documented reasons include falling transportation costs and rising
incomes. The more up-scale housing preferred by higher income people generally can
be provided most economically in suburbs, where land values are lower.
Less well studied, but probably important, is the interaction between suburbaniza-
tion of employment and housing. Manufacturing has long dispersed from central cities
and has been moving to distant edges of MAs and beyond since the 1950s.This stems
from technical progress that has reduced labor inputs even as manufacturing output has
grown. In addition, our superb interstate highway system enables many workers to com-
mute from MA suburbs to manufacturing jobs even well outside the MA. Factories line
the interstates leading from the Chicago MA. Many services, including finance, insur-
ance, real estate, retailing, and healthcare have moved to the suburbs massively in recent
decades, perhaps following their employees and customers as much as leading them.
Finally, increasingly stringent land use controls, especially since the mid-1970s, have
limited population densities to below competitive levels in both central cities and sub-
urbs, especially in large MAs. Chicago illustrates typical effects of suburbanization.
From 1980 to 2003, the urban population of the Chicago MA increased less than 30%,
but the land area increased more than 40% (see DeBrandt and Gaquin 2006). Two-thirds
of the population and 60% of employment are located in near-by suburbs.
Costs of moving people and commodities fall gradually, but costs of processing and
transporting information fall much more rapidly. Estimates are that the real cost of
doing a given arithmetic operation has fallen at a compound annual rate of 10–20% dur-
ing the last quarter or third of a century. Experts assert that no end of this technical
revolution is in sight. During the 1970s and 1980s, the costs of moving information—
anything that can be put on paper—fell dramatically. The cost of data transmission over
long distances has fallen because of improved small computers, fax machines, e-mail,
cheaper long distance telephone service, and computers especially designed to network.
Although careful studies do not exist, this revolution must have promoted growth
of the suburbs and development of edge cities. There is as yet little evidence of disper-
sion of service sector employment away from substantial centers, either downtown or in
suburbs. That suggests that access—inexpensive face-to-face contacts among people—
has been the driving force. My hypothesis is that subcenter development is proceed-
ing much the same way and for much the same reasons that downtown development
proceeded in earlier years. The difference is that cluster development is proceeding
faster outside of downtown areas in large MAs than in small MAs. Businesses in sub-
urban subcenters appear to interact little with downtown businesses. Naperville, nearly
30 miles west of downtown, is the quintessential edge city in the Chicago MA. It is an
edge city of 140 thousand residents in 2004, having more than tripled since 1980, and
is a thriving and independent community. In 2000, 65,000 jobs were located there, but
80 Edwin S. Mills
not everyone who worked there lived there, and everyone who lived there did not also
work there.
The rapid and extensive growth of the suburbs increasingly blurs the distinction
between metropolitan and rural. People who work in Naperville or who sell com-
modities and/or services there can easily live 20–40 miles west of Naperville, placing
them well beyond the limits of the Chicago MA. As edge cities become larger and more
self-contained, exurban locations become increasingly attractive. Indeed, in some
places an exurban location, say, no more than 50 miles from an MA downtown, may be
little farther from the downtowns of one or two other MAs. Are such locations rural or
metropolitan? The name is not crucial, but the effects may be very important. Twenty
years ago, these would have been distant rural areas. For many small towns and rural
places that have become edge cities or have come to have easy access to edge cities, such
developments have provided increased employment and large capital gains on farm-
land. For others, such developments have brought unwelcome newcomers and lifestyle
changes.
4. The Future of Large

Metropolitan Areas
I conclude with speculations about the future of large MAs. The only safe statement is
that the largest 5–10 US MAs in 2000 are almost certain to grow at slower rates than the
US population in coming decades. It would physically difficult for these 5–10 largest
MAs to grow much because they are located near other MAs. In 2004, the five largest
MAs contained 52.6 million people, 17.9% of the total US population and 22.4% of the
MA population.
I expect the MA share of total population to increase about 1 percentage point during
the decade or so after 2004. However, the MA share of total population is unlikely to rise
as far as 85% during the first few decades of the 21st century. The five largest MAs are
likely to grow slowly and to fall slightly as a share of total MA population.
This forecast is a conservative extrapolation of trends during the past 20 years. Why
might it be wrong? One common conjecture is that large MAs are increasingly unpleas-
ant places to live and do business and that people prefer small MAs anyway. I do not
believe that is a significant argument. For decades, people have told pollsters that they
prefer to live and work in small urban areas; 50,000–100,000 people is the most com-
mon range. Whatever such polls tell us, they do not forecast behavior. Population and
employment have continued to grow throughout the MA size distribution, and small
MAs have not grown much faster than middle-size MAs.
Living and working are not unpleasant in MAs; they are, to some extent, unpleas-
ant in some large MA cities. Population fell slowly in many large MA cities for several
decades, but the trend reversed slowly toward the end of the 1990s. Growth has focused
in suburbs, and nearly two-thirds of the MA population now lives in suburbs. As noted
earlier, many large suburban communities now have most of the advantages of central
cities: cultural, recreational, and the like. The polls tell us that many MAs are of the sizes
where many people like to live. Many suburban residents think that large suburban
communities are developing some of the disadvantages of central cities: traffic conges-
tion and crime, specifically. Because land use controls are, or can be, effective in keeping
out low-income people, and because traffic investments can be made, I do not think the
danger is great. In sum, I do not believe quality of life issues will cause people and jobs to
flee large MAs.
Suburbs have grown relative to large or inner cities for a variety of reasons. The result,
however, is clear: suburbanites have higher incomes and greater educational attainment
than inner-city residents. Inner cities have a greater mixture of racial and ethnic minori-
ties and an appalling concentration of alienated and poor black residents.
Studies indicate that school performance, illegitimacy, and crime all improve if
low-income minorities are somewhat dispersed instead of living together in low-income
neighborhoods. Role models appear to be the key causal factor. Large MAs have larger
fractions of their middle and upper middle-income populations living in exclusionary
suburbs than have small MAs. The ratio of suburban to central city income increases
with MA size. The result is more segregation of large groups of low-income minorities in
inner cities and greater alienation in large than in small MAs. An important part of the
solution of this peculiarly US inner-city problem is reduced government density con-
trols in both inner cities and suburbs, but it is not essentially a problem of MA size.
Why have poor minorities not followed jobs to suburbs? The answers are complex
and poorly understood. But one partial answer revealed by studies is government den-
sity controls. The poor are effectively zoned out of many suburbs. How many more
low-income and minority residents would live in suburbs, and how many would
perform better there, if land use controls were less of a barrier is impossible to know.
However, some simple calculations in Mills (1985) indicate that central cities would
contain more white residents, more residents altogether, and more jobs if low-income
and minority people were more evenly spread out among MA suburbs. The reason is
that, to some extent, high-income people locate in suburbs to avoid the “blight” that
results from the concentration of low-income and minority residents in central cities.
If it were easier for low-income and minority residents to disperse from the inner cit-
ies, there would be less real or perceived inner city blight. In sum, there would be fewer
places for higher income people to go and less to escape from (also see also Mills 2005).
Therefore, movement of upper income residents to the suburbs and the use of police
power to keep low-income people out of suburbs have caused movement to the suburbs
to be more extreme than it would be otherwise, and neither the private nor government
sectors in the inner cities perform as well as they are capable of performing.
More difficult to deal with is the second common conjecture: computerization. It is
certain that the compilation, analysis, and transmission of data over long distances will
become increasingly cheap and common in the coming years. Some conjecture this will
destroy the rationale for large collections of office-type activity. If information can be
82 Edwin S. Mills
transmitted electronically, why locate in an office center where land is many times as
expensive as it would be at a more isolated location?
The key issue is whether computers will destroy the need for face-to-face contact in
business communication. Inexpensive long distance electronic transmission of infor-
mation has been available for some years. It has long been possible to fly diskettes across
the country overnight by express delivery services, and fax messages and e-mail have
been widely used for more than two decades. Videophones and low long distance phone
rates have long been available as well. These innovations seem to have had almost no
effect in dispersing business activity beyond MA boundaries. During the 70s and 80s,
suburbanization probably was faster because of these technologies, but this seems to
have been caused more from gradually falling costs of moving people and goods rather
than from rapidly falling costs of moving data. Technology will soon be available that
will permit instantaneous interaction by voice, video, and printed documents over great
distances and at low cost, permitting meetings among people separated by long dis-
tances. They will be able to see and hear one another and transmit documents to each
other quickly and cheaply.
I have maintained that access to large numbers of businesses and households is the
essence of large MAs. If face-to-face meetings became obsolete, beyond a doubt, large
MAs would shrink dramatically within a decade or so. I have grave doubts whether that
will happen, but I have no crystal ball and I offer the following with an unusual dose of
humility.
I do not believe electronics will make face-to-face meetings obsolete. Anything that
can be spoken can be transmitted electronically. The issue is the benefits versus the costs
of electronic transmission as compared to face-to-face transmission. In Mills (1992a),
I distinguished between unambiguous and ambiguous information. Ambiguous infor-
mation is what is transmitted in early meetings between potential vendors and buyers
of a new product. Each side wants to explore the other side’s needs, wishes, abilities,
reliability, willingness-to-pay, and likely costs and speed of production and delivery.
It is what is transmitted when opposing attorneys in a case meet to discuss possible
settlement out of court. It is what is transmitted when members of a profession meet
for lunch. They all know they are competitors, and they all want to get more valuable
information than they give about technology, market trends, product innovations,
and the like. Yet they all know that they must give some information in order to get
some. Quintessentially, it is what is transmitted in an academic seminar. The essence
of a research seminar is that a group of people with a common vocabulary and body of
expertise come together to listen to a colleague discuss a half-baked idea. The result is
akin to a controlled free association exchange, the essence of the creative process.
My claim is that the exchange of ambiguous information is what face-to-face com-
munication has always been about and that electronic communication is a poor sub-
stitute. In such exchanges, it is disadvantageous to write too much down. In addition,
each participant wants to iterate in the information exchange. Finally, participants fre-
quently want to “feel each other out” prior to providing unambiguous information. The
exchange proceeds in ways that depend on the information set that participants bring
to the meeting and are willing to communicate. Such information can be known only
approximately prior to the meeting.
Experiments and scientific (mostly by sociologists and management specialists)
observations of electronic meetings confirm the above conjectures. Electronic meet-
ings induce people to bring prepared statements and take positions they then find
it awkward to modify or abandon. Those with supervisory responsibility may not
be able to manage those who are long-winded or get sidetracked. Closely related,
those with managerial responsibility find it difficult to monitor the productivity of
their supervisees when they are not on the same site. That has limited the spread of
work-at-home jobs. If it were not so, supervisees would be paid piece rates. Finally,
work on a common site stimulates employees by creating a competitive atmosphere.
This extends to schools and universities. An important advantage of a common site
for education is the stimulation, exchanges, and competition that students provide for
each other.
Academics should consider the possibility of an electronic university. It is now tech-
nically possible for me to live in Buena Vista, Colorado and to lecture, with voice, visual,
and written communications, students who are dispersed around the country or the
world. Communication can easily be interactive. Indeed, my research can be done the
same way. It is possible to bring on my computer screen any book or article that is in the
university library now, or, indeed, any data set stored in some central location. My work-
ing papers can be distributed to a worldwide audience, and seminars can be held using
the same computer network. My paycheck can, of course, be sent to me or my bank.
Approximations to such electronic universities already exist but show little sign of sub-
stituting for high-quality research institutions. And deans have shown little enthusiasm
for sending paychecks to distant bank accounts.
Electronic communication certainly has had, and will continue to have, important
effects. It permits increased specialization, downsizing, and efficiency among insti-
tutions. To take one example, each large bank, until recently, had its own economics
department that did forecasting and market analysis for bank management. Now, it is
possible to buy higher quality forecasts and analysis than the bank can undertake itself.
Such information can be transmitted electronically to any place in the world. That and
similar examples are, I believe, at the core of downsizing that has been and is under way
throughout the US economy. Entire layers of middle level employees who formerly
compiled, analyzed, and transmitted data are being replaced by electronic systems that
do the work, both domestically and internationally.
Electronics already has and will continue to facilitate suburbanization. Face-to-face
meetings have come to be needed less frequently, but they are still required. That process
permits businesses to be located in more distant suburbs than was previously economi-
cal. But it does not permit universities, law office, or similar organizations offices to be
dispersed among Chicago, Buena Vista, or Baja, California.
In conclusion, my forecast for the next 10–20 years is the continued rapid growth of
suburbs. I believe also there will be growth, but slower growth, of the large MAs than for
the population as a whole.
84 Edwin S. Mills
References
DeBrandt, K. A., and D. A. Gaquin (eds.). 2006. 2006 County and city extra. World Development.
Lanham, MD: Bernan Press.
Henderson, J. V. (ed.). 2005. New economic geography. Northampton, MA: Edward Elgar.
Mills, E. 1985. Open housing laws as stimulus to central city employment. Journal of Urban
Economics 17(2): 184–188.
Mills, E. 1992a. Sectoral clustering and metropolitan development. In Sources of metropolitan
growth, eds. Edwin S. Mills and John F. McDonald (3–18). New Brunswick, NJ: Center for
Urban Policy Research.
Mills, E. 1992b. Large metropolitan areas: Their function and prospects. National Rural Studies
Committee, a Proceeding, 94–100. Corvalis, OR: Oregon State University.
Mills, E. 2005. Why do we have urban density controls? Real Estate Economics 3(3): 571–585.
C HA P T E R 4
MODELING THE LAND USE

C HA N G E W I T H B I O F U E L S
M A DH U K HA N NA , DAV I D Z I L BE R M A N , A N D
C H R I ST I N E L . C R AG O
There is growing interest in increasing reliance on biofuels to reduce dependence

on foreign oil, mitigate climate change and stimulate rural economic development.
Increased biofuel production can change land use directly by diverting land away from
agricultural production and indirectly by affecting crop prices. Changes in crop prices
can create incentives to intensify agricultural production (by increasing yields per acre)
and to expand agricultural acreage. Land use changes due to an increase in biofuel
production in one country can affect land use throughout the globe, and have implica-
tions for food security and greenhouse gas emissions (Rajagopal and Zilberman 2007;
Searchinger et al. 2008; Khanna and Crago 2012).
Although first-generation biofuels are being produced primarily from food-based
crops and sugarcane, there is considerable policy support and research to develop
advanced or second-generation biofuels from cellulosic feedstocks, such as crop
and forest residues and dedicated energy crops. These biofuels typically have lower
life-cycle greenhouse gas (GHG) intensity compared to food-crop-based biofuels and
would divert less productive land from food production per unit of fuel produced since
they could be produced either from crop by-products or from energy crops that can
potentially be grown productively on low-quality land that is marginal for food crop
production. There is considerable variability in the land requirements, GHG intensity,
and costs of production among the different pathways for second-generation biofuels
(Huang et al. 2013).
86 Madhu Khanna, David Zilberman, and Christine L. Crago
The United States US and European Union (EU) are relying on mandates, tax credits,
and import tariffs to stimulate biofuel production. The advent of biofuels has raised sev-
eral research and policy questions: How much land will be required to meet the various
mandates for biofuels? How much of the additional demand for land for biofuels will be
met by changes at the intensive margin versus the extensive margin? Which feedstocks
are likely to be used for biofuel production? What economic, technological, and bio-
physical factors are likely to significantly influence land use choices to support biofuel
production? How does the land use effect of biofuels differ with various policy choices?
A number of economic models are being used to answer these questions. The purpose of
this paper is to examine the key assumptions and synthesize the major findings of these
models to develop an understanding of the drivers of land use change and the land avail-
ability constraints for biofuel expansion.
Land use changes are outcomes of decisions affected by returns to land under alter-
native activities. Collectively, these microlevel decisions affect the aggregate sup-
ply and costs of food and fuel. In turn, macrolevel variables, like demand, prices and
energy and climate policies influence decisions at the microlevel. Major economic
theories and concepts have been introduced to explain land use decisions, patterns of
trade, and the value of land and the economic benefits and ecosystem services it pro-
vides. These include the classic von Thünen (1966) model of regional land allocation,
which laid the foundation of “Urban Economics,” David Ricardo’s (1891) theory of
trade and the notion of rent, and John Krutilla’s (1967) “Conservation Reconsidered”
that emphasized the economic importance of ecological services. These bodies of lit-
erature provide the foundation for the development of models to study the drivers of
land use change for biofuels. Section 1 of this chapter provides a background on the
land economics literature and key principles that have emerged from it for under-
standing land use changes induced by biofuels.
The recent development of biofuels has integrated the energy and the agricultural
sectors. Prior to biofuels, energy prices have affected the supply side of agricultural
production, since energy is a key input. Now energy prices are also affecting the
demand for land and crops (used for biofuels). The nexus between energy markets
and land use has required adaptation of existing models and development of new
models of agricultural markets. These models tend to emphasize the heterogeneity
in land and to link biophysical models of biofuel feedstocks with economic models.
They differ in their structure, assumptions, data used, and the mix of biofuel feed-
stocks and policy choices considered. Section 2 of this chapter describes the impacts
of introducing biofuels on models used to analyze the agricultural sector, followed
by a description of the elements of an ideal model for analyzing the implications of
biofuels.
Section 3 of this chapter presents a description of different types of models being used
to study the implications of biofuel policies for land use, and classifies them into: static
partial equilibrium models, dynamic partial equilibrium models, and general equi-
librium models. It examines how the differences in model structure affect outcomes.
Section 4 of this chapter discusses the key drivers of land use change due to biofuels, and
Modeling the Land Use Change with Biofuels 87
Section 5 discusses ways to deal with multiple models. Section 6 presents the main find-
ings that emerge from these models and Section 7 concludes.
1. Overview of the Land Economics

Literature
There are several strands of economics literature relevant to understanding the impacts
of biofuel on land use changes. Some are conceptual models that recognize that land is
heterogeneous and that differences among parcels of land will affect their use and value.
von Thünen (1966) established a major principle that land will be used in the activity
in which it generates the most value. His work suggests that land use choices will differ
across locations and will change over time as technology and the climate change. This
literature provides insights that are useful for determining the location of biofuel feed-
stock production and refineries for biofuels.
Ricardo (1891) introduced the notion of rent, which is the residual left to landown-
ers after selling the output and paying for all inputs. When each landowner selects the
most profitable activity, rents and land use patterns can be derived given prices and
technological coefficients at each location. Since agricultural products are frequently
traded, with free trade, land use patterns will adjust to take advantage of distribution of
resources across locations (Heckscher and Ohlin 1991).
Another relevant literature expands the Hotelling approach of dynamic modeling
of utilization and pricing of nonrenewable resources over time to examine the effects
of introducing renewable energy as a backstop. Chakravorty, Magne, and Moreaux
(2008) use the Ricardian-Hotelling framework to analyze the dynamics of land alloca-
tion decisions for food and fuel production as available energy resources become scarce.
Xabadia, Goetz, and Zilberman (2006) developed a conceptual framework for opti-
mal allocation of resources over space and time. Tsur and Zemel (2005) incorporated
research and development (R&D) of alternative technologies in dynamic models ana-
lyzing nonrenewable resources.
The existing literature identifies the following factors as being important in driving
land use change in agriculture.
Technological Change and Innovation: R&D processes produce new innovations that
are adopted first at locations where they provide the most value (Sunding and Zilberman
2001) and may lead to expansion of farmland to areas that have not been previously uti-
lized (the extensive margin effect of adoption). Adoption of new technologies can also
increase yield per unit of land (the intensive margin effect), thus decreasing the land
requirement per unit of output (Gardner 1992). Thus, the net effect of technological
changes on land use in agriculture is an empirical question. The changes in output in the
extensive and intensive margins will affect the amount of land required to accommodate
increased crop production due to biofuels.
Risk: Producers are frequently averse to risk, and their land allocation choices among
crops and, in particular, adoption choices are affected by uncertainty about yields and
other variables (Feder, Just, and Zilberman 1985). Introduction of insurance policies
that reduce risk or institutions like futures markets or contracts may lead to increased
acreage of high-risk–high-reward activities. Risk would be a major factor influencing
land allocation to second-generation biofuel feedstocks.
Institutions and Policies: The perfectly competitive model does not fully capture
the institutions and policies that affect land use patterns. Feder and Feeny (1991)
argue that introduction of land titles removes uncertainty about landownership
and tends to increase investment in agricultural production and land productivity.
Commodity-support programs in the United States and Europe as well as building of
transport infrastructure have expanded agricultural acreage (Anderson et al. 2001).
Similarly, water use and energy subsidies have led to expansion of irrigated agriculture
(Schoengold and Zilberman 2007).
Environmental Considerations: Existence of externalities like pollution provides justi-
fication for government intervention such as taxation and zoning. Externality issues are
not restricted to pollution problems; land and nature provide valued ecosystem services
and consumers benefit from open space (Krutilla, 1967). Policies (zoning, permits, and
conservation preserve programs) have been introduced to protect these environmental
services. Irwin et al. (2009) demonstrate that environmental regulation indeed affects
land use and location of crops.
Changes in Consumer Preferences and Economic Growth: Demand for food is depen-
dent on food prices as well as on income. Poor individuals may consume mostly grains,
while higher income households may consume more meat. Since meat production
requires more land per calorie, economic growth in developing countries will increase
agricultural acreage. On the other hand, shifts away from a meat-rich diet in other parts
of the world may have the opposite effect.
Population Growth and Demographics: Population growth is likely to increase
demand, but the pattern of increased demand for food depends on where and when
these changes occur. Migration, especially from rural to urban areas, also affects land
use patterns. A shift from production for self-consumption in rural areas to production
for export to urban centers affects the composition of food portfolios, energy intensity
of food production, and productivity.
Renewable and Nonrenewable Resources: Agricultural productivity is dependent
on natural resources like water and soil quality, whose stocks may vary over time. The
depletion of groundwater or increase in cost of pumping strongly affects patterns of
land use (Schoengold and Zilberman 2007). Similarly, processes of soil erosion may also
affect what and how much can be grown at different locations. Finally, climate change
will affect land use patterns through its effect on precipitation, temperature, and sun-
light, among others (Mendelsohn and Dinar 2009).
2. Implication of Biofuels for

Land Use Analysis
The introduction of biofuel to agriculture has led to a new reality that challenged the
way agriculture and land use are analyzed and modeled. The agricultural sector was tra-
ditionally quite isolated, produced mostly food/feed products and was affected by sec-
toral policies. Studies analyzing the sector used specialized models that could focus on
the agricultural sector by itself. The introduction of biofuel has expanded the range of
activities conducted and the policies that affect the agricultural sector, and expanded
the importance of environmental issues in management and modeling of agriculture, as
discussed in more detail below.
First, biofuels have integrated the agricultural, livestock, and energy markets.
Biofuels have added a new demand to agricultural activities; this has led to diversion
of land from production of food to fuel (Rajagopal et al. 2007). Modeling of farmers’
choices now has to take into account not only relative food prices and traditional agri-
cultural policies, but energy prices and biofuel policies. Moreover, biofuels have linked
energy markets and livestock markets since some of the co-products of biofuels are
substitutes for traditional feed for livestock. Furthermore, biofuels have expanded the
range of spatial considerations in modeling farmers’ choices. The selection of where to
allocate land for food or fuel is affected not only by relative prices, but by the biophysi-
cal suitability of locations to produce biofuel crops and distance to refineries and end
users. Distance from a refinery and from livestock facilities will affect markets for corn
and the by-products of corn ethanol, which is used as animal feed and, therefore, land
use choices. Incorporating biofuels in land use models also links transportation choice
decisions with fuel choices and has implications for feedstock production and land use.
Since demand for biofuels is a derived demand, it depends on the demand for vehicle
kilometers traveled and on the substitutability between biofuels and gasoline (Khanna,
Ando, and Taheripour, 2008). With biofuels, land use choices are affected by vehicle fleet
structure, development of biofuel supply chain, and development of biofuel conversion
technologies. Moreover, these land use choices now impact fuel markets because biofuel
production displaces gasoline and can affect gasoline prices with consequent feedback
effects on demand for biofuels and costs of energy for the agricultural sector. Modeling
the implications of biofuels, therefore, requires determining market clearing prices in
the food and fuel sectors simultaneously (Khanna et al. 2011).
Second, different biofuel feedstocks expand the types of land that could be displaced
by biofuel production. Both corn and sugarcane ethanol, for the most part, requires
diversion of existing land in commodity production to production of crops for fuel. On
the other hand, introduction of new second-generation energy crops that can be grown
on marginal land may require conversion of land that is under pasture or forests into
agricultural production. This requires further modeling efforts to identify regions with
good potential to grow these crops.
Third, biofuel expands the range of policies that affect agriculture and land use. These
include agricultural policies as well as climate and energy policies. Policies such as bio-
fuel mandates, subsidies, and import tariffs on biofuels affect demand for biofuels and,
therefore, land use allocation and crop prices. Similarly, climate policies will not only
affect energy prices and the cost of agricultural inputs but will also affect demand for
biofuels and, therefore, crop prices and land use. These biofuel and climate policies
affect not only domestic land use but have indirect impacts on global land use because
they affect the prices of globally traded crops. The diversion of globally traded food/
feed crops for biofuel production and the competition for cropland induced by biofuel
production has the inevitable impact of raising world prices of biofuel feedstocks and
other crops that compete for land resources. The increase in world prices could induce
crop acreage expansion on native vegetation and forested land leading to indirect land
use changes that also contribute to greenhouse gas emissions (Khanna and Crago 2012).
Fourth, biofuels have expanded the type of technical change that affects land use. In
addition to changes in agricultural technologies that affect the productivity of biofuel
and related crops, technical change in the biofuel processing industry will also affect
land use. Technological breakthroughs that lower the cost of producing advanced biofu-
els from cellulosic feedstocks will affect the mix of biofuels and the amount and type of
land diverted from food and feed production to fuel production.
The introduction of biofuel and biofuel policies, thus, require significant additions to
existing ways of modeling agricultural markets. The linkage between agricultural and
energy markets imply the need for an integrated model of the food, feed, and fuel mar-
kets that endogenously determines food and fuel prices and their feedback effects on
demand for biofuels and allocation of land for food and fuel crops. Models that seek to
quantify the land use implications of biofuels and simulate the effects of biofuel poli-
cies need to integrate across many different scales. A global representation of relevant
markets is needed to capture the effect of biofuel-induced changes in land use and prices
on international trade and land use in other countries. At the same time, the assessment
of biofuel impacts requires a high degree of spatial resolution to account for heteroge-
neous land qualities, climate, land availability and ease of its conversion from one use to
another. These assessments need to be based not only on models that capture economic
behavior but also on models that incorporate crop production technologies, biophysi-
cal factors that affect crop productivity, and land suitability and availability constraints.
Finally, models should also take into account the market structure of the energy mar-
kets. The market imperfection in the oil market due to the presence of an oil cartel such
as OPEC could impact the change in fuel prices that results from the displacement of
gasoline with biofuel, and it could have feedback effects on the demand for biofuels
(Hochman, Rajagopal, and Zilberman 2010).
The production of biofuel has both direct and indirect land use impacts as shown in
Figure 4.1. The pathway by which biofuel production affects land use is described in the
upper set of boxes. The determinants of the magnitude of these effects are listed in the
lower set of boxes and discussed in Section 4. The direct land requirements for biofuel pro-
duction are simply the land on which biofuel feedstocks are grown and could be simply
Exogenous Change in
Direct Change in Indirect
policy biofuel mix
land use commodity land use
shock and
change prices change
production
• Policy mix • Commercially • Types of • Responsiveness • Reduction in

available types feedstocks of domestic and food demand
• Type and size
of biofuels international
of shock • Feedstock yield • Intensification
markets
• Substitutability • Conversion of agricultural
between • Substitutability production
efficiency from
biofuel and between
feedstock to • Productivity of
gasoline domestic and
fuel marginal lands
imported
• Production of goods • Land available
co-products for conversion
FIGURE 4.1 Pathway of land use changes due to biofuels.
measured by [biofuel quantity/(conversion efficiency of feedstock to fuel times the crop

yield)]. The magnitude of this direct land requirement is, therefore, critically dependent
on the yield of biofuel crops per unit of land and the conversion efficiency of feedstock to
fuel. This will depend on the type of feedstock that is being used and the technology for
converting it to biofuel. However, this is not the amount of additional land required due
to biofuel production, since some biofuels (like corn ethanol) produce co-products that
can replace other products that require land for their production. Biofuels also affect land
use indirectly because they increase competition for land and affect the price of land and
all land-using activities. In the case of biofuel crops that are tradable, the production of
biofuels reduces exports and increases both domestic and world price of biofuel crops.
This increase in prices for biofuel crops has four types of effects both domestically and in
the rest of the world. First, it can affect crop yields by inducing crop producers to adopt
improved technologies and management practices; these changes increase the intensity
of crop production and reduce the demand for additional land due to biofuel production.
Second, it can increase the value of land under biofuel crops and make it profitable for
landowners to substitute land from other crops to biofuel crops. This will reduce produc-
tion of substitute crops and increase their prices as well as those of biofuel crops. Third,
the increase in crop prices for tradable commodities leads to an increase in cultivated
land both domestically and in other regions of the world (extensive margin effect). This
increase could occur on land under grasses or on forestland and bring marginal/noncro-
pland into production. The expansion of biofuel crops to other cropland or to noncrop-
land could lower yields per hectare and increase the land required to meet demand for
biofuel crops; this will, at least partly, offset the intensive margin effect that raises crop
yields. These indirect land use changes have the potential to lead to the release of carbon
stored in those ecosystems and negate some of the greenhouse gas benefit of biofuels.
Lastly, the increase in price of biofuel crops and substitute crops will reduce demand for
these commodities both domestically and internationally. This will reduce the additional
land required due to biofuel production.
Thus, the net land use requirement for biofuel production after considering
co-products and indirect effects is likely to be smaller than the direct effect. For example
Hertel et al. (2010) show that the land required to meet the 57 million-liter corn-ethanol
mandate would naively be 15 million hectares if resources (land, labor, and capital) were
in perfectly elastic supply and there was no price response at all. The finite availabil-
ity of suitable land induces a price increase, which will lead to a reduction in demand
for food and nonfood (forestry) products and intensification of livestock, crop, and
forest-product production. The use of co-products of corn-ethanol production for
livestock feed also reduces demand for corn. As a result, the additional land require-
ment is reduced to 4.4 million hectares. It is further reduced to 3.8 million hectares due
to price-induced growth in baseline yields in the United States; however, this is offset
partly by the expansion of production on less productive land, which lowers yield. As
a result, the net increase in cropland conversion is estimated to be 4.2 million hectares,
which implies that each gross hectare of corn diverted to fuel use results in 0.28 hectares
of net land conversion for corn production.
Isolating the extent to which biofuel production affects land use is complicated and
difficult because it is likely to be distributed across multiple regions by global trade and
occur with significant time lags. This makes it difficult to separate the causal impact of
biofuels on ILUC from all the other factors affecting observed land use changes. With
multiple uses of land and possibilities for crop substitutions and displacements occur-
ring simultaneously, the only way to isolate land use changes due to biofuels is by using
regional or global models of agricultural markets that simulate the effect of an incre-
mental exogenous shock to biofuel production from some baseline level on equilibrium
prices and land use. These are compared, in a comparative static sense, to land require-
ments in a baseline or counterfactual state in the absence of biofuels to examine the
direct and indirect land use effects of biofuels.
3. Existing Models Being Used for

Analyzing the Impact of Biofuels
Various types of economic models are being used to examine the impact of biofuels.
These models can be classified broadly into partial equilibrium and general equilibrium
models. Each of these two types can be further classified into models that are static ver-
sus dynamic. Some of these models are global and can analyze both domestic and inter-
national land use changes, whereas others only analyze domestic land use changes in
the country producing biofuels. We describe a few examples of each of these types of
models in the next section, although there are many other models that are being used
to analyze the land use implications of biofuels (see reviews in Edwards, Mulligan, and
Marelli 2010; Prins et al. 2010).
Additionally, several biophysical models are being used to study the land use impacts of
biofuels. These models can provide the foundation for determining where certain crops
can be grown subject to biophysical constraints. They have been used to examine the
technical potential for biofuel production given land availability (see, for example, Cai,
Zhang, and Wang, 2011). We do not describe these biophysical models here in the interest
of brevity. Only some of the economic models below rely explicitly on biophysical models
for modeling heterogeneity in land suitability and crop productivity across locations.
3.1 Partial Equilibrium Models

Partial equilibrium models focus on a few sectors of the economy that are most closely
associated with biofuel production, namely the agricultural, forestry and fuel sectors.
Prices, production, and land allocation within these sectors are determined within the
model, and it is assumed that conditions in the rest of the economy remain unchanged
with biofuel production.
Static Equilibrium Models: These models include multiple markets represented by
demand and supply conditions; these markets are linked to each other by considering
own- and cross-price effects. A shock to demand or price in one market, for example
due to a biofuel policy, perturbs the market equilibrium and leads to changes in all mar-
kets to establish an instantaneous new equilibrium. Multimarket models usually do not
explicitly include a constraint on land availability.
The Food and Agricultural Policy Research Institute (FAPRI-CARD) model, devel-
oped at Iowa State University’s Center for Agricultural and Rural Development (CARD)
is a widely used multimarket model for analyzing the land use impact of biofuels
(Fabiosa et al. 2010). It includes world agricultural, food, livestock, fiber, and bioenergy
crop markets but does not explicitly include land constraints. The model currently only
considers corn ethanol as the source of biofuel. It also does not include a gasoline sec-
tor; a reduced form relationship is used to incorporate a feedback effect of biofuel pro-
duction on oil prices. Crop yields are responsive to prices over time both in the United
States and internationally. This price-induced yield increase is partially offset by the
reduced yields that result from expanding on to new crop acres. All non-US countries
are analyzed at the national level, with the exception of Brazil. The FAPRI-CARD model
is linked with the Brazilian Land Use Model (BLUM) to compute the impacts of sug-
arcane ethanol exports to the United States on land use and GHG emissions. BLUM
is developed by Brazil’s Institute for International Trade Negotiations (ICONE, 2011).
Land use change in BLUM occurs due to two effects: competition and scale. Different
activities compete for a given amount of land based on their net returns per acre and at
the same time returns to land determine the need for expansion of agricultural area over
natural vegetation (Gouvello 2010).
Dynamic Programming Models: Unlike static multimarket models that are reduced

form models of supply and demand, dynamic programming models are structural mod-
els that represent the behavior of utility maximizing consumers and profit maximizing
producers. They typically include fairly detailed biophysical data to model the dynamics
of crop yields, soil carbon changes, and GHG emissions. A programming model can be
a one-period or multiperiod model. These models use nonlinear programming meth-
ods to determine land allocation, equilibrium production, and prices that maximize the
discounted sum of consumer and producer surplus (net of externality costs) subject to
constraints on land, technology, and various material balances, and solves for endog-
enous output and factor prices. Landowners are assumed to choose allocation of land
among alternatives based on the net present value of the future returns, subject to cali-
bration constraints that prevent large deviations from historical land use patterns. Land
can shift from cropland to pasture based on relative returns, and equilibrium land prices
can vary across regions. In contrast to these intertemporal models, dynamic-recursive
models calculate results one period at a time. In all these models, traded and domesti-
cally produced goods are treated as homogenous and net trade flows are endogenously
determined as the difference between demand and supply.
Multiperiod models include the Forestry and Agricultural Sector Optimization
Model (FASOM), the Global Biomass Optimization Model (GLOBIOM) and the
Biofuel and Environmental Policy Analysis Model (BEPAM), which have a similar
structure but differ in terms of the sectors included, their geographic scope, the degree
of spatial heterogeneity and the time horizon considered. FASOM, developed at Texas
A&M University is a US-based model of the agricultural and forestry sectors (Adams
et al. 2005; Beach et al. 2009; Beach, Zhang, and McCarl 2012). The production activi-
ties incorporated include crop and livestock production and processing, bioenergy pro-
duction, and forest product production and processing. Biofuels can be produced from
several types of first-and second-generation feedstocks. In addition to biofuels, FASOM
contains a set of activities for replacing coal with biomass in electricity production.
Agricultural land can move between cropland pasture (marginal land) and cropland by
incurring conversion costs that are equal to the difference in land rental rates between
alternative uses based on the assumption of equilibrium in land markets. The model
assumes perfect foresight over a 100-year time horizon, and expected future prices are
identical to prices realized in the future. Technological change in crop production and
in biofuel conversion, which reduces biofuel production costs, is assumed to be exog-
enously given over time. FASOM also includes a comprehensive set of GHG mitigation
options, including biological sequestration of carbon in agricultural soils and forest
stands, alternative crop and livestock production practices to reduce emissions, and
bioenergy feedstock substitutes for fossil fuels (McCarl and Schneider 2001). The model
has recently been used to analyze the effect of biomass-storage costs for the mix of feed-
stocks used to produce second-generation biofuels and the land use changes due to bio-
fuel and carbon policies in the United States (Beach, Zhang, and McCarl 2012).
GLOBIOM, developed at the International Institute for Applied Systems Analysis
is similar to FASOM but models global agricultural, bioenergy, and forestry sectors
(Havlík et al. 2011). Although the regions are fairly aggregated, land use decisions are
examined at a much finer spatial resolution. The availability of land resources and their
productivity are determined using a biophysical model EPIC (Environmental Policy
Integrated Climate Model) and detailed geospatial data on soil, climate, and topog-
raphy, which is used to define homogenous simulation units at fine spatial scale. The
model assumes there is zero technological progress in crop improvement. Like FASOM,
this model also accounts for the major GHG emissions/sinks related to agriculture and
forestry. In GLOBIOM, ethanol is produced from corn and sugarcane, and biodiesel
from rapeseed and soybeans. Second-generation biofuels use forest products as feed-
stock. Bioenergy can also be used to generate heat and power. Demand for biofuels and
the share of first- and second-generation biofuels are fixed at exogenously given levels.
Unlike FASOM and BEPAM, which are US-based models, GLOBIOM examines land
use changes globally and it can, therefore, determine both domestic and international
land use changes due to biofuel production.
BEPAM, developed at the Energy Biosciences Institute at the University of Illinois
Urbana-Champaign differs from FASOM and GLOBIOM in that it integrates the agri-
cultural and fuel sectors (Chen, Huang, and Khanna 2011; Khanna et al. 2011; Chen
et al. 2012). Demand for gasoline and biofuels is derived from the demand for vehi-
cle miles traveled. The vehicle fleet structure is explicitly included and influences the
extent and type of biofuels that can be consumed. The model includes various first- and
second-generation biofuels, including from corn and imported sugarcane ethanol,
crop and forest residues, and perennial herbaceous grasses for biofuel production. It
includes the use of forest biomass residues but not the use of highly valued forest prod-
ucts as biofuel feedstocks. The mix of feedstocks used to produce biofuels and the share
of first-generation to second-generation biofuels is endogenously determined subject
to the restrictions and incentives provided by various biofuel and climate policies. The
model also distinguishes between domestically produced gasoline and imported gaso-
line. The imports and the price of gasoline in the United States is determined endoge-
nously; this allows for biofuel production in the United States to affect the world price of
gasoline and generate a feedback effect on the demand for biofuels in the United States.
The model incorporates spatial heterogeneity in yields and returns to land by consider-
ing decision making at a crop reporting district level. BEPAM includes life-cycle GHG
emissions from gasoline, diesel, and biofuel production and from all crop production
activities. Instead of assuming perfect foresight and extremely long time horizons for
decision making, BEPAM considers a 10-year rolling horizon for decision making based
on expectations about prices and land availability. These expectations are updated annu-
ally (for the following 10 years) after equilibrium market outcomes are realized each year
(see Chen et al. 2012). FASOM relies on historical crop mixes to generate results that are
consistent with farmers’ planting history, and allows crop acreage to deviate 10% from
observed historical mixes to accommodate new bioenergy crops and unprecedented
changes in future crop prices. BEPAM, instead uses the estimated own- and cross-price
crop elasticities to limit the flexibility of crop-acreage changes. Crop yields grow at an
exogenous rate over time and are also price responsive and thus partly determined
endogenously by the model. Another distinguishing feature of BEPAM is that it incor-

porates an experience curve for each type of biofuel, which allows for the costs of pro-
cessing feedstocks into biofuel to be endogenously determined based on the cumulative
production levels of the biofuel. Unlike other models, that consider cost-reducing tech-
nological change to be entirely determined by the passage of time, this approach incor-
porates the possibility of learning by doing and endogenous technological change. As a
result, policies that differ in their effect on the volume and mix of biofuels lead to vary-
ing levels of cost-reducing technological change in biofuel production.
The AGLINK-COSIMO model, used by the Organisation for Economic Co-operation
and Development and the Food and Agriculture Organisation is a dynamic-recursive
partial equilibrium model of world agricultural markets (OECD 2008). The model
treats most OECD member countries and their main trading partners as individual sup-
plying and demanding regions. The model considers fossil energy prices as exogenously
fixed. It includes first- and second-generation biofuels; the former is modeled endog-
enously, whereas the latter is treated as exogenous and is assumed to have no land use
implications. AGLINK-COSIMO does not simulate land use effects in Indonesia and
Malaysia; therefore, any land use impact resulting from land expansion is not included
in the quantified global arable land use change (Fonseca et al. 2010). Exogenous rates of
technical progress are assumed for first-generation biofuels and their by-products and
for crop-yield growth, based on past trends. Yields of major crops are price endogenous,
as in BEPAM. Land use constraints are not modeled explicitly but the implicit assump-
tion is that total agricultural land is fixed.
Dynamic programming models include more spatial detail and explicit land avail-
ability constraints compared to static multi-market models; the latter, however, are eas-
ily modified to consider noncompetitive behavior, and usually include many sectors and
cover a larger geographical area. The main appeal of both static and dynamic program-
ming partial equilibrium models is that they are generally well contained, and results
are easy to interpret and very intuitive. However, the main flaw of these models is that
they do not include some of the major feedbacks that are caused by changes in income
because of changes in commodity prices, which may in turn affect demand for com-
modities in other markets. To address these issues one needs a more inclusive model that
goes beyond the small number of sectors incorporated in partial equilibrium analysis.
3.2 Computable General Equilibrium Models

Unlike partial equilibrium models, computable general equilibrium (CGE) models sim-
ulate economy-wide effects of a biofuels shock and include intersectoral linkages and
constraints on labor and capital and determine all prices and incomes in the economy
simultaneously. These models are global in scope, represent multiple economic sectors
in each region, and include factor markets for labor and capital. They consider some of
the feedbacks that biofuel policy may have through its costs to the taxpayer and through
the employment possibilities that are generated through the supply chain as a result of
biofuel production. CGE models are especially suited to address a globally common
problem like climate change. Although broad in geographic and sectoral scope, many
CGE models have limited spatial resolution and usually partition the world into a few
large homogenous regions called agro-ecological zones (AEZs). Each region has a
regional representative household that allocates resources domestically and a represen-
tative producer that produces goods and services using consumer-owned endowments
as primary inputs. Each region interacts with other regions through trade. Consumers
maximize utility and producers maximize profits in a perfectly competitive market
setting, leading to endogenously determined prices and quantities of goods and fac-
tors of production. These models typically limit the number of agricultural products
considered by categorizing individual commodities into large groups (e.g., all coarse
grains) and imposing the same behavioral and market assumptions on the individual
components.
CGE models analyzing the effect of biofuel production on land use include the Global
Trade Analysis Project (GTAP), the Integrated Global System Model (IGSM), and the
Modeling International Relationships in Applied General Equilibrium (MIRAGE)
model. The three models are similar in that they are global in scope, and are multire-
gional, multisectoral, and multifactoral models. All three models are using essentially
the same database developed by GTAP but differ in the base year used for calibration.
IGSM and MIRAGE are both dynamic models, with varying time-steps and time hori-
zons, whereas the GTAP is an intrinsically static model (CARB 2009; Hertel et al. 2010).
IGSM and MIRAGE are different from GTAP, which considers only managed land,
in that they model the conversion of natural forests and grasslands into cropland or
pasture land.
In CGE models, land conversion occurs within an agro-ecological zone (AEZ) or grid
cell. The easier it is for land to be converted from one use to another, the greater the
potential is for biofuel production and land use change. In GTAP and MIRAGE, the
ease of land conversion from one use to another is governed by a Constant Elasticity of
Transformation (CET) frontier. The CET frontier is used to determine the supply of par-
ticular types of land (pasture, cropland, forest) given total availability of land. It is based
on the assumption that a landowner allocates land to different uses in order to maxi-
mize the total rents. The responsiveness of land in a particular use to a change in the
land rent influences the ease with which land is transformed from one use to another.1
The greater the elasticity of transformation parameter, the easier it is for land to be
converted from one use to another. IGSM uses a number of techniques to model land
conversion, including an “observed land supply response” approach—which is similar
to the CET method described earlier. Other approaches, such as the “pure conversion
1
The absolute value of the CET parameter depends on the elasticity of supply of land to a given use
of land in response to a change in its rental rate and the share of revenue from that land use in the total
revenue for all land. The value of CET ranges between 0 and 1. The more dominant a given use in total
land revenue is, the smaller the value of CET (since the potential for further changes in the amount of
land in that use are small, even if land rents increase).
cost response” are also used. In this case, land conversion occurs as long as the cost of
conversion is less than returns from clearing land for production. The advantage of the
latter approach is that it allows land rents to equalize across all uses and is, therefore,
consistent with long run equilibrium behavior.
The GTAP model, developed at Purdue University’s Center for Global Trade Analysis
considers first-generation biofuels from coarse grains (ethanol), edible oils (biodiesel),
and sugarcane (ethanol) (Golub et al. 2010). It uses difference in land rental rates within
each AEZ between cropland, pasture, and forests to determine which land will be con-
verted to cropland as a result of increasing biofuel demand. For modeling the competi-
tion between livestock and crop sectors, it uses the average coarse-grain yield in each
AEZ as representative of pasture land yields. Unmanaged land such as shrubland,
savanna, and grassland, is not allowed to be brought into productive use (EPA 2010).
The IGSM is developed by the Massachusetts Institute of Technology Joint Program
for the Science and Policy of Global Change (Gurgel, Reilly, and Peltsev 2007; Melillo
et al. 2009). The model integrates three components, a dynamic recursive CGE model
(Emissions Prediction and Policy Analysis or EPPA), a climate model and a land ecosys-
tems model (Terrestrial Ecosystem Model). The IGSM framework has the most sophis-
ticated emissions modeling, as it features full dynamic accounting of carbon fluxes in
vegetation and soils. In contrast, GTAP and MIRAGE use constant factor intensities of
conversion from one land use to another. However, the modeling of biofuel produc-
tion pathways is relatively coarse in the IGSM, whereas GTAP and MIRAGE have
fairly detailed modeling of first-generation biofuel-production pathways, including
co-products and interaction with the livestock industry. One of the distinguishing fea-
tures of IGSM and MIRAGE is that they incorporate both managed and unmanaged
land unlike GTAP, which includes only managed land. Another advantage of IGSM is
that it also considers feedback effects between the climate and the economy, and can
examine the land use effects of various climate stabilization policies.
The MIRAGE model developed at the International Food and Policy Research
Institute is a dynamic recursive CGE model that considers two main biofuel sectors,
ethanol and biodiesel from first-generation feedstocks (Al-Riffai, Dimaranan, and
Laborde 2010; Laborde 2011). Feedstocks for ethanol include wheat, sugarcane, sugar
beet, and maize, and those for biodiesel include palm oil, soybean oil, sunflower oil
and rapeseed oil. It combines a bottom-up approach for the biofuel sector to include
production costs and volume, by-products and input requirements. It improves on the
GTAP model, which only includes values and not physical quantities, by linking land
value and volume. It allows for intensive margin effects, through increased use of fertil-
izers to increase crop yields, exogenous technical change and endogenous factor based
intensification (land combined with more labor and capital). Extensive margin effects
are considered by differentiating between different types of AEZs and allowing for sub-
stitution of land among different crops and expansion of arable land using different
land-responsiveness coefficients in different AEZs. Allocation of land expansion among
different types of unmanaged land (grasslands, shrublands, etc.) is based mostly on his-
torical data and on remote sensing data for some countries.
4. Drivers of Land Use Change in Economic

Models of Biofuels
Based on the models discussed in the previous section, we identify several demand and
supply side factors that affect the estimated land use change impact of biofuels.
4.1 Demand-Side Determinants of Land Use Change

Ease of Substitution Between Liquid Fossil Fuels and Biofuels: The most important
demand side considerations are the technological constraints to substitution of bio-
fuels for gasoline (that is the blend wall), the substitution of one type of biofuel for
another type and the price of substitute goods (namely, liquid fossil fuels). In addi-
tion to the vehicle fleet, the infrastructure for distributing ethanol and blending
it could also be a constraint to its usage. Models differ in the way they incorporate
demand for ethanol. The partial equilibrium models, with the exception of BEPAM
and the FAPRI-CARD model consider ethanol and gasoline to be perfect substitutes.
Unlike BEPAM, other models such as FASOM and the FAPRI-CARD model represent
demand for biofuels either as a mandated quantity or as a result of an oil price shock
(which creates demand for its substitute good, ethanol). CGE models include an elas-
ticity of substitution between biofuels and gasoline and between different types of bio-
fuels but differ in the extent to which they consider these fuels to be substitutes. Earlier
versions of FAPRI-CARD ignored the effect of biofuels on global petroleum markets
and its feedback effect on the demand for biofuels and other bottlenecks to consume
biofuels. As a result, the oil-price shock considered by Searchinger et al. (2008) led
to a much greater increase in biofuel production and indirect land use change than it
would if these bottlenecks are considered (as discussed in Dumortier et al. 2011). The
analysis using BEPAM shows that increasing the elasticity of substitution between
biofuels and gasoline in the presence of volumetric tax credits for biofuels leads to a
significant increase in biofuel production.
Size and Type of Policy Shock: Biofuel production can be induced by various pol-
icies, such as technology mandates and subsidies as well as by climate policies like a
Low Carbon Fuel Standard or carbon cap and trade policy. The type of policy stimu-
lus as well as the magnitude of the stimulus will affect the demand for biofuels and the
mix of first- and second-generation biofuels induced and, thus, the direct and indirect
land use changes. Studies differ in their assumptions about the nature and magnitude
of the policy shock considered. The size of the policy shock matters because the rela-
tionship between land use change and biofuel production is nonlinear. An increase in
the volume of biofuel increases land requirements more than proportionately. This is
because greater pressure for biofuel production from a higher target results in increas-
ing use of less efficiently produced feedstock. Similarly the adverse effect on indirect
land use changes increases nonlinearly with an increase in biofuel production. Chen
et al. (2012) also show that a carbon tax policy will lead to a modest increase in corn
ethanol consumption at low carbon prices and will not induce any second-generation
ethanol production. Carbon prices that are over $150 per ton of CO2 will be needed to
induce production of second-generation biofuels. The extent to which price-based poli-
cies like a carbon tax or a biofuel subsidy will induce demand for biofuels depends on
the responsiveness of demand for fuel or of vehicle kilometers traveled to its price.
Mix of Policies: Countries are typically using multiple policy instruments to support
biofuels. Chen et al. (2012) and Khanna et al. (2011) show (using BEPAM) that the land
use impacts of the Renewable Fuel Standard (RFS), can be significantly modified by
accompanying tax credits for second-generation biofuels that can increase their com-
petitiveness relative to corn ethanol. Since these second-generation biofuels have higher
yields per unit of land, this policy induced shift in the mix of biofuels lowers the demand
for corn and the land needed to meet the RFS. Moreover, it has implications for the type
of land used for biofuel production: the use of noncropland for second-generation bio-
fuels increases, whereas the use of cropland for corn production decreases.
Land use is also sensitive to trade barriers. The US import tariff on biofuels reduces
the competitiveness of sugarcane ethanol and increases the land use impact of the RFS.
Trade liberalization would increase the volume of sugarcane ethanol used to meet the
mandate and lower the pressure on diversion of land to biofuel production domestically
in the United States. Using BEPAM, Chen and Khanna (2012) show that the removal of
the tax credit for ethanol and the tariff on imports of sugarcane ethanol to the United
States could significantly alter the mix of biofuels and increase reliance on sugarcane
ethanol to meet the RFS with implications for land use under corn and sugarcane in the
United States and Brazil, respectively.
Al-Riffai, Dimaranan, and Laborde (2010) consider the implications of the EU
Renewable Energy Directive (RED) with and without trade liberalization in the EU
and find that trade liberalization significantly changes the impact on biofuel produc-
tion in the EU. They find that the removal of tariffs on ethanol would lead to a surge
in European imports of sugarcane ethanol. In 2020 ethanol production would increase
by 157% in the EU under the EU RED in the absence of trade liberalization, whereas
it would decrease by 48% in the event of the full liberalization scenario because of
increased imports from Brazil.
4.2 Supply-Side Determinants of Land Use Change

Conversion Efficiencies and Co-Products: Conversion efficiencies differ across feedstocks
(Huang et al. 2013) with some energy crops yielding more than twice as many liters of
biofuels per hectare as corn ethanol. Therefore, a change in the mix of biofuels toward
sugarcane or high-yielding energy crops would significantly lower the land use impact
of biofuels.
The land requirements for biofuels also depend on their co-products, which can
replace other products in the market place, reducing the net quantity of food or feed
displaced. Thus, the amount of additional or net land required to produce these bio-
fuels is less than the total amount of land on which the biofuel crop is produced. Most
first-generation biofuels produce co-products that substitute for products that would
otherwise require land. This is particularly the case for corn ethanol, which produces
DDGS that can substitute for corn meal and soymeal used for animal feed. Taheripour
et al (2010) introduced by-products in the GTAP model and showed that it reduced the
need for cropland conversion due to US and EU biofuel mandates by 27%.
Land Productivity in Intensive and Extensive Margins: Assumptions about the rate
of growth of crop productivity affect the land use impact of biofuels. Yield increases
through the application of nonland inputs on currently utilized land will lower the rate
of land conversion to cropland. Most models specify an exogenous rate of growth for
yields for crops and conversion efficiency. Some models, such as BEPAM, GTAP and
MIRAGE, allow for the possibility of price-induced yield growth. The effect of a policy
shock on land use also depends on differences in yields between land under crop pro-
duction and marginal land. Keeney (2010) reports that estimates of the ratio of marginal
to average land productivity range from 0.47 to 0.9 in the literature. MIRAGE assumes
that the productivity of marginal land is half of the average productivity in existing
cropland for all regions, except in Brazil where the value is 0.75 (Laborde 2011). GTAP
assumes this ratio is 0.66 globally (Hertel et al. 2010). Data on the productivity of mar-
ginal land, particularly in developing countries is limited, and much more research is
needed in this area. There is some evidence that yields may not be much lower on newly
converted lands on the agricultural frontier; Babcock and Carriquiry (2010) found that
regions in Brazil experiencing faster expansion of soybeans did not have lower soybean
yields or yield growth.
Ease of Substitution of Land from one use to Another: As prices change,
profit-maximizing producers change the mix of crops they produce and may bring
noncropland into crop production. However, land use changes are costly. For example,
former pasture or forestland is expected to be less productive for crop production com-
pared to existing cropland. Models like GTAP and MIRAGE have reflected the cost of
converting land from one use to another by specifying a CET value that represents the
ease of substitutability between crops and other uses. This approach introduces nonlin-
earity in the ease of conversion of land with the implicit costs of conversion increasing as
more land is converted. This creates greater pressure at the extensive margin to expand
cropland as the demand for land for biofuels increases. A similar mechanism applies to
pasture and forestland that is converted to cropland. Substitution possibilities are lim-
ited and nonlinear due to the CET effect.
In other models, land conversion is based on the net returns to land and a conversion
cost that is incurred by converting unmanaged land to managed land. This may lead to a
lower elasticity of transformation across land uses in the short run but a larger elasticity
of transformation in the long run. The advantage of the approach based on net returns
and conversion cost is that it allows rents to equalize across all uses and is, therefore,
consistent with long-run equilibrium. Using IGSM, Gurgel, Reilly, and Paltsev (2007)
show that global bioenergy production is 10–20% greater when land conversion is based
only on net returns as compared to when it is based on the elasticity of transformation,
which tends to limit market response to follow observed historical trends. Chen, Huang,
and Khanna (2011) show that limits on the amount of idle/marginal land that can be
converted to energy crops in BEPAM can raise the costs of producing energy crops
while reducing the ease of conversion of land across different conventional crops can
raise the costs of producing corn ethanol.
Technological Factors: Technological change in the biofuel industry will be a signifi-
cant driver of land use change. The cost-effectiveness of second-generation biofuels and
the type of technological development that occurs for conversion of feedstocks to liquid
fuel will influence the amount of land that is converted to energy crops. High initial
costs of producing advanced biofuels and low learning rates will reduce their competi-
tiveness relative to first-generation biofuels and require larger diversion of land from
food to fuel production to meet given biofuel targets. Development of new technologies
for harvesting biomass, for collecting crop residues in one pass over the field and meth-
ods for establishing energy crops (using seeds or rhizomes) can have a significant effect
on the mix of feedstocks that are produced and land required to produce them (Chen
et al. 2012). There are very few studies that analyze the implications of varying levels of
technological development and costs of new biofuel technologies on land use. Using
BEPAM, Chen et al. (2012) show that higher processing cost for cellulosic biofuels and
low learning rates relative to the benchmark case would significantly affect the mix of
biofuels and reduce land under energy crops while increasing acreage under corn for
ethanol and total land under crop production. Beach, Zhang, and McCarl (2012) show
that high storage costs for feedstocks like crop residues and energy crops can reduce
their competitiveness relative to corn, which has a well-developed and low-cost infra-
structure for storage and marketing.
Ease of Transmission of Price Shocks in World Markets: The impact of increased biofuel
production on land use changes in the rest of the world depends on the ease with which
price shocks are transmitted from domestic markets to the rest of the world. This, in turn,
depends on assumptions about the ease with which goods can be traded across coun-
tries. Two approaches are currently used in the models described here—the Armington
approach, used in GTAP and MIRAGE, differentiates otherwise homogenous goods by
country of origin. In contrast, the Integrated World Model (IWM) used in IGSM2 and
in Searchinger et al. (2008) assume that there is one world price for homogenous goods
and goods will be produced where it is least costly to do so. IWM allows for an easier
transmission of a shock throughout the world economy. However, Golub et al. (2010)
note that using the IWM could result in “unrealistic” trade patterns. For example in
Searchinger et al. (2008), a lot of agricultural production and land conversion occur in
2
The IGSM model uses a Heckscher-Ohlin model for biofuels, which is similar to IWM, and it uses
Armington for other goods.
India due to favorable growing conditions, even though historically, India has not been
a major exporter of agricultural commodities. The Armington approach leads to results
that follow observed trade patterns. Countries or regions first decide on the sources of
their imports, and then, based on the composite import price, decide on the allocation
between domestic production and imports. A potential pitfall of this approach is that it
allows price differentials for homogenous goods, such as imported ethanol and domes-
tic ethanol, to persist.
5. How to Deal with Multiple Models
Having multiple models may be a source of confusion, but also a source of extra insight
and increased reliability for policy design. Different models are introduced for different
purposes or built under different assumptions, but when they address similar phenom-
ena they can provide a range of answers and a complementary insight that will allow
better decision making. The outcome of these models may differ due to differences in
model specifications, the counterfactual baseline considered, the policy scenarios ana-
lyzed the sectoral and geographic scope included. The results of these models should
be used to determine directions and ranges for outcomes and orders of magnitudes for
effects.
There are several approaches to deal with multiple models. Policy analysis can use tri-
angulation to synthesize the results from several models. When several models address
the same phenomenon, they provide a distribution of estimates. These distributions can
provide either a range of values that determine the impact or yield a weighted statis-
tical estimate based on all the studies that may contain much more information. The
second approach for researchers that are choosing which model to use is nesting two
or more models or linking them off-line. The same problem may have many dimen-
sions that have to be addressed at different degrees of detail. For example, when assess-
ing the impact of introducing a new feed crop in a certain region, a good understanding
of where the new crop can be produced and reliable parameters of the distribution of
yield and cost, are needed. Obtaining this information may require a very detailed bio-
physical model. The information that this model generates can be used as an input in
an economic model that can allocate land use over space and time based on economic
criteria. A third approach is modularity that takes the nesting approach much further
and involves incorporating subroutines of one model in another model. Developing a
network of models that can be linked with each other is quite challenging because dif-
ferent models use different softwares, operate on different time and geographical scales,
and so forth. However, a system of models that speak to one another can allow us to take
advantage of all the different components so that the total will be bigger than the sum of
the parts.
6. Key Findings on Land Use

Impacts of Biofuels
6.1 Land Requirement for Biofuels

In the near term, Hertel et al. (2010) find that 3.5 million hectares of land is needed to
meet the US corn ethanol mandate. This estimate depends on the baseline used, and
assumptions about growth in yield and population. Chen et al. (2011) find that 12 mil-
lion hectares are required in the United States to meet the total RFS, of which about
4.7 million hectares is the additional land needed to produce corn to meet the RFS.
Using MIRAGE, Laborde (2011) finds that the EU RED is expected to increase total
cropland globally by 1.73 million hectares without trade liberalization and 1.83 with
trade liberalization. The additional cropland comes primarily from pasture and man-
aged forests, with 80% of the land use change taking place on managed land.
The preceding estimates are for total land requirement, which includes direct and
indirect land use changes. Some studies specifically focus on estimating the indirect
land use change associated with biofuels. Khanna and Crago (2012) review these studies
and find that estimates for corn ethanol in the United States range from 20 to 430 hect-
ares per million liters. These estimates are sensitive to the scale of biofuel production, to
the counterfactual baseline, to the mix of policies and biofuels considered, and to varia-
tions in the parametric assumptions of the models.
In general, these studies show that the additional area planted to biofuel feedstocks
will come from a reduction in cropland and pasture area, as well as some conversion
of idle or natural areas to agricultural production. The results from the different mod-
els suggest that some competition with agricultural production is inevitable, even if
there is greater production of second-generation biofuels that rely on crop residues or
energy crops that can be grown on marginal land (Chen, Huang, and Khanna 2011).
In addition, conversion of forests to cropland is also to be expected. The impact of bio-
fuel feedstock production on deforestation and competition for land could be mini-
mized by increasing biomass yields and conversion efficiencies of biomass to liquid fuel.
Encouraging biofuels that have lower land requirements will also ease the pressure on
land supply.
6.2 Location of Biofuel Production

Most of the studies reviewed point to significant feedstock production in Latin America,
specifically Brazil, and Africa. The greater feedstock productivity and land availability
in these regions have the potential to make them the largest producers of biofuels in the
world in the long run. The United States is likely to be the third largest producer of biofu-
els (Reilly, Gurgel, and Paltsev 2008). Analysis of the EU RED using AGLINK-COSIMO
shows that RED requirements are met by significant increases in biofuels from Brazil
and Southeast Asia (Fonseca et al. 2010). In contrast, the analysis by Laborde (2011)
using MIRAGE shows significant land use change in sub-Saharan Africa. Within
Brazil, studies suggest that expansion is expected to occur in the center-south region
where sugarcane production has been traditionally grown (de Souza Ferreira Filho and
Horridge 2011; ICONE 2011). Models of land use change in Brazil show that an increase
in sugarcane production leads to land use change, but these changes are mostly due to
conversion of other cropland to sugarcane production. Deforestation occurs due to the
expansion of the agricultural frontier, but the land use change due to deforestation is
minimal. In the study by de Souza Ferreira Filho and Horridge (2011), deforestation
accounts for less than 3% of land expansion associated with increased biofuel produc-
tion from 2006–2020; the majority comes from a reduction in pasture and forest plan-
tations. They estimate the ILUC effect to be 8%, that is, 0.08 hectare is deforested per 1
hectare increase in sugarcane production. The results for biofuel production are mixed
for China and India. The production of biofuel feedstocks in these countries will depend
on the growth of domestic demand for food, which competes with biomass production
for land resources, and on government policy about enforcing mandates and providing
economic incentives for biofuel production. In the case of India, a 20% blend mandate
could be met by diverting about 1 million hectare (< 1% of cropland) from food crops
to sugarcane production without significant impact on the production of other crops
(Khanna et al. 2013).
The findings here are based on the assumption that production will occur in areas
with land availability as well as low cost of land and production. However, these models
do not account for other factors such as political stability that may deter production in
the African region. In addition to production cost, the exchange rates between biofuel
exporting and importing countries will also be an important determinant of biofuels
trade and production location (Crago et al. 2010).
6.3 Mix of Feedstocks

If biomass is to be produced on a grand scale without compromising food supply,
feedstock from nonfood sources like crop and forest residues, woody biomass, and
dedicated energy crops will dominate the biomass supply landscape in the coming
decades. Among energy crops, the incentives to use switchgrass and miscanthus for
producing second-generation biofuels have been analyzed in considerable detail using
FASOMGHG and BEPAM. The relative yields per hectare and conversion efficiency of
feedstock to liquid fuel differ considerably among these feedstocks and are crucial deter-
minants of which specific feedstocks will be used (EPA 2010; Beach, Zhang, and McCarl
2012; Chen et al. 2012; Huang et al. 2013) Miscanthus has a much higher biomass yield
per hectare of land than switchgrass or corn stover. Using BEPAM, which includes mis-
canthus as a feedstock option, Chen et al. (2011) find that miscanthus has a significant
potential as a feedstock compared to other feedstocks due to its superior yields, longer
lifetime, and low input requirements. Earlier versions of FASOM, without miscanthus as
an energy crop, predicted that corn stover and switchgrass will be the main feedstocks
used to meet the RFS (EPA 2010). Recent analysis that includes miscanthus as a feed-
stock option in FASOM also find that miscanthus is the dominant feedstock to produce
second-generation biofuels (Beach, Zhang, and McCarl 2012). Biofuel subsidy policies
that are typically paid per unit of volume or per ton of biomass also encourage high yield-
ing feedstocks and would create further incentives to increase acreage under the higher
yielding miscanthus as compared to switchgrass in the United States (Khanna et al. 2011).
7. Summary and Implications
This chapter presents an overview of alternative modeling approaches to assess the

factors that will determine land use changes associated with the introduction of bio-
fuel and their implications for the location of land use changes and the types of land
that will be converted to biofuel production. Numerical models used to predict land
use changes simulate future scenarios of biofuel production using economic param-
eters on market behavior, in particular elasticities of demand and supply in relevant
markets, biophysical data on land characteristics and suitability, as well as projections
about future biofuel technologies. These models vary in their scope of operation: some
are regional or national and others are global. They also vary in their assessment of the
feedbacks associated with the introduction of biofuels through the agricultural and fuel
markets. Nevertheless, the general qualitative findings of various models are consis-
tent, although there is a high degree of variability in the numerical outcomes reflect-
ing differences in assumptions regarding the behavior of economic agents, features of
the technology, model structure and policy scenarios. These studies show that biofuels
have the potential to meet a significant percentage of liquid fuel demand and demand
for renewable electricity in the future and that current targets being set in the United
States and EU for renewable fuels are feasible with moderate increases in crop prices.
To reach these targets, however, requires the use of a diverse portfolio of biofuel feed-
stocks that includes first-generation feedstocks, in particular, corn and sugarcane,
second-generation feedstocks like miscanthus and switchgrass, and multiple forest
products and crop residues. High-yielding feedstocks are more likely to be economically
viable and are critical to achieve policy targets with minimal adverse impacts on crop
prices and indirect land use change.
Model results also show that large-scale production of bioenergy will entail some
trade-offs. Because biomass production competes with agricultural production for land,
crop prices will increase, although the effect is likely to be modest if the biomass is pro-
duced from crop/forest residues and from dedicated energy crops. These crops are more
likely to be economically viable on marginal land. Thus, expansion of second-generation
biofuels is likely to result in expansion of agricultural land for biofuel production
as fallow and low-quality pasture land is brought into production. This could reduce
some areas under native grasses or permanent pastureland unless regulated by policy.
Some policies, such as large tax credits for second-generation biofuels, may even make
energy crops competitive on cropland. In this case the substitution between food and
fuel cannot be completely avoided, but the price effects are likely to be smaller because
the amount of land that will need to be diverted to meet given biofuel targets will be
smaller than with first-generation biofuels. Moreover, the productivity of both tradi-
tional agricultural crops as well as of biofuel feedstock will crucially determine the link-
age between food and fuel production and prices, and will have implications for land use
changes. Higher productivity of traditional crops and biofuel crops reduces the conflict
between biofuel and food production and the adverse impact of biofuel production on
deforestation and GHG emissions. Increasing biomass yields and conversion efficien-
cies will lower the land requirement for biofuel production. Intensifying livestock pro-
duction could also ease the competition for land.
Acknowledgments
The authors are grateful for funding provided by the BP Environmental Sustainability
Challenge Project and the Energy Biosciences Institute, University of California,
Berkeley, for this research.
References
Adams, D., R. Alig, B. McCarl, and B. C. Murray. 2005. FASOMGHG conceptual structure,
and specification: Documentation. http://agecon2.tamu.edu/people/faculty/mccarl-bruce/
FASOM.html.
Al-Riffai, P., B. Dimaranan, and D. Laborde. 2010. Global trade and environmental impact
study of the EU Biofuels Mandate. International Food Policy Research Institute Report for
the Directorate General for Trade of the European Commission, Washington, DC.
Anderson, K., B. Dimaranan, J. Francois, T. Hertel, B. Hoekman, and W. Martin. 2001. The cost
of rich (and poor) country protection to developing countries. Journal of African Economies
10: 227–257.
Babcock, B., and M. Carriquiry, M. 2010. An exploration of certain aspects of Carb’s approach
to modeling indirect land use from expanded biodiesel production. Center for Agricultural
and Rural Development. Ames: Iowa State University.
Beach, R. H., A. J. Daigneault, B. A. McCarl, and S. Rose. 2009. Modeling alternative policies
for forestry and agricultural bioenergy production and GHG mitigation. Paper presented
at the Association of Environmental and Resource Economists Workshop. Energy and the
Environment, June 18–20, 2009, Washington, DC.
Beach, R. H., Y. W. Zhang, and B. A. McCarl. 2012. Modeling bioenergy, land use, and GHG
emissions with FASOMGHG: Model overview and analysis of storage cost implications.
Climate Change Economics 3(3): 1250012.
Cai, X., X. Zhang, and D. Wang. 2011. Land availability for biofuels. Environmental Science and
Technology 45: 334–339.
California Air Resources Board (CARB). 2009. Proposed regulation to implement the low carbon
fuel standard, Vol. I. California Environmental Protection Agency Air Resources Board.
http://www.arb.ca.gov/fuels/lcfs/030409lcfs_isor_vol1.pdf.
Chakravorty, U., B. Magne, and M. Moreaux. 2008. A dynamic model of food and clean energy.
Journal of Economic Dynamics and Control 32: 1181–1203.
Chen, X., H. Huang, and M. Khanna. 2011. Land use and greenhouse gas implications of biofu-
els: Role of technology and policy. Climate Change Economics 3(3): 1250013.
Chen, X., H. Huang, M. Khanna, and H. Onal. 2012. Meeting the mandate for biofu-
els: Implications for land use and food and fuel prices. In The intended and unintended effects of
U.S. agricultural and biotechnology policies, eds. J. G. Zivin and J. Perloff. Chicago: University
of Chicago Press.
Chen, X. and M. Khanna. 2013. Food vs. fuel: Role of biofuel policies. American Journal of
Crago, C., M. Khanna, J. Barton, E. Giuliani, and W. Amaral. 2010. Competitiveness of Brazilian
sugarcane ethanol compared to US corn ethanol. Energy Policy 38: 7404–7415.
De Souza Ferreira Filho, J. B., and M. Horridge. 2011. Ethanol expansion and indirect land use
change in Brazil. Clayton, Australia: Centre of Policy Studies, Monash University.
Dumortier, J., D. J. Hayes, M. A. Carriquiry, F. Dong, X. Du, A. Elobeid, J. Fabiosa, and S.
Tokgoz. 2011. Sensitivity of carbon emission estimates from indirect land-use change.
Applied Economic Perspectives and Policy 33: 428–448.
Edwards, R., D. Mulligan, and L. Marelli. 2010. Indirect land use change from increased biofu-
els demand. JRC Scientific and Technical Reports JRC 59771.
EPA. 2010. Renewable fuel standard program regulatory impact analysis. Washington,
DC: Assessment and Standards Division, Office of Transportation and Air Quality, U.S.
Environmental Protection Agency.
Fabiosa, J. F., J. C. Beghin, F. Dong, A. Elobeid, S. Tokgöz, and T. Yu. 2010. Land allocation
effects of the global ethanol surge: Predictions from the International FAPRI Model. Land
Feder, G., and D. Feeny. 1991. Land tenure and property rights: Theory and implications for
development policy. The World Bank Economic Review 5: 135–153.
Feder, G., R. E. Just, and D. Zilberman. 1985. Adoption of agricultural innovations in develop-
ing countries: A survey. Economic Development and Cultural Change 32: 255–298.
Fonseca, M., A. Burrell, H. Gay, M. Henseler, A. Kavallari, R. M’barek, I. Domínguez, and A.
Tonini. 2010. Impacts of the EU biofuel target on agricultural markets and land use: A com-
parative modelling assessment. European Commission Joint Research Centre, Institute for
Prospective Technological Studies. http://ftp.jrc.es/EURdoc/JRC58484.pdf.
Gardner, B. L. 1992. Changing economic perspectives on the farm problem. Journal of Economic
Literature 30: 62–101.
Golub, A., T. Hertel, F. Taheripour, and W. Tyner. 2010. Modeling biofuels policies in general
equilibrium: Insights, pitfalls, and opportunities. In New developments in computable gen-
eral equilibrium analysis for trade policy (Frontiers of Economics and Globalization, Vol. 7),
153–187. Bingley, UK: Emerald Group Publishing.
Gouvello, C. 2010. Brazil low-carbon country case study: The World Bank Group. Washington,
DC: The International Bank for Reconstruction and Development, The World Bank.
Gurgel, A., J. Reilly, and S. Paltsev. 2007. Potential land use implications of a global biofuels
industry. Journal of Agricultural and Food Industrial Organization Special Issue: Explorations
in Biofuels Economics, Policy and History 5(2), Article 9.
Havlík, P., U. A. Schneider, E. Schmid, H. Böttcher, S. Fritz, R. Skalský, K. Aoki, S. D. Cara, G.

Kindermann, F. Kraxner, S. Leduc, L. Mccallum, A. Mosnier, T. Sauer, and M. Obersteiner.
2011. Global land-use implications of first and second generation biofuel targets. Energy
Policy 39: 5690–5702.
Heckscher, E. F., and B. Ohlin. 1991. Heckscher-Ohlin trade theory, eds. Harry Flam and M. June
Flanders. Cambridge, MA: MIT Press.
Hertel, T. W., A. A. Golub, A. D. Jones, M. O’Hare, R. J. Plevin, and D. M. Kammen. 2010.
Effects of US maize ethanol on global land use and greenhouse gas emissions: Estimating
market-mediated responses. BioScience 60(3): 223–231.
Hochman, G., D. Rajagopal, and D. Zilberman. 2010. The effect of biofuels on crude oil markets.
AgBioForum 13: 112–118.
Huang, H., M. Khanna, H. Onal, and X. Chen. 2013. Stacking low carbon policies on the renew-
able fuel standard: Economic and greenhouse gas implications. Energy Policy 56: 5–15.
Instituto de Estudos do Comércio e Negociações Internacionais (ICONE). 2011. Simulating
land use and agriculture expansion in Brazil: Food, energy, agroindustrial and environmen-
tal impacts. Internal report, ICONE, Brazil. http://www.iconebrasil.org.br/arquivos/noti-
cia/2258.pdf.
Irwin, E., K. Bell, N. Bockstael, D. Newburn, M. Partridge, and J. Wu. 2009. The economics of
urban-rural space. Annual Review of Resource Economics 1: 435–459.
Keeney, R. 2010. Yield response and biofuels: Issues and evidence on the extensive margin.
Paper presented at the World Congress of Environmental and Resource Economists. June
28–July 2, 2010, Montreal.
Khanna, M., A. W. Ando, and F. Taheripour. 2008. Welfare effects and unintended consequences
of ethanol subsidies. Applied Economic Perspectives and Policy 30: 411–421.
Khanna, M., X. Chen, H. Huang, and H. Önal. 2011. Land use and greenhouse gas mitiga-
tion: Effects of biofuel policies. University of Illinois Law Review 2: 549–588.
Khanna, M., and C. L.Crago. 2012. Measuring indirect land use change with biofu-
els: Implications for policy. Annual Review of Resource Economics 4: 161–184.
Khanna, M., H. Onal, C. L. Crago, and K. Mino. 2013. Can India meet biofuel policy tar-
gets? Implications for food and fuel prices. American Journal of Agricultural Economics
95(2): 296–302.
Krutilla, J. V. 1967. Conservation reconsidered. The American Economic Review 57: 777–786.
Laborde, D. 2011. Assessing the land use consequences of EU biofuel policies. Specific Contract
No SI2. 580403 implementing Framework Contract No. TRADE/07/A2. Final Report for
ATLASS Consortium. http://www.ifpri.org/sites/default/files/publications/biofuelsre-
portec2011.pdf.
McCarl, B. A., and U. A. Schneider. 2001. Greenhouse gas mitigation in U.S. agriculture and
forestry. Science 294: 2481–2482.
Melillo, J. M., J. M. Reilly, D. W. Kicklighter, A. C. Gurgel, T. W. Cronin, S. Paltsev, B. S. Felzer,
X. Wang, A. P. Sokolov, and C. A. Schlosser. 2009. Indirect emissions from biofuels: How
important? Science 326: 1397–1399.
Mendelsohn, R., and A. Dinar. 2009. Land use and climate change interactions. Annual Review
of Resource Economics 1: 309–332.
OECD. 2008. Biofuel support policies: An economic assessment. Organisation for Economic
Co-operation and Development.
Prins, A., E. Stehfest, K. Overmars, and J. Ros. 2010. Are models suitable for determining ILUC
factors? Netherlands Environmental Assessment Agency (PBL) 12. http://www.pbl.nl/en/
publications/2010/Are-models-suitable-for-determining-ILUC-factors.html
Rajagopal, D., S. E. Sexton, D. Roland-Holst, and D. Zilberman. 2007. Challenge of bio-
fuel: Filling the tank without emptying the stomach? Environmental Research Letters 2,
044004.
Rajagopal, D., and D. Zilberman. 2007. Review of environmental, economic and policy aspects
of biofuels. Policy Research Working Paper Series: 4341. Washington, DC: The World Bank.
Reilly, J., A. Gurgel, and S. Paltsev. 2008. Biofuels and land use change. In Proceedings of Farm
Foundation Conference: Transition to a bioeconomy: Environmental and rural development
impacts, eds. M. Khanna, October 15–16, 2008, St. Louis, MO, 1–17.
Ricardo, D. 1891. Principles of political economy and taxation. London: G. Bell and Sons.
Schoengold, K., and D. Zilberman. 2007. The economics of water, irrigation, and development.
In Handbook of agricultural economics, Vol. 3, eds. R. Evenson and P. Pingali, 2933–2977.
Amsterdam: Elsevier.
Searchinger, T., R. Heimlich, R. A. Houghton, F. Dong, A. Elobeid, J. Fabiosa, S. Tokgoz, D.
Hayes, and T.-H. Yu. 2008. Use of U.S. croplands for biofuels increases greenhouse gases
through emissions from land-use change. Science 319: 1238–1240.
Searchinger, T. D. 2010. Biofuels and the need for additional carbon. Environmental Research
Letters 5.
Sunding, D., and D. Zilberman. 2001. The agricultural innovation process: Research and tech-
nology adoption in a changing agricultural sector. In Handbook of agricultural economics,
Vol. 1. eds. B. L. Gardner and G. C. Rausser, 207–261. North-Holland, Amsterdam: Elsevier.
Taheripour, F., T. Hertel, W. Tyner, J., Beckman, and D. Birur. 2010. Biofuels and their
by-products: Global economic and environmental implications. Biomass and Bioenergy
34: 278–289.
Tsur, Y., and A. Zemel. 2005. Scarcity, growth and Rand. Journal of Environmental Economics
and Management 49: 484–499.
Von Thünen, J. 1966. Isolated state: An English edition of Der Isolierte Staat by C.M. Wartenberg,
ed. P. Hall. Oxford: Pergamon Press.
Xabadia, À., R. U. Goetz, and D. Zilberman. 2006. Control of accumulating stock pollution by
heterogeneous producers. Journal of Economic Dynamics and Control 30: 1105–1130.
C HA P T E R 5
MODELING THE
D E T E R M I NA N T S O F
FA R M L A N D VA LU E S I N T H E
U N I T E D S TAT E S
C Y N T H IA J. N IC K E R S ON A N D
W E N D ONG Z HA NG
Although once distributed for free to the earliest settlers in the United States, land has
long been traded in private markets. For most of the past 100 years, real estate (land and
structures) has comprised a significant portion of the wealth of many landowners. This
is particularly true for the farming sector, which also is a major user of land—51% of the
US land base in 2007 was in agricultural use (Nickerson et al. 2012). Valued at $1.85 tril-
lion in 2010, farm real estate accounted for 85% of total US farm assets (US Department
of Agriculture, Economic Research Service [USDA-ERS] 2012). Because it comprises
such a significant portion of the balance sheet of US farms, changes in the value of farm
real estate have an important bearing on the farm sector’s financial performance. Farm
real estate also represents the largest single investment item in a typical farmer’s invest-
ment portfolio; as a principal source of collateral for farm loans and a key component of
many farmers’ retirement funds, changes in its value can affect the financial well-being
of landowners.
Because of the longstanding significance of land values to both the farming sector
and landowners, understanding the determinants of farmland values has been the sub-
ject of a great deal of economic research. Although the earliest studies date back well
more than 100 years, most methodological and empirical advances in the study of farm-
land values have occurred more recently. The farmland valuation models developed
and tested in the ensuing decades have generally evolved to help explain changes in
farmland values that began to diverge from trends in returns to farming. The foci of the
research have shifted over time partly due to recognition that existing models were not
very well explaining significant swings in farmland values observed both at national and
112 Cynthia J. Nickerson and Wendong Zhang
regional levels. The direction of research has also been influenced by the types of data
available for empirical analysis, with the availability of increasingly detailed data spawn-
ing new opportunities to explain the determinants of farmland values and changes in
those values.
In this chapter, we provide a comprehensive overview of significant developments
in modeling farmland values. In doing so, we cover a wide variety of models and give
particular attention to methodological challenges and recent modeling innovations.
We begin by outlining the capitalization model, which has been—and continues to
be—widely used as the theoretical basis in economic studies on this topic. We next
discuss modeling efforts to address perceived shortcomings of this basic model in
the context of farmland values. Dynamic modeling approaches using aggregate data
to explain changes in farmland values have been heavily used for this purpose. We
then turn attention to cross-sectional hedonic models that use spatially disaggregate
or parcel-level data to examine the influence of particular determinants on farmland
values, which in recent decades have become the mainstay of modeling techniques
in the farmland values literature. We describe estimation issues that arise in hedonic
modeling of farmland values, devoting most attention to those methodological issues
that deserve special consideration in the context of farmland values, including spa-
tial dependence and sample selection bias. In the course of doing so, we focus less on
the specific findings of the studies (of which there are many) and more on the models
themselves.
Because many of the advances in the study of farmland values occurred due to
changes in farmland markets over time and to the applications of new modeling tech-
niques, it is instructive to proceed in a more or less linear fashion, beginning with the
earliest models, and describe the conditions that induced changes in modeling. We con-
clude with the most recent advances in modeling the determinants of farmland values
and a discussion of what we perceive to be promising future research directions.
1. The Basic Capitalization Model
David Ricardo’s (1817) formulation of an economic theory of rent, which was originally
developed in the context of the value of farmland, is an important theoretical corner-
stone in the basic model of land rents and land values. Ricardo’s key insight was that land
that differs in quality and is limited in supply generates rents that arise from the pro-
ductive differences in land quality or from differences in location. Ricardo’s work and
that of others (e.g., Malthus’ concept of residual surplus and von Thünen’s theory of rent
differentials arising from distance from a central market) form the basis of our modern
understanding of land rents and land values (Barlowe 1986).
In the basic model, farmland is recognized as a fixed factor of production. Farmland
prices are comprised of the discounted stream of economic returns generated by the
Modeling the Determinants of Farmland Values 113
land, where returns are defined as the return above all variable factors of production.
Formally, the model is written as
∫ A (t )e
− rt
Pt = i dt (1)
t =0
where Pt is the price of farmland in period t, Ai is annual net returns from farming, and
r is the discount rate. The use of this basic model underlies not only farmland values
research but also is used to model landowner decisions about land use choices.
Throughout the early decades of the 1900s, even though commodity prices experi-
enced both rapid increases and significant declines, farmland prices and net returns
remained relatively closely correlated. Farmland values began diverging from net
returns in the 1950s, with farmland values increasing fourfold relative to farm income
between 1952 and 1964 (Chryst 1965). Around this time, several studies attempted
to model farmland values in a simultaneous equations framework (e.g., Herdt and
Cochrane 1966; Tweeten and Martin 1966). However, this direction of research was
short-lived, due primarily to concerns about identifying classic supply equations in a
market with inelastic farmland quantities (e.g., Falk 1991), and subsequent research that
determined the ability of these models to explain changes in farmland prices was very
sensitive to the time period of the data (Pope et al. 1979).
2. Developments Using Time Series Models

and Aggregate Data
Dramatic changes in farmland prices occurred in the following decades, with rapid
appreciation in the 1970s followed by large declines in the 1980s. These changes raised
a number of questions about the usefulness of the basic capitalization model in explain-
ing changes in farmland values. In addition to assuming that land is valued only for its
economic returns (which are known with certainty), the model assumes a constant dis-
count rate, risk neutrality, and no effects from capital gains, inflation, transaction costs,
and taxes. These issues lend themselves to examination using dynamic approaches, and
many of the ensuing studies used time series techniques used to study stock price move-
ments to test empirically these and other assumptions. These studies also used highly
aggregated data in most cases—often state-level averages—due at least in part to a lack
of more disaggregated, high-quality data for farmland.
An issue receiving early attention was the specification of A, net returns. Melichar
(1979) pointed out that net farm income may not be the best measure of returns because
it includes returns to all productive assets, labor, and management time. As a result,
many subsequent papers used net rents instead of net farm income as the measure of
returns. However, other studies support the use of imputed returns (Mishra et al. 2004).
2.1 Distributed Lag and Vector Autoregressive Models

Several time-series studies used distributed lag models to test the relative effects of
returns and inflation on farmland price movements. Because returns anticipated in the
future are not observable, these models used observed returns in previous years to proxy
for expected returns. The models placed less weight on returns earned in the most recent
years than in earlier years, reasoning that changes are capitalized into land values only if
they persist. Using different specifications of distributed lags with different aggregated
data, both Alston (1986) and Burt (1986) found returns to be the major explanation
of land prices and the effects of inflation to be small at most. Alston’s study used data
from eight US Midwestern states between 1963 and 1982, whereas Burt used data from
Illinois over a similar time period. A study by Moss (1997) suggested that the relative
effects of returns and inflation vary by region, with returns providing more explanatory
power in regions relying more heavily on government payments.
Vector autoregressive (VAR) techniques were also used to test the basic capitaliza-
tion model. These models capture interdependencies by defining an equation for each
variable that is based on own-value lags, as well as on lags of the other variables in the
model. An often cited study is that by Featherstone and Baker (1987), who simultane-
ously estimated equations for farmland values, returns, and interest rates to examine the
time path of farmland value adjustments to changes in returns and interest rates. Using
annual data on US farmland values for 1910–1985, their results suggest that specula-
tive factors seem important: that is, farmland values overreact to shocks in values, real
returns, or interest rates, and the reaction lasts for up to six years. Others have used VAR
methods to test whether the discount rate in the capitalization model was time-varying
(e.g., Falk 1992). Assuming the trend series was difference-stationary rather than sta-
tionary, Falk and Lee (1998) used VAR and Iowa data from 1922–1994 and concluded
that the capitalization model explained farmland price movements in the long run; in
the short run, however, they concluded that overreactions to temporary shocks caused
deviations between prices and predictions of the capitalization model.
2.2 Cointegration Analysis

Advances in the study of time series data led to challenges of the stationarity assump-
tions used in traditional time series representations. A number of ensuing studies were
influenced by the work of Campbell and Shiller (1987), which showed that if the PV
model were to hold, (1) land prices and rents must both have the same time-series
properties, and (2) certain restrictions were required on the VAR representation of the
changes in rents and the spread between rents and land prices (see Falk 1991, 3–4). These
studies used cointegration analysis to overcome spurious results that could occur when
using traditional time series approaches with data characterized by nonstationarity and
unit roots. A number of these studies reject the present value model on the basis of an
inability to find that farmland prices and rents are cointegrated (e.g., Falk 1991; Clark et
al. 1993; Tegene and Kuchler 1993). However, Gutierrez et al. (2007) argue that this lack
of support may be due to previous studies’ not taking into account structural breaks and
also assuming that states’ data are independent of each other—which they point out is
unlikely to hold, given the common boom-bust cycles in the data typically employed.
Using recent advances in modeling nonstationary panel data and data from 31 US states
over 1960–2000, they find that, by controlling for structural breaks, they cannot reject
the present value model. Using a cointegration approach and error-correction models,
Erickson, Mishra, and Moss (2003) also found support for the present value model, but
note that the results are sensitive to the specification of the economic returns to land.
Cointegration analysis has also been used to examine whether discount rates vary by
income source. Weersink et al. (1999) found government payments tended to be dis-
counted less than market-based returns in Ontario. Schmitz (1995) found the opposite
in Saskatchewan, which Weersink et al. (1999) posit is a result of farmers viewing gov-
ernment payment programs in the former province as a more stable source of income
than the ad hoc transfers that are more characteristic of payments in the latter.
2.3 Structural Models
The conflicting evidence these studies find on the role of expectations, inflation,
time-varying discount rates, and other factors is attributed by some to the use of econo-
metric approaches that examine possible influences in isolation and which use specifi-
cations that are not based on economic theory (e.g., Just and Miranowski 1993; Chavas
and Thomas 1999; Weersink et al. 1999). In a seminal paper, Just and Miranowski (1993)
developed a comprehensive structural model to examine the multidimensional effects
of inflation on capital and savings-return erosion and real debt reduction, as well as of
changes in the opportunity cost of capital, while accounting for risk preferences and
transaction costs. Using state-level pooled cross section data from 1963–1986, they
found increased returns to farming, inflation, and opportunity cost were major expla-
nations of the large increases in farmland prices in the 1970s, whereas only the latter
two factors primarily explained subsequent large declines in the 1980s. Their results also
suggest that inflation and opportunity cost explained the tendency of changes in land
prices to exceed changes in rents (Featherstone and Baker 1987; Falk 1991). They did
not find the results were sensitive to the expectations regime used. Although the study
did not account for nonstationarity of the data as pointed out by Lence (2001), a subse-
quent study that did and which used very similar data found similar results (Awokuse
and Duke 2006).
In another particularly notable paper, Chavas and Thomas (1999) developed a

model at the microeconomic level that incorporates risk aversion, transaction costs,
and dynamic preferences. Recognizing that time series data have been available almost
exclusively only at an aggregate level, they described the conditions necessary for main-
taining consistency between microlevel decision rules and aggregate price data—and
the particular challenges for empirical modeling of the role of transaction costs. Using
data on US farmland values over 1950–1996, they found that both risk aversion and
transaction costs affected land prices and helped explain the inadequacies of the static
present value model.
2.4 Other Dynamic Modeling Approaches

Other dynamic modeling approaches have been employed in the farmland value lit-
erature, although they have not been adopted as widely as the models just discussed.1
Several of these techniques were utilized to specifically examine the influence of govern-
ment payments. Because agricultural payment programs in the United States have been
in place since the Agricultural Adjustment Act of 1933, several studies using dynamic
modeling approaches considered the impact of government payments on changes in
farmland values. Several studies found that US government payments had little effect on
annual changes in farmland prices in the United States (e.g., Just and Miranowski 1993;
Gardner 2003), attributing the findings of limited impacts on price fluctuations to the
stabilizing effects of the payments. Studies using cointegration techniques suggest the
relative responsiveness of land values to changes in government payments in Canada
may depend on the proportion of government payments to total income (Weersink
et al. 1999).
Estimating the impacts of government programs with precision in a dynamic model-
ing framework is challenging because these programs have been subject to change dur-
ing the course of Farm Bill reauthorizations that occur approximately every five years,
and the complexity of farm policies has increased over time. For many years, payments
were tied to production or market conditions, so payment amounts could vary sub-
stantially across Farm Bill periods. Changes in the programs also mean that estimated
effects of past farm programs may not be representative of effects of current farm pro-
grams. In particular, through 1950, commodity programs provided relatively little sup-
port, but during the next 15 years or so new programs were introduced that provided
more support (Gardner 2003). Farm legislation in the 1980s and 1990s shifted away
from market-distorting policies, with the addition of income-supporting (as opposed to
price-supporting) commodity loan programs in 1985 and the introduction of planting
flexibility on acres qualifying for commodity program payments in 1990. The Federal
Agricultural Improvement and Reform Act of 1996 (i.e., the 1996 Farm Bill) eliminated
1 For example, see papers included in Moss and Schmitz, eds. (2003).
all cropping restrictions; commodity payments previously tied to current planting deci-
sions were decoupled from current production decisions and replaced with payments
based on historical production choices (Nelson and Schertz 1996).
A few studies accommodated these program complexities by using different empiri-
cal techniques to model explicitly whether the land value effects of US commodity
payment programs have varied across Farm Bill periods. Gardner (2003) used pooled
county-level data between 1950 and 1992 and found only weak evidence that the rate of
growth in farmland values in counties with substantial amounts of program crops was
higher than it would have been in the absence of commodity programs (i.e., compared
to “non-program crop” counties). Gardner (2003) posits that the evidence was not
stronger because farmland may benefit more uniformly from the existence of commod-
ity programs (i.e., if farms are not enrolled, the value attached to the option to enroll
would be capitalized into the value of the land). Also, although payment impacts may be
evident in the short run, the effect could be dampened in the long run if a larger share of
program benefits goes to commodity buyers.
Using a recursive model to account for identification issues arising from the
counter-cyclical nature of some farm program payments, Shaik et al. (2005) find that
farm program payments may have increased farmland values by as much as 30–40%
during 1940–1980, but that the effect declined to 15–20% during 1980–2002. Mishra
et al. (2011) used an information measure and found that impacts on land value
changed after passage of the 1996 Farm Bill, noting less divergence between the dis-
tributions of farmland values and government payments in the post-1996 Bill period.
Nonetheless, a challenge continues to be that modeling the impacts of government
payments with aggregate data is problematic. That, coupled with the recognition that
government payments are likely to also affect input and output markets, helps explain
a shift in modeling the incidence of policies away from the effects on prices (Sumner
et al. 2010).
Collectively, studies employing dynamic modeling techniques demonstrate that these
approaches offer several benefits in the context of modeling farmland values. Among
the most important are that these models inform on the relative importance of macro-
economic factors, such as interest rates and inflation, whose identification requires tem-
poral variation. The contributions they provide to informing farmland value forecasting
models are also important (Erickson et al. 2003). Criticisms include a lack of a behav-
ioral basis, as well as the potential for aggregation bias; a continuing challenge is obtain-
ing consistent results. Although recent advances in nonstationary panel techniques may
help improve consistency or the identification of some impacts (e.g., Gutierrez et al.
2007), and extensions that incorporate demands for land in alternative uses could be
useful (Moss and Katchova 2005; Shaik et al. 2005), they may not fully address the criti-
cisms noted above.
3. Developments Using Cross-Sectional

Models and Spatially Explicit Data
In more recent decades, the increasing availability of cross-sectional and spatially disag-
gregated data provided new opportunities to model the determinants of farmland val-
ues with data at a scale that more closely matched economic behavioral decisions (Irwin
et al. 2010). A strain of farmland values literature evolved that exploited these increas-
ingly disaggregate data and adapted property value modeling approaches that were com-
mon in the urban economics literature. In particular, application of these techniques to
farmland markets in urbanizing areas became widespread. This occurred in part due to
the recognition that, in many regions, farmland can earn returns not just from agricul-
tural production and government payments, but also from “nonfarm” sources. Principal
among the nonfarm sources of returns first considered was the expected future rent
increases arising from returns from future development for residential or commercial
uses for farmland in close proximity to urban areas. Capozza and Helsley’s (1989) semi-
nal work laid the theoretical foundation for this literature and showed how the value of
expected future rent increases could be quite large, especially in rapidly growing cities.
That is, in such areas, farmland values are represented by (setting aside uncertainty):
∫ A (x , t )e
− rt
Pt = i i dt + Ri ( xi , u)e − ru (2)
t =0
where Pt is the price of farmland in period t, Ai is annual net returns from farming, Ri is
the one-time net returns from converting the land to an urban use at the optimal con-
version time u, xi is a vector of exogenous parcel characteristics, and r is the discount
rate. In this specification, farming returns are no longer earned once time u arrives. The
returns to conversion are represented as a one-time payment to reflect the typical lump
sum payment that landowners receive when land is converted to an urban use. This
model could also be expanded to include other sources of nonfarm income—income
from hunting leases, for example—that generate a stream of payments that are earned in
addition to farming returns.
Hedonic models quickly became the most widely used property value model in the
study of the determinants of farmland values. Because of its extensive use, we provide
an overview of the basic model and issues that require attention when estimating the
model. We note that hedonic models are not the only models used to explain non-
farm influences. For example, Hardie et al. (2001) adapt an urban growth model and
used a simultaneous equations approach with county-level data to explain residen-
tial and farm real estate prices. Others used ordinary least squares (OLS) regressions
with farm-level survey data to study the impacts of both various forms of government
payments (disaggregated by program type) and potential returns from future develop-
ment (Goodwin et al. 2003a, 2003b).
3.1 Hedonic Models: Conceptual Approach

Hedonic models are a revealed preference technique based on the notion that the price
of a good observed in the marketplace is a function of its attributes or characteristics.
A seminal article by Rosen (1974) developed the model for differentiated consumer
products (as noted by Palmquist [1989]; Freeman [1974] also developed a similar
model). These models provide the theoretical underpinnings for empirical models that
estimate marginal prices for a product’s characteristics. The theory of hedonic property
value models is thoroughly described in Freeman (1993) and in Palmquist (2006); how-
ever, those models were confined to residential properties. Under the assumption of
perfect competition, the hedonic price function represents an equilibrium price sched-
ule that is comprised of the market-clearing bid-and-offer curves of heterogeneous
agents (Rosen 1974). This equilibrium price of a property is a function of property attri-
butes and location characteristics, and each characteristic is valued by its implicit price.
Although studies have shown that these implicit prices could be used to identify mar-
ginal willingness-to-pay (MWTP) functions in the second stage estimation of hedonic
models (e.g., Freeman 1993), most current studies only focus on the first stage estima-
tion of implicit prices due to potential endogeneity concerns (Bartik 1987; Epple 1987;
Bishop and Timmins 2011).
The equilibrium conditions of the hedonic model have been criticized because they
require instantaneous adjustment in demand or supply. In particular, when market
forces are moving continuously in one direction (or are expected to move in one direc-
tion), the imperfect adjustments of the market to changing conditions of supply and
demand might introduce bias in the estimates of MWTP using observed implicit prices
from hedonic regressions (Freeman 1993). As a result, researchers should be especially
cautious in applying hedonic models when markets are changing rapidly. However, in
most circumstances, divergence from hedonic equilibrium will only introduce random
errors, and, even in cases of rapidly changing markets, hedonic estimates could still
serve as the upper (or lower) bound of the MWTP estimates and provide useful infor-
mation to infer the direction of biases.
In a seminal paper, Palmquist (1989) adapted the model for differentiated factors of
production and applied it in the context of farmland rental markets. That paper assumes
farmland owners and buyers are profit-maximizing farmers who own and buy land
strictly for its productive capacity. Palmquist and Danielson (1989) discussed mod-
ifications needed in models using farmland sales as opposed to rent data but did not
explicitly model them. Specifically, they note that the interpretation of the coefficients
can differ depending on whether rents or sales prices are used in the hedonic model.
Differences can arise when the marginal value of a characteristic differs in a short
amount of time (within the length of the rental lease) relative to a longer period that
would be capitalized into the value of the land. For example, being adjacent to a national
park might reduce the rental price of farmland due to potential wildlife damage of crops
but could increase the sales price if close proximity is expected to provide positive ben-
efits in the more distant future.
The Palmquist and Danielson framework also does not account for the fact that, for
many farm parcels, the land provides benefits beyond the net returns earned from farm-
ing, such as the value associated with the option to convert the land to residential use
at some point in the future as modeled in (2) above, and benefits from close proxim-
ity to open space or other natural amenities that do not contribute specifically to the
land’s productive capacity. Indeed, US Department of Agriculture data reveal that most
farmland owners in 1999 (the most recent data available on farmland ownership) did
not operate farms as their primary business (US Department of Agriculture, National
Agricultural Statistics Service [USDA-NASS] 2001). Some farmland owners farm on a
part-time basis, but about 25% of farmland in 2007 was farmed by operators who were
retired or operated a farm primarily for residential or lifestyle reasons (Hoppe and
Banker 2010). The point that farmland has value both as a factor of production and as a
consumption good has been recognized by some (e.g., Henneberry and Barrows 1990;
Ma and Swinton 2012), although it appears that most researchers who estimate hedonic
models in all but the most rural areas cite Rosen’s theory related to consumer goods.
Many of the early applications of hedonic models to farmland markets used the
approach to estimate the marginal value of both farm and nonfarm characteristics of
farmland in urbanizing areas. One of the earliest and most well-cited papers is Chicoine
(1981), who used sales data on unimproved farmland parcels in Will County, Illinois
and found that the influence of factors affecting potential development returns R were
far greater than soil productivity, the sole characteristic included in A as a proxy for farm
returns. Numerous subsequent studies have also modeled the impact of urban prox-
imity on farmland values; in areas that are more urbanized or have rapid population
growth, these studies find that the demand for land for urban uses is the most signifi-
cant nonfarm factor affecting farmland values (e.g., Shi et al. 1997; Plantinga et al. 2002;
Huang et al. 2006; Guiling et al. 2009).
Hedonic models have also been used to examine the role of environmental factors
and recreational opportunities on farmland prices. In response to concerns about farm-
land erosion resulting from the 1970s agricultural export boom and increases in non-
point water pollution, a number of studies during the 1980s examined the effect of soil
erodibility, as well as drainage, on farmland values (e.g., Miranowski and Hammes 1984;
Ervin and Mill 1985; Gardner and Barrows 1985; Palmquist and Danielson 1989). Ervin
and Mill (1985) also noted that such studies are useful for identifying the extent to which
private markets capture the value of changes in a land characteristic that have impli-
cations for both on-site productivity and off-site environmental quality. Other studies
examined the impact of wildlife recreation opportunities (e.g., Henderson and Moore
2006) and other amenities (see Bergstrom and Ready 2009 for a review), as well as the
impact of restrictions on land uses, such as zoning (e.g., Chicoine 1981; Henneberry and
Barrows 1990), agricultural district and greenbelt designation (Vitaliano and Hill 1994;
Deaton and Vyn 2010), and farmland protection easements (e.g., Nickerson and Lynch
2001; Lynch et al. 2007). Several recent studies have considered the impact of bioenergy
policies by analyzing the impact of proximity to ethanol plants on farmland values (e.g.,
Henderson and Gloy 2009; Blomendahl et al. 2011; Zhang et al. 2012).
3.2 Empirical Issues in Hedonic Modeling of Farmland Prices

A number of well-known econometric problems may arise when estimating hedonic
models. One issue that has particular significance in the context of farmland markets
relates to the geographic extent of the market. A key assumption of the equilibrium
hedonic price schedule is that sales transactions are drawn from a single market. This
assumption is particularly restrictive in studies using farmland price data, since the his-
torical thinness of the market limits the number of transactions within narrowly defined
geographic areas. Indeed, recent surveys reveal that less than 2% of farmland is sold
annually (Sherrick and Barry 2003; Duffy 2011). Previous studies have utilized transac-
tions data at various levels, from a single county (e.g., Chicoine 1981; Henneberry and
Barrows 1990), to a single state (e.g., Guiling et al. 2009), and to entire regions (e.g.,
Barnard et al. 1997; Roka and Palmquist 1997). However, the appropriate size will likely
vary depending on the topic of the study. Studies on the value of farmland in urbanizing
areas could arguably have markets covering a much smaller geographic area compared
to studies on farmland values in rural areas, for example.
The historical thinness of farmland markets also raises two other important issues
unique to farmland values studies. The first is about the construction of the dependent
variable, given the fact that sales prices reflect the value of both land and structures in
the presence of farm structures, residential dwellings, or both. Previous researchers have
included a dummy variable indicating the presence of structures (e.g., Palmquist and
Danielson 1989), subtracted the value of improvements from the total sales price (e.g.,
Guiling et al. 2009; Zhang et al. 2012), or simply excluded the parcels with structures
(e.g., Chicoine 1981). Although information on the attributes or even presence of farm
buildings is rarely available, including the value of structures in the dependent variable
is not inconsistent with theory (Freeman 1993). The other issue relates to the choice of
the data source. Whereas use of survey data (e.g., Roka and Palmquist 1997; Henderson
and Gloy 2009) can yield more observations than microlevel sales transaction data, it
raises a question about how well survey respondents’ assessments of farmland values
represent true market prices.2
A particularly important empirical issue that requires consideration in farmland
value hedonic studies is omitted variable bias, in which the correlation of observed
2 Ma and Swinton (2012) found tax assessor estimates of farmland values were particularly likely to
underestimate the value of surrounding natural amenities.
variables and unobserved attributes lead to biased estimates of the implicit prices of
characteristics of a property, a land parcel, or a product (Palmquist 2006). Bias resulting
from spatial dependence and sample selection due to observables and unobservables
are two distinct types of omitted variable bias that researchers have begun address in
recent farmland value studies. Agricultural land parcels are essentially spatially ordered
data, and achieving unbiased and efficient estimates requires addressing the inherent
spatial dependence (Anselin 1988). This dependence has long been recognized in the
areas of regional science and geography and was nicely summarized in Tobler’s (1970,
236) First Law of Geography—“everything is related to everything else, but near things
are more related than distant things.” In the presence of spatial dependence, the stan-
dard OLS assumptions of uncorrelated error terms and independent observations are
violated, and thus the parameter estimates from the standard hedonic regressions will
be biased and inefficient. A sample selection problem occurs when a nonrandomly
selected sample used to estimate behavioral relationships is not representative of the
desired population (Heckman 1979), which could arise from selection on the unobserv-
ables (Heckman 1979) or on the observed characteristics (Heckman and Robb 1985).
If left uncontrolled, the sample selection problem may result in biased parameter esti-
mates of the hedonic models.
Two other well-known problems that may affect any hedonic study are the func-
tional form of the empirical model and multicollinearity. Although the choice of
functional form can affect both the magnitude and significance of coefficients, as
noted by many studies, economic theory offers little guidance regarding model spec-
ification and restrictions on functional form. In practice, data availability and the
goodness of fit often dictate the choice among different functional forms; farmland
value studies have used a variety of forms, including transcendental, linear, semi-log,
and double-log; some researchers prefer the flexibility afforded by the Box-Cox
functional form, which lets the data determine the appropriate form (Palmquist and
Danielson 1989; Roka and Palmquist 1997; Nivens et al. 2002). Another key speci-
fication issue in hedonic models is the multicollinearity that often arises from the
attempt to control for all relevant characteristics of the land. This problem arises at
least in part from difficulties in obtaining enough data for ideal model specifications,
which is challenging given the thinness of farmland markets. As noted by Freeman
(1993), including collinear variables increases the variance of coefficient estimates
and affects inference.
Substantial research effort has been devoted to alleviating all of these econometric
problems imbedded in hedonic models. In the context of research on farmland val-
ues, recent econometric developments have largely been focused on addressing biases
arising from spatial dependence and addressing sample selection bias due to observ-
ables and unobservables. Our discussion of these techniques in the following sections
describes these developments. We also draw on the wider hedonics literature, in which
several developments are sufficiently recent that they have not been often embraced in
models of farmland values.
3.3 Recent Developments in Addressing Spatial

Autocorrelation and Spatial Heterogeneity
To account for spatial dependence in hedonic models of farmland values, two paramet-
ric spatial econometric models are primarily applied: spatial lag (spatial autoregres-
sive) models and spatial error (spatial autocorrelation) models. Spatial lag dependence
means the dependent variable in one location is affected by independent variables in
that location and other locations. The standard spatial lag model solves this problem by
adding a weighted average of nearby values of the dependent variable as an additional
set of explanatory variables, which instead of the traditional model y = Xβ + u yields
y = ρWy + Xβ + u = (I − ρW )−1 ( Xβ + u)
(3)
= (I + ρW + ρ2W 2 + …)( Xβ + u)

where W is an n × n spatial weight matrix, and the scalar ρ is the spatial coefficient.
As can be seen in the last equation of (3), the Leontief inverse reduced form, spatial
lag of the dependent variable implies a spatial diffusion process or a so-called “spatial
multiplier” effect, in which each observation is potentially influenced by all other obser-
vations (Anselin 2001), and such influence decays with the increase in distance between
observations.
Spatial error dependence or spatial autocorrelation, in which the correlation of error
terms is across different spatial units, is typically caused by measurement error or omit-
ted spatial variables, or by a modifiable areal unit problem (i.e., results differ when the
data are aggregated in different ways) (Griffith 2009). In contrast with the spatial lag
model, in which the spatial interaction is the process of interest, the spatial error model
offers a more common and direct treatment of the spatial dependence among error
terms of the observations, in which the spatial dependence is a nuisance:
y = X β + u, with u = θWu + e
(4)
y = Xβ + ( I − θW )−1 e

where W is an n × n spatial weight matrix, and the scalar is the spatial coefficient.
Opportunities to account explicitly for spatial dependence among observations
in farmland values studies have grown in recent years, due to increased availability of
spatially explicit data on farmland, the explosive diffusion of Geographic Information
System software, and the dramatic increase in the ability of statistical packages to handle
large spatial matrices. Using county-level data in the Corn Belt, Benirschka and Binkley
(1994) offer one of the first treatments of spatial autocorrelation in studies of the rela-
tionship between agricultural land price variations and distances to markets, in which
the spatial correlation of error terms across counties was represented by a standard
spatial error model specification, with W being a simple binary continuity matrix. In
a spatial lag, serially correlated hedonic pricing framework, Huang et al. (2006) further
controlled for serial correlation using a first-order autoregressive process along with the
assumed time-invariant spatial lag dependence using a Kronecker product of the spa-
tial matrix W and a T × T identity matrix. A similar spatiotemporal weight matrix is
also used by Maddison (2009). In a study of effects of natural amenities on Michigan
farmland values, Ma and Swinton (2012) used a spatial error specification to account for
spatial dependence, in which the spatial weights matrix was defined using the inverse
distance formula with a cutoff distance of 600 meters from the parcel centroids beyond
which no correlation is assumed. The spatial error model structure was determined
through diagnosis and tests of the structure of spatial correlation.
Due to improved computational speed and functional simplicity, spatial lag and
spatial error models have become routine fixes for nearly any model misspecification
related to space (McMillen 2012). However, these standard spatial econometric models
are far from problem-free. In particular, most spatial econometric models face an ironic
paradox that their very use is an admission that the true model structure is unknown,
yet the common estimation technique of maximum likelihood relies heavily on know-
ing the true structure in advance (McMillen 2010). Other criticisms include identifi-
cation problems and usually exogenously imposed spatial weights matrix, which can
result in biased parameter estimates if misspecified.3
As emphasized by McMillen (2010, 2012), standard spatial econometric models are
simply another form of spatial smoothing, and they should be viewed as additional
statistical tools for model specification tests and convenient robustness checks, rather
than as the primary means of analyzing spatial data. In general, applications of spa-
tial models should be guided by economic theory (e.g., Brueckner 2006) and by actual
empirical questions (Pinkse and Slade 2010). Instead of focusing solely on spatial lag
and spatial error models, researchers have advocated alternatives, such as semiparamet-
ric and nonparametric approaches (McMillen 2010), and “experimentalist paradigm”
approaches, such as instrumental variables (IV) and spatial differencing (Gibbons and
Overman 2012).
These alternative approaches have gained popularity in residential real estate valua-
tion studies, for which spatially explicit data has traditionally been more readily avail-
able than farmland data. Two recent studies using these approaches are worth noting.
The first is a nonparametric analysis of capitalization of proximity to rapid transit lines
in residential house prices in Chicago, in which McMillen and Redfearn (2010) illus-
trate that, unlike standard parametric spatial models, nonparametric estimators control
for spatial variations in marginal effects and spatial autocorrelation while using highly
flexible functional forms, without imposing an arbitrary weight matrix. The second
is a study that identifies the influence of spatial land use spillovers on housing values.
3 See Pinkse and Slade (2010), McMillen (2010, 2012), Gibbons and Overman (2012), and Brady and
Irwin (2012) for further discussions of the criticisms of standard spatial econometrics models.
Carrión-Flores and Irwin (2010) exploited a natural discontinuity in the data and show
that a partial population identification strategy solves the endogeneity problem and is
a superior alternative to the common spatial error model for eliminating spatial error
autocorrelation and identifying spatial interactions.
Some progress in addressing spatial autocorrelation and spatial heterogeneity has
also been made in studies of farmland values beyond the spatial lag and error models.
Cotteleer et al. (2011) tried to resolve specification uncertainty in selecting explanatory
variables and weighting matrices in parametric spatial econometric models by employ-
ing Bayesian Model Averaging in combination with Markov chain, Monte Carlo model
composition. In this framework, no single correct model specification is assumed and
learning from the data is allowed, but prior information is needed. Using parcel-level
data in Northern Ireland, Kostov (2009) generalized the linear spatial lag model by
employing a flexible, semiparametric IV quantile regression approach, which not only
allowed for varying effects of the hedonic attributes, but also varying degrees of spatial
dependence. In two similar Northern Ireland studies, Kostov et al. (2008) and Kostov
(2010) employed two different nonparametric approaches and found that buyer char-
acteristics and personal relationships exert nonuniform and nonlinear effects on the
implicit prices of farmland characteristics. Using intramunicipal-level French data,
Geniaux et al. (2011) extended Capozza and Helsley’s (1989) model to account for
uncertainty in future land use zoning and used mixed geographically weighted regres-
sion estimations of a spatial hedonic model to recover intramunicipally heterogeneous
impacts of land use conversion anticipation on farmland prices.
3.4 Recent Developments in Addressing Sample

Selection Bias
Sample selection problems may arise from a variety of selection mechanisms, including
self-selection by the data units (Heckman 1979) and the so-called incidental truncation
problem, in which data on a key variable are available only for a clearly defined subset
(Wooldridge 2002); for example, farmland rental rates can only observed for those land
that are actually rented. In such cases, unobserved factors determining inclusion in the
subsample are correlated with unobservables influencing the variable of primary inter-
est, leading to biased parameter estimates of the hedonic models.
Heckman’s 1979 seminal paper offers the first and the most widely applied correction
model of sample selection (or selectivity) bias. The sample selection problem is charac-
terized by two latent variable equations, the selection or participation equation and the
outcome equation, which are allowed to have correlated errors. Correction of the sample
selection bias is commonly achieved through a limited-information two-step estima-
tion procedure (Greene 2012), in which the inverse Mills ratios are formulated from the
estimated parameters of the first-stage probit selection equation to control for selectivity
bias. This Heckman-style selection model has become a standard solution to the sample
selection problem in various fields of economics, especially in the literature of program
evaluation. In the context of research on land values, especially farmland values, this
model is also widely applied. In a study of residential land value functions in which land
use is determined by zoning, McMillen and McDonald (1989) find evidence of selectivity
bias for undeveloped and multifamily residential land uses in which the “self-selectivity”
arises when local governments use land values to guide zoning decisions. However, in the
context of farmland markets, sample selection was not detected in two recent studies that
addressed it using a Heckman selection model (Nickerson and Lynch 2001; Kirwan 2009).
The Heckman selection models address selection on the unobservables; however, in
a broader sense, sample selection could also occur when the unobserved disturbance in
the outcome function is correlated with the observed explanatory variables in the selec-
tion model, which is introduced as “selection on the observables” by Heckman and Robb
(1985). As a result, when estimating the average treatment effect, the assumptions about
the distributional equality of the covariates across the treatment and control subsam-
ples imposed by hedonic regressions could be problematic, and the differences between
covariates among treatment and control units may need to be adjusted for (Imbens and
Wooldridge 2009). Matching offers a straightforward and effective way to balance these
differences, which facilitates the identification of the causal treatment effect. Intuitively,
matching solves the sample selection on the observables by selecting treated observa-
tions and comparison observations with similar characteristics, by covariates X (e.g.,
Rubin 1980), or by propensity score p (e.g., Rosenbaum and Rubin 1983).
In this section, we focus on propensity score matching (PSM) methods, which use
propensity scores (the probability of selection into treatment conditional on covariates)
in matching, because these methods are most commonly used and have been shown to
be reliable under certain regularity conditions (Todd 2007). PSM presents several key
advantages over the least squares hedonic approach. Most importantly, PSM does not
require a parametric model linking outcomes and program participation (Dehajia and
Wahba 2002; Smith and Todd 2005; Ravallion 2007). In addition, unlike standard regres-
sion methods, PSM ensures that observations in treatment and control groups share
the common support (Ravallion 2007), and, finally, unlike Heckman selection model,
PSM does not assume a particular functional form for the price equation (Heckman and
Navarro 2004). Matching estimators such as PSM are justified if the selection is only on
the observables (Imbens and Wooldridge 2009), and the performance of PSM depends
crucially on the set of covariates included in the estimation (Heckman et al. 1998; Todd
2007). However, instead of elaborating on the methodological and implementation
details on PSM, we aim to highlight specific applications of PSM in farmland values.
The reader is referred to Caliendo and Kopeinig (2005), Smith and Todd (2005), Todd
(2007), Zhao (2004), and Towe, Lewis, and Lynch in this handbook (Chapter 18) for
detailed discussions on the matching methods.
PSM has become a popular approach to estimate causal treatment effects and has been
used in some recent studies of farmland values. In an analysis of the selection problem
due to the voluntary nature of farmland easement programs analyzed also in Nickerson
and Lynch (2001), Lynch et al. (2007) used a PSM approach in which observed vari-
ables closely related to the future development option values, and variables affecting
eligibility or probability of program participation are included as conditioning vari-

ables. Specifically, in contrast with results from hedonic models but consistent with
findings by Nickerson and Lynch (2001), they find little evidence that preserved par-
cels sell for a significantly lower price than nearby unrestricted land. Using a sample of
UK cereal farms, Sauer et al. (2012) incorporated the PSM approach with a production
theory based multi-output, multi-input directional distance function framework and
find that different agri-environmental schemes significantly affect production behavior
at farm level.
However, systematic differences in unobservables may still bias these PSM estima-
tors. Various extensions have been proposed as a response, including combining PSM
and linear regression (Imbens and Wooldridge 2009), allowing for selection on unob-
servables by imposing a factor structure on the errors and estimating the distribution
of unobserved errors (e.g., Carneiro et al. 2003), and controlling for time-invariant
unobserved heterogeneity using a difference-in-difference (DID) matching estimator as
defined in Heckman et al. (1997). Here, we focus on the DID PSM estimator, which has
attracted more interest in the farmland value literature. When estimating average treat-
ment effect, the conditional DID PSM estimator compares the conditional before-after
outcomes of treated units with those of nontreated units. This DID PSM estimator is
attractive because it permits selection to be based on potential program outcomes and
allows for selection on unobservables (Heckman et al. 1997). A study by Ciaian et al.
(2011) is worth mentioning because, rather than using the conventional binary PSM
estimator to identify the effects of European Union government programs, it employed
a generalized propensity score (GPS) method proposed by Hirano and Imbens (2005),
which allows for estimation of the capitalization rates into farmland values for differ-
ent levels of government payments as multivalued, continuous treatments. Two very
recent farmland value studies have used a DID PSM estimator to identify the impact of
an expanding ethanol market. Using a panel dataset of US farmland parcels from 2001
to 2007, Towe and Tra (2013) investigated the differential impacts of the construction of
new ethanol facilities before and after the Renewable Fuel Standard legislation passed.
Their results suggest that the RFS created expectations of higher returns to agricul-
ture, beyond those derived from higher commodity prices. Zhang et al. (2012) instead
combined the regular binary PSM estimator with DID regressions and applied them
on parcel-level agricultural land sales data in Ohio 2001–2010 and find evidence of a
structural change in the marginal value of the proximity to ethanol plants induced by
the 2007 residential housing market bust and concurrent expansion of ethanol facilities.
3.5 Addressing Omitted Variable Bias Using Instrumental

Variables Approach
To address the omitted variable bias and endogeneity concerns other than sample selec-
tion bias, some recent studies have employed the standard instrumental variables (IV)
approach to identify the impact of governmental subsidies. Although land studies in
this area are all on farmland rental rates, the techniques are very amenable to examining
the impact of government payments in the context of farmland value studies. Using US
farm-level data, Kirwan (2009) designed an IV strategy to overcome the attenuation bias
induced by the expectation error, which is the difference between actual agricultural
subsidies and expected subsidies. Specifically, he instrumented the 1992–1997 subsidy
change using the post-FAIR Act 1997 subsidy level and addressed the measurement
error problem with a second instrument, the county-level average subsidy per acre.
Following Lence and Mishra (2003) and using data in Northern Ireland, Patton et al.
(2008) adopted an IV strategy combined with GMM technique to recognize the fact that
payments are not known when rental contracts are determined and therefore instru-
ments using lagged realizations of the “pre-2002 SAP” payments are needed in the pres-
ence of expectation error. Using a rich dataset of pooled cross-sections at the farm level,
Goodwin et al. (2010) instrumented the expected payment benefits using a four-year
historical average of real payments per farm acre in the county where the farm is located.
They argued that this measure better represents the long-run potential benefits associ-
ated with agricultural policy, whereas the common measure, realized payments, may, in
contrast, reflect individual policy choices and characteristics of the farms.
4. Conclusion and Future Research

Directions
The continued significance of farmland values to both the farm sector and to many farm
households means that understanding the key determinants of farmland prices will
remain of perennial interest. In this chapter, we have sought to identify major model-
ing approaches used to model farmland values and to describe recent innovations. As
this chapter highlights, both dynamic time series and static cross-sectional approaches
have been utilized by a large number of studies, with each contributing unique insights.
In this section, we identify several areas in which future research may yield the highest
return both in terms of advances in modeling and in terms of topics of interest to policy
makers.
Dynamic models reveal important information about macroeconomic factors affect-
ing rates of change in farmland values. However, criticisms of ad hoc econometric
specifications that could contribute to misleading results have plagued many of these
studies. A natural direction for these studies would be to utilize some of the more recent
advances in time series techniques in ways that are supported by an underlying struc-
tural model that is both consistent with individual behavior and that captures critical
market relationships (along the lines of Just and Miranowski 1993). In particular, a better
(or at least more current) understanding is needed of how expectations by landowners
are formed over prices, costs, and other key variables. Also, how changes in determi-
nants are transmitted through expectations as suggested by Just and Miranowski (1993)
could be useful, especially if the studies can illuminate how quickly farmland values
react to changes in its determinants.
Furthering our understanding of the dynamics of farmland markets in these ways
seem useful for at least three reasons. First, the rapid onset of and large (double-digit)
annual increases in farmland values that we have witnessed in recent years is occurring
under different conditions than increases that occurred in the 1970s, so the primary
drivers of change are different. In particular, studies that consider the formation and
role of price expectations, market relationships, and incidence may help inform deci-
sion makers about how quickly high farmland values could erode (or could be further
enhanced) due to policy changes under their control (e.g., government farm program
payments, bioenergy policies that increase demand for biofuel crops like corn and soy-
bean, and macroeconomic policies such as interest rates). Second, nonfarm influences
on farmland are growing, and models that incorporate these influences can help inform
on how changes in related land markets are influencing farmland values.4 Finally,
advances in these areas could help inform efforts to link farmland value models and
models of land use and land use change. We return to this last point below.
In terms of future directions in cross-sectional hedonic studies, we note several
compelling opportunities to better address omitted variable bias—which is arguably
among the most important econometric issues requiring treatment in farmland value
studies using disaggregated, parcel-level data. Exploiting the ever-widening range of
new spatially explicit modeling approaches allows researchers to reveal the rich spa-
tial heterogeneity of the influences of determinants of farmland values with fewer
restrictive assumptions. These approaches include the nonparametric approaches,
quasi-experimental (QE) designs, and structural econometric models, many of which
we mentioned in Section 4 in this chapter. In the following sections, we highlight some
examples relevant for farmland values research.
Minimizing the bias and inefficiency caused by untreated spatial dependence in
cross-sectional studies has spurred the adoption of a variety of techniques in land values
studies. Although largely applied in land markets in or near urbanizing areas, the inher-
ent spatially correlated processes underlying many farmland value determinants means
the results of farmland valuation studies that do not consider spatial dependence are
likely to be suspect. Standard spatial lag and spatial error models have yielded insights
regarding the magnitude of the bias that can result if spatial dependence is left untreated.
However, future research using spatially ordered farmland transactions data would
likely benefit by embracing newer techniques that avoid the restrictive assumptions of
these models. In particular, these newer techniques enable researchers to control for
spatial dependence without imposing a certain spatial structure a priori. Approaches
such as those relying on quasi-randomness, such as the “partial population identifier”
4 We also note that the increasing influence of urban demands on farmland raises questions about
whether time series properties differ between farmland subject to urban influence and farmland that
is not.
used in Carrión-Flores and Irwin (2010), and semiparametric and nonparametric

approaches employed by McMillen and Redfearn (2010) seem particularly fruitful in
this regard. However, the standard spatial econometric models still serve as a useful
toolbox for model specification tests and robustness checks, and a spatial lag model is
still justified if the objective is to identify the effects of neighboring values on the depen-
dent variable and the empirical model rests on economic theory (McMillen 2010).
In contrast with the standard hedonic models, QE designs popular in labor and
regional economics, such as matching approaches and regression discontinuity design
(RDD), present some interesting alternatives. By controlling for observable covari-
ate differences and time-invariant unobserved heterogeneity, the DID PSM estima-
tors illustrated in Section 4.4 can yield more plausible results than traditional hedonic
estimators, if correctly implemented. Researchers may also benefit by using matching
estimators other than PSM. A good candidate is covariate matching, including the com-
mon Mahalanobis metric (e.g., Rubin 1980), or the recently developed genetic matching
method (Diamond and Sekhon 2013).
Although, to our knowledge, RDD has not yet been applied in the studies of farmland
values, it has been enthusiastically embraced in the literatures of political science, epide-
miology, and other fields of economics, such as real estate studies.5 Future farmland val-
ues studies could benefit by explicitly considering RDD, especially when estimating the
impact of state or local governmental programs and the effects of strict agricultural zon-
ing policies. However, caution must be exercised regarding the potential spatial spill-
over problems when geographic borders are used in RDD, in which case a robustness
check using matching estimators may be helpful.
The importance of addressing sample selection is a well-known empirical issue in
the farmland values literature. Given that a wide array of government policies and pro-
grams support the agricultural sector and the increasing reliance on mechanisms with
voluntary participation, advances in addressing this issue could be particularly fruit-
ful. However, current applications of Heckman selection models in farmland values
research are limited to the original Heckman (1979) model, which has a rather limited
structure and is highly parameterized (Vella 1998). Besides the aforementioned QE
approaches, future research may adopt a broader view and consider more generalized
selection models with less restrictive modeling assumptions, such as those used by Lee
(1982, 1983), Heckman and Robb (1985), and Puhani (2000). Other methods, such
as control functions, could also prove to be beneficial in certain circumstances (e.g.,
Heckman and Navarro 2004; Imbens and Wooldridge 2007; Navarro 2008).
As we mentioned earlier in this section, more work can be done in the farmland values
literature to inform on efforts to uncover the structural parameters of the demand and
supply of farmland, which helps link changes in farmland values with land use change
5 The reader is referred to van der Klaauw (2008), Imbens and Lemieux (2008), and Lee and Lemieux
(2010) for excellent reviews of RDD, and to Black (1999), Chay and Greenstone (2005), Greenstone and
Gallagher (2008), and Grout et al. (2011) for applications of RDD in urban housing market studies.
models described (see, for instance, Chapter 13 by Irwin and Wrenn in this handbook).
Modeling dynamic aspects that take into account the formulation of expectations by
farmland owners over prices, costs, and other key variables is crucial to estimating the
supply of farmland and necessitates a dynamic modeling approach for the structural
estimation of farmland supply. Current reduced-form models, such as hedonics and QE
designs, are static, and they do not take these dynamics into account. However, as illus-
trated in Chapter 13 by Irwin and Wrenn in this handbook, the complexity of dynamic
discrete choice models makes it sometimes infeasible empirically. Nevertheless, incor-
porating feedback or forward-looking expectations in structural hedonic models of
farmland markets remains a crucial unsolved issue. In the hedonics literature, some
notable advances have been made to identify the marginal willingness-to-pay functions,
including the IV approach by Ekeland et al. (2004), the new econometric inversion
estimation by Bishop and Timmins (2011), and the dynamic hedonic model by Bishop
and Murphy (2011), which allows for forward-looking behavior of decision-makers.
However, as mentioned in Section 4, researchers need to be cautious about using
hedonic approaches when market forces are changing rapidly (Freeman 1993).
The ability of researchers to move forward on many of these fronts will be contin-
gent on the increasing availability of spatially disaggregated data. Previous studies on
agricultural land values that have employed aggregate data often mask important differ-
ences in the spatially disaggregated determinants of farmland values, such as distance
from urban centers and proximity to agricultural delivery points like ethanol plants,
grain elevators, and agricultural terminals. Aggregate data also hinder the application
of new modeling approaches from related fields such as residential land/housing values
research to studies on farmland values. A data challenge will continue to be the cost of
developing parcel-level panel datasets via surveys and the thinness of farmland markets
of developing pooled parcel-level sales data over time. Nonetheless, with more spatially
explicit data available and techniques like nonparametric approaches and panel data
analysis, researchers will have improved opportunities to analyze spatial variation as
well as potential structural changes in certain determinants of farmland values.
Acknowledgments
The authors wish to thank Elena G. Irwin for insightful comments and a critical review
of an earlier draft. The views in this chapter are attributable to the authors and not to
the USDA.
References
Alston, J. M. 1986. An analysis of growth of U.S. farmland prices, 1963–82. American Journal of
Agricultural Economics 68(1):1–9.
Anselin, L. 1988. Spatial econometrics: Methods and models. Dordrecht: Kluwer Academic.

Anselin, L. 2001. Spatial Econometrics. In A companion to theoretical econometrics, ed. B.
H. Baltagi, 310–330. Malden, MA: Blackwell.
Awokuse, T. O., and J. M. Duke. 2006. The causal structure of land price determinants. Canadian
Journal of Agricultural Economics 54:227–245.
Barlowe, R. 1986. Land resource economics: The economics of real estate, 4th ed. Upper Saddle
River, NJ: Prentice Hall.
Barnard, C. H., G. Whittaker, D. Westenbarger, and M. Ahearn. 1997. Evidence of capitalization
of direct government payments into US cropland values. American Journal of Agricultural
Economics 79(5): 1642–1650.
Bartik, T. J. 1987. The estimation of demand parameters in hedonic price models. Journal of
Political Economy 95(1): 81–88.
Benirschka, M., and J. Binkley. 1994. Land price volatility in a geographically dispersed market.
Bergstrom, J. C. and R. C. Ready. 2009. What have we learned from over twenty years of farm-
land amenity valuation research in North America? Review of Agricultural Economics
31(1): 21–49.
Bishop, K. C., and A. D. Murphy. 2011. Estimating the willingness to pay to avoid violent
crime: A dynamic approach. American Economic Review 101(3): 625–629.
Bishop, K. C., and C. Timmins. 2011. Hedonic prices and implicit markets: Estimating mar-
ginal willingness to pay for differentiated products without instrumental variables. NBER
Working Paper No. 17611.
Black, S. 1999. Do better schools matter? Parental valuation of elementary education. Quarterly
Journal of Economics 114: 577–599.
Blomendahl, B. H., R. K. Perrin, and B. B. Johnson. 2011. The impact of ethanol plants on sur-
rounding farmland values: A case study. Land Economics 87(2): 223–232.
Brady, M., and E. G. Irwin. 2011. Accounting for spatial effects in economic models of land
use: Recent developments and challenges ahead. Environmental Resource Economics
48: 487–509.
Brueckner, J. K. 2006. Strategic interactions among governments. In A companion to urban
economics, eds. R. J. Arnott and D. P. McMillen, 332–347. Malden, MA: Blackwell.
Burt, O. R. 1986. Econometric modeling of the capitalization formula for farmland prices.
American Journal of Agricultural Economics 68(1): 10–26.
Caliendo, M., and S. Kopeinig. 2005. Some practical guidance for the implementation of pro-
pensity score matching. IZA Discussion Paper No. 1588.
Campbell, J. Y., and R. J. Shiller. 1987. Cointegration and tests of present value models. Journal
of Political Economy 95(4):1062–1088.
Capozza, D. R., and R. W. Helsley. 1989. The fundamentals of land prices and urban growth.
Journal of Urban Economics 26: 295–306.
Carneiro, P., K. T. Hansen, and J. J. Heckman. 2003. Estimating distributions of treatment effects
with an application to the returns to schooling and measurement of the effects of uncertainty
on college choice. International Economic Review 44(2): 361–422
Carrión-Flores, C., and E. G. Irwin. 2010. Identifying spatial interactions in the presence of spa-
tial autocorrelation: An application to land use spillovers. Resource and Energy Economics
32: 135–153.
Chavas, J.-P., and A. Thomas. 1999. A dynamic analysis of land prices. American Journal of
Chay, K., and M. Greenstone. 2005. Does air quality matter: Evidence from the housing market.
Journal of Political Economy 113: 376–424.
Chicoine, D. L. 1981. Farmland values at the urban fringe: An analysis of sale prices. Land
Economics 57(3): 353–362.
Chryst, W. E. 1965. Land values and agricultural income: A paradox. Journal of Farm Economics
47(5): 1265–1273.
Ciaian, P., A. Kancs, and J. Michalek. 2011. SPS capitalization into land values: Generalized
propensity score evidence from the EU. LICOS Discussion Paper No. 293.
Clark, J. S., M. Fulton, and J. T. Scott, Jr. 1993. The inconsistency of land values, land rents, and
capitalization formulas. American Journal of Agricultural Economics 75(1):147–155.
Cotteleer, G., T. Stobbe, and G. C. van Kooten. 2011. Bayesian model averaging in the context
of spatial hedonic pricing: An application to farmland values. Journal of Regional Science
52(2): 172–191.
Deaton, B. J., and R. J. Vyn. 2010. The effect of strict agricultural zoning on agricultural
land values: The case of Ontario’s greenbelt. American Journal of Agricultural Economics
92(4): 941–955.
Dehajia, R. H., and S. Wahba. 2002. Propensity score-matching methods for nonexperimental
causal studies. The Review of Economics and Statistics 84(1): 151–161.
Diamond, A., and J. S. Sekhon. 2013. Genetic matching for estimating causal effects: A general
multivariate matching method for achieving balance in observational studies. The Review of
Economics and Statistics 95(3): 932–945.
Duffy, M. 2011. The current situation on farmland values and ownership. Choices 26(2). http://
www.choicesmagazine.org/magazine/pdf/cmsarticle_24.pdf
Ekeland, I., J. J. Heckman, and L. Nesheim.2004. Identification and estimation of hedonic mod-
els. Journal of Political Economy 112(1): S60–S109.
Epple, D. 1987. Hedonic prices and implicit markets: Estimating demand and supply functions
for differentiated products. Journal of Political Economy 95: 59–80.
Erickson, K., A. K. Mishra, and C. B. Moss. 2003. Cash rents, imputed returns, and the valuation
of farmland revisited. In Government policy and farmland markets: The maintenance of farmer
wealth, eds. C. B. Moss and A. Schmitz, 223–235. Ames: Iowa State University Press.
Ervin, D., and J. Mill. 1985. Agricultural land markets and soil erosion: Policy relevance and
conceptual issues. American Journal of Agricultural Economics 67: 938–942.
Falk, B. 1991. Formally testing the present value model of farmland prices. American Journal of
Falk, B. 1992. Predictable excess returns in real estate markets: A study of Iowa farmland values.
Journal of Housing Economics 2: 84–105.
Falk, B., and B. -S. Lee. 1998. Fads versus fundamentals in farmland prices. American Journal of
Featherstone, A. M., and T. O. Baker. 1987. An examination of farm sector real estate dynam-
ics: 1910–85. American Journal of Agricultural Economics 69(3): 532–546.
Freeman, A. M., III. 1974. On estimating air pollution control benefits from land value studies.
Journal of Environmental Economics and Management 1: 74–83.
Freeman, A. M., III. 1993. The measurement of environmental and resource values: Theory and
methods. Washington, DC: Resources for the Future.
Gardner, B. 2003. US commodity policies and farmland values. In Government policy and farm-
land markets: The maintenance of farmer wealth, eds. C. B. Moss and A. Schmitz, 81–96.
Ames: Iowa State Press.
Gardner, K., and R. Barrows. 1985. The impact of soil conservation investments on land prices.
Geniaux, G., J.-S. Ay, and C. Napoléone. 2011. A spatial hedonic approach on land use change
anticipations. Journal of Regional Science 61 (5): 967–986.
Gibbons, S., and H. G. Overman. 2012. Mostly pointless spatial econometrics? Journal of
Regional Science 52(2): 172–191.
Goodwin, B. K., A. K. Mishra, and F. N. Ortalo-Magnè. 2003a. What’s wrong with our models of
agricultural land values? American Journal of Agricultural Economics 85: 744–752.
Goodwin, B. K., A. K. Mishra, and F. N. Ortalo-Magnè. 2003b. Explaining regional differences
in the capitalization of policy benefits into agricultural land values. In Government policy
and farmland markets: The maintenance of farmer wealth, eds. C. B. Moss and A. Schmitz,
97–114. Ames: Iowa State University Press.
Goodwin, B. K., A. K. Mishra, and F. Ortalo-Magnè. 2010. The buck stops where? The distribu-
tion of agricultural subsidies. NBER Working Paper No. 16693.
Greene, W. H. 2012. Econometric analysis. 7th ed. Upper Saddle River, NJ: Prentice Hall.
Greenstone, M., and J. Gallagher. 2008. Does hazardous waste matter? Evidence from the hous-
ing market and the Superfund Program. Quarterly Journal of Economics 123(3): 951–1003.
Griffith, D. A. 2009. Spatial autocorrelation. Unpublished. http://www.elsevierdirect.com/bro-
chures/hugy/SampleContent/Spatial-Autocorrelation.pdf
Grout, C. A., W. K. Jaeger, and A. J. Plantinga. 2011. Land-use regulations and property val-
ues in Portland, Oregon: A regression discontinuity design approach. Regional Science and
Guiling, P., B. W. Brorsen, and D. Doye. 2009. Effect of urban proximity on agricultural land
values. Land Economics 85(2): 252–264.
Gutierrez, L., J. Westerlund, and K. Erickson. 2007. Farmland prices, structural breaks and
panel data. European Review of Agricultural Economics 34(2): 161–179.
Hardie, I. W., T. A. Narayan, and B. L. Gardner. 2001. The joint influence of agricultural and
nonfarm factors on real estate values: An application to the Mid-Atlantic region. American
Journal of Agricultural Economics 83(1):120–132.
Heckman, J. J. 1979. Sample selection bias as a specification error. Econometrica 47(1): 153–161.
Heckman, J. J., H. Ichimura, J. A. Smith, and P. E. Todd. 1998. Characterizing selection bias
using experimental data. Econometrica 66(5): 1017–1098.
Heckman, J. J., H. Ichimura, and P. E. Todd. 1997. Matching as an econometric evaluation esti-
mator: Evidence from evaluating a job training programme. Review of Economic Studies
64(4): 605–654.
Heckman, J. J., and S. Navarro. 2004. Using matching, instrumental variables and control func-
tions to estimate economic choice models. Review of Economics and Statistics 86(1): 30–57.
Heckman, J. J., and R. Robb. 1985. Alternative methods for evaluation the impact of interven-
tions. Journal of Econometrics 30: 239–267.
Henderson, J., and B. A. Gloy. 2009. The impact of ethanol plants on cropland values in the
Great Plains. Agricultural Financial Review 69: 36–48.
Henderson, J., and S. Moore. 2006. The capitalization of wildlife recreation income into farm-
land values. Journal of Agricultural and Applied Economics 38(3): 597–610.
Henneberry, D. M., and R. L. Barrows. 1990. Capitalization of exclusive agricultural zoning into
farmland prices. Land Economics 66: 249–258.
Herdt, R. W., and W. W. Cochrane. 1966. Farmland prices and technological advance. Journal
of Farm Economics 48: 243–263.
Hirano, K., and G. W. Imbens. 2005. The propensity score with continuous treatments. In
Applied Bayesian modeling and causal inference from incomplete-data perspectives: An essen-
tial journey with Donald Rubin’s statistical family, eds. A. Gelman and X.-L. Meng, 73–84.
Chichester, UK: John Wiley & Sons.
Hoppe, R. A., and D. E. Banker. 2010. Structure and finances of US family farms: Family Farm
Report, 2010 Edition. EIB-66, USDA-Economic Research Service, July.
Huang, H., G. Y. Miller, B. J. Sherrick, and M. I. Gomez. 2006. Factors influencing Illinois farm-
land values. American Journal of Agricultural Economics, 88(2): 458–470.
Imbens, G. W., and T. Lemieux. 2008. Regression discontinuity designs: A guide to practice.
Journal of Econometrics 142(2): 615–635.
Imbens, G. W., and J. M. Wooldridge. 2007. Control function and related methods. In NBER
Summer Institute What’s new in econometrics? Lecture 6. http://www.nber.org/WNE/
lect_6_controlfuncs.pdf
Imbens, G. W., and J. M. Wooldridge. 2009. Recent developments in the econometrics of pro-
gram evaluation. Journal of Economic Literature 47(1): 5–86.
Irwin, E. G., A. M. Isserman, M. Kilkenny, and M. Partridge. 2010. A century of research on rural
development and regional issues. American Journal of Agricultural Economics 92(2): 522–553.
Just, R. E., and J. A. Miranowski. 1993. Understanding farmland price changes. American
Journal of Agricultural Economics 75(1): 156–168.
Kirwan, B. E. 2009. The incidence of US agricultural subsidies on farmland rental rates. Journal
of Political Economy 117(1): 138–164.
Kostov, P. 2009. A spatial quantile regression hedonic model of agricultural land prices. Spatial
Economic Analysis 4(1): 53–70.
Kostov, P. 2010. Do buyers’ characteristics and personal relationships affect agricultural land
prices? American Journal of Agricultural Economics 86(1): 48–65.
Kostov, P., M. Patton, and S. McErlean. 2008. Non-parametric analysis of the influence of buy-
ers’ characteristics and personal relationships on agricultural land prices. Agribusiness 24
(2): 161–176.
Lee, L. F. 1982. Some approaches to the correction of sample selection bias. Review of Economic
Studies 49(3): 355–372.
Lee, L. F. 1983. Generalized econometric models with selectivity. Econometrica 51(2): 507–512.
Lee, D. S., and T. Lemieux. 2010. Regression discontinuity designs in economics. Journal of
Economic Literatures 48: 281–355.
Lence, S. 2001. Farmland prices in the presence of transaction costs: A cautionary note.
Lence, S. H., and A. K. Mishra. 2003. The impacts of different farm programs on cash rents.
Lynch, L., W. Gray, and J. Geoghegan. 2007. Are farmland preservation program easement
restrictions capitalized into farmland prices? What can propensity score matching tell us?
Review of Agricultural Economics 29: 502–509.
Ma, S., and S. M. Swinton. 2012. Hedonic valuation of farmland using sale prices versus
appraised values. Land Economics 88(1): 1–15.
Maddison, D. 2009. A spatio-temporal model of farmland values. Journal of Agricultural
Economics 60(1): 171–189.
Melichar, E. 1979. Capital gains versus current income in the farming sector. American Journal
of Agricultural Economics 61(5): 1085–1092.
McMillen, D. P. 2010. Issues in spatial data analysis. Journal of Regional Science 50: 119–141.
McMillen, D. P. 2012. Perspectives on spatial econometrics. Journal of Regional Science
52(2): 192–209.
McMillen, D. P., and J. E. McDonald. 1989. Selectivity bias in urban land value functions. Land
McMillen, D. P., and C. Redfearn. 2010. Estimation and hypothesis testing for nonparametric
hedonic house price functions. Journal of Regional Science 50: 712–733.
Miranowski, J. A., and B. D. Hammes. 1984. Implicit prices of soil characteristics for farmland
in Iowa. American Journal of Agricultural Economics 66(4): 745–749.
Mishra, A. K., C. B. Moss, and K. E. Erickson. 2004. Valuing farmland with multiple quasi-fixed
inputs. Applied Economics 36: 1669–1675.
Mishra, A. K., G. T. Livanis, and C. B. Moss. 2011. Did the Federal Agriculture Improvement
and Reform Act of 1996 affect farmland values? Entropy 13: 668–682.
Moss, C. B. 1997. Returns, interest rates, and inflation: How they explain changes in farmland
values. American Journal of Agricultural Economics 79(4): 1311–1318.
Moss, C. B., and A. Schmitz, eds. 2003. Government policy and farmland markets: The mainte-
nance of farmer wealth. Ames: Iowa State Press.
Moss, C. B., and A. L. Katchova. 2005. Farmland values and asset performance. Agricultural
Finance Review 65(2): 119–130.
Navarro, S. 2008. Control function. In The new Palgrave dictionary of economics, 2nd eds., eds.
S. N. Durlauf and L. E. Blume, 208–212. London: Palgrave Macmillan Press.
Nelson, F. J., and L. P. Schertz. 1996. Provisions of the Federal Agriculture Improvement and
Reform Act of 1996. AIB-729, USDA-Economic Research Service.
Nickerson, C. J., and L. Lynch. 2001. The effect of farmland preservation programs on farmland
prices. American Journal of Agricultural Economics 83(2): 341–351.
Nickerson, C. J., M. Morehart, T. Kuethe, J. Beckman, J. Ifft, and R. Williams. 2012. Trends in
farmland values and ownership. EIB-92, USDA-Economic Research Service, February.
Nivens, H. D., T. L. Kastens, K. C. Dhuyvetter, and A. M. Featherstone. 2002. Using satel-
lite imagery in predicting Kansas farmland values. Journal of Agricultural and Resource
Economics 27(2): 464–480.
Palmquist, R. 1989. Land as a differentiated factor of production: A hedonic model and its
implications for welfare measurement. Land Economics 65(1):23–28.
Palmquist, R. B. 2006. Property value models. In Handbook of environmental economics, Vol. 2,
eds. K. G. Mäler and J. R. Vincent, 763–819. North Holland: Elsevier.
Palmquist, R. B., and L. E. Danielson. 1989. A hedonic study of the effects of erosion control and
drainage on farmland values. American Journal of Agricultural Economics 71: 55–62.
Patton, M., P. Kostov, S. McErlean, and J. Moss. 2008. Assessing the influence of direct payments
on the rental value of agricultural land. Food Policy 33: 397–405.
Pinkse, J., and M. E. Slade. 2010. The future of spatial econometrics. Journal of Regional Science
50: 103–117.
Plantinga, A. J., R. N. Lubowski, and R. N. Stavins. 2002. The effects of potential land develop-
ment on agricultural land prices. Journal of Urban Economics 52 (3): 561–581.
Pope, R., R. A. Kramer, R. D. Green, and D. B. Gardner. 1979. An evaluation of econometric
models of US farmland prices. Western Journal of Agricultural Economics 4(1):107–118.
Puhani, P. A. 2000. The Heckman correction for sample selection and its critique. Journal of
Economic Surveys 14(1): 53–68.
Ravallion, M. 2007. Evaluating anti-poverty programs. In Handbook of development economics,
Vol. 4, eds. T. Schultz and J. Strauss, 3787–3846. North-Holland: Elsevier.
Ricardo, D. 1996. The principles of political economy and taxation. Amherst, NY: Prometheus
Books [originally published in 1817].
Roka, F. M., and R. B. Plamquist. 1997. Examining the use of national databases in a hedonic anal-
ysis of regional farmland values. American Journal of Agricultural Economics 79: 1651–1656.
Rosen, S. 1974. Hedonic prices and implicit markets: Product differentiation in pure competi-
tion. Journal of Political Economy 82(1): 34–55.
Rosenbaum, P. R., and D. B. Rubin. 1983. The central role of the propensity score in observa-
tional studies for causal effects. Biometrika 70: 41–55.
Rubin, D. B. 1980. Bias reduction using Mahalanobis-metric matching. Biometrics 36: 293–298.
Sauer, J., J. Walsh, and D. Zilberman. 2012. Producer behaviour and agri-environmental poli-
cies: A directional distance based matching approach. Selected paper prepared for presenta-
tion at the 2012 Agricultural & Applied Economics Association Annual Meeting, Seattle,
WA, August 12–14, 2012.
Schmitz, A. 1995. Boom-bust cycles and Ricardian rents. American Journal of Agricultural
Economics 77(5): 1110–1125.
Shaik, S., G. A. Helmers, and J. A. Atwood. 2005. The evolution of farm programs and their
contribution to agricultural land values. American Journal of Agricultural Economics
87(5): 1190–1197.
Sherrick, B. J., and P. J. Barry. 2003. Farmland markets: Historical perspectives and contempo-
rary issues. In Government policy and farmland markets: The maintenance of farmer wealth,
eds. C. B. Moss and A. Schmitz, 27–52. Ames: Iowa State Press.
Shi, Y. J., T. T. Phipps, and D. Colyer. 1997. Agricultural land values under urbanizing influence.
Land Economics 73(February): 90–100.
Smith, J. A., and P. E. Todd. 2005. Does matching overcome LaLonde’s critique of nonexperi-
mental estimators? Journal of Econometrics 125: 305–353.
Sumner, D. A., J. M. Alston, and J. W. Glauber. 2010. Evolution of the economics of agricultural
policy. American Journal of Agricultural Economics 92(2): 403–423.
Tegene, A., and F. Kuchler. 1993. A regression test of the present value model of US farmland.
Tobler, W. 1970. A computer movie simulating urban growth in the Detroit region. Economic
Geography 46(2): 234–240.
Todd, P. E. 2007. Evaluating social programs with endogenous program placement and selec-
tion of the treated. In Handbook of development economics, Vol. 4, eds. T. P. Schultz and J.
Strauss, 3847–3894. North-Holland: Elsevier.
Towe, C., R. Lewis, and L. Lynch. 2014. Applying experiments to land economics: Public infor-
mation and auction efficiency in ecosystem service markets. In The Oxford handbook of
land economics, eds. J. M. Duke and J. Wu, 481–510. New York: Oxford University Press.
Towe, C., and C. I. Tra. 2013. ‘Vegetable spirits’ and energy policy. American Journal of
Tweeten, L. T., and J. E. Martin. 1966. A methodology for predicting US farm real estate price
variation. Journal of Farm Economics 48(2):378–393.
U.S. Department of Agriculture, Economic Research Service. 2012. Balance sheet of the farm-
ing sector, 2008–2012F. USDA-ERS Farm Income and Wealth Statistics Topic Page. http://
www.ers.usda.gov/data-products/farm-income-and-wealth-statistics.aspx
U.S. Department of Agriculture, National Agricultural Statistics Service. 2001. 1997 Census of
Agriculture. vol. 3, Special Studies; Part IV: Agricultural and Land Economics Ownership
Survey. AC97-SP-4. December.
Vitaliano, D. F., and C. Hill. 1994. Agricultural districts and farmland prices. Journal of Real
Estate Finance and Economics 8(3): 213–223.
van der Klaauw, W. 2008. Regression-discontinuity analysis: A survey of recent developments
in economics. Labour 22(2): 219–245.
Vella, F. 1998. Estimating models with sample selection bias: A survey. Journal of Human
Resources 33(1): 127–169.
Weersink, A., S. Clark, C. G. Turvey, and R. Sarker. 1999. The effect of agricultural policy on
farmland values. Land Economics 75(3): 425–439.
Wooldridge, J. M. 2002. Econometric analysis of cross section and panel data. Cambridge,
MA: MIT Press.
Zhang, W., E. G. Irwin, and C. J. Nickerson. 2012. The expanding ethanol market and farm-
land values: Identifying the changing influence of proximity to agricultural delivery points.
Selected paper prepared for presentation at the 2012 Agricultural & Applied Economics
Association Annual Meeting, Seattle, WA, August 12–14, 2012.
Zhao, Z. 2004. Using matching to estimate treatment effects: Data requirements, matching met-
rics, and Monte Carlo evidence. The Review of Economics and Statistics 86(1): 91–107.
C HA P T E R 6
L A N D U S E A N D S U S TA I NA B L E
ECONOMIC DEVELOPMENT
Developing World
E DWA R D B. BA R BI E R
Land use change in developing countries is critically bound up with the pattern of eco-
nomic development in these countries. Most developing economies, and certainly the
majority of the populations living within them, depend directly on natural resources.
For many of these economies, primary product exports account for the vast majority of
their export earnings, and one or two primary commodities make up the bulk of exports
(Barbier 2005b, chapter 1). Agricultural value added accounts for an average of 40%
of GDP, and nearly 80% of the labor force is engaged in agricultural or resource-based
activities (World Bank 2008). Further adding to these disparities, by 2025, the rural
population of the developing world will have increased to almost 3.2 billion, placing
increasing pressure on a declining resource base (Population Division of the United
Nations 2008).
Over the past 50 years, the pattern of land use change in developing as opposed
to developed economies has been dramatically different (Fischer and Heilig 1997;
Ramankutty and Foley 1999; FAO 2006; World Bank 2008; Barbier 2011). In developed
countries, cropland area slowed its growth, eventually stabilized, and is now declin-
ing. As a result, the decline of forest and woodland has halted in developed countries
in aggregate, and since 1990, total forest area has increased (FAO 2006). Not only has
primary forest area recovered but also the growth in plantations has been strong. In con-
trast, in developing economies, cropland area has continued to expand. In the develop-
ing regions of Africa, Asia, and Latin America, tropical forests were the primary sources
of new agricultural land in the 1980s and 1990s (Gibbs et al. 2010). Almost one-fifth of
new crop production in developing countries from 1990 to 2050 is expected to rely on
expanding cultivated area, and two-thirds of this new land will come from conversion of
forests and wetlands (Fischer and Heilig 1997). In some regions, such as tropical Latin
America, livestock grazing is also projected to cause extensive deforestation in the near
future (Wassenaar et al. 2007).
140 Edward B. Barbier
However, although historically such land use changes leading to cropland expansion
may have been associated with successful resource-based development, this is less likely
for most developing countries today (Barbier 2011). The main purpose of this chapter is
to offer an economic explanation about why this might be the case. That is, development
in low and middle-income economies is accompanied by substantial resource conver-
sion, especially the expansion of the agricultural land base through the conversion of
forests, wetlands and other natural habitat, but this pattern of land use is generating less
economy-wide benefits than in previous eras. The main reason is that the current pro-
cess of land use and expansion has two unique structural features.
First, considerable land expansion in ecologically fragile areas is serving mainly
as an outlet for the subsistence and near-subsistence needs of the rural poor (Barbier
2005b, 2010). A substantial proportion of the population in low and middle-income
countries is concentrated in marginal areas and on ecologically fragile land, such
as converted forest frontier areas, poor quality uplands, converted wetlands, and
so forth (Comprehensive Assessment of Water Management in Agriculture 2007;
World Bank 2003). Households on these lands not only face problems of land degra-
dation and low productivity but also tend to be some of the poorest in the world. Yet,
population increases and other economic pressures are driving many of the rural
poor to bring yet more marginal land into production (Chen and Ravillion 2007;
Population Division of the United Nations 2008). The result is that such marginal
land expansion continues to be the main basis for absorbing the growing number
of the rural poor in developing economies (Pichón 1997; Coxhead et al. 2002; Carr
2009; Barbier 2011).
Second, marginal land expansion may be an important outlet for the rural poor, but
it may not be the main cause of overall land conversion and use in developing coun-
tries. Recent evidence suggests that commercially oriented economic activities are
responsible for much of the land expansion that is occurring in low- and middle-income
economies. For example, the main “agents of deforestation” globally are now plantation
owners, large-scale farmers, ranchers, and timber and mining operations, assisted by
government policies (FAO 2001, 2003, 2006; Chomitz et al. 2007; Rudel 2007; DeFries
et al. 2010; Boucher et al. 2011). Large-scale capital investments, which include planta-
tion agriculture, ranching, forestry and mining activities, often result in export-oriented
extractive enclaves with little or no forward and backward linkages to the rest of the
economy (Bridge 2008; Barbier 2005b, 2011; van der Ploeg 2011). The result is that
development in low- and middle-income economies is accompanied by substantial
resource conversion, especially the expansion of the agricultural land base through the
conversion of forests, wetlands, and other natural habitat. At the same time, most devel-
oping economies remain highly dependent on the exploitation of natural resources and
are unable to diversify.
The consequence of these two structural features of land use and expansion in devel-
oping economies is that they are symptomatic of a dualistic frontier economy. The clas-
sic definition of a frontier is “a geographic region adjacent to the unsettled portions of
the continent in which a low man/land ratio and unusually abundant, unexploited,
Land Use and Sustainable Economic Development 141
natural resources provide an exceptional opportunity for social and economic better-
ment” (Billington 1966, 25). To exploit these resources, processes of frontier expansion
or frontier-based development are “characterized by the initial existence of abundant
land, mostly unoccupied, and by a substantial migration of capital and people” (di
Tella 1982, 212). For heavily resource-dependent developing economies, that is, those
that have 75% or more of primary production to total exports, such a process of
frontier-based development may characterize nearly the entire economy. For other low
and middle-income economies, in addition to the frontier economy, there may also be
burgeoning industrial and service sectors.
However, the main structural feature of the frontier economy in most developing
countries is that it is inherently dualistic. The frontier economy contains both a tra-
ditional sector that converts and exploits available land to produce a nontraded agri-
cultural output, and a fully developed, commercially oriented sector that converts and
exploits available land and natural resources for a variety of traded outputs. The lat-
ter could include plantation agriculture, ranching, forestry and mining activities. In
addition, the traditional agricultural sector is dominated by farm holdings that occupy
marginal or ecologically fragile land with poor land quality and productivity poten-
tial. Although these two types of economic activities differ significantly and may also
be geographically separated, they are linked by labor use, as the rural poor on marginal
land form a large pool of surplus unskilled labor that can be employed in commercial
frontier activities. This linkage is important not only to the dynamics of land expansion
and use within developing economies but also to the pattern of overall economic devel-
opment (Hansen 1979; Píchón 1997; Coxhead et al. 2002; Barbier 2005a; Maertens et
al. 2006; Carr 2009).
To set the stage, this chapter first describes in more detail the dualistic frontier econ-
omy and processes of land expansion that typify many developing economies. A model
of the dualistic frontier economy is then developed to explore its main economic impli-
cations for economic development in many low- and middle-income countries today.
These implications lie at the core of why land use and expansion in developing econo-
mies may not be generating greater economy-wide benefits.
To summarize the key results, in the dual frontier economy, because there are no
diminishing returns to labor in the use of marginal land for agricultural production, real
wages are invariant to rural employment. As long as there remains abundant marginal
land to absorb more farmers and employment, the use of land relative to labor on this
land will determine nominal wages throughout the dual frontier economy. The impli-
cation is that, with given international prices for the marketed-oriented activities, the
real wage and thus the amount of unskilled labor employed by these activities will be
fully determined. The pool of surplus labor on marginal lands is essentially a barom-
eter of frontier-based development. As long as there are abundant marginal lands for
cultivation, they serve to absorb rural migrants, population increases, and displaced
unskilled labor from elsewhere in the economy. On the other hand, expanding com-
mercial activities that exploit more resources and land on the frontier will absorb more
workers from the pool of surplus labor existing on marginal frontier lands. Although the
latter outcome may seem beneficial, it has the tendency to promote boom and bust cycles
of economic development (Wunder 2003, 2005; Barbier 2005a, 2005b, 2007, 2011; Ha
and Shively 2008; Agergaard et al. 2009; Barney 2009; Rodrigues et al. 2009; Hall 2011;
Knudsen and Folds 2011). Such cycles are reinforced by a policy environment that, on
the one hand, encourages frontier commercial activities to remain as isolated enclaves
and, on the other, fails to ensure that the resource rents generated by these activities lead
to greater economic diversification (Barbier 2005b, 2007).
1. Land Use and the Dualistic

Frontier Economy
Since 1950, many economies with abundant endowments of land, mineral and fossil
fuel resources have had difficulty in achieving successful resource-based development
(Barbier 2005b, 2011; van der Ploeg 2011). For example, Gylfason (2001) has examined
the long-run growth performance of 85 resource-rich developing economies since 1965.
Only Botswana, Malaysia, and Thailand managed to achieve a long-term investment rate
exceeding 25% of GDP and long-run average annual growth rates exceeding 4%, which
is a performance comparable to that of high income economies. Malaysia and Thailand
have also managed successfully to diversify their economies through re-investing the
financial gains from primary production for export. Botswana has yet to diversify its
economy significantly but has developed favorable institutions and policies for man-
aging its natural wealth and primary production for extensive economy-wide benefits.
Although many other developing countries still depend on finding new reserves of land
and other natural resources to exploit, very few appear to have benefited from such
resource-based development. This poses an intriguing paradox: Why should economic
dependence on natural resource exploitation and land expansion be associated with
“unsustainable” resource-based development in many low and middle-income coun-
tries today, especially because historically this has not always been the case?
One reason is that the unique pattern of frontier land expansion emerging in devel-
oping economies appears to be inimical to successful economy-wide development. An
early criticism of this pattern was the hollow frontier hypothesis, which James (1969)
first used to describe the expansion of the coffee frontier in southern and central Brazil
during the mid-20th century. Although these areas were originally settled by small-
holders, they were later displaced to more remote regions by wealthy landowners
through property aggregation, which lead to a relatively depopulated and “hollow”
frontier. Evidence of this process has been found in the Brazilian Amazon not only
for coffee but also for ranching and other forms of large-scale commercial agricul-
ture (Casetti and Gauthier 1977; Wood 1983; Aldrich et al. 2006; Morton et al. 2006;
Browder et al. 2008).
In addition, if institutions and economic policies encourage large profits from fron-
tier expansion, then “well-capitalized interests, including land speculators and ranch-
ers, consolidate the properties of subsistence farmers through market transactions or
outright expulsions” (Aldrich et al. 2006, 272). However, such large-scale capital invest-
ments, which include plantation agriculture, ranching, forestry and mining activities,
often result in export-oriented extractive enclaves with little or no forward and back-
ward linkages to the rest of the economy (Barbier 2005b, 2011; Bridge 2008). As pointed
out by Bunker (1989, 607): “Overconfidence in the linkage potential of extractive econ-
omies can lead directly to public investments aimed at capturing the linkages near the
mouth of the mine when in fact the locational disadvantages are so great that only under
extraordinary circumstances would these investments be competitive.” The result is a
vicious cycle, whereby policies and institutions continue to favor, subsidize, and sup-
port capital investments to create abnormal profits for mineral and large-scale agricul-
tural projects in the frontier, yet the lack of linkages to the rest of the economy simply
reinforce the tendency of these investments to create commercially oriented extractive
enclaves (Barbier 2005b, 2011). These enclaves are more tied to the “global produc-
tion network” that focuses on exploitation of agricultural and mineral resources for the
world market or domestic consumption in urban and industrial centers (Bridge 2008).
Government policies have actively promoted capital investment in commercially
oriented frontier agricultural and extractive activities. For example, in the Brazilian
Amazon, “spatial differentiation in the pattern of development would be largely influ-
enced by the State, in its infrastructure investment decisions (e.g., roads and utility
extensions into the frontier) and in fiscal incentive policies targeted to specific regions
that would invite capital investment there” (Browder et al. 2008, 1472). State programs
to improve property rights and the efficiency of land markets increase land values and
attract additional frontier investments. As Gould et al. (2006) illustrate with a case study
of the Petén, Guatemala, such land administration and privatization policies can have
the unintended consequence of increasing the incentive for land speculation rather than
investment in productive agricultural activities. Similarly, Bromley (2008, 561) shows
that, in Africa, “an exclusive focus on the property relations of isolated villages and their
commons will necessarily fail if development programs ignore the institutional archi-
tecture of markets and market processes throughout the entirety of a nation.”
Government policies have supported the expansion of large-scale soybean cultiva-
tion and mechanized agriculture in Amazonia (Hecht 2005; Bulte et al. 2007; Killeen
et al. 2007; Carr 2009; Walker et al. 2009); oil palm, coffee, and other cash crops in
Asia (Coxhead et al. 2002; Agergaard et al. 2009; Barney 2009; Curry and Koczberski
2009; Hirsch 2009; McCarthy and Cramb 2009); cocoa, cotton, and other cash crop
frontiers in Africa (Mosley 2005; Bromley 2008; Knudsen and Fold 2011); ranching in
Latin America (Walker 2003; Bulte et al. 2007; Killeen et al. 2007; Wassenaar et al. 2007;
Caviglia-Harris and Harris 2008; Schmook and Vance 2009; Walker et al. 2009); and
extractive frontiers globally (Hyndman 1994; Wunder 2003, 2005; Akpalu and Parks
2007; Bridge 2008; Campbell 2009).
Frontiers are also the means for marginal land expansion as a “safety valve” outlet for
the rural poor. As noted by Coxhead et al. (2002, 345), “the land frontier has long served
as the employer of last resort for underemployed, unskilled labor.” This process was
fostered by colonial policies in many developing regions yet has continued unabated
since the 1950s (James 1969; Hansen 1979; Foweraker 1981; Bunker 1984; Williamson
2002, 2006; Austin 2007; Etter et al. 2008; Barbier 2011). The result has been a large
concentration of the rural poor on low quality land for agriculture, characterized by
traditional farming methods with negligible marginal productivity, zero land rents
or profits, and informal or nonexistent land-tenure arrangements, inadequate trans-
port and infrastructure, and other market imperfections (Mueller 1997; Coxhead et al.
2003; Barbier 2005b; Gould et al. 2006; Jepson 2006; Maertens et al. 2006; Carr 2009;
Schmook and Vance 2009).
In sum, long-term land use trends and economic development in many low and mid-
dle income countries has evolved a dualistic frontier economy. This outcome was first
highlighted by Hansen (1979) to describe colonial land use in developing regions, and
then by Wood (1983, 259) to characterize frontier development in Amazonia: “A cen-
tral feature of the contemporary settlement of the Brazilian Amazon is the simultane-
ous expansion into the region of capitalist enterprises and peasant farmers. The dual
character of the frontier is, to a large extent, a consequence of the development policies
adopted by the state.” As noted by Aldrich et al. (2006, 72) the outcome of this dualistic
process of frontier expansion is often frontier stratification: “Although the smallholders
who initiate frontier settlement are poor, they share their poverty in relative equality
until the aggregation of property causes the distribution of land to be skewed and drives
social stratification.”
The result is an inherently dualistic economy. Coexisting in most frontiers are
highly developed, modern, and profitable commercial economic activities along
with more traditional, relatively poor agricultural activities on marginal lands.
That is, “according to the dualist model the frontier is comprised of two different
economies: the traditional, non-capitalist sector, which is subsistence-oriented
and has minimal ties to the marketplace; and the modern, capitalist sector, which
is market-oriented and follows the logic of profit maximization” (Wood 1983, 263).
Although these two types of economic activities differ significantly and may also be
geographically separated, they are linked by labor use. This linkage is important to
the dynamics of frontier expansion, because it means that rural poor on marginal
land form a large pool of surplus unskilled labor that can be employed in commercial
frontier activities (Hansen 1979; Píchón 1997; Coxhead et al. 2002; Barbier 2005a;
Maertens et al. 2006; Carr 2009).
The dualistic frontier economy has important implications for economic develop-
ment in many low- and middle-income countries today. To explore these implications
more fully, it is useful to develop a model depicting land use and labor allocation in the
dualistic frontier economy.
2. A Model of the Dualistic Frontier

Economy
Following the discussion above, it is assumed that the dualistic frontier economy com-
prises two sectors: (1) a fully developed, commercially oriented sector that converts and
exploits available land and natural resources for a variety of traded outputs, and (2) a
traditional sector that converts and exploits available land to produce a nontraded agri-
cultural output, which is dominated by farm holdings that occupy marginal or ecologi-
cally fragile land with poor land quality and productivity potential. Although these two
sectors comprising the frontier economy differ significantly and may also be geographi-
cally separated, they are linked by labor use. That is, the rural poor on marginal land
form a large pool of surplus unskilled labor that can be employed in commercial frontier
activities, and the wage rate is determined by the dynamics of land expansion within
the frontier economy (Hansen 1979; Píchón 1997; Coxhead et al. 2002; Barbier 2005a;
Maertens et al. 2006; Carr 2009).
For heavily resource-dependent economies (i.e., those that have 75% or more of pri-
mary production to total exports), the commercial and traditional frontier sectors may
comprise nearly the entire economy. For other low- and middle-income economies,
there may also be a burgeoning industrial and/or service sectors. For the purposes of
the model, it does not matter whether the dualistic frontier is an enclave within a larger
developing economy or whether it comprises the entire economy.
2.1 Sector 1: Commercial Primary Production

Production of the traded primary product (plantation crops, timber, beef, mineral,
etc.) is modeled in a similar way as in Findlay and Lundahl (1994). Primary produc-
tion depends directly on inputs of land and/or natural resources (N1)and labor (L1), and
indirectly on capital (K1). The sector imports capital from either the rest of the econ-
omy or abroad, and this capital consists of both reproducible capital (machines, equip-
ment, tools, etc.) and the skilled labor, or human capital, required to maintain and run
such durable goods. As domestic and foreign claims on capital are perfect substitutes as
stores of wealth, and the open economy is small in relation to the world economy, capi-
tal is available in perfectly elastic supply at the international interest rate, r (Barro and
Sala-I-Martin 2004). Thus, the accumulation of capital in the commercial activity sector
has a negligible impact on the interest rate, which can be treated as exogenous.
Primary production, Q1, is determined by a function with the normal concave prop-
erties and is homogeneous of degree one
Q1 = f (N1 , L1 ), fi > 0, fii < 0, i = N , L (1)

The commercial activity can obtain more land or natural resources (hereafter referred
to as “resources”) for primary production, but only by employing and allocating more
capital for this purpose. It is assumed that increasing N1 incurs a rising input of K1
K1 = c(N1 ), c ′ > 0, c ″ > 0 (2)

where c′ (N1) is the marginal capital requirement of obtaining and transforming a unit
of the resource input, which is a convex function of the amount of N1 appropriated. As
c′ (N1) represents the “marginal cost” of obtaining land and resources, c″>0 implies that
these costs are rising as more appropriation occurs.
Letting p1 be the price of primary production output, r the interest rate, and w the
wage rate, it follows that total profits are
π = L1 p1 f (n1 ) − rc(N1 ) − wL1n1 = N1 / L1 (3)

Profit-maximizing leads to
w
f (n1 ) − f N n1 = (4)
p1
p1 f N (n1 )
=r (5)
c′
Condition (4) is the normal value marginal product conditions for use of labor in pro-
duction. Condition (5) determines the optimal use of natural resources, and indicates
that the rate of return from appropriating N1 for primary production must be equal to
the interest rate. The rate of return consists of the marginal rent per unit of N1 divided by
the marginal cost of converting it for use in primary production.
2.2 Sector 2: Traditional Agriculture on Marginal Land

Production of nontraded agricultural output involves two inputs, land (N2) and labor
(L2); any capital input is fixed and fully funded out of normal profits. Both land and
labor are required for traditional agricultural production, Q2, which is determined by a
function with the normal concave properties and is homogeneous of degree one
Q2 = g (N 2 , L2 ), g i ≥ 0, g ii < 0, i = N , L (6)

Note that the marginal productivity of land is not necessarily positive. This Ricardian
surplus-land condition follows from the assumption that poor quality marginal land is
unproductive in cultivation (Hansen 1979). That is, for traditional agriculture on mar-
ginal land, gN = 0 and, thus, equilibrium is determined by
w Nm
g ′(n2 ) = 0, g (n2 ) = , n2 = n2m = m2 (7)
p2 L2

The result of this outcome is that there are no diminishing returns to labor in the use
of marginal land for agricultural production. Real wages are invariant to rural employ-
ment (the number of farmers and/or labor input on marginal land) and determined by
the average product of labor. Moreover, the condition of zero marginal productivity
fixes the land/labor ratio on marginal land, which can be designated as nm2 . Finally, given
the average product of labor relationship in Equation (7), the fixed land/labor ratio will
determine the nominal wage rate w for any predetermined output price p2. Thus, the
best that farmers and their families on marginal land can do is either sell their labor to
each other and obtain an equilibrium real wage w/p2, or alternatively, farm their own
plots of land and earn the same real wage. Since there is little advantage in selling their
labor, farmers will tend to use their and family labor to farm their own land. Hence,
under this marginal land condition, small family farms will predominate. Unless the
population increases, no more land will be brought into production and there will be
surplus land.1
Finally, the total labor force in the frontier economy is given, and is
L = L1 + Lm2 (8)
2.3 Equilibrium
Because the fixed land/labor ratio on marginal land determines the nominal wage rate,
the model of the frontier economy is fully recursive. With w determined, condition
(4) indicates that to each value of p1 there corresponds a unique value of the resource/
labor ratio n1 in commercial primary production. As r is also given, Equation (5) can
now be solved for the equilibrium amount of natural resources appropriated and used
N1. With n1 and N1 known, L1 follows. Finally, Q1 can be determined from Equation (1)
for primary production.
1
Although the agricultural production of the traditional sector is a nontraded good, any surplus
produced in excess of subsistence consumption is likely to be sold in competitive local markets. The
standard assumption is that the resulting output price p2 is predetermined in such markets, which is
the general observation for traditional agriculture in frontier economies, whether its output is wholly
consumed for subsistence or any surplus is locally traded (see Hansen 1979; Mueller 1997; Píchón 1997;
Coxhead et al. 2002; Barbier 2005a; Maertens et al. 2006).
With L1 known, Lm2 can be found as a residual from Equation (8). As the fixed
land/labor ratio nm2 is already known, Nm2 follows. From Equation (6) it is now pos-
sible to determine traditional agricultural production Qm2 from marginal land.
From Equation (4) and the concavity conditions of Equation (1), it follows that
dn1 w
= 2 < 0 and thus
dp1 p1 nf ′′
n1 = n1 ( p1 ), n1 ′ < 0 (9)
As a rise in p1 leads to a fall in n1, the numerator of Equation (5) will increase. Given
that c″> 0, then N1 must rise in order for equilibrium condition (5) to continue to hold.
Consequently,
N1 = N1 ( p1 ), n1 ′ > 0 (10)
It follows from Equations (2) and (10) that K = c-1 (N1(p1)) and K′ > 0. Also, from
Equations (9), (10) and (1),
L1 = L1 ( p1 ), L1 ′ > 0, Q1 = Q1 ( p1 ), Q1 ′ > 0 (11)
Aggregate supply of primary products from the frontier is positively sloped.

From condition (7) on marginal lands, all output is consumed wLm2 = p2Qm2. However,
the outputs, Q1, from primary production are traded goods that are exported to the
rest of the economy or abroad. It follows that total income, Y, in the frontier economy,
excluding subsistence income from marginal land, is determined by p1
Y = wL1 + rK + p1 f N (n1 ) (12)

Assuming that consumers have identical and homothetic preferences, define the
demand function for primary products as
∂ Q1D ∂ Q1D
Q1D = Q1D ( p1 , Y ), < 0, >0 (13)
∂ p1 ∂Y

From Equations (11) and (13), the excess supply function for primary products is,
therefore,
EQ1 ( p1 ) = Q1 ( p1 ) − Q1D( p1 ) > 0, EQ1 ′ > 0 (14)

Because primary products are traded, the excess supply is used to import goods and
services, either from the rest of the economy or abroad; that is,
1
EQ1 ( p1 ) = M ( ), M ′ < 0 (15)
p1

2.4 An Increase in Population

An increase in population is tantamount to the increase in the total labor force L in
Equation (8). However, because neither nominal nor real wages change, the increase in
population must occur solely on marginal land. In order for the land/labor ratio nm2 to
remain fixed, there must be an equal increase in Nm2 to absorb the rise in Lm2 . The demand
for labor from converting and cultivating marginal land is almost infinitely elastic.
Because nm2 is unchanged, nominal and real wages remain the same. Consequently, the
effect of an increase in population is to increase labor use, cultivated area, and, thus,
aggregate agricultural output Qm2 on marginal land.
As total and rural population has increased in developing countries, marginal lands
have served as an important outlet. In 1950, the rural areas of the developing world con-
tained 1.8 billion people, which, by 2005, had almost doubled to 3.4 billion. From 1950
to 1975, annual rural population growth in these regions was 1.8%, and from 1975 to
2007 it was just over 1.0% (Population Division of the United Nations 2008). Since 1950,
the number of people on marginal land in developing economies has doubled, reaching
nearly 1.3 billion today (World Bank 2003). The result is that marginal land expansion,
especially in frontier areas, continues to be the main basis for absorbing the growing
number of the rural poor in developing economies (Pichón 1997; Coxhead et al. 2002;
Carr 2009; Barbier 2011).
2.5 An Increase in the Price of Primary Products

If p1 rises, then real wages in this sector w/p1 fall. The result is to increase the demand
for labor L1 used in primary production. Also, from (9) and (10), it follows that the
resource/labor ratio for primary production will decline, and resource inputs will
rise. More resource conversion will in turn attract additional capital to the sector.
Condition (5) confirms that resource use must rise in primary production. The fall in
the resource/labor ratio causes marginal rent p1 fN, which is the numerator of condition
(5), to rise. Because the interest rate is unchanged, N1 must rise to increase c′ and main-
tain condition (5) in equilibrium.
In order for n1 to fall, the rise in L1 must exceed the increase in N1. Given Equation
(8), the increase in L1 must come from reducing labor on marginal land Lm2 . The fall in
Lm2 must be accompanied by an equivalent decline in Nm2 in order to keep the fixed land/
labor ratio on marginal land. Thus, the increase in employment, capital, resource use,
and output in the primary production sector in response to the rise in p1 will reduce
labor, cultivation, and production on marginal land. As Equation (14) indicates, excess
supply of primary production increases, and the resulting exports allow more goods and
services to be imported to the frontier.
Of course, if the price of primary products from the frontier falls, the opposite
occurs. The export-oriented primary sector contracts, and the resulting surplus labor
is absorbed on marginal land. The result is more land conversion and a larger share of
the population cultivating less favorable land. Rural poverty invariably increases as a
result.
These effects of price increases have been observed for coffee, ranching, large-scale
commercial agriculture in the Brazilian Amazon (Casetti and Gauthier 1977; Wood
1983; Aldrich et al. 2006; Morton et al. 2006; Browder et al. 2008; Rodrigues et al. 2009),
as well as for cocoa, coffee, oil palm, and shrimp in Southeast Asia and Africa (Ha and
Shively 2008; Agergaard et al. 2009; Barney 2009; Hall 2011; Knudsen and Folds 2011).
Oil price booms have interacted with agricultural expansion and deforestation in a
range of tropical countries, but with only short-lived economy-wide gains (Wunder
2003, 2005). As will be discussed later, as price rises for primary products are often
short-lived, they tend to promote “boom and bust” cycles of economic development in
many frontier areas of developing countries.
2.6 An Increase in the Price of Traditional Agricultural

Products
Because real wages are invariant to the number of farmers or workers employed on
marginal land, an increase in p2 must translate into a proportional increase in money
wages w. That is, the land/labor ratio must stay fixed at nm2, and so despite the rise
in agricultural prices, the real wage remains constant at w/p2. However, with the rise
in w, real wages in primary production w/p1 go up. As a result, the amount of labor
employed in this sector L1 declines. The resource/labor ratio increases, but this causes
marginal rents to fall. As the interest rate in Equation (5) is unchanged, N1 must also
decrease to maintain the equilibrium. In order for n1 to increase, L1 must decline more
than N1.
The unemployed labor on the frontier has to be absorbed through additional conver-
sion of marginal land. As Lm2 increases, Nm2 must rise proportionately in order to keep the
land/labor ratio fixed. Thus, the effect of the price rise is to expand cultivation and pro-
duction on marginal land, whereas the export-oriented primary production sector on
the frontier contracts. A fall in p2 would have the opposite outcome. Some evidence of
these effects of changes in the price of traditional products on marginal land expansion
is available for the uplands in Southeast Asia (Coxhead et al. 2002; Maertens et al. 2005;
Ha and Shively 2008).
2.7 An Increase in the Interest Rate

If the interest rises, then the rate of return from appropriating N1 for primary produc-
tion, the left-hand side of Equation (5), must also increase. However, since real wages in
primary production w/p1 are unchanged, the resource/labor ratio and, thus, marginal
rent p1 fN remain the same. Thus, only the denominator c′ in Equation (5) can fall, and
this means that fewer resources N1 are converted and used for primary production. This
is the expected outcome. An increase in the interest rate raises the cost of employing
capital to obtain more resources for primary production.
In order for n1 to remain constant, the contraction in N1 must be accompanied by a
proportionate decline in L1. This displaced labor from primary production must find
employment through additional conversion of marginal land. Once again, the increase
in Lm2 occurs with a proportionate rise in Nm2 to keep nm2 constant.
In sum, a rise in the interest rate causes a reduction in the capital employed in the pri-
mary production sector. Resource use, employment, and output decline in this sector,
and excess supply for export also falls. The unemployed labor will be absorbed through
expanded marginal land cultivation. In contrast, a fall in the interest rate will lead to a
contraction in marginal land use and an expansion of the export-oriented primary pro-
duction sector. Unfortunately, there appear to be little empirical evidence of the effects
of an increase of interest rate on dualistic frontier economic conditions.
2.8 Technical Progress on Marginal Land

The introduction of new inputs, such as fertilizers or improved varieties, and other tech-
nical improvements on marginal land may be neutral, or biased in favor of either land
or labor. However, if any such technical progress fails to affect the zero marginal pro-
ductivity condition indicated in Equation (7), then the land/labor ratio for production
on marginal land must, thereby, remain the same. However, the average productivity
of labor (nm2) can rise as a result of technical improvements on marginal land, and if
that is the case, real wages w/p2 will increase. Since p2 is fixed, this implies a rise in the
nominal wage.
The rise in the nominal wage leads to an increase in real wages w/p1 in commercial pri-
mary production activities. Labor employment L1 declines and the resource/labor ratio
increases. Marginal rents fall, but as the interest rate in (5) is fixed, N1 must also decrease
to maintain the equilibrium. In order for n1 to rise, L1 must decline more than N1. Thus,
the effect of technical progress on marginal land is a contraction in export-oriented pri-
mary production. The resulting unemployed labor must be absorbed through greater
cultivation of marginal land. As Lm2 increases, Nm2 must rise proportionately in order to
keep the land/labor ratio fixed.
Note, though, that this outcome hinges critically on the assumption that any technical
progress on marginal land does not alter the zero marginal productivity condition in Equation
(7). In contrast, empirical evidence of technical change and public investments in fron-
tier economies indicates that any resulting land improvements that do increase the value of
homesteads can have a positive effect on both land rents and reducing agricultural expansion
(Coxhead et al. 2002; Maertens et al. 2006; Sills and Caviglia-Harris 2008; Dercon et al. 2009).
3. Conclusion: Implications for Economic

Development
In the dual frontier economy found in many developing countries, real wages are invari-
ant to rural employment, because there are no diminishing returns to labor in the use of
marginal land for agricultural production. As long as there remains abundant marginal
land to absorb more farmers and employment, the use of land relative to labor on this
land will determine nominal wages throughout the dual frontier economy. The implica-
tion is that, with given international prices for the marketed-oriented activities, the real
wage and thus the amount of unskilled labor employed by these activities will be fully
determined. The pool of surplus labor on marginal lands is essentially a barometer of
frontier-based development. As long as there are abundant marginal lands for cultiva-
tion, they serve to absorb rural migrants, population increases, and displaced unskilled
labor from elsewhere in the economy. On the other hand, expanding commercial activi-
ties that exploit more resources and land on the frontier will absorb more workers from
the pool of surplus labor existing on marginal frontier lands.
Since 1950, the estimated population in developing economies on “fragile lands” has
doubled (World Bank 2003). These fragile environments are prone to land degradation,
and consist of upland areas, forest systems and drylands that suffer from low agricul-
tural productivity, and areas that present significant constraints for intensive agricul-
ture. Today, nearly 1.3 billion people—almost a fifth of the world’s population—live in
such areas in developing regions (Barbier 2011, Table 9.10). Almost half the people in
these fragile environments (631 million) consist of the rural poor, who, throughout the
developing world, outnumber the poor living on favored lands by 2 to 1 (Comprehensive
Assessment of Water Management in Agriculture 2007, Table 15.1).
The result is that marginal land expansion in frontier areas continues to be the main
basis of absorbing numbers of rural poor, whether they are displaced from more favor-
able lands or simply growing in number (Pichón 1997; Carr 2009; Barbier 2011). This
process is described eloquently by Pichón (1997, 707–708) for “marginal farmers” in
the Ecuadorian Amazon: “Most forest intervention in the region has come at the hands
of colonist farmers attempting to establish land claims along transport routes originally
constructed to aid in petroleum exploration and exploitation. These are farmers who
formerly have made a living in long-established farmlands and who, for various reasons
(population pressures, pervasive poverty, maldistribution of farmland, lack of inputs
for intensive cultivation, lack of nonagrarian livelihood opportunities, and generally
inadequate rural development) have been increasingly squeezed out of their home-
lands. A marginal person by virtue of his low socioeconomic and political status, the
farmer often perceives no way to sustain his family other than by seeking a livelihood on
the marginal environments of tropical rain forests.”
Equally, the poor on marginal lands serve as a pool of surplus low-wage labor for
commercial activities, including those in frontier regions. For example, in Southeast
Asia, agricultural and extractive activities in the lowlands rely on labor from marginal
uplands, and thus technological and economic changes in lowland agriculture signifi-
cantly impacts agricultural expansion and deforestation in the uplands (Coxhead et al.
2002; Maertens et al. 2006; Barney 2009). Oil palm expansion on the Malaysian and
Indonesian frontiers has depended on off-farm labor provided by agricultural small-
holders and poor migrants (McCarthy and Cramb 2009). If such employment opportu-
nities are sufficiently large and sustained, they can actually reduce long-term marginal
land expansion. For example, in Colombia, since 1970 high-input, intensified, highly
mechanized cropping on the most suitable land, as well expansion in cattle grazing has
drawn labor from more traditional agriculture, so that “areas of marginal land are slowly
being abandoned and left to revegetate (Etter et al. 2008, 17).
However, the continuing encouragement of commercial activities to exploit frontier
land and natural resources is impacting environmental change, especially deforestation.
For example, the main “agents of deforestation” globally are now plantation owners,
large scale farmers, ranchers and timber and mining operations, assisted by government
policies (FAO 2001, 2003; Chomitz et al. 2007; Rudel 2007; DeFries et al. 2010; Boucher
et al. 2011). For example, according to Rudel (2007, 40), “to facilitate their plans for
expansion, large landowners lobbied for the construction of improved and expanded
networks of roads. Local politicians and bankers joined the landowners to form ‘growth
coalitions’ that lobbied federal and provincial governments for improved infrastruc-
ture.” These governments were soon “won over by powerful interest groups of landown-
ers whose agendas involved agricultural expansion at the expense of forests.”
There are nevertheless important regional differences (FAO 2001). In Africa, much
deforestation (around 60%) is due to the conversion of forest for the establishment
of small-scale permanent agriculture, whereas direct conversion of forest cover to
large-scale agriculture, including raising livestock, predominates in Latin America and
Asia (48% and 30%, respectively). As well as directly causing forest degradation and
loss, many large-scale resource-extractive activities, such as timber harvesting, min-
ing, ranching, and plantations, initially open up previously inaccessible forested fron-
tier areas to permanent agricultural conversion (Wunder 2003, 2005; Barbier 2005b;
Wassenaar et al. 2007). Small-scale farmers usually follow because forest and other
land are now available and more accessible for conversion (Walker 2003; Verburg et al.
2004).2
2 Wassenaar et al. (2007, 101) note that “Amazonian cropland expansion hot spots in Brazil and
Bolivia for example are adjacent to current large soybean production zones, the creation of which,
Dualistic frontier expansion also promotes boom and bust cycles of economic devel-
opment (Wunder 2003, 2005; Barbier 2005a, 2005b, 2007 and 2011; Ha and Shively
2008; Agergaard et al. 2009; Barney 2009; Rodrigues et al. 2009; Hall 2011; Knudsen
and Folds 2011). State-sponsored promotion of commercial activities often ensures
that frontier expansion occurs rapidly and generates growth in marketable outputs.
However, this initial “economic boom” is invariably short-lived. Once the frontier is
“closed” and the valuable land and natural resources have been fully exploited or con-
verted, some economic retrenchment is inevitable. Under certain conditions, the
“bust” may start even before profitable frontier opportunities are exhausted.3 Such
boom and bust cycles associated with rapid frontier expansion are further exacerbated
if the commercial activities are isolated enclaves, as any production and profits gener-
ated will have limited impacts on economy-wide investment, innovation and growth.
The short-term windfall benefits of a commodity price rise will further reinforce this
outcome. In addition, during the expansion phase, commercial activities may generate
employment opportunities for unskilled labor and off-farm work on the frontier, but
with the inevitable bust and contraction, marginal land expansion once again becomes
the main outlet for absorbing the rural poor. As cultivation of such lands generates little
rents and productivity gains, economic livelihoods and incomes are not improved sig-
nificantly in the long run.
Such boom and bust patterns of frontier expansion have occurred for cocoa, coffee,
oil palm and shrimp in Southeast Asia and Africa (Ha and Shively 2008; Agergaard et al.
2009; Barney 2009; Hall 2011; Knudsen and Folds 2011). Oil price booms have interacted
with agricultural expansion and deforestation in a range of tropical countries, but with
only short-lived economy-wide gains (Wunder 2003, 2005). Long-run agricultural land
expansion and oil and natural gas proved that reserve expansion appear to be associated
with boom and bust cycles in a number of low- and middle-income countries (Barbier
2007). Finally, a study of 286 municipalities in the Brazilian Amazon found a consistent
boom and bust pattern in levels of human development (Rodrigues et al. 2009). Relative
standards of living, literacy, and life expectancy increase initially as forest conversion for
cattle ranching, logging, and agriculture proceed. However, these improvements appear
to be transitory; development levels decline in the postfrontier municipalities to lev-
els similar to those in prefrontier municipalities. As the authors conclude, “this ‘bust’ is
likely to reflect the exhaustion of the natural resources that supported the initial ‘boom’,
compounded by the increasing human population. Accordingly, per capita timber,
largely driven by increasing animal feed needs, has caused large scale deforestation in the recent
past.” Walker (2003) describes a similar process linking the road building by loggers in the Brazilian
Amazon and the subsequent “infilling” of the landscape by smallholder migrants. Barbier (2005b)
and Wunder (2003, 2005) provide numerous case studies of the links between mineral, energy, and
timber developments across the tropics and initially opening inaccessible frontier areas for subsequent
agricultural conversion.
3 For an economic model of such a boom and bust pattern of economic development in a
resource-dependent small open economy, see Barbier (2005a, 2005b).
cattle and crop production also exhibit boom-and-bust patterns across the deforesta-
tion frontier” (Rodrigues et al. 2009, 1436).
A number of important research issues emerge from this review of land use and dual-
istic frontier economic conditions in developing countries. First, this chapter points to
the need for better data on the geographical location of the rural poor. We require more
reliable data on the distribution of populations and poor households in least favored and
ecologically fragile areas in developing countries and more long-term monitoring of the
economic livelihoods of such populations. Second, such evidence would assist greatly in
testing two important hypotheses that emerges from this review: first, whether the pool
of surplus labor on marginal lands is essentially a barometer of frontier-based devel-
opment in low- and middle-income economies, and second, whether dualistic frontier
expansion leads to boom and bust cycles of economic development. Finally, this chapter
has also shown that patterns of land use change in developing countries are fundamen-
tal to their overall economic development. Yet, very few studies examine this link more
closely. Hopefully, future economics research will take more seriously how land use
change may influence sustainable economic development in low- and middle-income
economies.
Acknowledgments
I am grateful for comments by JunJie Wu on an earlier draft of this chapter.
References
Agergaard, J., N. Fold, and K. Gough. 2009. Global-local interactions: Socioeconomic and spa-
tial dynamics in Vietnam’s coffee frontier. The Geographical Journal 175: 133–145.
Akpalu, W., and P. Parks. 2007. Natural resource use conflict: Gold mining in tropical rainforest
in Ghana. Environment and Development Economics 12: 55–72.
Aldrich, S., R. Walker, E. Arima, and M. Caldas. 2006. Land-cover and land-use change in the
Brazlian Amazon: Smallholders, ranchers, and frontier stratification. Economic Geography
82: 265–288.
Austin, G. 2007. Resources, techniques, and strategies south of the Sahara: Revising the factor
endowments perspective on African economic development, 1500–2000. Economic History
Review 1–38.
Barbier, E. 2005a. Frontier expansion and economic development. Contemporary Economic
Policy 23(2): 286–303.
Barbier, E. B. 2005b. Natural resources and economic development. Cambridge, UK: Cambridge
University Press.
Barbier, E. B. 2007. Frontiers and sustainable economic development. Environmental and
Resource Economics 37: 271–295.
Barbier, E. B. 2010. Poverty, development and environment. Environment and Development
Barbier, E. B. 2011. Scarcity and frontiers: How economies have developed through natural
resource exploitation. Cambridge, UK: Cambridge University Press.
Barney, K. 2009. Laos and the making of a ‘relational’ resource frontier. The Geographical
Journal 175: 146–159.
Barro, R. J., and X. Sala-I. Martin. 2004. Economic growth, 2nd ed. Cambridge, MA: MIT
Press.
Billington, R. 1966. America’s frontier heritage. New York: Holt, Rinehart and Winston.
Boucher, D., P. Elias, K. Lininger, C. May-Tobin, S. Roquemore, and E. Saxon. 2011. The root
of the problem: What’s driving tropical deforestation today? Cambridge, MA: Union of
Concerned Scientists.
Bridge, G. 2008. Global production networks and the extractive sector: Governing
resource-based development. Journal of Economic Geography 8: 389–419.
Browder, J., M. Pedlowski, R. Walker, R. Wynne, P. Summers, A. Abad, et al. 2008. Revisiting
theories of frontier expansion in the Brazilian Amazon: A survey of colonist farming popu-
lation in Rondônia’s post-frontier, 1992–2002. World Development 36: 1469–1492.
Bromley, D. 2008. Resource degradation in the African commons: Accounting for institutional
decay. Environment and Development Economics 13: 539–563.
Bulte, E., R. Damania, and R. López. 2007. On the gains of committing to inefficiency: Corruption,
deforestation and low land productivity in Latin America. Journal of Environmental
Economics and Management 54: 277–295.
Bunker, S. 1984. Modes of extraction, unequal exchange, and the progressive underdevelop-
ment of an extreme periphery: The Brazilian Amazon, 1600–1980. American Journal of
Sociology 89: 1017–1064.
Bunker, S. 1989. Staples, links, and poles in the construction of regional development theories.
Sociological Forum 4: 589–610.
Campbell, B. (ed.) 2009, Mining in Africa: Regulation and development. London: Pluto Press.
Carr, D. 2009. Population and deforestation: Why rural migration matters. Progress in Human
Geography 33: 355–378.
Casetti, E., and H. Guathier. 1977. A formalization and test of the “hollow frontier” hypothesis.
Economic Geography 53: 70–78.
Caviglia-Harris, J., and D. Harris. 2008. Integrating survey and remote sensing data to analyze
land use scale: Insights from agricultural households in the Brazilian Amazon International
Regional Science Review 31: 115–137.
Chen, S., and M. Ravallion. 2007. Absolute poverty measures for the developing world, 1981–
2004. Proceedings of the National Academy of Sciences of the USA 104(43): 16757–16762.
Chomitz, K., P. Buys, G. De Luca, T. Thomas, and S. Wertz-Kanounnikoff. 2007. At logger-
heads? agricultural expansion, poverty reduction, and environment in the tropical forests.
Washington, DC: The World Bank.
Comprehensive Assessment of Water Management in Agriculture. 2007. Water for food, water
for life: A comprehensive assessment of water management in agriculture. London: Earthscan
and International Water Management Institute, Colombo, Sri Lanka.
Coxhead, I., G. Shively, G. Shuai, and X. Shuai. 2002. Development policies, resource con-
straints, and agricultural expansion on the Philippine land frontier. Environment and
Development Economics 7: 341–363.
Curry, G., and G. Koczberski. 2009. Finding common ground: Relational concepts of land
tenure and economy in the oil palm frontier of Papua New Guinea. Geographical Journal
175: 98–111.
DeFries, R., T. Rudel, M. Uriarte, and M. Hansen. 2010. Deforestation driven by urban
population growth and agricultural trade in the twenty-first century. Nature Geoscience
3: 178–801.
Dercon, S., D. O. Gilligan, J. Hoddinott, and T. Woldehanna. 2009. The impact of agricultural
extension and roads on poverty and consumption growth in fifteen Ethiopian villages.
di Tella, G. 1982. The economics of the frontier. In Economics in the long view , eds. C.
Kindleberger and G. di Tella, 210–227. London: Macmillan.
Etter, A., C. McAlpine, and H. Possingham. 2008. Historical patterns and drivers of landscape
change in Colombia since 1500: A regionalized spatial approach. Annals of the Association
of American Geographers 98: 2–23.
Findlay, R., and M. Lundahl. 1994. Natural resources, “vent-for-surplus,” and the staples theory.
In From classical economics to development economics: Essays in honor of Hla Myint, ed. G.
Meier, 68–93. New York: St. Martin’s Press.
Fischer, G., and G. K. Heilig. 1997. Population momentum and the demand on land and water
resources. Philosophical Transactions of the Royal Society Series B 352(1356): 869–889.
FAO. 2001. Forest resources assessment 2000: Main report. FAO Forestry Paper 140. Rome: Food
and Agricultural Organization.
FAO. 2003. State of the world’s forests 2003. Rome: Food and Agricultural Organization.
FAO. 2006, Global forest resources assessment 2005, main report. Progress towards sus-
tainable forest management. FAO Forestry Paper 147. Rome: Food and Agricultural
Organization.
Foweraker, J. 1981. The struggle for land: A political economy of the pioneer frontier in Brazil from
1930 to the present day. Cambridge, UK: Cambridge University Press.
Gibbs, H. K., A. S. Ruesch, F. Achard, M. K. Clayton, P. Holmgren, N. Ramankutty, and J.
A. Foley. 2010. Tropical forests were the primary sources of new agricultural lands
in the 1980s and 1990s. Proceedings of the National Academy of Sciences of the USA
107: 16732–16737.
Gould, K., D. Carter, and R. Shrestha. 2006. Extra-legal land market dynamics on a Guatemalan
agricultural frontier: Implications for neoliberal policies. Land Use Policy 23: 408–420.
Gylfason, T. 2001. Nature, power, and growth. Scottish Journal of Political Economy
48: 558–588.
Ha, D., and G. Shively. 2008. Coffee boom, coffee bust and smallholder response in Vietnam’s
central highlands. Review of Development Economics 12: 312–326.
Hall, D. 2011. Land control, land grabs, and Southeast Asian crop booms. Journal of Peasant
Studies 38: 837–857.
Hansen, B. 1979. Colonial economic development with unlimited supply of land: A Ricardian
case. Economic Development and Cultural Change 27(4): 611–627.
Hecht, S. 2005. Soybeans, development and conservation on the Amazon frontier. Development
and Change 36: 375–404.
Hirsch, P. 2009. Revisiting frontiers as transitional spaces in Thailand. Geographical Journal
175: 124–132.
Hyndman, D. 1994. A sacred mountain of gold: The creation of a mining resource frontier in

Papua New Guinea. Journal of Pacific History 29: 203–221.
James, P. 1969. Latin America, 4th ed. New York: Odyssey Press.
Jepson, W. 2006. Producing a modern agricultural frontier: Firms and cooperatives in eastern
Mato Grasso, Brazil. Economic Geography 82: 289–316.
Killeen, T., V. Calderon, L. Soria, B. Quezada, M. Steininger, G. Harper, et al. 2007. Thirty years
of land-cover change in Bolivia. Ambio 36: 600–606.
Knudsen, M., and N. Folds. 2011. Land distribution and acquisition practices in Ghana’s cocoa
frontier: The impact of a state-regulated marketing system. Land Use Policy 28: 378–387.
Maertens, M., M. Zeller, and R. Birner. 2006. Sustainable agricultural intensification in forest
frontier areas. Agricultural Economics 34: 197–206.
McCarthy, J., and R. Cramb. 2009. Policy narratives, landholder engagement, and oil palm
expansion on the Malaysian and Indonesia frontiers. Geographical Journal 175: 112–123.
Morton, D. C., R. S. DeFries, Y. E. Shimabukuro, L. O. Anderson, E. Arai, F. del Bon Espirito-Santo,
et al. 2006. Cropland expansion changes deforestation dynamics in the southern Brazilian
Amazon. Proceedings of the National Academy of Sciences of the USA 103: 14637–14641.
Mosley, W. 2005. Global cotton and local environmental management: The political ecology
of rich and poor small-holder farmers in southern Mali. Geographical Journal 171: 36–55.
Mueller, B. 1997. Property rights and the evolution of the frontier. Land Economics 73: 42–57.
Pichón, F. 1997. Colonist land-allocation decisions, land use, and deforestation in the
Ecuadorian frontier. Economic Development and Cultural Change 45: 707–744.
Population Division of the United Nations Secretariat. 2008. World urbanization prospects: The
2007 revision: Executive summary. New York: United Nations.
Ramankutty, N., and J. A. Foley. 1999. Estimating historical changes in global land
cover: Croplands from 1700 to 1992. Global Biogeochemical Cycles 13: 997–1027.
Rodrigues, A., R. Ewers, L. Parry, C. Souza, A. Verissimo, and A. Balmford. 2009. Boom-and-bust
development patterns across the Amazonian deforestation frontier. Science 324: 1435–1437.
Rudel, T. 2007. Changing agents of deforestation: From state-initiated to enterprise driven pro-
cess, 1970–2000. Land Use Policy 24: 35–41.
Schmook, B., and C. Vance. 2009, Agricultural policy, market barriers, and deforestation: The
case of Mexico’s southern Yucatán. World Development 37: 1015–1025.
Sills, E., and J. Caviglia-Harris. 2008. Evolution of the Amazonian frontier: Land values in
Rondônia, Brazil. Land Use Policy 26: 55–67.
van der Ploeg, R. 2011. Natural resources: Curse or blessing? Journal of Economic Literature
49: 366–420.
Verburg, P., K. Overmers, and N. Witte. 2004. Accessibility and land-use patterns at the forest
fringe in the northeastern part of the Philippines. Geographical Journal, 170: 238–255.
Walker, R. 2003. Mapping process to pattern in the landscape change of the Amazonian fron-
tier. Annals of the Association of American Geographers 93: 376–398.
Walker, R., J. Browder, E. Arima, C. Simmons, R. Pereira, M. Caldas, et al. 2009. Ranching and
the new global range: Amazônia in the 21st century. Geoforum 40: 732–745.
Wassenaar, T., P. Gerber, P. H. Verburg, M. Rosales, M. Ibrahim, and H. Steinfeld. 2007.
Projecting land use changes in the neotropics: The geography of pasture expansion into for-
est. Global Environmental Change 17: 86–104.
Williamson, J. 2002. Land, labor and globalization in the Third World, 1870–1914. Journal of
Economic History 62: 55–85.
Williamson, J. 2006. Globalization and the poor periphery before 1950. Cambridge, MA: MIT
Press.
Wood, C. H. 1983. Peasant and capitalist production in the Brazilian Amazon: A conceptual
framework for the study of frontier expansion. In The dilemma of Amazonian development,
ed. E. F. Moran, 259–277. Boulder, CO: Westview Press.
World Bank. 2003. World development report 2003. Washington, DC: The World Bank.
World Bank. 2008. Word development indicators 2008. Washington, DC: The World Bank.
Wunder, S. 2003. Oil wealth and the fate of the forest: A comparative study of eight tropical coun-
tries. London: Routledge.
Wunder, S. 2005. Macroeconomic change, competitiveness and timber produc-
tion: A five-country comparison. World Development 33: 65–86.
PA R T I I
E N V I RON M E N TA L
AND
S O C IOE C ON OM IC
C ON SE QU E N C E S OF
L A N D U SE A N D L A N D
U SE C HA N G E
C HA P T E R 7
THE ECONOMICS OF
W I L D L I F E C O N S E RVAT I O N
DAV I D J. L EW I S A N D E R I K N E L S ON
Wildlife populations have been adversely impacted by a multitude of human activi-

ties, although most ecologists argue that the clearing of forests and grasslands for urban
areas and agriculture has had the greatest impact (Sala et al. 2000; Wilcove et al. 2000;
Millennium Ecosystem Assessment [MEA] 2005). The economic argument for conserv-
ing wildlife is largely a public goods argument. A private landowner lacks the incentive
to provide adequate habitat for wildlife species because much of the use value (hunt-
ing, bird-watching, ecosystem service regulation, etc.) and nonuse value (existence
of species) produced on the landowner-provided habitat will accrue to other people.
Therefore, government policies or nongovernmental organization (NGO) programs
that encourage the conservation of wildlife habitat may improve the efficiency of land
use patterns on landscapes.
This chapter covers several economic issues pertinent to wildlife conservation efforts.
Wildlife conservation activities include climate change mitigation, limits on freshwa-
ter withdrawals from watersheds, and efforts to reduce the spread of invasive species.
However, the dominant wildlife conservation activity undertaken globally is the set-
ting aside of land to provide wildlife habitat. Here, we focus on unresolved economic
issues related to the three primary means of establishing habitat set-asides: government
regulation, direct appropriation or purchase of habitat by governments or NGOs, and
payments to landowners for voluntarily altering land use activities to be more wildlife
friendly. Rather than provide a comprehensive literature review, our approach is to pro-
vide an in-depth discussion of representative economic research related to these three
habitat conservation approaches. The research we review is selected to highlight what
we believe are some of the outstanding economic issues in wildlife conservation that
deserve future research attention. Our main arguments are illustrated with several sim-
ple extensions to prior studies.
Government regulation is one approach to conserving habitat and is typified by the
US Endangered Species Act (ESA). Under the ESA, a species determined to be at risk of
164 David J. Lewis and Erik Nelson
extinction is listed and afforded regulatory protection. For example, the ESA generally
gives the US government the authority to regulate timber harvesting if it is expected
that unmitigated harvest activity would threaten the persistence or habitat of a listed
species. We provide a simple extension to previous theoretical models to show that
regulatory designs similar to the ESA can drive a wedge between privately and socially
preferred behavior. Furthermore, it can create cases in which society in general prefers
harming wildlife populations. Effective regulatory design must address the tensions that
approaches like the ESA can create between societal wildlife goals and individual prefer-
ences. To that end, the ESA must integrate rigorous ex post evaluations of conservation
outcomes, and regulators must be willing to act on uncovered shortcomings.1
The direct purchase of habitat by governments and conservation organizations is an
alternative to government regulation of wildlife populations. The purchase of habitat for
set-asides can take several forms. For example, Norway has bought and retired Peruvian
government debt in exchange for the establishment of a reserve area in Peru (Hansen
1999). In fiscal year 2010, The Nature Conservancy spent $204 million on the purchases
of conservation land and easements across the globe (The Nature Conservancy [TNC]
2010). At the heart of direct purchase programs is the problem of selecting which land
to purchase when conservation funds are scarce and not all desirable habitat can be
protected. The literature devoted to finding the best use of funds for some biological
objective has been termed “reserve-site selection (RSS)” or “systematic conservation
planning (SCP)” and has been developed by both economists and conservation biolo-
gists. Recent efforts to more accurately measure the biological benefit created by reserve
networks have been dubbed return-on-investment (ROI) for conservation. We develop
a new US-wide reserve selection model and use it to argue that existing reserve selection
approaches must (1) properly specify the conservation benefits from a reserve system
and (2) incorporate realistic expectations of landscape dynamics outside of the selected
network.
The final approach to setting aside habitat is to offer voluntary payments to landown-
ers to alter their land use practices. This approach is typified by Costa Rica’s 1996 national
forest law and the US Wildlife Habitat Incentives Program, both of which pay landown-
ers directly for improved habitat provision. Two dynamics make efficient design of vol-
untary payment programs difficult: landowners’ willingness to accept (WTA) payments
is private information, and habitat benefits are spatially dependent, meaning benefits
are a function of the spatial pattern of conservation across large landscapes of multiple
landowners. The configuration of conservation across a landscape is difficult for agen-
cies to control when WTA information is private because it is unclear ex ante which
landowners will accept payments. Furthermore, when benefits of habitat conservation
are spatially dependent, it is difficult for agencies to identify ex ante the benefits that
1 Here we ignore another major type of government regulation associated with wildlife conservation,
the direct appropriation of land. For example, in 1982 the Uganda government evicted approximately
4,500 families from land that became Lake Mburo National Park (Emerton 1999).
The Economics of Wildlife Conservation 165
will result from a particular payment program. As such, we develop a simple example to
argue that efficient conservation of wildlife with incentives must overcome the problem
of eliciting private information on landowners’ WTA. New empirical evidence from the
state of Oregon is used to illuminate the importance of WTA information and to illus-
trate the large efficiency gains from solving this information problem.
1. Command-and-Control Approach and

the US Endangered Species Act
The original version of the ESA, passed in 1973, prohibited an individual, corporation,
or government agency from killing or destroying the habitat of a species listed under
the Act (a “taking”).2 According to the law’s original language, the imperative of sav-
ing endangered public goods trumped the private economic interests of landowners
(McAnaney 2006). Therefore, just like the original versions of the US Clean Air and
Clean Water Acts, the initial version of the ESA was a command-and-control policy
with little regulatory flexibility and no compensation for landowner economic losses
due to regulatory actions. Since 1978, however, the ESA has been amended several times
and has become a more flexible or permissive policy than its original incarnation, espe-
cially when dealing with habitat on private land (Scott et al. 2006).
The ESA’s private land policies are vital to the success and cost of the Act because data
suggests that more than half of all listed species have at least 80% of their habitat on pri-
vate property (Innes and Frisvold 2009). One example of this increased regulatory flex-
ibility is the availability of permits that allow landowners or developers to destroy listed
species or its habitat as long as the applicant can convince the permitting wildlife agency
that the “take” will not appreciably reduce the species’ likelihood of recovery. In many
cases, incidental take permits are only granted if the applicant agrees to install conser-
vation measures somewhere on their land or contribute to a general conservation fund
(Thompson 2006). Landowner activities necessary to acquire an incidental take permit
are laid out in a Habitat Conservation Plan (HCP).
In Section 1.1, we extend Polasky and Doremus’s (1998) model of landowner-wildlife
agency relationships to consider incidental take permits and HCPs and we argue that
command-and-control regulation for wildlife conservation can create situations in
which both individuals and society prefer harming wildlife populations. In Section
1.2, we review evidence of the effectiveness of the ESA and argue that it is reason-
able for society to expect robust and recovering wildlife populations to result from
2 An area is defined as habitat for a species if the species has been observed feeding or breeding on
that land in the immediate past (Lueck and Michael 2003).
command-and-control regulation, given the well-documented welfare losses associated

with such policy approaches. Unfortunately, evidence of the efficacy of the ESA is mixed.
1.1 Landowner–Wildlife Agency Relationships Under the ESA

Polasky and Doremus (1998) model the interplay between a landowner who is contem-
plating developing her land and an endangered species regulating agency, in which the
agency is uncertain whether the landowner’s plot contains listed species or their habi-
tats. Although Polasky and Doremus consider several potential relationships between
the landowner and agency, including payments for conservation, here we mention the
two cases that most resemble the relationship under the current version of the ESA. In
one case, the agency forces a landowner to set aside his or her land for conservation if
the agency can prove that the value to society of this action, given by S, is greater than
the private economic value that will accrue to the landowner after development, given
by D (land development costs are netted out of D). Assume S falls to 0 if the land is
developed (we will relax this assumption in a modification of the Polasky and Doremus
model below) and U indicates private returns from conservation (if any) where D > U.
In this case, the burden of determining S lies with the regulating agency. Assuming S
can only be calculated by inspection of the land and that private property holders have
the right to prevent any inspections, the landowner has no incentive to allow agency
representatives on his land. This blanket refusal of inspection, although always privately
optimal given that D > U, may be inefficient from society’s perspective. Before develop-
ment, a survey would be warranted from society’s perspective if the expected net social
benefit generated by paying for information on S exceeds the social benefits generated
without the information,
(1 − p)(S + U ) + pD − C > D ⇒ (1)
(1 − p)(S + U − D) > C , (2)
where 1−p is the probability that the survey will find S ≥ D and C indicates the cost of the
survey.
Under another relevant landowner-agency relationship framework explored by
Polasky and Doremus, the landowner must prove D > S before he or she can develop, or
otherwise pay a development fine F where F > D. First, the landowner will never develop
without a survey, otherwise private net returns will be negative (D−F < 0). Therefore,
a utility-maximizing landowner will commission a survey before development if the
expected net private benefit of doing so outweighs the private benefit of not doing so,
pD + (1 − p)U − C > U ⇒ (3)

p(D − U ) > C , (4)
where p is the probability that the survey will find D > S. However, from society’s per-
spective, a survey is only welfare enhancing if it is expected to reveal that D is signifi-
cantly larger than S,
pD + (1 − p)(S + U ) − C > S + U ⇒ (5)
p(D − S − U ) > C. (6)
According to inequalities (4) and (6), the landowner is more likely to find it in his best
interest to survey than society would.3 In both of these cases, a wedge exists between pri-
vate and socially preferred behavior.
1.1.1 Landowner-Wildlife Agency Relationships with a Habitat

Conservation Plan (HCP)
The current version of the ESA differs from the species conservation policies considered
by Polasky and Doremus in several ways (also see Innes and Frisvold [2009] for a revi-
sion of the Polasky and Doremus model). First, a finding of a listed species or its habitat
on a parcel does not mean it cannot be developed; instead, it may only mean a restric-
tion on certain development activities in the parcel. Second, regulators are supposed to
limit habitat destruction on a parcel no matter the expected social value of the conserva-
tion behavior. Finally, a landowner can choose to limit some species harm and/or habi-
tat damage when developing in exchange for an incidental take permit.
We modify the Polasky and Doremus model to reflect the current agency-landowner
relationship. We assume a landowner knows that her land contains listed species or its
habitat but the agency does not necessarily know this. She can choose to fully develop the
land and risk regulatory penalties, take the steps necessary to gain an incidental take per-
mit, or not develop. Ex ante the private economic value generated by a developed parcel
with an incidental take permit is uncertain because the landowner does not know what
the regulating agency will require in exchange for a permit. An incidental take permit
will require the landowner to institute some conservation action or implement a land use
that results in less value than unfettered development. In addition, the landowner will
incur some HCP negotiation and implementation costs. Let 0 to N indicate the range of
private economic value generated on the parcel with a permit, less private permit trans-
action costs, where N < D. Let n indicate the expected private value of the developed par-
cel with an incidental take permit. Further, let t and w indicate the expected public and
3 Specifically, D has to be C/p + U units greater than S for private and social incentives to align.
private conservation value, respectively, generated by an HCP on the parcel. Because an

HCP allows for some development the conservation or nonmarket value of a parcel with
an HCP is not as great as the nonmarket values on an undeveloped parcel: t < S and w < U.
By seeking an incidental take permit the landowner signals to the regulator that she
has a listed species or habitat on her land. Even if the landowner does not signal the
presence of a listed species or its habitat, development action by a landowner could trig-
ger regulatory scrutiny and a judgment of a taking. Let pS indicate the probability that
the agency will become aware that a listed species does use or has used the parcel or the
parcel does contain or did contain a listed species’ habitat during parcel development.
In this initial setup, we will assume that pS is known to the parcel owner and it cannot be
affected by parcel owner behavior. In other words, pS will be determined by habitat dis-
tribution across space and the regulating agency’s competence, budget, and other fac-
tors.4 Again, let D indicate the private economic value of parcel development, where any
land development costs are netted out.
If the parcel owner develops and the agency becomes aware of a taking, then fine F is
levied and we assume the agency will enforce a redevelopment plan similar to the one
that would have been generated under an incidental take permit negotiation process.
Therefore, the profit-maximizing landowner chooses his development path according
to the following,
max{(1 − pS )D + pS (n + w − F ), n + w , U }. (7)

where the first term is the expected net private economic value associated with not
approaching the agency to cooperate on an HCP, 5 the second term is the expected net
private economic value of approaching the agency to cooperate on an HCP, and the
third term is the nonmarket return to the landowner from not developing her land (we
assume the private economic value of undeveloped land is 0). The utility-maximizing
landowner will approach the agency to cooperate on the design of an HCP if,
 p 
n≥ D −w − S  F (8)
 1 − pS 

and
n ≥ U − w. (9)
4
There is some question as to how aggressively the ESA actually enforces takings on private land. In
reality, pS may essentially be 0 for many private landowners.
5
We assume that w can be reached on a piece of land that was developed but then was forced to
institute some conservation due to the discovery of a taking. In reality, the private nonmarket benefit
on a parcel that was caught in a taking may not be reasonably restored to a nonmarket benefit level
associated with the use of an HCP from the beginning.
In words, equation (8) indicates the landowner will only come forward to develop
an HCP in conjunction with the agency if pS and F are large enough to bridge the gap
between D (the value of unfettered development) and n (the value of development with
an HCP). Monetary compensation for cooperating landowners would enter inequality
(8) on the left-hand side, making cooperation on an HCP more likely. For simplicity,
we assume from here on out that n is always larger than U−w, or the expected value
of development with an HCP is greater than the incremental private nonmarket value
from no development versus development with an HCP.
Conversely, landowner initiative on an HCP is socially efficient only if the social
returns of this decision are greater than expected social benefits of unfettered
development,
n + t + w − C ≥ (1 − pS )D + pS (n + t + w − C ) ⇒,
(10)
n ≥ D − t − w + C , (11)
and the social benefits of nondevelopment,
n + t + w − C ≥ S + U ⇒, (12)
n ≥ S + U − t − w + C , (13)
where C is the regulatory agency’s HCP finding, planning, and implementation costs.6
For simplicity, we assume from here on out that n is always larger than the incremental
benefit of not developing at all, plus the regulatory agency’s HCP planning and imple-
mentation costs (i.e., n > S + U − t − w + C). Therefore, social and landowner incentives
on landowner initiated
 pS   p 
HCPs are aligned when t − C =   F . Otherwise, if t − C > (<)  S  F , then
 1 − pS   1 − pS 
society is more likely (less likely) to prefer landowner initiative on HCPs than the private
landowner.
6
Development fine F is not a social cost, just a redistribution of funds. We assume that t can be
reached on a piece of land that was developed but then was forced to institute some conservation due to
the discovery of a taking. In reality, the public nonmarket value created by a parcel that was caught in a
taking may not be reasonably restored to a nonmarket value level associated with the use of an HCP from
the beginning.
1.1.2 The “Shoot, Shovel, and Shut Up” Incentive

As Polasky (2001), Lueck and Michael (2003), and others have pointed out, the prob-
ability of the regulatory agency detecting a taking can be lower than pS for several rea-
sons. For example, the parcel owner can attempt to prevent the wildlife agency from
gleaning information about his land prior to development by blocking access, or he can
destroy or alter potential habitat on his land (“shoot, shovel, and shut up”) prior to regu-
latory attention. Specifically, let pS be the landowner-influenced probability that the
wildlife agency will become aware of listed species or its habitat on the parcel in the pro-
cess of development, and c( ps , ps ) indicates the landowner’s cost of obtaining ps where
ps ≤ ps , c( ps , ps ) = 0 if ps = ps , and c( ps , ps ) increases as the landowner reduces ps . Let
p∗s indicate the ps that maximizes the expected value of full development on the plot.
Now the landowner’s problem is the same as (7) except the first term in (7) is subtracted
by c( ps , ps ) and the landowner determines the optimal “shoot, shovel up, and shot up”
behavior before solving the maximization function. The landowner will approach the
agency to design an HCP if,
 p∗  c( pS , pS∗ )
n≥ D−w+ S ∗ F − .
 1 − pS  1 − pS∗ (14)

Because of lower odds of a takings discovery, the fine F that may have been large
enough to convince the landowner to seek an HCP with exogenous ps (inequality [8]‌)
may not be high enough to engender the same reaction with endogenous p∗s ; it will
c( pS , pS∗ )
depend on the size of . Again, the inclusion of landowner compensation in
1 − pS∗
an HCP would make conservation cooperation much more likely because the left-hand
side of inequality (14) would be larger.
Finally, we can show that, under certain conditions, privately optimal “shoot, shovel,
and shut up” behavior under the ESA, given by p∗s , can generate higher net social ben-
efits than when the landowner does not influence pS. Ex ante society will prefer “shoot,
shovel up, and shut up” behavior on the part of the landowner if it is expected to generate
more in net social benefits than not engaging in it,
(1 − pS )D + pS (n + t + w − C )
 
(1 − pS∗ )D + pS∗ (n + t +w − C ) − c( pS , pS∗ ) ≥ L ×  n+t +w −C  (15)
 S + U 
 
where L is a 1 x 3 vector that has a value of 1 in the first element if the solution to prob-
lem (7) is development, has a value of 1 in the second element if the solution to problem
(7) is an HCP, or has a value of 1 in the third element if the solution to problem (7) is no
development. Further, the two elements that are not equal to 1 are equal to 0. If,
(1 − pS∗ )D + pS∗ (n + t + w − C ) − U − S ≥ c( pS , pS∗ ), (16)

then inequality (15) holds for all permutations of vector L and “shoot, shovel up, and
shut up” behavior unconditionally generates higher net social benefits than having the
landowner not influence PS.7 In other words, the lower that pS∗ can be driven at a rea-
sonable cost, and the higher that the unfettered development value is compared to the
social returns from an HCP, the more likely it is that optimal “shoot, shovel, and shut up”
behavior is preferred by both the landowner and society in general.
To summarize, there are two main points from this section. First, under the
current version of the ESA, the regulating agency can encourage conservation
cooperation by levying high fines for a taking by the landowner (or compensating
landowners for lost private economic value). However, there is a point at which the
fine becomes too large from society’s point of view because it encourages the devel-
opment of an HCP that generates less in expected net social benefit than an unco-
operative landowner. Second, because the social benefits of an HCP can be small
compared to the value of development, net social benefits can be higher when the
landowner reduces the odds of finding an HCP optimal or being punished for
avoiding one (“shoot, shovel, and shut up”). The fact that net social benefits can be
higher with such perverse landowner behavior than without it highlights the mis-
alignment of private, social, and regulatory incentives under the current version of
the ESA.
1.2 How Effective Is the ESA?

Despite these incentive compatibility issues on private land, whether the Act as cur-
rently constituted is, as a whole, creating more societal benefit than social cost is an
open question and can only be determined by adding up all the market and nonmarket
values created by the regulation and comparing these to the sum of the opportunity
costs generated by the Act’s restrictions (Rachlinski 1997). However, this monumen-
tal cost-benefit analysis (CBA) has not yet been undertaken by researchers. Given the
difficulty in accurately monetizing nonmarket benefits, a CBA of the entire ESA may
be impossible. An alternative measure of regulatory success is given by progress on
regulatory goals. And, in cases where cost of achieving these goals can be monitized,
cost-effective goal achievement would mean meeting goals at least cost (Shogren et al.
1999; Naidoo et al. 2006).
The goal of ESA regulators is to list species that might go extinct without intervention
and then take actions such that these species can eventually be taken off the list due to
sufficiently reduced extinction probabilities. Up to now, the ESA has failed miserably
7 There are other contingent cases in which it is socially preferable for the landowner to engage
in “shoot, shovel up, and shut up” behavior. Inequality Equation (15) also always holds if unfettered
development or an HCP solves problem (7) and (1 − pS∗ )(D + C − n − t − w ) > c( pS , pS∗ ) . Inequality
(15) also holds if an HCP solves problem (7)and ( pS − pS∗ )(D + C − n − t − w ) > c( pS , pS∗ ) . Contact
author Nelson at [email protected] or http://www.bowdoin.edu/faculty/e/enelson/ for a more
detailed proof.
on this goal. As of June 2012, 2,000 animals and plant species8 were listed as endangered
or threatened (607 of these species inhabit ranges completely outside of US territories).
Since 1973, only 21 species have been delisted due to recovery (US Fish and Wildlife
Service [FWS] 2009).
Of course, the lack of recovered species does not mean that the Act has not had ben-
eficial effect. Some have argued that many more listed species would have gone extinct
without regulatory coverage (e.g., Schwartz 1999). It could also be that recovery suf-
ficient for a delisting takes several generations of regulatory attention. If so, short-term
progress toward delisting could be measured by change in the status of species over time,
a metric tracked by the US Fish and Wildlife Service (FWS) (Rachlinski 1997; Male and
Bean 2005; Kerkvliet and Langpap 2007). If we assign a 1 to species whose population is
in decline, a 2 to species whose population is stable, and a 3 to species whose population
is improving or recovered, then the average status score across 255 listed vertebrates was
1.71 in 1990, 1.74 in 1994, 1.75 in 1998, and 1.68 in 2002 (Kerkvliet and Langpap 2007).
This trend seems to suggest that ESA protection has done little to improve the overall
status of these 255 species.
Other than some landowners having incentive to reduce the persistence probabilities
of listed species (see argument in Section 1.1), scarce progress on delisting could also be
explained by too little spending on listed species’ recovery activities (Miller et al. 2002).
There is evidence that increased spending on listed species’ recovery activities does pro-
mote progress toward delisting. For example, Kerkvliet and Langpap (2007) find that
increased spending on a species is correlated with a lower likelihood that the FWS will
classify that species’ status as extinct or declining. However, the direction of causal-
ity is unclear: does increased spending lower the risk of extinction, or is more money
being directed to species that are less likely to go extinct? Taylor et al. (2005) argue that
increased recovery spending is likely to promote delisting because the activities that they
found most explain species’ progress towards delisting—published recovery plans, des-
ignated critical habitat, length of time listed, and the like—are positively correlated with
more recovery spending, all else equal. Further, Ferraro et al. (2007) find that, on average,
the conservation status of listed species with substantial recovery funding has improved
over time compared to the contemporaneous conservation status of species with similar
characteristics that are only candidates for listing and therefore are not subject to ESA
protections and recovery funding. Provocatively, Ferraro et al. also find that the average
conservation status of listed species with little or no recovery funding has deteriorated
overall compared to the average status of similar candidate species. Why unfunded regu-
latory protection could lead to worse outcomes than no protection at all is still a matter
of conjecture. Some argue that this trend can in part be explained by the incentives that
private landowners have to engage in “shoot, shovel, and shut up” behavior (Ruhl 1998).
8 Some listed species are actually subspecies, whereas others are distinct populations of species (e.g.,
gray wolf populations in the northern Rockies versus Great Lakes). Here, we refer to all listed entities as
“species.”
If it is true that “shoot, shovel, and shut up” behavior mainly impacts lightly funded spe-
cies then then this would suggest that better funded species are more closely monitored
and tracked on private land and this deters landowners from destroying the habitat of
the better funded species.
Presuming ESA funding will never be great enough to implement all or even most rec-
ommended listed species’ recovery activities, an endangered species-regulating agency
has two reasonable constrained maximization objectives to choose from. One approach
would be to spend recovery funds to maximize the number of species that are delisted
(Mann and Plummer 1995). In this case, recovery funds would be directed toward spe-
cies that could conceivably recover enough for delisting with limited funding. This
choice likely would leave little money for other listed species and, therefore, could lead
to an increased listed species extinction rate. An alternative approach would be to dis-
tribute recovery funds such that the sum of increase in persistence probabilities across all
listed species is maximized. For many researchers, this is the definition of cost-effective
biological conservation (e.g., Possingham et al. 2002; Polasky et al. 2008). Although this
approach may not lead to many delistings, it should limit the number of extinctions.
Figure 7.1 illustrates both approaches.
Is there any evidence to suggest that either of these two constrained maximization objec-
tives have been adopted by the FWS and other ESA regulatory agencies? Recovery fund-
ing is unequally distributed across listed species (see Figure 7.2), so there does appear to
be some pattern to funding. Cash (2001) does find that species that are considered by sci-
entists more likely to recover have received more in recovery funding, all else equal. If we
assume that these types of species are like species A in Figure 7.1—recovery curves that
increase rapidly and meet the delisting criteria with limited funding—then this observed
funding pattern supports an effort to prioritize delisting of a few species. However, at the
same time, Cash (2001) also finds that species whose recovery is more likely to cause con-
flict with economic development goals have received more in funding, all else equal. Such a
funding pattern is at odds with the basic tenants of cost-effective goal achievement. Metrick
and Weitzman (1998) suggest that there is a strong preference among regulatory agencies
for funding the recovery of charismatic species above and beyond what is warranted by
recovery science. Such a funding pattern is consistent with the political economy story that
regulators attempt to curry emotional support for the Act from the US public rather than
demonstrate efficiency. In addition, the allocation of up to 75% in recovery funds has been
dictated by line items in appropriations legislation from Congress (Miller et al. 2002), and
listed species’ funding has been shown to depend on whether their geographic range falls
within political districts represented by Congressional representatives on the Department
of Interior Subcommittees (Cash 2001; DeShazo and Freeman 2003, 2006).
1.3 Discussion
Forty years after its passage, opinion on the effectiveness and the net returns created
by the ESA vary greatly. In 2003, then Assistant Secretary of the US Department of the
A
Delisting
threshold
B
Persistence probability
XB XC XD XA X Recovery funding
FIGURE 7.1 Potential recovery funding distribution strategies across listed species. Assume
there are four listed species, named A, B, C, and D. Each curve represents how a species
responds to recovery funding where the height of the curve indicates the species’ indefinite
persistence probability. In this case, the marginal persistence value of recovery funding is
diminishing across the entire range of funding. (In some conservation contexts, the persis-
tence probability curves may initially increase in recovery funding and, after some thresh-
old, begin to decrease in recovery funding; see Lamberson et al. 1992 and Wu et al. 2000.)
When persistence probability becomes high enough, a species is delisted. In this case, even
with an unlimited budget, the agency could only fund the delisting of two species, A and
B. Here, assume the wildlife agency only has X dollars to spend on listed species recovery
activities. Suppose X, if entirely spent on species A’s recovery, would be just enough to fund
its delisting. If the agency’s objective is to maximize the number of species delisted, it will
provide X in recovery spending for species A. If the agency’s objective is to fund as much of
an increase in aggregate persistence probability as possible, it will give to species such that
the marginal persistence value for each species is the same and the budget is exhausted. In
this illustrative example, this occurs at the funding levels xA, xB, xC, and xD, where xA + xB
+ xC + xD = X.
Interior Craig Manson “said the 30-year-old environmental law is ‘broken’ and should
no longer be used to give endangered plants and animals priority over human needs.”9
Manson argues that the Act does not give regulators enough flexibility to balance eco-
nomic and environmental tradeoffs. In addition, the listing process has been embroiled
in lawsuits over the past decade. Environmental groups that have brought the law-
suits argue that the US government is not fulfilling its regulatory obligation to list all
9 Julie Cart, “Species protection act ‘broken’: A top interior officer says the law should be revised
to give economic and other interests equal footing with endangered animals and plants,” L.A. Times,
November 14, 2003.
1,400
1,200
1,000
Millions of $
800
600
400
200
0
1 200 400 600 800 1000 1200 1400
Number of species
FIGURE 7.2 Cumulative recovery funding by US federal and state agencies across listed spe-
cies in fiscal year 2009 (not including land acquisition costs). Listed species are arraigned on
the x-axis in order of recovery funding. The top 10 and 50 listed species in terms of fiscal
year 2009 recovery funding received 34% and 85%, respectively, of all spending that year
(US FWS 2009).
endangered species. Recently, the US government has, in response to environmental

group pressure, agreed to decide whether listing is appropriate for 757 additional species
by 2018 (pending approval by a federal judge).10 Advocates of the Act, such as the Center
for Biological Diversity, argue that the Act is essential and deserves strengthening.
Recent estimates indicate that US urban area will increase by 33 million hectares from
2001 to 2052 (Radeloff et al. 2012). Climate models predict accelerated climate change
in the lower 48 states, which has the potential to drastically alter habitat and species
geographic ranges on a large scale (Lawler et al. 2009). Whether the ESA—and similar
command-and-control regulatory approaches to species protection—can be effective in
a rapidly developing and evolving landscape is questionable. To work, the ESA will need
to provide landowners with a stronger incentive to conserve habitat than what is cur-
rently in place. We show that a landowner compensation system (or strongly enforced
fine system) is one approach to providing this incentive under the ESA’s current inci-
dental take system. However, the more incentive that landowners are given to cooperate
with authorities on listed species conservation, the more likely that net benefits to soci-
ety will decrease if the decision to regulate private land activities is not a function of the
opportunity cost of conservation. Our second argument is that any wildlife conservation
program like the ESA must be subject to some type of ex post evaluation and adjustment
10 Matthew Brown, “Deal struck to protect imperiled plants, animals,” July 12, 2011, Associated Press.
to compensate for some of the efficiency losses generated by a command-and-control

policy. Such a process, however, requires researchers to develop appropriate counter-
factual scenarios regarding how a species would fare in the absence of being covered in
a conservation program. The recent econometric literature on program evaluation (e.g.,
Ferraro et al. 2007) has potential in this regard.
2. Purchasing Habitat for

Conservation with Complete Information:
Reserve-Site Selection
An alternative to government regulation of wildlife conservation is the purchase of

existing habitat from private landowners. For a government or conservation organi-
zation involved in buying habitat, a basic question is which land should be purchased
when conservation funds and/or available land is scarce. In this section, we present the
basic reserve site selection (RSS) problem and highlight two largely unresolved issues of
economic importance: (1) specifying a quantitative environmental benefit function and
(2) how to incorporate baseline outcomes in the absence of reserve siting. We highlight
these issues with a ROI approach using conservation siting across the United States.
Whereas species form the set of the decision units under the ESA, selecting a set
of undeveloped sites for habitat protection is the focus of RSS problems (RSS is often
referred to as systematic conservation planning in the conservation biology literature; see
Margules and Pressey 2000). In the rudimentary RSS problem, the social planner selects
a set of undeveloped sites to purchase such that the network of selected sites will provide
additional habitat for as many species as possible, given an area or habitat protection
cost constraint (Ando et al. 1998).
max s j ∑∑s x
i j
j ij
(17)

Subject to
∑c s
j
j j ≤B
(18)

s j ∈{0, 1}
(19)
where sj equals 1 if site j is selected for habitat protection and equals 0 otherwise, j indexes
all sites on the landscape, xij equals 1 if species i is known to use site j for breeding or
feeding activities, cj is the area of site j or cost of purchasing and maintaining or establish-
ing habitat on site j, and B is the social planner’s areal or monetary budget. If site j that
contains species x is selected, then the species is considered “covered” or represented by
the selected reserve network, and the objective function value increases by one. Solutions
to (17)–(19) typically include sites that are strongly complementary with one another in
terms of species composition, not necessarily the sites that contain the most species (e.g.,
two neighboring sites that contain many species may contain the same species, making
the selection of only one of the sites optimal). Because solving binary integer problems
over a large choice set can be computationally difficult, heuristic methods for solving
(17)–(19) and related problems have been developed. For example, a simulated annealing
heuristic that can approximate solutions to a problem like (17)–(19) has been codified in
a software package called MARXAN (Ball et al. 2009). MARXAN is a widely used in the
conservation planning community.
Early work on RSS formulated (17)–(19) as an area-constrained problem in which the
planner was constrained by total land area rather than budget (Camm et al. 1996; Church
et al. 1996; Dobson et al. 1997). Ando et al. (1998) relax the implicit assumption of uni-
form costs in the area-constrained problem and show that by setting cj equal to the expected
cost of purchasing an acre of habitat in county j and B equal to a conservation budget,
the same number of species can be covered for less aggregate cost than Dobson et al.’s
area-constrained solution to (17)–(19).
2.1 Issues in Reserve-Site Selection

Dobson et al. (1997) and Ando et al. (1998) assume that all species found in a selected
county would benefit from a representative protected area within the county. However,
this is unrealistic, given the size of counties and the disparate habitat preferences of spe-
cies located within a county. Therefore, over time, the rudimentary RSS problem (17)–
(19) has seen substantial refinement in representing the conservation benefits gained
by selecting a site. One approach is to reduce the size of potential sites j so that each
site contains uniform habitat (Haight et al. 2005; Polasky et al. 2008). Alternatively, the
objective function (17) can be respecified such that additional habitat contributes to
a more biologically complex metric. These more complex metrics include the sum of
individual species persistence probabilities (Polasky et al. 2008) and some variant of the
well-known biological species-area relationship (SAR) from the discipline of conser-
vation biology (Rosenzweig 1995). Common across all of these approaches is that the
biological score is more than just a sum of species covered by habitat area. Furthermore,
in many second-generation RSS problems, the biological metric is also a function of the
portions of the landscape that are not protected.11
Although all RSS problems assume that more habitat on a landscape increases the
value of the objective function, the rate and shape of objective value increase can
vary substantially across second-generation RSS problems. For example, Wu and
Boggess (1999) and Wu and Skelton-Groth (2002) argue that returns to resource
conservation (e.g., improving water quality for recreational purposes, adding habi-
tat to the landscape to increase biodiversity persistence) tend to display a “∫”-shape.
Under such an assumption, rapid increases in the benefits provided by additional
resource conservation only occur once a threshold of minimum conservation has
been reached; prior to that point, resource conservation only has a small effect on
objective. Polasky et al. (2008) use such a “∫”-shape when explaining species’ per-
sistence probabilities across the Willamette Basin of Oregon. Eventually, species
response to additional habitat becomes saturated. At fairly high levels of habitat on
the landscape, additional habitat is relatively worthless. Conversely, the SAR, which
specifies the number of species found on the landscape (richness) as a function of
habitat provision, is strictly convex in conservation. In SAR-based RSS, the first few
units of habitat on the landscape add the most to the biodiversity objective maxi-
mization. Therefore, using a “∫”-shape objective function versus a SAR curve in an
RSS problem over the same landscape with the same policy parameters can gener-
ate fairly different outcomes and/or solutions. In some cases, the same pattern of
habitat conservation is selected by both types of objective functions, but the gain in
the relevant biodiversity score will be much more impressive with the SAR objec-
tive function. Or, in other cases, the threshold effect in “∫”-shaped objective func-
tions will mean very different patterns of habitat conservation when compared to
the SAR-generated landscape. For example, the threshold for rapid increases in spe-
cies persistence is reached more quickly in Polasky et al. (2008) if the initial habitat
is clumped spatially on the landscape due to spatial dependencies in habitat value.
A SAR-based analysis of the same landscape would not necessarily reward habitat
clumping to the same degree.
Another issue that has seen recent attention is uncertainty across the RSS’s parame-
ters. For example, Haight et al. (2005) maximized expected species coverage given prob-
abilistic geographic range maps. Such an approach can account for probabilistic shifts
in species’ range due to ongoing climate change (Araújo et al. 2004; Pyke and Fischer
2005; Ando and Mallory 2012). In addition to biological uncertainty, some recent analy-
ses have accounted for uncertain opportunity costs of conservation (e.g., Nelson et al.
11 See Margules and Pressey (2000), Cabeza and Moilanen (2003), Moilanen et al. (2005), and
Newbold and Siikamaki (2009) for other examples of RSS problems with more biologically meaningful
objective functions.
2008; Carwardine et al. 2010; Lewis et al. 2011) and uncertainty in which habitat sites
are actually available for purchase.
Landscape dynamics have also been incorporated in RSS problems. Much of the
RSS literature assumes that sites not selected for protection will be lost to develop-
ment. Although this may be true in the long run, it is not true in the short run: valu-
able habitat not selected for protection can persist indefinitely if current market
conditions do not give the site’s owner incentive to develop. In fact, optimal dynamic
RSS analyses have shown efficiency gains by purchasing habitat sites that are imme-
diately threatened by development, even if they are not as biologically valuable as less
threatened sites (Costello and Polasky 2004). Conversely, biologically valuable highly
unlikely to be developed in the near future may be best left unprotected indefinitely
since there is limited expected value in protecting them immediately.12
2.2 An Empirical Illustration Using the Return on

Investment (ROI) Approach
Recently, biological conservation journals have published a number of articles on the
so-called ROI problem (e.g., Murdoch et al. 2007). The objective of the ROI problem is
generally the same as the RSS problem: maximize the return (in biological terms) per
unit investment (generally a conservation budget). However, ROI improves on sev-
eral features of the fundamental RSS problem. First, ROI is more explicit in incorpo-
rating baseline or business-as-usual threats to undeveloped area over time. Second, to
make this approach applicable to conservation organizations and their rapid funding
cycles, ROI approaches may select the sites that are expected to generate the greatest
ROI at each decision time step and not the trajectory of site selection that will maximize
ROI over the problem’s entire time frame. Third, in addition to habitat purchases, ROI
approaches can allocate effort across other conservation actions that can increase spe-
cies persistence in an area (e.g., invasive species removal, fire suppression, fuel reduc-
tion, etc.; Murdoch et al. 2007). Finally, the ROI approach acknowledges some of the
realities in conservation implementation, including the risk that purchased protected
areas and the species they host may be lost due to unforeseen events, species extinction
may reduce the biological value of a protected site in the future, conservation organiza-
tions with different objectives compete for the same sites (Bode et al. 2011), and that
funds are not fully fungible across regions.
In this section, we present an illustrative application that combines some of the con-
servation complications addressed in the dynamic RSS and ROI approaches with the
traditional RSS problem given by (17)–(19). Similar to the early work of Ando et al.
12 Examples of stochastic dynamic RSS are found in Costello and Polasky (2004), Snyder et al. (2005),
and Haight et al. (2005).
(1998), we solve equations (17)–(19) using the set of US counties for the establishment
of 1,000-hectare conservation reserves. Our dataset includes 1,066 vertebrate species
in the continuous US, with detailed range maps. The data used are fully described in
Withey et al. (2012). Finally, like most ROI literature, we assume the biological objective
is convex in habitat.
Our goal is to illustrate the effects of (1) diminishing “biological returns” to con-
serving land, (2) a baseline or business-as-usual future in which not all land is at risk
of development immediately, and (3) incorporating scale into a measure of biological
benefits on solutions to the rudimentary RSS problem. First, we incorporate diminish-
ing returns by favoring the selection of counties with less land already protected as of
the late 2000s (The Conservation Biology Institute [CBI] 2010). Define Pj as one plus
the proportion of county j protected, where higher Pj indicates greater existing protec-
tion (a completely unprotected county would have a score of 1 and a completely pro-
tected county would have a sore of 2). Second, the degree to which habitat is threatened
is accounted for by targeting counties with higher rates of expected future habitat losses.
Let Tj be a metric equal to one minus the proportion of natural land cover in county j
developed between 1992 and 2001 (Fry et al. 2009). Therefore, Tj varies between 0 and
1, where a lower value means that development of habitat has been rapid in the immedi-
ate past and presumably will continue to be intense in the near future. Finally, given the
large size of US counties, we account for the fact that a 1,000-hectare reserve is unlikely
to cover the range of relevant habitats within each county. We address this scale issue by
selecting counties that have relatively more homogeneous land cover. Define Dj to range
from 1 (high diversity of natural land cover types in county j) to 2 (no diversity of natu-
ral land cover types in county j) as of 2001 (Comer et al. 2003).
There are several ways we could add these selection criteria to the traditional
RSS problem. For example, weight wi could be added to objective function (17),
max s j ∑∑w s x
i j
ij j ij , such that species that range over counties with lower P, lower T,
and higher D values have higher w. Instead, our approach is to simply add constraints
to the RSS problem such that the selected network has average P and T values equal to
or less than 30th percentile values for P and T across all counties (1.0013 and 0.9748,
respectively) and an average D value equal to or greater than the 70th percentile value
for D across all counties (0.8460). This means selected networks will have very little
protected area already, are expected to experience significant development pressure in
the immediate future, and will have much less natural land cover diversity than other
counties. The more we lower average P and T and increase average D in the selected
network, the more consistent the network is with ROI principles. The budget con-
straint is given by,
∑s A C
j =1
j j j ≤ B, (20)

(a)
800
b
700 a
600
B
500
Species covered
A
400
300
200
ROI-influenced RSS
100
Traditional RSS
0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Network cost (millions of $)
(b)
800
700
600 d
500 c
Species covered
400 D
300 C
200
ROI-influenced RSS
100
Traditional RSS
0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00
Network cost (millions of $)
FIGURE 7.3 Species covered in networks selected by a traditional and ROI-influenced RSS

problem at various budget levels. In Figure (a), a species is considered covered by a network
if a 1,000-hectare site is selected in at least one county that the species is known to range in.
The traditional RSS networks represented by the frontier in Figure (a) were found by solv-
ing problem (17)–(19). The ROI-influenced networks represented by the frontier in Figure
(a) were found by solving the problem (17)–(19) with additional constraints that increase the
network’s ROI. In Figure (b), a species is considered covered by a network if a 1,000-hectare
site is selected in at least two counties that the species is known to range in (unless the spe-
cies is endemic to a county, then it is covered if its home county is selected). Otherwise, the
traditional and ROI-influenced networks are selected in the same way as before. The dashed
lines in Figure (b) are the frontier solutions from Figure (a).
where sj = 1 if area in county j is selected and equals 0 otherwise, Aj is the area selected in
j for protection and is equal to 1,000 ha for all j in this case, Cj is the average per hectare
cost of undeveloped land in j as of 2001 (Withey et al. 2012), and B is the budget. For
comparison, we also solve the traditional RSS problem (17)–(19).
In Figure 7.3 and Table 7.1, we present the comparison for various budget levels. If the
networks that form the traditional RSS curve score poorly on average P, T, and D values that
are associated with high ROI, then the vertical gap between the ROI-influenced and tra-
ditional RSS curves can be interpreted as a measure of untenable species protection if one
assumes that the ROI-influenced RSS networks are much more likely to increase the persis-
tence probabilities of the covered species than those species covered by the traditional RSS
networks. Consider the highlighted points in Figure 7.3(a). Points “A” and “a” on the two
graphed curves represent reserve networks that cost approximately $1,025,000. Table 7.1
indicates that the traditional RSS solution (“a”) places conservation in areas that have experi-
enced, on average, recent habitat loss of less than 1% (T = 0.994), that have about 10% of their
land already protected (P = 1.098), and that have a fairly diverse land cover (D = 0.634). In
contrast, the ROI-influenced RSS (“A”) places conservation in areas that have experienced,
on average, recent habitat loss of more than 3% (T = 0.969), that have no existing protected
land (P = 1), and that have less land cover diversity (D = 0.85). An interpretation is that by
misspecifying conservation benefits, the traditional RSS drastically overestimates the num-
ber of “protected species” by failing to account for diminishing returns and a baseline in
which much of the unprotected habitat remains on the landscape for the indefinite future.
An additional issue in specifying conservation benefits is the fact that effective cover-
age of species on a landscape is likely to require more than one additional habitat site.
This is similar to Wu and Boggess’s (1999) argument that returns to resource conserva-
tion tend to display a “∫” shape. To begin to explore the ramifications of requiring more
sites for species coverage, we rerun the ROI-influenced and traditional RSS problems
in which a species that has geographic range over two or more counties needs to have
range in at least two selected counties to be considered covered. Figure 7.3(b) gives the
cost curves for this more restrictive approach. The requirement of a second site reduces
the species covered by one-half to a third across the modeled budget levels. Requiring
two counties versus one for coverage does little to change the overall characteristics of
counties selected for protection at various budget levels (compare the average values
of D, P, and T between solutions “A” and “C”, “a” and “c”, “B” and “D”, and “b” and “d” in
Table 7.1). As with the one-county problem, the traditional RSS networks score poorly
on the diminishing returns, threat, and natural land cover indices that indicate strong
ROI. Therefore, the gap between the two frontiers in Figure 7.3(b) is likely to be indica-
tive of overestimated species protection under the traditional RSS networks.
2.3 Discussion
The problem of where to site nature reserves under a budget constraint has become a clas-
sic economic problem. In this section, we optimally locate reserves across US counties to
Table 7.1 Selected solutions to the two versions of the RSS problem. See Figure 7.3
for the location of the selected reserve networks (‘A’, ‘a’, ‘B’, ‘b’, etc.) on the various
frontiers
No. of
Reserve No. of species counties
network covered Cost w/ a site Average D Average P Average T
One-county cover
A 439 1,034,800 5 0.850 1.000 0.969
a 656 1,022,429 6 0.634 1.098 0.994
B 522 1,913,140 8 0.861 1.001 0.971
b 743 1,909,807 9 0.545 1.141 0.993
Two-county cover
C 255 821,313 6 0.858 1.000 0.977
c 461 814,650 8 0.514 1.090 0.993
D 376 1,799,383 9 0.850 1.001 0.976
d 583 1,799,004 11 0.579 1.137 0.993
All US counties NA NA Mean 0.658 1.039 0.979
Std. dev. 0.231 0.081 0.026
highlight two important features of the RSS problem that deserve further research atten-
tion. First, the solution of where to site reserves is greatly influenced by the specification of
conservation benefits. Although ecologists understand many principles about desirable
wildlife habitat, much work remains on understanding how the conservation benefits of
reserve creation are influenced by factors such as diminishing returns, spatial dependen-
cies in habitat value, and species range considerations (the benefits of reserve creation
become even more difficult to model if it is assumed that returns to habitat provision are
increasing over one range of provision and decreasing over another). Second, reserve sit-
ing is greatly influenced by how the analyst treats baseline outcomes in the absence of
reserve siting. Many regions are likely to see little loss of habitat in the absence of reserve
creation, and so siting reserves in such areas is likely to be inefficient in the usual case of
a scarce conservation budget. Continued emphasis on modeling and incorporating base-
line landscape dynamics into RSS would generate substantial research value.
3. Conserving Wildlife with Voluntary

Incentive-Based Payments
Many countries and government entities attempt to conserve wildlife and other
ecosystem services through nonregulatory means, with voluntary payment pro-
grams (often termed payments for ecosystem services) being among the most pop-
ular approaches. In the United States the multibillion dollar annual budget of the
Conservation Reserve Program is an example, whereby owners of agricultural land

are offered voluntary payments to undertake conservation activities on their land. The
efficient design of voluntary payments for wildlife conservation must overcome two
principal challenges. First, a landscape’s ability to provide the habitat resources neces-
sary to sustain a wildlife population is likely dependent on the spatial configuration
of that habitat across many independent landowners. Second, landowners have pri-
vate information regarding their willingness to accept (WTA) payments in exchange
for adopting conservation measures on their land, and profit-maximizing landowners
typically have no incentive to truthfully reveal their WTA to a conservation agency. It
is the combination of spatial dependencies and private WTA information that makes
designing efficient payment programs challenging and will be the focus of this section.
The underlying argument of this section is that efficient design of voluntary incentives
for wildlife conservation is essentially a problem of obtaining private information on
landowners’ WTA.
3.1 A Simple Example with Spatial Dependencies

A 1 × 4 parcel landscape is used to demonstrate the challenges of designing an effi-
cient payment program with spatial dependencies in conservation benefits and private
WTA information. Figure 7.4 illustrates this landscape. The WTA of each landowner
to place his or her parcel in conservation is indicated at the top of the parcel, whereas
the wildlife benefit in biophysical units of conservation is indicated along the bottom.
The first number is the benefit when the parcel is conserved but no adjacent neighbor
is conserved. The second number is the benefit when the parcel is conserved and one
adjacent neighbor is conserved as well. The third number is the benefit when the parcel
is conserved and two adjacent neighbors is conserved as well.. Each parcel is assumed to
generate biological benefits of zero if not conserved. On this landscape, wildlife benefits
exhibit spatial dependencies, and conservation costs are heterogeneous.
Conservation costs may be heterogeneous because land quality varies across par-
cels or because landowners’ WTA reflects different land management skills or other
attributes associated with how they value their land. Wildlife benefits are heteroge-
neous across parcels because natural habitat may vary across the landscape, the abil-
ity to restore natural habitat may vary across the landscape, or the geographic range of
some species may only comprise a subset of the four parcels. Finally, wildlife benefits
exhibit spatial dependencies. In other words, benefits are increasing in the number of
conserved neighbors. In general, species prefer larger contiguous patches of habitat than
isolated, smaller patches.
If we assume that the value of biophysical benefits is $1/unit, the conserved landscape
that maximizes net benefits can be determined by enumerating the total net benefits
from all possible configurations. The maximum net benefit generated by parcel conser-
vation on this landscape is $2, and it arises when adjacent parcels A, B, and C are con-
served ($8 + $9 + $5 − $5 − $8 − $7). The marginal benefit of conservation generated by
Parcel A Parcel B Parcel C Parcel D
$5 $8 $7 $8
6 8 4 6 9 4 5 6 4 6
FIGURE 7.4 An example landscape with costs (top number in $) and biophysical benefits that
depend on having zero, one, or two conserved neighbors (bottom numbers, in biophysical units).
parcel i in the optimal landscape configuration can be determined by calculating the

total benefit from optimal conservation less the total benefit without parcel i in the
conservation network. For example, the total benefit of conserving parcels A, B, and
C (the optimal network) is $22, whereas the total benefit would only be $14 ($8 + $6) if
parcel C were not conserved. Therefore, the marginal benefit of conserving parcel C is
equal to $8 ($22 − $14), which is larger than parcel C’s opportunity cost of $7. Parcel C is
optimally conserved. If we measured parcel C’s marginal benefit outside of the optimal
network, for example a network in which only parcels B and C are conserved, then C’s
marginal benefit in conservation has decreased to $7 ($11 − $4 = $7). In this particular
case, society is now indifferent to conserving C because its marginal benefit equals its
WTA. When benefits are spatially dependent, the full marginal benefit of a conserved
parcel can only be determined once the optimal landscape is known.
If the price of biophysical benefits is not known, an alternative formulation of the con-
servation problem would be to maximize biophysical benefits under a cost constraint.
For example, all four parcels in Figure 7.4 would be optimally conserved under a cost
constraint of $28. In the cost-constrained formulation of the problem, the marginal
benefit of an optimally conserved parcel is a function of the cost constraint. For exam-
ple, the marginal benefit of parcel C is 8 biophysical units under a cost constraint of $20
(8 + 9 + 5 − 8 − 6)13 and 11 units under a cost constraint of $28 (8 + 9 + 6 + 6 − 8 − 6 − 4).
Regardless of whether the problem is formulated as a net-benefit maximization or
cost-constrained optimization problem, an important conservation question is how
to implement the optimal landscape with voluntary payments when WTA is known
by the landowners but not by the conservation agency. As we have just seen, under the
net-benefit maximization problem, the optimal pattern is to conserve parcels A, B, and
C. Let us say a uniform payment of $8 was offered to parcel owners on the landscape in
order to entice the owner of parcels A, B, and C to conserve his land. However, a pay-
ment of $8 could also induce parcel D to enroll, which is not optimal. As an alternative to
a uniform payment program, an “agglomeration bonus” has been proposed (Parkhurst
et al. 2002; Parkhurst and Shogren 2007) as a means of giving a bonus payment to
those landowners who jointly conserve their land along with a neighbor. However, the
13 Biophysical benefits are maximized at a budget of $20 when conserving adjacent parcels A,
B, and C.
optimal size of the “bonus” would have to vary when marginal benefits from conserved
land are heterogeneous across the landscape, as they are in the illustrative example here.
In such cases, an agglomeration bonus program would require offering a menu of con-
tracts in which each landowner’s bonus would depend on the exact configuration of the
landscape. As an example of how complex the menu could get even with small land-
scapes, a 4 × 4 landscape has 65,535 possible conservation configurations. A Piguovian
subsidy is a third implementation strategy. If the price of biophysical services is $1/
unit, a Pigouvian approach would entail offering landowners payments equal to their
land’s marginal benefit from conservation. If there were no spatial dependencies, each
parcel would be offered a monetary payment equal to the first number in the row of
numbers that indicate the parcel’s benefit in conservation. Only parcel A would accept
this payment, and the optimal landscape in the absence of spatial dependencies could
be conserved. However, this approach doesn’t work with spatial dependencies: without
information about landowners’ WTA, which we had in our illustrative examples, the
regulator cannot solve the optimal landscape and determine each parcel’s marginal ben-
efit in equilibrium.
3.2 Empirical Analysis of Incentive Policies Under Spatial

Dependencies and Asymmetric Information
We use data from a real landscape and extend the empirical analysis by Lewis et al.
(2011) to illustrate two points discussed in Section 3.1. First, we show how sensitive spa-
tially dependent marginal benefits can be to changes in the optimal landscape. Second,
we illustrate the importance of WTA information by examining the poor performance
of second-best conservation policies. The recent study by Lewis et al. provides an empir-
ical analysis of the efficiency of a series of second-best policies that operate when private
WTA information is combined with spatially dependent benefits of conservation. The
authors combine econometrically generated distributions of landowners’ WTA with
biological models of species persistence. The WTA distributions are estimated from
observed plot-level land use decisions over a 15-year period in the Willamette Basin of
Oregon (Figure 7.5). The estimated WTA distributions are used to simulate landowner
responses to a variety of incentive policies, whereby landowners know their WTA, but
the conservation agency does not. The biological model uses spatial landscape patterns
generated by the econometric models and information on species’ range and habitat
compatibility as inputs and returns the sum of estimated persistence probabilities across
a set of 24 terrestrial species of conservation concern (the landscape’s biological score
is normalized on a 0–100 scale, where 100 means all species have a persistence prob-
ability of 100). In this application, the response of species persistence probabilities to
additional conservation on the landscape is “∫”-shaped. The authors are able to compare
outcomes from second-best policies with a first-best optimal policy. The optimal pol-
icy is estimated within a simulation by taking a random draw from the WTA distribu-
tions from each parcel, treating the draw as a known WTA value, and then selecting the
FIGURE 7.5 The Willamette Basin of Oregon.
conservation pattern that maximizes the biological score for a given level of opportunity
cost. The opportunity cost constraint is treated as the sum of WTA across all conserved
parcels.
The top row of maps in Figure 7.6 presents the landscapes that maximize the bio-
logical score for a given opportunity cost. The conservation budget sums the WTA for
each conserved parcel, where the landscape of random draws from the estimated WTA
distributions for each parcel is held fixed. The five landscapes differ in terms of their
conservation budget. As seen in the maps of Figure 7.6, small changes in the budget con-
straint can imply fairly large changes in the optimal conservation pattern, whereby par-
cels are both added and subtracted from the optimal conservation pattern as the budget
constraint is relaxed. These results are driven by the spatial dependencies in the biology
model and the particular landscape of WTA values. Different draws from the WTA dis-
tribution would also change the optimal conservation pattern.
Figure 7.7 presents marginal benefits (expressed as marginal biological scores) for a
select set of 16 parcels that are included in one or more of the optimal landscapes from
Figure 7.6. As in the simple example in Section 3.1, the marginal benefits of conserva-
tion for optimally conserved parcel i can be evaluated by examining the optimal biology
score minus the landscape score without parcel i being conserved. The main point to
Biological Score 49.75 52.71 53.69 54.56 55.58

Opportunity $16.61 $18.70 $18.92 $26.41 $27.99
Cost
Fraction of 500-Hectare Hexagon in Conservation
A B C D E
0.0 - 0.2
0.2 - 0.4
0.4 - 0.6
0.6 - 0.8
B Less A C Less B D Less C E Less D
0.8 - 1.0
< –0.30
–0.30 to 0.00
0.00 to 0.30
> 0.30
FIGURE 7.6 Optimal conservation for the Willamette Basin. The top row gives conservation
patterns that maximize the biology score (scaled from 0–100) for a given opportunity cost
budget. Each mapped unit is a 500-hectare hexagon and is comprised of nonuniform parcels.
The darker the shade of a hexagon in the top row of maps, the greater the fraction of the parcel
space in the hexagon that is conserved. The bottom row of maps is the difference between two
maps; map B less map A, etc., which shows how the distribution of conserved area changes
from one landscape to the next. Areas with the two lightest shade of gray represent hexagons
that lost conserved area vis-à-vis the previous landscape, and hexagons with the two darkest
shade of gray have gained more conserved area. For example, if a hexagon has a score of –0.31,
its fraction of conserved area has fallen by 0.31; in other words it has lost 0.31 × 500 = 155
hectares of conserved area (the greatest decline is 99%; the greatest increase is 98%).
be taken from Figure 7.7 is the fact that marginal benefits greatly depend on the opti-
mal landscape and so will differ as the budget constraint is changed. The other striking
feature of Figure 7.7 is the magnitude of changes in marginal benefits that result from
seemingly small changes in the conservation budget. This result falls from the highly
nonlinear nature of the spatially dependent biological benefit function (“∫”-shaped per-
sistence probability function) and the potential of turning a fairly fragmented network
of conservation sites into a much more connected network by strategically placing a few
more conserved parcels on the landscape. All of this is indicative of the complexity of
examining optimal landscapes for wildlife conservation.
39,653 3.14
0.97
39,138 0.62
0.26
0.05
0.007
35,127
34,217
7,486
32,340
14,806
18,583
Parcel ID
31,989 21,205
22,658
23,470
29,615 25,583 24,833
28,373
28,373 29,615
31,989 32,340
25,583
34,217
24,833 35,127
39,138
39,653
23,470
22,658
21,205
18,583
14,806
7,486
49.75 52.71 53.69 54.56 55.58
Landscape biology score
FIGURE 7.7 Marginal benefits for a select set of optimally conserved parcels. Each column of
bubbles gives the parcel’s marginal biological score on a given optimal landscape (indicated
by the biology score at the bottom of the figure). Blank cells either mean that the parcel
was not part of the landscape’s conservation network or its marginal biodiversity score was
so small that it cannot even be represented by a visible point. The map on the right of the
bubble diagram indicates each parcel’s location on the landscape.
Using the optimal landscape biology scores as a benchmark, Lewis et al. (2011) exam-
ine the performance of several alternative policy designs in which landowner WTA is
assumed unknown by the regulator. First, a set of “least-cost” policies are evaluated in
which uniform per-acre payments are offered to all landowners who meet particular
eligibility requirements based on habitat type and size characteristics and are offered
an agglomeration bonus. Relative to a baseline, none of the policies achieved even 25%
(55%) of the optimal increase in the biology score at low budget levels (high budget
levels). Second, a set of “benefit-cost” policies are evaluated, in which “benefit” indices
were constructed using the same set of habitat type/size and agglomeration character-
istics just considered. Although none of the “benefit-cost” policies achieved even 28%
of the optimal increase in the biology score at low budget levels, the best-performing
policy did achieve a more respectable 87% of the optimal increase at very high budget
levels. Of course, it must be pointed out that one can conserve all available land as habi-
tat if the budget is high enough. The underlying lesson from this analysis is that efficient
conservation with spatially dependent benefits is extremely difficult in the absence of
information on landowner WTA, and so efficient wildlife conservation with voluntary
incentives should be treated as an information problem.
3.3 Related Literature

Several related literatures evaluate and shed light on issues in conserving wildlife with
voluntary incentives. Parkhurst and Shogren (2002, 2007) use experimental methods
with students to examine possibilities regarding the agglomeration bonus. This set
of papers generally finds that an agglomeration bonus can encourage clustered habi-
tat, although the evaluated settings consist of only two to four landowners. Lewis et al.
(2009) examine a second-best approach that divides landscapes into geographic sec-
tions consisting of multiple landowners each, whereby uniform afforestation payments
are offered to all landowners within sections, whereas the payment amount differs by
section. Their findings emphasize the optimality of corner solutions, whereby it is opti-
mal to either conserve all land in a section or none. Finally, there is a related literature on
conservation auctions (Latacz-Lohmann and Van der Hamsvoort 1997; Stoneham et al.
2003; Cason and Gangadharan 2004; Kirwan et al. 2005; Schillizzi and Latacz-Lohmann
2007). Although this auction literature focuses on information asymmetry issues with
conservation programs, none of the auction designs examined is aimed at achieving
truthful revelation of landowner WTA. In a new paper, Polasky et al. (2013) develop an
auction mechanism that pays a landowner the full marginal benefit generated by con-
serving their land and provides incentives for landowners to truthfully reveal cost infor-
mation. This auction is unique in the literature it that it allows the conservation agency
to implement the optimal provision of spatially-dependent ecosystem services under
asymmetric information.
4. Conclusion
Slowing the rate of decline in wildlife populations presents a significant public goods
provision challenge to economists. The benefits from wildlife are generally non-
market and largely accrue to individuals who do not own land that contains habitat.
Governments and NGOs have addressed the conservation of wildlife habitat largely
through land use regulation, habitat purchases, and payments for voluntary conserva-
tion. This chapter synthesizes a set of outstanding economic issues that are necessary to
understand the efficient design of wildlife conservation. Although we highlight many of
the issues that have been the focus in the literature over the past 15 years, we argue that
many important issues remain to be explored in the economics literature. First, land use
regulatory design must provide direct conservation incentives for landowners or habitat
destruction can be socially preferable, and researchers need to develop better methods
for empirically evaluating regulatory outcomes and appropriately adjusting policy to
partially compensate for the efficiency costs of regulation. Second, solving the problem
of spending scarce conservation dollars on habitat purchases must devote more atten-
tion to the specification of a conservation benefit function and the specification of base-
line landscape outcomes in the absence of habitat reserves. Finally, the efficient design
of voluntary conservation payments must solve the problem of how to elicit landowner
opportunity costs of conservation because there are no current auction methods that
have been successfully developed for this problem.
Acknowledgments
The authors acknowledge funding from the National Science Foundation’s Collaborative
Research Grants No. 0814424 (Lewis) and No. 0814628 (Nelson). Senior authorship is
shared.
References
Ando, A., J. Camm, S. Polasky, and A. Solow. 1998. Species distributions, land values, and effi-
cient conservation. Science 279: 2126–2128.
Ando, A. W., and M. L. Mallory. 2012. Optimal portfolio design to reduce climate-related con-
servation uncertainty in the Prairie Pothole Region. PNAS published ahead of print March
26, 2012. doi: 10.1073/pnas.1114653109.
Araújo, M. B., M. Cabeza, W. Thuiller, L. Hannah, and P. H. Williams. 2004. Would climate
change drive species out of reserves? An assessment of existing reserve-selection methods.
Global Change Biology 10: 1618–1626.
Ball, I. R., H. P. Possingham, and M. Watts. 2009. Marxan and relatives: Software for spatial
conservation prioritisation. In Spatial conservation prioritisation: Quantitative methods and
computational tools, eds. A. Moilanen, A., K. A. Wilson, and H. P. Possingham, 185–195.
Oxford: Oxford University Press.
Bode, M., W. Probert, W. R. Turner, K. A. Wilson, and O. Venter. 2011. Conservation planning
with multiple organizations and objectives. Conservation Biology 25: 295–304.
Cabeza, M., and A. Moilanen. 2003. Site-selection algorithms and habitat loss. Conservation
Biology 17: 1402–1413.
Camm, J. D., S. Polasky, A. Solow, and B. Csuti. 1996. A note on optimal algorithms for reserve
site selection. Biological Conservation 78: 353–355.
Carwardine, J., K. A. Wilson, S. A. Hajkowicz, R. J. Smith, C. J. Klein, M. Watts, and H.
P. Possingham. 2010. Conservation planning when costs are uncertain. Conservation
Biology 24: 1529–1537.
Cash, D. W. 2001. Beyond cute and fuzzy: Science and politics in the US Endangered Species
Act. In Protecting endangered species in the United States. Biological needs, political
realities, economic choices, eds. J. F. Shogren and J Tschirhart, 106–137. New York: Cambridge

University Press.
Cason, T., and L. Gangadharan. 2004. Auction design for voluntary conservation programs.
Church, R. L., D. M. Stoms, and F. W. Davis. 1996. Reserve selection as a maximal coverage
problem. Biological Conservation 76: 105–112.
The Conservation Biology Institute (CBI). 2010. PAD-US 1.1 (CBI Edition). Corvallis, OR.
http://consbio.org/products/projects/pad-us-cbi-edition
Comer, P., D. Faber-Langendoen, R. Evans, S. Gawler, C. Josse, G. Kittel, S. Menard, M. Pyne, M.
Reid, K. Schulz, K. Snow, and J. Teague. 2003. Ecological systems of the United States: A work-
ing classification of US terrestrial systems. Arlington, VA: NatureServe.
Costello, C., and S. Polasky. 2004. Dynamic reserve site selection. Resource and Energy
DeShazo, J. R., and J. Freeman. 2003. The congressional competition to control delegated power.
Texas Law Review 81: 1443–1520.
DeShazo, J. R., and J. Freeman. 2006. Congressional politics. In The Endangered Species Act
at thirty. Renewing the conservation promise, Vol. 1, eds. D. D. Goble, J. M. Scott, and R.
W. Davis, 68–71. Washington, DC: Island Press.
Dobson, A. P., J. P. Rodriguez, W. M. Roberts, and D. S. Wilcove. 1997. Geographic distribution
of endangered species in the United States. Science, 275(5299): 550–553.
Emerton, L. 1999. Balancing the opportunity costs of wildlife conservation for communities
around Lake Mburo National Park, Uganda. Evaluating Eden Series discussion paper No.
5. The International Institute for Environment and Development (IIED). http://pubs.iied.
org/pdfs/7798IIED.pdf.
Ferraro, P. J., C. McIntosh, and M. Ospina. 2007. The effectiveness of the US Endangered Species
Act: An econometric analysis using matching methods. Journal of Environmental Economics
Fry, J. A., M. J. Coan, C. G. Homer, D. K. Meyer, and J. D. Wickham. 2009. Completion of the
National Land Cover Database (NLCD) 1992–2001 Land Cover Change Retrofit product.
U. S. Geological Survey Open-File Report 2008–1379.
Haight, R. G., S. A. Snyder, and C. S. Revelle. 2005. Metropolitan open-space protection with
uncertain site availability. Conservation Biology 19: 327–337.
Hansen, S. 1999. Debt for nature swaps—Overview and discussion of key issues. Ecological
Innes, R., and G. Frisvold. 2009. The economics of endangered species. Annual Review of
Kerkvliet, J., and C. Langpap. 2007. Learning from endangered and threatened species recov-
ery programs: A case study using US Endangered Species Act recovery scores. Ecological
Kirwan, B., R. N. Lubowski, and M. J. Roberts. 2005. How cost-effective are land retirement
auctions? Estimating the difference between payments and willingness to accept in the
Conservation Reserve Program. American Journal of Agricultural Economics 87: 1239–1247.
Lamberson, R. H., R. McKelvey, B. R. Noon, and C. Voss. 1992. A dynamic analysis of Northern
Spotted Owl viability in a fragmented forest landscape. Conservation Biology 6(4): 505–512.
Latacz-Lohmann, U., and C. Van der Hamsvoort. 1997. Auctioning conservation con-
tracts: A theoretical analysis and application. American Journal of Agricultural Economics
79(2): 407–418.
Lawler, J. J., S. L. Shafer, D. White, P. Kareiva, E. P. Maurer, A. R. Blaustein, and P. J. Bartlein. 2009.
Projected climate-induced faunal change in the western hemisphere. Ecology 90: 588–597.
Lewis, D. J., A. J. Plantinga, E. Nelson, and S. Polasky. 2011. The efficiency of voluntary incentive
policies for preventing biodiversity loss. Resource and Energy Economics, 33(1): 192–211.
Lewis, D. J., A. J. Plantinga, and J. Wu. 2009. Targeting incentives to reduce habitat fragmenta-
tion. American Journal of Agricultural Economics 91(4): 1080–1096.
Lueck, D., and J. A. Michael. 2003. Preemptive habitat destruction under the Endangered
Species Act. Journal of Law and Economics 46: 27–60.
Male, T. D., and M. J. Bean. 2005. Measuring progress in US endangered species conservation.
Ecology Letters 8: 986–992.
Mann, C., and M. Plummer. 1995. Noah’s choice. New York: Alfred A. Knopf.
Margules, C. R., and R. L. Pressey. 2000. Systematic conservation planning. Nature 405: 243–253.
McAnaney, A. P. 2006. Remembering the spirit of the Endangered Species Act: A case for nar-
rowing agency discretion to interpret ‘significant portion’ of a species’ range. Golden Gate
University Law Review 36: 6. http://digitalcommons.law.ggu.edu/ggulrev/vol36/iss3/6.
Metrick, A., and M. L. Weitzman. 1998. Conflicts and choices in biodiversity preservation. The
Journal of Economic Perspectives 12: 21–34.
Millennium Ecosystem Assessment (MEA). 2005. Living beyond our means: Natural assets and
human well-being. Washington, DC: Island Press.
Miller, J. K., J. M. Scott, C. R. Miller, and L. P. Waits. 2002. The Endangered Species Act: Dollars
and sense? BioScience 52: 163–168.
Moilanen, A., A. M. A. Franco, R. I. Early, R. Fox, B. Wintle, and C. D. Thomas. 2005. Prioritizing
multiple-use landscapes for conservation: Methods for large multi-species planning prob-
lems. Proceedings of the Royal Society: Biological Sciences 272: 1885–1891.
Murdoch, W., S. Polasky, K. A. Wilson, H. P. Possingham, P. Kareiva, and R. Shaw. 2007.
Maximizing return on investment in conservation. Biological Conservation 139: 375–388.
Naidoo, R., A. Balmford, P. J. Ferraro, S. Polasky, T. H. Ricketts, and M. Rouget. 2006. Integrating
economic costs into conservation planning. Trends in Ecology & Evolution 21: 681–687.
Nelson, E., S. Polasky, D. Lewis, A. Plantinga, E. Lonsdorf, D. White, D. Bael and J. Lawler. 2008.
Efficiency of incentives to jointly increase carbon sequestration and species conservation on
a landscape. Proceedings of the National Academy of Sciences of the USA 105(28): 9471–9476.
Newbold, S. C., and J. Siikamaki. 2009. Prioritizing conservation activities using reserve site
selection methods and population viability analysis. Ecological Applications 19: 1774–1790.
Parkhurst, G. M., and J. F. Shogren. 2007. Spatial incentives to coordinate contiguous habitat.
Ecological Economics 64: 344–355.
Parkhurst, G. M., J. F. Shogren, C. Bastian, P. Kivi, J. Donner, and R. B. W. Smith. 2002.
Agglomeration bonus: An incentive mechanism to reunite fragmented habitat for biodiver-
sity conservation. Ecological Economics 41: 305–328.
Polasky, S. 2001. Investment, information collection, and endangered species conservation on
private land. In Protecting endangered species in the United States. Biological needs, political
realities, economic choices, eds. J. F. Shogren and J. Tschirhart, 312–325. New York: Cambridge
University Press.
Polasky, S., and H. Doremus. 1998. When the truth hurts: Endangered species policy on private land
with imperfect information. Journal of Environmental Economics and Management 35: 22–47
Polasky, S., D. J Lewis, A. J. Plantinga, and E. Nelson. 2013. Implementing the Optimal
Provision of Ecosystem Services. Economics Department Working Paper Series. Paper 10.
http://digitalcommons.bowdoin.edu/econpapers/10.
Polasky, S., Erik Nelson, Jeff Camm, Blair Csuti, Paul Fackler, Eric Lonsdorf, Claire Montgomery,
Denis White, Jeff Arthur, Brian Garber-Yonts, Robert Haight, Jimmy Kagan, Anthony
Starfield, and Claudine Tobalske. 2008. Where to put things? Spatial land management to
sustain biodiversity and economic returns. Biological Conservation 141: 1505–1524.
Possingham, Hugh P., Sandy J. Andelman, Mark A. Burgman, Rodrigo A. Medellin, Larry
L. Master, David A. Keith. 2002. Limits to the use of threatened species lists. Trends in
Ecology & Evolution 17: 503–507.
Pyke, C. R., and D. T. Fischer. 2005. Selection of bioclimatically representative biological reserve
systems under climate change. Biological Conservation 121: 429–441.
Rachlinski, J. J. 1997. Noah by the numbers: an empirical evaluation of the Endangered Species
Act. Cornell Law Review 82: 356–389.
Radeloff, V. C., E. Nelson, A. J. Plantinga, D. J. Lewis, D. Helmers, J. J. Lawler, J. C. Withey,
F. Beaudry, S. Martinuzzi, V. Butsic, E. Lonsdorf, D. White, and S. Polasky. 2012.
Economic-based projections of future land use in the conterminous United States under
alternative policy scenarios. Ecological Applications 22: 1036–1049.
Rosenzweig, M. L. 1995. Species diversity in space and time. New York: Cambridge
University Press.
Ruhl, J. B. 1998. The Endangered Species Act and private property: A matter of timing and loca-
tion. Cornell Journal of Law and Public Policy 8: 37–53.
Sala, O. E., F. S. Chapin, III, J. J. Armesto, E. Berlow, J. Bloomfield, R. Dirzo, E. Huber-Sanwald,
L. F. Huenneke, R. B. Jackson, A. Kinzig, R. Leemans, D. M. Lodge, H. A. Mooney, M.
Oesterheld, N. L. Poff, M. T. Sykes, B. H. Walker, M. Walker, and D. H. Wall. 2000. Global
biodiversity scenarios in the Year 2100. Science 287: 1770–1774.
Schillizzi, S., and U. Latacz-Lohmann. 2007. Assessing the performance of conservation auc-
tions: An experimental study. Land Economics 83(4): 497–515.
Schwartz, M. W. 1999. Choosing the appropriate scale of reserves for conservation. Annual
Review of Ecology and Systematics 30: 83–108.
Scott, J. M., D. D. Goble, and F. W. Davis. 2006. Introduction. In The Endangered Species Act
at thirty: Renewing the conservation promise, Vol. 1, eds. D. D. Goble, J. M. Scott, and R.
Shogren, J. F., J. Tschirhart, T. Anderson, A. W. Ando, S. R. Beissinger, D. Brookshire, G.
M. Brown Jr., D. Coursey, R. Innes, S. M. Meyer, and S. Polasky, 1999. Why economics mat-
ters for endangered species protection. Conservation Biology 13: 1257–1261.
Snyder, S., R. Haight, and C. ReVelle. 2005. A scenario optimization model for dynamic reserve
site selection. Environmental Modeling and Assessment 9: 179–187.
Stoneham, G., V. Chaudri, A. Ha, and L. Strappazzon. 2003. Auctions for conservation con-
tracts: An empirical examination of Victoria’s Bush Tender Trial. The Australian Journal of
Taylor, M. F. J., K. F. Suckling, and J. J. Rachkinski. 2005. The effectiveness of the Endangered
Species Act: A quantitative analysis. BioScience 55: 360–367
The Nature Conservancy (TNC). 2010 Annual report: Roots of innovation. http://www.nature.
org/aboutus/ouraccountability/annualreport/index.htm.
Thompson, B. H., Jr. 2006. Managing the working landscape. In The Endangered Species Act
at thirty. Renewing the conservation promise, Vol. 1, eds. D. D. Goble, J. M. Scott, and R.
US Fish and Wildlife Service (FWS). 2009. 2009 Expenditure report. http://www.fws.gov/
endangered/esa-library/index.html.
Wilcove, D. S., D. Rothstein, J. Dubow, A. Phillips, and E. Losos. 2000. Leading threats to bio-
diversity: What’s imperiling US species. In Precious heritage: The status of biodiversity in
the United States, eds. B. A. Stein, L. S. Kutner, and J. S. Adams, 239–254. Oxford: Oxford
University Press.
Withey, J. C., J. J. Lawler, S. Polasky, A. J. Plantinga, E. Nelson, P. Kareiva, C. B. Wilsey, C. A.
Schloss, T. M. Nogeire, A. Ruesch, J. Ramos Jr., and W. Reid. 2012. Maximizing return on
conservation investment in the conterminous USA. Ecology Letters 15: 1249–1256.
Wu, J., R. M. Adams, and W. G. Boggess. 2000. Cumulative effects and optimal targeting of
conservation efforts: Steelhead trout habitat enhancement in Oregon. American Journal of
Wu, J., and W. G. Boggess. 1999. The optimal allocation of conservation funds. Journal of
Wu, J., and K. Skelton-Groth. 2002. Targeting conservation efforts in the presence of threshold
effects and ecosystem linkages. Ecological Economics 42: 313–331.
C HA P T E R 8
C O N N E C T I N G E C O S YS T E M
S E RV I C E S T O L A N D U S E
Implications for Valuation and Policy
ROBE RT J. JOH N STON , ST E PH E N K . S WA L LOW,

DA NA M A R I E BAU E R , E M I U C H I DA , A N D
C H R I STOPH E R M . A N DE R S ON
Ecosystem goods and services (henceforth, “services”)1 may be defined as the outputs
of natural systems that benefit society (Daily 1997; Millennium Ecosystem Assessment
2005) or “the flows from an ecosystem that are of relatively immediate benefit to
humans and occur naturally” (Brown et al. 2007, 334). Although economists have long
sought to quantify the market and nonmarket benefits humans derive from natural sys-
tems, the concept of ecosystem services has gained recent attention among natural sci-
entists, policy makers, and advocacy groups. Among the factors that distinguish this
work from traditional economic analysis, at least in principle, is a more fundamental
multidisciplinary focus, including an emphasis on both ecological production2 and
economic value.
Much of the recent research and policy emphasis on ecosystem services has targeted
those services linked in some way to land use and cover (henceforth, “use”), includ-
ing those flowing from agriculture, forests, wetlands, rangelands, and other terrestrial
systems (Bauer and Johnston 2013). Changes in land use can affect multiple ecosystem
services, many of which are not traded in markets and hence lack direct signals of value
(Polasky et al. 2011). Among the primary motivations for research in this area is the pro-
vision of information to quantify tradeoffs and promote optimal, or socially efficient,
1
We define ecosystem services to include both goods and services provided directly by ecosystems,
including nonmarket goods that are often titled “cultural” or “social” benefits by the ecosystem services
literature (Brown et al. 2007; Bateman et al. 2011). These may include aesthetic benefits.
2 Bioeconomic researchers have long been concerned with ecological production functions (Clark
1976; Wilen 1985; Conrad and Clark 1987).

Connecting Ecosystem Services to Land Use 197
management. Within agricultural policy, for example, the nonmarket and often unrec-
ognized value of ecosystem services and disservices is recognized as among the primary
causes of market failure (Dale and Polasky 2007; Kroeger and Casey 2007; National
Research Council 2010; Ribaudo et al. 2010). Many decisions are potentially informed
by quantification and valuation of ecosystem services, including those related to restora-
tion programs, land set-asides, and conservation easements or purchases; assessments
of the equivalency of market credits or habitat mitigation; development of regulatory or
incentive programs to motivate changes in agricultural or land use practices; and devel-
opment of ecosystem service markets (Johnston and Duke 2007; Swinton et al. 2007;
Swallow et al. 2008; Duke and Johnston 2010; Wainger et al. 2010).
Examples of the many conceptual, theoretical, and empirical publications linking
ecosystem services to land use include Bateman et al. (2011), Dale and Polasky (2007),
Heal and Small (2002), Johnston et al. (2002a, 2002b), Nelson et al. (2009), Polasky et al.
(2011), Priess et al. (2007), Ricketts et al. (2004), Swinton et al. (2007), and Wainger et al.
(2010). Although categorizations of ecosystem services vary and often double count
contributions to welfare (Fisher et al. 2009), commonly cited services include the pro-
duction of flora, fauna, and natural (bio)diversity; provision of water (quantity and qual-
ity); regulation of climate (e.g., through carbon sequestration or microclimate, such as
through shading or heat islands in absence of shading); regulation of hazards (e.g., flood
and erosion mitigation); breakdown and detoxification of waste; purification processes
(e.g., of air and water); and the generation and maintenance of socially valued places and
landscapes (Hanley and Barbier 2009; Balmford et al. 2011; Bateman et al. 2011; Polasky
et al. 2011). Beyond the provision of food, fiber, and fuel, often cited examples of services
related specifically to agricultural land use include nutrient cycling, pollination, wildlife
habitat, biodiversity, carbon sequestration, aesthetic services, and recreational services
(Millennium Ecosystem Assessment 2005; Swinton et al. 2007; Fisher et al. 2009); these
can include services received by or that benefit agriculture (e.g., pollination services),
services provided by agricultural land uses (e.g., open-space aesthetics), and ecosystem
disservices or decreases in ecosystem services caused by agricultural production (e.g.,
animal waste generating odors or fostering insect pests) (Johnston et al. 2001; Ready
and Abdalla 2005; Zhang et al. 2007).
Despite the relevance of ecosystem services for policy and recent enthusiasm for the
concept, “there have been relatively few attempts to define the concept clearly to make it
operational” (Fisher et al. 2008, 2051; 2009). “[W]‌hile progress is being made in the inte-
gration of economics and ecological sciences for understanding ecosystem services, this
is a field still in its nascent stage” (Fisher et al. 2011, 152). The literature is dominated by
works proposing frameworks, typologies, and perspectives. At the same time, research-
ers seeking to evaluate ecosystem services and link them to land use changes face empir-
ical challenges (Bateman et al. 2011). Within this context, the validity and precision of
ecosystem service evaluations (including quantification, prediction, and valuation) are
largely determined by three overarching factors: (1) what services are evaluated—the
conceptual and theoretical foundations of ecosystem service analysis and how these
relate to the specific set of ecosystem conditions, functions, and outcomes chosen for
198 Robert J. Johnston et al.
analysis; (2) how services are evaluated—the validity and precision of the economic and
ecological methods used to quantify and value selected services; and (3) at which scopes
and scales are services evaluated—the magnitude of changes considered and the geo-
graphic scale over which evaluations are conducted.
This chapter describes methods, challenges, and prospects involved in linkages
between ecosystem services and land use. We begin with a discussion of the current
state of the literature devoted to land use–related ecosystem services. This is followed by
a review of relationships between methods used for ecosystem service evaluations and
the accuracy and precision of empirical results. We conclude with illustrative applica-
tions that elucidate some of the challenges faced when linking ecosystem services to
land use, as well as the use of resulting information to guide policy.
1. Ecosystem Services and Land Use
The idea that ecological processes provide outputs valued by human society or that these
outputs extend beyond direct products typically sold in markets is not new (Krutilla
1967; Daily et al. 2009). Many of the precepts of nonmarket valuation are grounded in
this idea, extending back more than four decades (Champ et al. 2003; Freeman 2003).
The capacity of land to provide market and nonmarket benefits—including those
related to ecological composition, structure, and function—has long been a part of eco-
nomic analysis and discourse related to land use policy (Swallow 1996b; Johnston and
Swallo 2006; Bergstrom and Ready 2009). Valuation of ecosystem services is grounded
in the theoretical structure that underpins all economic welfare analysis (Freeman 2003;
Just et al. 2004), although many empirical applications apply methods that violate eco-
nomic theory required for valid welfare estimation (see discussions in Toman 1998;
Bockstael et al. 2000; Tallis and Polasky 2009, 268–269; Bateman et al. 2011, particu-
larly 188–196). Valuation methods applicable to ecosystem services are summarized by
numerous works, including those by Holland et al. (2010), Bateman et al. (2011), US
Environmental Protection Agency (US EPA 2009), Swinton et al. (2007), and Hanley
and Barbier (2009). Depending on the type of services involved and mechanisms
through which they enhance welfare, options for valuation can include factor input
methods, ecological productivity methods, and a wide range of nonmarket revealed and
stated preference methods.
Recognizing these foundations, research linking ecosystem services to land use is
most appropriately considered an application of existing methods rather than a novel
methodological approach (Daily et al. 2009). Within agricultural policy for example,
Gardner’s 1977 American Journal of Agricultural Economics article elucidated the eco-
nomic justification for public investment in farmland protection. The provision and
preservation of nonmarket environmental amenities—many akin to what we now
call ecosystem services—were among the primary motivations. Bergstrom and Ready
(2009) review two decades of research estimating the value of agricultural amenity
benefits in the United States, many of these related to what would now be labeled ecosys-
tem services. Similarly, the multifunctional agriculture movement in the United States
and Europe has at its core a recognition that agriculture provides nonmarket benefits
beyond traditional food and fiber, including those related to the ecological functions of
agro-ecosystems (Josling 2002; Batie 2003; Boody et al. 2005; Duke and Johnston 2010;
National Research Council 2010). Similar themes appear in research related to the pres-
ervation of nonagricultural lands (Johnston and Swallow 2006).
Among the main distinguishing features of the ecosystem services movement is
greater attention to the formal linkages between ecology and economics necessary to
provide valid estimates of the human benefits. This includes an emphasis on the ben-
efits of ecosystem structure and function realized through ecological production func-
tions, rather than solely on the end products that influence welfare (Daily et al. 2009; cf.,
Collins et al. 2010). In the case of valuation, although the economic valuation methods
are often identical, the emphasis is on the values provided directly or indirectly by eco-
system functions, rather than on the values provided by market or nonmarket goods. As
noted below, this shift in emphasis—although perhaps subtle—can lead to nontrivial
empirical challenges. These include steps required to account for, quantify, or appropri-
ately disentangle multiple interacting services provided jointly by a single set of ecosys-
tem functions (Nelson et al. 2009; Johnston and Russell 2011; Polasky et al. 2011).
Accordingly, and unlike some of the past research in nonmarket valuation, the lit-
erature linking ecosystem services to land use includes substantial participation of the
natural science community. Models are often characterized by heavy use of geographic
information systems (GIS) and attention to ecological modeling, although the charac-
terization of ecological production relationships is one of the areas in which empirical
results are often lacking (Bateman et al. 2011; Polasky et al. 2011). Empirical examples
are diverse, with applications to such issues as nationwide agriculture (Bateman et al.
2011), statewide land use (Polasky et al. 2011), mangroves (Hanley and Barbier 2009),
wetlands and intertidal habitats (Johnston et al. 2002a, 2002b; Boyer and Polasky
2004), carbon sequestration and species conservation (Nelson et al. 2009), biodiversity
(Naidoo and Ricketts 2006), deforestation and pollination (Priess et al. 2007), invasive
species (Wainger et al. 2010), and deforestation (Tallis and Polasky 2009).
Despite the diversity of empirical applications and divergent terminology applied
across the ecosystem services literature, most research connecting ecosystem services
to land use relies on a similar underlying conceptual framework, summarized in Figure
8.1.3 This figure adapts and coordinates concepts found in works such as Brown et al.
(2007), Fisher et al. (2009), Dale and Polasky (2007), and Johnston and Russell (2011).
Within this framework, ecosystem properties, including structure and function at vari-
ous trophic levels, contribute to biophysical outputs, which directly enhance the welfare
3 Collins et al. (2010) provide an analogous framework that is more comprehensive with respect to
social sciences other than economics.

Ecosystem Biogeophysical
properties outputs
Direct influence on utility
Final services
Intermediate services Produced Utility
Structure goods &
Direct outputs of services Benefits realized by
Co
ecosystems of humans, reflected

n
mp
tio immediate benefit Production using in ecosystem

osi
nc
to humans additional human service values
Fu
tio
and built capital

n
inputs
Feedbacks Human
Human activity benefits
& production
Natural system Human system
FIGURE 8.1 General linkages among ecosystem properties, ecosystem services, human pro-
duction, and benefits.
of at least one human beneficiary (Johnston and Russell 2011). These outputs are defined
as final ecosystem services (Boyd and Banzhaf 2007; Brown et al. 2007; Turner and Daily
2008; Fisher et al. 2009; Johnston and Russell 2011). Intermediate services, in contrast,
are ecological conditions or processes that only benefit humans through effects on other,
final services. These may be viewed as inputs into the production of final services. The
status of an ecological condition or process as a final versus intermediate service may
vary across beneficiaries (Johnston and Russell 2011).
As shown in Figure 8.1 (and discussed by Brown et al. [2007], Bateman et al. [2011],
Fisher et al. [2009], and Johnston and Russell [2011]), ecosystem services only some-
times influence human welfare directly. An example would be nonuse benefits flowing
directly from the provision of a biophysical output, such as wildlife abundance or diver-
sity. However, more commonly, benefits are “generated by ecosystem services in combina-
tion with other forms of capital like people, knowledge, or equipment, e.g., hydroelectric
power utilizes water regulation services of nature but also needs human engineering, con-
crete, etc.” (Fisher et al. 2008, 2052). Once human labor or capital is applied to transform
a biophysical output into something else, the result is no longer an ecosystem service but
rather the result of human production (Johnston and Russell 2011); this is reflected by the
“Produced Goods & Services” box in Figure 8.1. Most, if not all, agricultural commodities,
for example, fall into this category. Human activities such as agricultural production, how-
ever, often jointly produce feedbacks that affect the ecosystem structures and functions that
provide the initial ecosystem services (Zhang et al. 2007).
Based on this model, it is clear that valuation of market commodities produced using
ecosystem services is not the same as valuation of ecosystem services (Bateman et al. 2011;
Johnston and Russell 2011). In many cases, the final product sold in a market, and hence
most easily valued, does not represent an ecosystem service. Rather, these services serve
as inputs to market production. Although this distinction may render valuation of the
ecosystem services themselves more difficult, it is required to consistently evaluate the ben-
efits humans derive from ecosystems and distinguish these from the benefits that people
obtain from human production (Johnston and Russell 2011).
For reasons such as this, it may sometimes be counterproductive to insist on separate and
distinct values for ecosystem services. Land use and land use change affect ecosystem struc-
ture and function, intentionally and unintentionally. Swallow (1996a) places this obser-
vation within the context of basic economic theories of resource use, noting that the role
of land in ecosystem production, both in its developed state, as well as in its “natural” (or
undeveloped) state, affects its net value (cf., Swallow [1994] for an application). Often, land
use change is focused on human-produced goods, some of which may be built on inter-
mediate inputs or goods provided by ecosystems. The inseparability of human-produced
values from at least some ecosystem-produced values recommends a careful understand-
ing of derived demand and the view of ecosystems as one critical contributor to overall
human well-being, but one that may not operate alone, in the absence of human ingenuity
and productivity.
Beyond the general conceptual approach outlined in Figure 8.1, there have been many
proposed typologies of ecosystem services; these seek to provide templates that may
be used to identify and categorize services across applications (e.g., de Groot et al. 2002;
Millennium Ecosystem Assessment 2005; Wallace 2007; Balmford et al. 2011). The rele-
vance of such taxonomies, however, is limited. As noted by Fisher et al. (2008, 2051), for
example, “while [the typology in the Millennium Ecosystem Assessment] is useful as a
heuristic tool, it can lead to confusion when trying to assign economic values to ecosystem
services.”4 Numerous authors have highlighted the double counting implicit in many typol-
ogies (Wallace 2007; Fisher et al. 2008, 2009; Johnston and Russell 2011), and Johnston and
Russell (2011) note the impossibility of a universal typology that applies to all beneficiaries.
The ubiquity of frameworks and typologies within the literature may also obscure
the multiple valid approaches that one may take toward quantification and valuation.
Many market and nonmarket goods are joint products of multiple components of eco-
system structure and function. Within this context, the analyst should exercise caution
before concluding that there is a single, proper point at which to conduct quantification
and valuation. As discussed later in this chapter, there are many advantages to valua-
tion focused on end products (e.g., final goods). However, there may be instances in
which—for policy or analytical purposes—valuation of intermediate products might
be desirable (Johnston et al. 2011).5 In cases where values of intermediate outcomes
4
Indeed, the category of “supporting services” is nearly tailor-made to induce double counting
(Johnston et al. 2011). The concept seems closely associated with the standard economic concept of
derived demand (see Swallow 1997); the careless analyst might forget that the value of inputs (ecological
structures and functions) that “support” production of ecosystem service outputs of direct relevance to
human well-being is fully incorporated (already) upon measuring the value of final services.
5 For example, in some instances final welfare-producing products may be unobservable to the
analyst or difficult to quantify, whereas intermediate outcomes may be more easily quantified or
observed.
are estimated, the analyst may face increased risk of double counting or omission of
welfare-relevant outcomes, particularly when a comprehensive understanding of rela-
tionships between intermediate and final services is unavailable. Although welfare the-
ory enables valuation of either intermediate or final outcomes, there can be empirical
challenges to both.
In summary, despite the foundation of contemporary ecosystem services research and
the recent explosion of publications seeking to link land use to ecosystem service provi-
sion and value, this area of study remains immature. Improved methods—or at least a
more transparent delineation of assumptions, caveats, and implications—are required if
information on ecosystem services is to be used broadly to guide land use policy (Bauer
and Johnston 2013). Challenges are particularly evident in empirical work.
2. The Validity and Precision of Research

Linking Ecosystem Services to Land Use
The characteristics and quality of empirical methods within the ecosystem services lit-
erature vary. Despite this heterogeneity, the validity and precision of empirical ecosys-
tem services research is strongly related to three often interlinked factors. These include
(1) what services are evaluated, (2) how these services are evaluated, and (3) at which
scopes and scales evaluations are conducted.6
2.1 The Conceptual Basis for Ecosystem Service Values: What

Services Are Evaluated?
Consistent estimates of ecosystem service values require careful definition of the
specific services under consideration and how these contribute to human welfare.
Differentiation of intermediate ecosystem functions from final ecosystem services,
where feasible, also ensures that the benefit of each distinct ecosystem condition or
process, to each human beneficiary,7 is counted once and only once (Fisher et al. 2008;
US EPA 2009; Johnston and Russell 2011). The distinction between intermediate and
final ecosystem services may be defined within the framework described by Boyd and
Krupnick (2013), Fisher et al. (2009), Turner and Daily (2008), and Johnston and Russell
6
In addition, many ecological studies advance the frontiers of ecological knowledge but miss
opportunities to publish quantitative models to which economists could attach their own models. The
corollary is also true, with economists’ disciplinary focus failing to facilitate connections for ecologists.
7 Following Johnston and Russell (2011), we define a “beneficiary” as a person or group operating
in a particular role (e.g., bird watcher, farmer) whose welfare in that role is improved by a particular
ecosystem service and is therefore willing to pay for improvement or to avoid reduction in the service.
(2011), among others, which characterizes the provision of natural outputs (goods and
services) in terms of systems of ecological production. The final outputs of these sys-
tems—final ecosystem services—are biophysical outcomes that directly enhance the
welfare of at least one human beneficiary.8 Intermediate services are conditions or pro-
cesses that only benefit humans through effects on other, final services.
Failure to recognize the value of intermediate ecosystem services encourages actions
that lead to suboptimal provision of these services. Conversely, summation of values
for both intermediate and final services is also misleading because it double counts the
contribution of the intermediate services to welfare, which also promotes suboptimal
provision relative to human welfare. Such double counting is common (Johnston and
Russell 2011).
As a purely conceptual matter, valuation of intermediate services is straightforward.
Within economic theory, one can value either changes in inputs or the correspond-
ing change in the final output; the value of the change in output reflects the sum of the
value of changes in all inputs used in production. However, in the context of ecosystem
services, valuing changes in intermediate services often presents empirical challenges
(Johnston et al. 2011). What is often missing is a quantifiable linkage between changes
in ecosystem structure or function, implying changes in intermediate services, and how
such changes affect the value of final services.
For example, revealed preference analyses rarely account for the influence of inter-
mediate ecological inputs on final ecosystem goods and services that (directly) influ-
ence observed behaviors. Hence, revealed preference analyses generally provide values
for final ecosystem services only.9 Additional biophysical information is required to
estimate values for associated intermediate services, such as water filtration services of
undeveloped land; this is often unavailable (Johnston et al. 2011). Stated preference (SP)
research has also given little attention to distinctions between intermediate and final
services, or more broadly to the definition of ecosystem services, leading to the potential
for welfare biases (Fisher et al. 2009; Johnston et al. 2011; Boyd and Krupnick 2013).
Existing works in the economics literature have sought to clarify the challenge and
provide solutions (Boyd and Banzhaf 2007; Brown et al. 2007; Wallace 2007; Fisher and
Turner 2008; Fisher et al. 2009; Boyd and Krupnick 2013). Solutions suggested by these
works, however, have yet to gain broad adoption outside of the economics literature.
Johnston and Russell (2011) argue that the emphasis of the ecosystem services literature
on complex “one size fits all” classifications exacerbates the confusion over final versus
intermediate services. In response, they propose a set of guidelines that can be used to
clarify the set of final ecosystem services that benefit any given beneficiary and distin-
guish these from intermediate services.
8 Boyd and Krupnick (2013) define a closely related concept of ecological endpoints.
9 An exception is ecological productivity methods, in which empirical welfare estimates are grounded
in an explicit model of ecological production (e.g., Johnston et al. 2002b).
Even given such guidelines, cataloguing of ecosystem services may prove difficult,
because it often requires the analyst to characterize complex, dynamic, and interrelated
processes responsible for the production of final ecosystem goods and services. These
services may be subject to unknowns concerning both ecological and household pro-
duction functions. Indeed, in a world of joint production (by ecosystems and by house-
holds using ecosystems), a set of separate and distinct service values and value estimates
may prove impossible without potentially arbitrary allocation rules.10 Here, the most
critical need is often not for a complete cataloguing of intermediate and final services
for any beneficiary but rather a consistent framework to ensure that double counting is
mitigated. Some have also argued that the valuation of jointly produced ecosystem ser-
vices is best accomplished within a general equilibrium framework (National Research
Council 2005; Carbone and Smith 2008, 2010).11
Regardless of the specific mechanisms applied, research evaluating ecosystem ser-
vices from land use requires careful attention to consistency and theoretical validity,
including methods to identify the ecosystem services to be valued (Fisher et al. 2008;
Bauer and Johnston 2013). These concerns are particularly relevant for broad-scale
analyses that attempt to simultaneously quantify values for multiple, often interrelated
or bundled ecosystem services, particularly when aggregated benefits from these mod-
els are used to inform land use policy.
2.2 The Methodological Basis for Ecosystem Service

Values: How Are Services Evaluated?
Although methodological advances have been made in research linking ecosystem
services to land use and quantifying associated values, required “[i]‌nterdisciplinary
research, combining economics and other social sciences with the natural sci-
ences. . . remains a relatively immature area of study” (Bateman et al. 2011, 209). Among
“the most serious problems facing the effective and robust valuation of ecosystem ser-
vices are gaps in our understanding of the underpinning science relating those ser-
vices to the production of goods and the paucity of valuation studies and available data
regarding the values of these goods” (Bateman et al. 2011, 193). Research linking ecosys-
tem services to land use not infrequently incorporates substandard economic or ecolog-
ical methods or unrealistic assumptions; “most modeling to date has been unavoidably
10
For example, some authors characterize agricultural production as a final ecosystem service
because the growth of crops is largely an ecological process (Fisher et al. 2009). Others characterize
agricultural production as the result of a combination of human labor and capital with ecosystem
services—such that the agricultural production itself is not a final ecosystem service (Brown et al.
2007; Johnston and Russell 2011). Neither perspective is incorrect; they merely reflect subtly different
definitions of a final ecosystem service. What is most important is transparency and consistency in
whatever classifications are applied.
11 Finnoff et al. (2008) and Finnoff and Tschirhart (2008) illustrate a provocative approach.
simplistic” (Balmford et al. 2011). This simplicity arises, in part, because disciplinary
segregation may not lead ecologists to produce quantitative models of links between
drivers and outputs of relevance to human welfare and may not lead economists to pro-
duce models with sufficient foundation in ecological production processes.
Johnston et al. (2012), for example, detail ecological shortcomings of SP analyses of
ecosystem service values, including the tendency of the literature to quantify values for
“ecosystem services” that have no precise ecological definition. Simpson (1998) details
similar shortcomings applicable to economists’ treatment of ecology more broadly.
Swallow (1996b) demonstrates that the dynamic ecological context of habitat develop-
ment can invalidate standard economic intuition regarding the ranking of social value
from new development in a heterogeneous, ecologically dynamic landscape.
At the same time, there are gaps in the underlying natural science knowledge neces-
sary to link changes in ecosystem structure and function to effects on quantifiable eco-
system services (Bateman et al. 2011); we are often unable to predict, for example, how
the flow of services will be affected by changes in land use or management (Daily et al.
2009; Polasky et al. 2011). Although large-scale, data-intensive models such as InVEST
(Integrated Valuation of Ecosystem Services and Tradeoffs) have been developed to
model the production and value of ecosystem services, these models generally operate
over large scales, exceeding the scale of common decisions on land use or land devel-
opment (Nelson et al. 2009; Tallis and Polasky 2009; Polasky et al. 2011). The ability of
such tools to quantify changes in ecosystem service flows and values related to land use
also depends on the validity of the underlying ecological and economic models and the
coherence with which these models are combined.
Although established methods are available to quantify ecosystem service values
linked to land use, many commonly cited analyses are grounded in economic methods
that “generate misleading and potentially biased results” (Balmford et al. 2011, 167).
Frequent shortcomings include (1) extrapolating from a small number of unrepre-
sentative studies to entire biomes (Bockstael et al. 2000; Balmford et al. 2011); (2) the
common use of economic measures, such as replacement costs that “bear little resem-
blance to the values they [are used to] approximate” (Bockstael et al. 2000; Holland et al.
2010; Bateman et al. 2011, p. 191); (3) overlooking the dependence of economic values
on policy scope and scale (see discussion in Section 3 below), including analyses that
attempt to value entire ecosystems rather than marginal changes and that ignore con-
cepts such as diminishing marginal utility (Toman 1998; Bockstael et al. 2000; Fisher
et al. 2008; Rolfe et al. 2011); and (4) the use of simplistic unit-value benefit transfers
and other approaches that fail to meet minimal standards (Johnston and Rosenberger
2010; Bateman et al. 2011).
Whether these issues are of concern for policy applications depends on the level of
validity and accuracy deemed necessary in particular policy contexts (Navrud and
Pruckner 1997). As many of these shortcomings lead to quantitative value estimates that
have little basis in economic theory (reducing or eliminating validity) and incorporate
errors of unknown sign and magnitude (undermining credibility), economists tend to
view resulting estimates with skepticism. The challenge for the field moving forward is to
progress beyond illustrative analyses with fundamental economic or ecological flaws to

empirical works grounded in a more rigorous coordination of natural and social sciences.
2.3 Recognizing Marginality: At What Scopes and Scales Are

Ecosystem Services Evaluated?
For the sake of clarity, we here define scope as the quantity or quality of an ecosystem
service under consideration, whereas scale is defined as the geographic area over which
an analysis is conducted. Ecosystem service assessments are often conducted at large
scopes and scales; for example, for many ecosystem services, over large geographical
areas, often concerning nonmarginal changes or total values (Fisher et al. 2008). This is
distinct from much of the prior nonmarket valuation literature, which tends to address
resources at smaller scopes and scales.
The archetypal example is Costanza et al. (1997), which seeks to quantify and value
ecosystem services at a planetary scale. Unfortunately, “there is little that can be usefully
done with a serious underestimate of infinity” (Toman 1998, 58). Moreover,
One needs a specified baseline, a specified measure of changes, and a set of criteria
for evaluating and comparing these changes. A simple point aggregation of “every-
thing,” . . . give[s]‌no insights into either the directions of current changes in ecosys-
tems and their services or the relative urgency of different changes.
(Toman 1998, 58)
Statewide and nationwide analyses are also common (Nelson et al. 2009; Bateman
et al. 2011; Polasky et al. 2011), reflecting a tendency of the ecosystem services litera-
ture to seek results relevant to large scopes and scales (Turner and Daily 2008; Daily
et al. 2009). Although the grounding of analyses in economic theory and measurable
ecosystem service changes may ameliorate some of the concerns with these models, the
concerns of Toman (1998) remain relevant.
In contrast, the most precise empirical assessments in both economics and ecology
tend to involve evaluations at small scopes and scales. These address a few ecosystem
services, over small geographical areas, for marginal changes. The divergence between
the smaller scopes or scales at which research is most precise and the larger scopes or
scales at which ecosystem service information is often requested has contributed to
evaluations that have applied questionable, or at best oversimplified, methods to pro-
vide results over large scopes and scales (Toman 1998; Bockstael et al. 2000; Fisher et al.
2008; Rolfe et al. 2011). This emphasis on large scale, nonmarginal changes is often puz-
zling to economists because the role of such information in policy formation is ques-
tionable and analytical results may fail validity criteria (Toman 1998; Fisher et al. 2008).
Large-scale analyses may motivate an attitude change among political forces, potentially
fueling rhetoric for advocacy, but fail to provide insight necessary for better decisions or
to establish priorities for action requiring policy-scale input.
Many of the challenges identified in the preceding sections of this chapter are related,
at least indirectly, to the preoccupation of the ecosystem services literature on large
scopes and scales. Valid economic valuation methods, for example, quantify values for
clearly specified marginal changes (Bockstael et al. 2000; Fisher et al. 2008), thus meet-
ing Toman’s (1998) standard to work from a specified baseline and specified change.
Among other advantages, the evaluation over small, marginal changes allows the use
of partial equilibrium analyses that hold most factors in the outside world constant,
beyond those closely related to the change being valued (Just et al. 2004). As the defined
margin becomes larger, the accuracy of economic forecasts often declines (Fisher et al.
2008). Similarly, nonlinearities in ecological relationships can have important implica-
tions for ecosystem service value. The effects of these nonlinearities on ecosystem ser-
vice values are often more difficult to model and predict for nonmarginal ecological
changes, particularly in the presence of complicating factors such as ecological resil-
ience and thresholds (Hanley and Barbier 2009; Bateman et al. 2011).
Attempts to link ecosystem services to land use over large scopes and scales has also
contributed to a tendency to overlook heterogeneity in ecosystems, populations, and
policy contexts and to apply often rudimentary benefit transfers that overlook the criti-
cal influences of these differences on values (Bateman et al. 2011; Rolfe et al. 2011). For
example, the value of wetlands in filtering excess nutrients may be substantial when
located upstream of a source water reservoir, but may be inconsequential—or at least
valued for different reasons—in an area remote from human population centers. Recent
applications linking ecosystem services to land use have attempted to use somewhat
more advanced benefit transfer approaches that, in coordination with GIS and other spa-
tial tools, enable at least somewhat improved adjustments for differences in ecosystems,
populations, and policy contexts when conducting large-scale analyses (Bateman et al.
2011; Polasky et al. 2011). However, even these analyses require strong assumptions.
The perspective of marginality is often criticized for failing to recognize that nonlin-
earities may manifest as thresholds, thus creating potential shifts in ecological struc-
ture and function. Even in such situations, the disciplined analyst still establishes a
clearly defined baseline from which change might occur, assessing the value of potential
changes from that baseline. An analysis tailored to the level of the marginal decision
can be organized within an economic framework to facilitate informed decision mak-
ing, even if results of the decision cross nonmarginal thresholds within the ecological
sphere.
3. Connecting Ecosystem Services to

Land Use in Practice
Various analytical frameworks may be applied when evaluating ecosystem services.

One of the most obvious is valuation; this is often the ideal if science (ecology and
economics) can adequately support needed models. However, the data and modeling
requirements for valuation can sometimes be prohibitive. Moreover, in some instances,
valuation is not necessary to improve policy; what is needed instead is insight into rela-
tionships between ecosystem structures and functions and desired ecosystem outputs,
whether or not these outputs have quantified values (Wainger and Mazzotta 2011). This
leads to alternative approaches to research that link ecosystem services to land use.
This section attempts to present illustrative points, not necessarily endpoints, on a
continuum of approaches that link ecosystem services to land use. The illustrated case
studies purposefully involve nonvaluation frameworks that often receive less attention
in the ecosystem services literature. The first application outlines the use of a bioeco-
nomic model to land use controls. This approach occurs in the absence of quantified
knowledge about values a community might hold for ecosystem services, proceeding
on the presumption that the community has already established, perhaps through a col-
lective process, a desire to preserve particular services. Here, the crucial question is not
whether to preserve services or to what extent they are valued, but how preservation is
best accomplished. The second application illustrates potential ways to incorporate eco-
system service values into landowner choices. This case illustrates the potential to move
beyond valuation into market creation.
Returning to the three factors detailed in Section 2 (i.e., what services are evaluated,
how services are evaluated, and the scopes and scales at which services are evaluated),
the two case studies are explicit in terms of the outcomes that are evaluated (amphibian
metapopulations; hayfields supporting bobolink nesting), the ways in which outcomes
are evaluated (bioeconomic models, experimental markets), and the (small) scopes
and scales at which evaluations are implemented (amphibian metapopulations within
wetland patches in Richmond, Rhode Island; specific agricultural fields in Jamestown,
Rhode Island). The case studies do not, however, specify the full set of linkages through
which these specific ecological outcomes support different intermediate and final eco-
system services.
This potential shortcoming warrants additional emphasis—when is it necessary clar-
ify all linkages between intermediate and final services? Among the primary reasons to
clarify these linkages is the mitigation of double counting and development of compre-
hensive welfare estimates. As a result, clarification of these linkages is critical when wel-
fare estimation (i.e., valuation) is the primary goal. In the case studies illustrated here,
however, welfare estimation is not the goal. Hence, whether the ecological outcomes in
question reflect intermediate or final ecosystem services, or both, for particular benefi-
ciary groups is irrelevant to the research questions at hand—such information would
not directly enhance the relevance or validity of the research results.
This distinction highlights a critical point in analyses that link ecosystem services to
land use. That is, the variation in possible research and policy contexts implies that any
set of unyielding guidelines for analysis, including some of those presented earlier, may
have exceptions. What is most crucial in these heterogeneous contexts is the application
of methods that are transparent and valid from the perspective of the underlying natu-
ral and social sciences. The case study examples here are meant to illustrate different,
nonvaluation ways in which rigorous insight can be provided on ecosystem services

connected to land use without sacrificing the validity of the underlying economics or
ecology.
3.1 A Bioeconomic Modeling Approach

Bauer et al.’s (2010a, 2010b) work assesses whether familiar approaches to wetland
conservation, based on treating wetland patches as sacrosanct, achieves the highest
level of ecosystem services for a given set of development restrictions. Using an indica-
tor species of amphibians, the work shows that integration of opportunity costs into
policy design might either enable a higher level of ecosystem services at a given social
cost or enable a lower cost to achieve a targeted level of ecosystem services. Related
approaches are illustrated by Jiang et al. (2007), Montgomery et al. (1994), and Nalle
et al. (2004).
Wetland policy is often based on regulations that restrict development of habitat
patches, perhaps including a buffer zone, while leaving few restrictions on upland
development. Recent work in conservation biology questions the wisdom of this
approach, with authors such as Semlitsch (1998, 2007) and Semlitsch and Bodie
(1998) raising questions within the context of a patch-corridor model motivated by
metapopulation dynamics and concern for amphibian species that depend on sea-
sonal wetlands. Species that depend on dispersed habitat patches, such as a network
of wetlands, may exhibit a metapopulation structure whereby the subpopulations
occupying different patches (at least for breeding) depend on connectivity between
patches to enable long-term survival of the overall population; this connectivity
enables patches to serve as sources of colonists should environmental uncertainty
cause extinction of a subpopulation. The collection of such subpopulations comprises
the metapopulation.
Bauer et al. (2010a, 2010b) develop a baseline landscape modeled from the town of
Richmond, Rhode Island, drawing parameters to model a metapopulation of amphib-
ians both from ecological data from the study area and from the literature more broadly.
Hereafter, we will refer to the Bauer et al. papers as the BSP case study. The case study
area is approximately 100 km2 of predominantly mixed deciduous forest interspersed
with residential development and small village centers. The actual town is 24% devel-
oped land, 33% protected, and 43% undeveloped and unprotected, incorporating 214
biologically distinct vernal pool clusters (or seasonal wetland ponds). For the case study,
30 existing zoning districts were condensed to 16 neighborhoods, each one zoned for
one of five land uses (high, medium, and low density residential use; commercial and
industrial land; and agricultural land).
This context presumes that the community, albeit through a state-level statute, has
expressed a desire to maintain an ecologically viable network of wetlands capable
of maintaining species dependent on vernal pools (fish-free, seasonal ponds). The
existence of wetland protection laws constitutes an institutional or legal reaction to a

community value for and desire to maintain ecosystem services.12
Within this context, the BSP case study was designed to assess how wetland policy
might better reflect the role of both protected and developed land in landscape struc-
ture and function. BSP use a metapopulation model inspired by Hanski (1994) and
co-authors (e.g., Moilanen and Hanski 1998; Hanski and Ovaskainen 2000; Ovaskainen
and Hanski 2001) to measure ecosystem health using the landscape’s metapopulation
size, an analog to ecological carrying capacity based on the average probability that a
habitat patch is occupied by amphibians in a steady state. With this measure of ecosys-
tem quality, BSP assess the value of development that must be foregone to maintain a
targeted level of ecosystem quality. This approach enables a consideration of the ecologi-
cal and economic heterogeneity of the study area, thereby evaluating how flexibility in
wetland policy might improve the cost-effective provision of ecosystem services, mea-
sured using amphibians’ metapopulation capacity as an indicator.
Summarized in conceptual equations outlined below, the approach models ecosys-
tem connections across land parcels and how the intensity of development alters these
connections, thereby diminishing potential ecosystem health. The model’s output then
traces the link between ecosystem health and costs measured as the monetary value of
lost development opportunities. Such information enables a community to evaluate its
own values (subjectively) for ecosystem health relative to the costs of alternative sets of
land use restrictions that it might adopt. These costs are measured using current (2005)
real estate values and against a defined baseline, such as the maximum extent of devel-
opment permitted under current wetland and land use regulations.
3.1.1 Conceptual Model
The model includes details of the structural and functional roles served by land in habi-
tat patches and the land intervening between these patches, where the latter is described
as dispersal matrix, which determines connectivity between patches. The analyst first
identifies the effective area gi of habitat patch i as a function of the patch size Ai (acres);
its quality Hi, as indicated by field measurements and modeling of existing breeding
populations at patch i (e.g., Eagan and Paton 2004); and the quantity of development QAi
within the patch (which cannot exceed patch size):
g i = g ( Ai , Hi , QAi ). (1)

12
The specific services desired by the community may be defined with varying levels of precision.
Stated preference work in this community and others within Rhode Island shows that the residents
are motivated by the maintenance of biodiversity; quality of surface and ground waters, with purposes
including outdoor recreation; sustaining open space; and preserving the rural character of their
community (Johnston et al. 1999, 2002c, 2002d, 2003).
Larger area, higher quality, and less development increases the effective area of a hab-
itat patch. Connectivity between patches is critical to metapopulation survival and is
influenced by the distance din between patch i and any other patch n; the amount djin of
that distance that crosses matrix unit j; and the size Zj of dispersal matrix j. Along with
the quantity QZj0 of existing development, these factors establish the baseline conditions
for the extent to which the landscape presents amphibians with barriers between patches
i and n, and, in turn, these barriers are enhanced (dispersal matrix land is degraded) by
any additional quantity QZj of development. Thus, connectivity between patches i and n,
fin, depends on the permeability of the intervening matrix lands, which may be shared
across several land management units in the overall dispersal matrix, such that
fin = f ( Bin , din ), (2a)

with
Bin = b(dijn / din , QZj 0 , Z j ). (2b)

This does not imply that absolute barriers exist, but rather that the model captures
the degree to which the passage through matrix land is impermeable to migrating
juvenile amphibians; in this model, more development implies a lower rate of suc-
cessful dispersals. Connectivity is greater for lower distance between patches, with
that distance passing through matrix land that is more permeable to migrants; per-
meability decreases (barriers increase) when migrants must cross more hostile seg-
ments of dispersal matrix, which occurs with more existing or new development as a
proportion of total land within a management-unit j of dispersal matrix land of area
Zj. A mathematical matrix quantifying landscape structure brings the components of
the metapopulation model together. This mathematical landscape structure matrix
includes elements min that capture the impact of habitat patch n as a contributor of
immigrants to patch i when patch i is empty. This is based on species-specific immi-
gration and emigration rate parameters applied to the effective areas of both patches,
gi and gn, and the connectivity measure fin:
min = m( g i , g n , fin ). (3)
Here, patches of larger effective area produce more emigrants, receive more immi-
grants, and support a larger subpopulation which, in turn, reduces the probability
that the subpopulation goes (temporarily) extinct. The metapopulation model sup-
ports estimation of metapopulation size, Sm, under alternative configurations of
development, particularly new development, in patches (QAi) and dispersal matrix
units (QZj):
Sm = s(QAi , QZj , g i , g n , fin ) (4)

where this representation places the emphasis on the management controls of allowing
or limiting development within patches and dispersal matrix land units.
The challenge for managers is to identify restrictions that minimize the loss of devel-
opment opportunities while sustaining the ecosystem services of interest, as measured
via (or indexed by) metapopulation size. The decision maker minimizes the cost of allo-
cating land (perhaps through regulatory control, rather than purchase) to preservation,
considering both habitat patches and matrix lands:
min Opportunity Cost = ∑R

i =1, N
Ai QAi + ∑R
j =1, J
Zj QZj , (5)

where RAi and RZj represent the potential Ricardian rent or land value that a developer
could realize by conversion of a unit within habitat patches i and dispersal matrix units
j, respectively; and QAi and QZi represent the quantity of land chosen to be protected
within these patches and matrix units. Within the zoning context, land protection would
occur as a proportion of the respective undeveloped lands available, so that the sum
of existing and new development plus protection does not exceed the total area avail-
able in a patch (Ai) or dispersal matrix unit. In this model, development is distributed
evenly within a habitat patch or matrix unit, but spatial identity is maintained across
these units.
3.1.2 Evaluating Results for Policy Application

By minimizing opportunity costs in (5), subject to establishing a minimum level of eco-
system health via a minimum metapopulation size Sm0 using (4), this approach traces
out a minimum cost in relation to alternative metapopulation sizes. Figure 8.2 provides
a stylized illustration of the key results; the figure is illustrative and not quantitatively
definitive. In Figure 8.2, each point on the curve represents the minimum cost (vertical
axis) of providing for a given metapopulation size. Point A (Figure 8.2) represents the
metapopulation size that would exist if Richmond developed every legally allowed par-
cel under 2005 zoning, keeping in mind that development in Richmond generally main-
tains a wooded landscape, albeit with increased fragmentation, but sustaining some
connectivity and permeability to migrating amphibians. The model places metapopula-
tion size at 0.804 under this “full build-out” scenario, interpreted as an 80% chance that
a habitat patch is occupied in steady state, with an opportunity cost of zero (since the
baseline of “full build-out” would imply no new protection).
Model results suggest that a steep increase in costs occurs at approximately a 0.95
value of metapopulation size, reflecting the challenge of mitigating development
impacts when pristine conditions no longer apply. Nonetheless, a political or social
decision process might establish a much more stringent level of metapopulation capac-
ity, compared to 0.804 achieved under full build out. The common approach to protec-
tion of wetlands treats all wetlands the same, setting an across-the-board standard. In
the present model, one such standard would be to establish a 165 m buffer around each
Opportunity
Cost, $
B C
• •
D
A
•
•
Metapopulation
Size, Sm
FIGURE 8.2 Stylized results of bioeconomic model estimating the opportunity cost of achiev-
ing a given metapopulation size. Bauer et al. (2010b) provide an empirical example of this
concept diagram, whereas Bauer et al. (2010a) provide estimated traces of the opportunity
cost curve under various policy assumptions.
wetland and protect all land within that buffer from any further development, consis-
tent with biological delineation of key habitat, for example, following Semlitsch (1998).
Such a policy would leave aside the issue of development in matrix lands. Bauer et al.
(2010b) calculate the implications of such a policy (derived without consideration of
economic factors), obtaining a result analogous to point B in Figure 8.2, which would
achieve a metapopulation size of 0.924 at an opportunity cost of around $101 million.
The key observation of the analysis is that a point such as B leaves at least two ways to
do better for human well-being and ecosystems: (1) one may achieve the same 0.924
ecological quality at substantially lower cost of foregone development benefits (perhaps
just 10% of $101 million), moving to a point like D in Figure 8.2; or (2) one may achieve
greater ecological quality if the community chooses to endure the given cost, moving
to a point like C in Figure 8.2, perhaps approaching the limit of 0.979 under a policy of
fully protecting all remaining patch and matrix lands (Bauer et al. 2010b).13
Moving down and to the right, toward the minimum cost curve between C and D
(Figure 8.2) would improve in both dimensions of cost and metapopulation size.
However, it is also important to recognize that the model could allow a zoning con-
straint to be established either separately for matrix lands and land in habitat patches
(as defined by a buffer zone) across the board or separately for patches and matrix land
13 Estimated to incur an opportunity cost of $463 million (Bauer et al. 2010b).

within different management units. For example, Bauer et al. (2010a, 805–806) show
that, compared to the full build-out (no protection) scenario, protecting 50% of habitat
patches (within a 165 m buffer around vernal pools) would achieve a higher metapopu-
lation size (nearly 0.90 versus about 0.87) at lower opportunity cost (about $50 million
vs. more than $100 million) as compared to protecting 25% of undeveloped dispersal
matrix (recognizing that total land area in the matrix is larger than total land in these
patches). The model allows assessment of further tradeoffs between protecting land
in different roles. For example, protecting 25% of both patch and matrix land yields a
metapopulation size of 0.922 at a cost of $163 million; moving from that base position to
100% protection of patch lands along with 25% of matrix lands increases metapopula-
tion size by 0.036 to 0.958 at an additional cost of $95 million. In comparison, moving
to 100% protection of dispersal land while remaining with 25% protection of patch land
raises metapopulation size by 0.044 to 0.966 at an additional cost of $395 million.
Such examples may oversimplify the perspective of human benefits from environ-
mental quality. Bauer et al. (2010b) provide an example in which a metapopulation size
of 0.95 occurs with partial or full protection of a cluster of habitat patches in the center
of Richmond, leaving the vast majority of Richmond in 15 other zones (outside this core
area) with little or no additional land conservation over current policy; they provide
a modified model establishing a metapopulation constraint separately for five subsec-
tions of the town in an effort to assess alternatives across the town’s landscape. Adding
these additional constraints raises the cost-minimizing options at each metapopulation
size (from around $12 million to approaching $50 million at Sm of 0.95), but if the com-
munity establishes goals to assure a broad, geographic distribution of opportunities for
residents to live within a landscape that produces ecosystem services at a minimal level,
these higher costs might prove acceptable.
Throughout this example, the specific ecosystem services provided are those associ-
ated with land conservation that promotes the longevity of amphibian metapopulations.
The ecosystem service benefits, however, include many beyond those associated solely
with amphibians. These may include services associated with a lower density of devel-
opment and broader geographic distribution of undeveloped lands. Benefits may also
include provision of habitat for a variety of other species, including birds or wildlife that
cohabitate the forested patches and undeveloped matrix, or conservation of groundwa-
ter recharge through preserved wetlands, among others. A full catalogue of these ser-
vices would improve the information base for land use decisions. Yet, even without this
information, the bioeconomic model illustrates opportunities to enhance both the eco-
nomic and ecological outcomes of land preservation as realized through the provision
of ecosystem health and services.
3.2 Beyond Valuation, Toward Markets for Ecosystem Services

Beyond the valuation of ecosystem services, there is increasing interest in ways to inte-
grate these values directly into the economy through markets, regulatory means, or
other mechanisms. Indeed, society as a whole spends little time seeking to quantify the
value of market goods because markets automatically establish prices effectively, even in
the presence of variation in local supply and demand. If policy innovation could estab-
lish similar markets for currently nonmarket ecosystem services, many of the challenges
of valuation might vanish; society could make more efficient decisions about ecosystem
services if markets more fully incorporated ecosystem service values. The second appli-
cation illustrates research that addresses this market challenge.
Swallow et al. (2008) describe experimental work that uses incentive mechanisms
and rules of exchange to generate revenues for ecosystem services specifically linked to
the management of farm hayfields as nesting bird habitat. Their application concerns
the conflict between the nesting season for bobolinks (Dolichonyx oryzivorus), a neo-
tropical migrant, and the peak nutritional value of hay as feed for livestock in north-
eastern US farms. In this study, the investigators served as brokers between farmers in
Jamestown, Rhode Island, and nonfarm households in this rural fringe community.
They established contracts through which farmers would agree to modify their harvest
during late May to early July in order to avoid impacts on bobolink nesting. Farmers
would give up at least one hay harvest so that nests in that field could produce fledglings.
Harvesting during nesting and prior to fledging of young birds has been shown to gen-
erate nest mortality rates of nearly 100% (e.g., Bollinger et al. 1990). In turn, the investi-
gators asked Jamestown residents to pay for these contracts. The study assigned groups
of households to a particular hayfield. Within each group, the independent decisions of
members regarding whether to buy into the contract determined whether the farmer for
that field was paid for bobolink management or whether the contract was dropped, allow-
ing the farmer to proceed with his or her normal harvesting plans.
The study thus implemented a voluntary market exchange. The researchers served the
role of brokers, raised revenues from town residents, and paid farmers for bird-friendly hay-
field management. In practice, such an exchange might be developed through independent
private action, by a for-profit or not-for-profit broker, or it could be aided by a government
subsidy providing a share of costs for a contract but requiring the remaining share to be gen-
erated from revenues contributed by local residents. An exchange such as this could reveal
at least some of the value for ecosystem services. Even with imperfect rules of exchange and
incentives (e.g., that allow some beneficiaries to free ride), the approach begins to develop
market mechanisms for ecosystem services in a way that improves human welfare.
Within this study, marketing materials placed the habitat services of a 10-acre hay-
field during a single season for nesting bobolinks as the focal point, including the slo-
gan “Their home, your hometown.” Marketing materials also noted the potential of
hayfields to maintain agrarian heritage or open-field views, as well as noting potential
linkages to water quality and carbon sequestration, and some Jamestown residents may
have considered these or other aspects of ecosystem services. However, what is clear
is that households were asked to contribute toward the specific action of bird-friendly
management of a specified size hayfield. A strength of this example is that the hayfield
contracts represent a specific action, or good, that households could consider buying
into for the provision of ecosystem services linked to a particular land use; the change
was clearly defined. The ability to at least partially capture resident’s willingness to pay
(WTP)—and thereby improve socially valuable outcomes—does not necessarily require
analysts to define beforehand the particular ecosystem services or quantities of these
services potentially affected, although the failure to do so may influence the ability of the
market-maker to identify an optimal level of provision.
The research was designed to identify the value of bobolink contracts to residents of
Jamestown and to identify a set of rules of exchange that might lead residents to offer a
higher percentage of their full value as a financial contribution to support a contract with
a farmer. Drawing on the experimental economics literature, the bobolink contracts were
presented under a variety of provision point mechanisms (Poe et al. 2002; Spencer et al.
2009). In contrast to a pure, open-ended donations approach used by charitable organiza-
tions, provision point mechanisms establish a minimum, target level of aggregate fund-
ing that must be achieved before provision of a unit of the public good is assured. In this
study, the unit was a 10-acre hayfield contracted for bird-friendly management during a
single summer nesting season. If individuals from a group of households failed to make
aggregate contributions that meet or exceed the cost of the farm contract14 (the provision
point), the farm field would be released from the contract, and all contributions would
be refunded to the households providing the funds. Rondeau et al. (1999, 2005; cf., Poe
et al. 2002) show that the provision point reduces incentives to free-ride15 by establishing
a rudimentary threat of nonprovision, as well as by assuring that contributions pay for
a specified unit of the good; Swallow et al. (2008) implemented this provision point by
defining a finite group of households to determine the outcome for each field.
The experimental market also incorporated marginal incentives to reduce the advan-
tages of free-riding. A primary example is a proportional rebate method tested in labo-
ratory experiments by Marks and Croson (1998; Spencer et al. 2009). The experimental
market tested other rules of exchange, all involving a provision point, but each differing
in the marginal incentives created by their rebate rule. Other rebate rules established a
“price cap” or uniform price that was endogenously determined by aggregate contribu-
tions within a group as the lowest amount that could be used as maximum charge while
still meeting the provision point, such that individuals making high offers would receive
a rebate of money offered above the price cap. The proportional rebate and uniform price
mechanisms were intended to raise revenues for actual provision of hayfields by private
entrepreneurs. An alternative mechanism, the “pivotal mechanism,” was not designed
to raise revenues but rather to estimate the potential willingness to pay of participants.16
14
In the experiments of Swallow et al. (2008), the actual provision points in some cases reflected a
subsidy from funds obtained outside of the contributions of households. However, those subsidies were
determined prior to solicitations to households, and the provision point reflected the net cost of the farm
contract after any subsidy.
15
Under a free-riding strategy, the beneficiary contributes zero or “cheap rides” by contributing less
than their full value in an effort to benefit from the contributions that others make to facilitate provision
of a public good, such as habitat services from hayfields.
16 The mechanism is in a class that is weakly incentive compatible. Kawogoe and Mori (2001) note the
advantages and disadvantages of this class of mechanisms.

Percentage of 2008 participants WTP > Amount, in Discrete Choice
100
Proportional
90 rebate
80 Pivotal
Percentage of participants
mechanism
70
Uniform price
60 auction
Uniform price
50 cap
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Amount ($)
FIGURE 8.3 Proportion (%) of participants willing to offer a fixed amount for a 10-acre hay-
field to be managed for one nesting season (unpublished data and analysis of S. Swallow,
C. Anderson, and E. Uchida).
Figure 8.3 shows the proportion of participants in the Jamestown bobolink experi-
ment who might be willing to offer a specified amount of money to support a single hay-
field for one nesting season. Each of these “willingness to offer” curves corresponds to
one of the rebate rules being tested in the experiment. Although these curves are based
on an econometric model, their distribution may not necessarily prove to be statistically
different. Therefore, Figure 8.3 should be considered illustrative rather than definitive.
In this context, a participant is an individual who responded with a definitive answer to
a direct mail solicitation; about 10–12% of Jamestown residents responded, with about
two-thirds of participants making a particular offer amount via personal check or credit
card authorization. Here, depending on the marginal incentives presented to a partici-
pant, 70% of participants appear to be willing to offer between $52 and $62 to support
the seasonal contract under the best performing marginal incentives (about $43 under
the worst performing marginal incentives).
Figure 8.4 shows how these results might translate to revenue for an entrepreneur
attempting to optimize her price point. These results suggest that price points of
between $50 and $62 might generate the highest level of revenue for a hayfield, account-
ing for the number of participants who might decline to pay that amount but who may
have paid a lower amount if given the opportunity. Based on these data, approximately
125 to 150 participants could cover the provision point on the marginal farm contract
in Jamestown (requiring a revenue of about $5,000 if unsubsidized), whereas in other
farm communities, such as in upstate New York or in Vermont, these provision points
might be substantially lower based on land use allocations and alternative availability of
feedstock.
5000
Proportional
4500 rebate
4000 Pivotal
mechanism
3500
Uniform price
3000 auction
Revenue ($)
Uniform price
2500
cap
2000
1500
1000
500
0
0 10 20 30 40 50 60 70 80 90 100
Amount ($)
FIGURE 8.4 Point estimates of projected revenue from 100 participants in the Jamestown
bobolink experiment versus a given fixed amount requested from all participants (unpub-
lished data and analysis of S. Swallow, C. Anderson, and E. Uchida).
This illustration demonstrates ways in which one might potentially capitalize on

economic insights and methods to improve the efficiency of ecosystem service provi-
sion without the need to necessarily quantify, value, or categorize the specific ecosys-
tem services. It shows an example, more at the extreme end of the continuum of market
approaches, that attempts to develop revenue-generating methods for ecosystem ser-
vices. Services such as these are traditionally (under)provided either as the fortuitous
by-products of normal land use, through government regulatory action, or perhaps
through government incentive programs such as the USDA’s conservation reserve
program payments. In contrast to approaches that depend on centralized decisions or
generic ties between consumer decisions and ecosystem outcomes, such as through
eco-certification programs for traditional market goods, the market approach repre-
sented here has the advantage of linking consumer values and payment to specific actions.
Successful approaches to marketing ecosystem services could obviate much of
the need for valuation of ecosystem services, thus leveraging entrepreneurial action
to more completely recognize the contribution of ecosystems to human welfare.
Unfortunately, this avenue of research faces its own challenge, including that of over-
coming the incentives that beneficiaries have to free-ride on the contributions of oth-
ers. Those incentives form a classic basis for market failure, which itself has spawned
the need for valuation research. As noted by Kroeger and Casey (2007), ecosystem ser-
vices related to land use often lack characteristics necessary for markets to assign prices
linked to social benefits. Experience with markets and payments for these services pro-
vides has yet to provide clear evidence that such approaches can enhance ecosystem
service provision on a broad scale, beyond services that would have been provided in
the absence of these programs (Kroeger and Casey 2007; Claassen et al. 2008, Bauer
and Johnston 2013 ).
4. Conclusion
The topics presented in this chapter may be consolidated into two broad themes.
First, the ecosystem services perspective and associated research can provide signifi-
cant insight to inform land use and policy. Ecosystem services research is grounded in
decades of work by economists, ecologists, and others seeking to model interactions
between humans and ecosystems. Much of the best ecosystem services research can be
considered a well-grounded evolution of these existing methods. From this perspective,
the validity of the underlying methods has already been established, and the key chal-
lenge is the recasting of results within an ecosystem services framework.
The second theme is that the integration of economics and ecology required by the
ecosystem services framework can impose nontrivial challenges. These challenges
expand as one moves beyond well-defined, marginal analyses to more revolutionary
attempts to characterize all services, linkages, and values in large-scale systems. Even
though the underlying economics and ecology may be well-developed, the integration
and scale of these models can require methodological sacrifices that threaten validity.
Given such tradeoffs, an ecosystem services framework may not always be the most
informative. Analysts must consider when and where the framework is appropriate ver-
sus those cases in which alternative methods are sufficient. The theoretical sections of
this chapter have sought to identify some key considerations in the validity of ecosystem
services research as a means to help answer these questions. Ecosystem services per-
spectives should be used when they can enhance the guidance that economists and oth-
ers can provide to the policy process. When this is not the case, these perspectives may
be unnecessary or even counterproductive. There are many cases in which careful use
of integrated economic and ecological thinking alone can improve the foundation for
decisions or in which market-based solutions can ameliorate society’s undervaluation
of ecosystems, even in the absence of operational ecosystem service frameworks.
Grounded in these themes, the two case studies illustrate both the insights provided
by, as well as the empirical challenges of, research linking ecosystem services to land use.
These illustrations reflect evolutionary rather than revolutionary work; for example, the
case studies fall short of providing a comprehensive linkage among land use, ecosystem
services, and human welfare. Such patterns are common in contemporary research that
seeks to establish linkages between ecosystem services and land use; a substantial pro-
portion of research in the field is best characterized as “proof of concept.” Among the
initial challenges in improving on research “still in its nascent stage” (Fisher et al. 2011)
is the establishment of improved coordination across the natural and social sciences,
from project initiation to completion, to promote empirical work widely accepted as
both valid and relevant (Bauer and Johnston 2013). We also purposefully avoid illustra-
tions focused on explicit valuation, instead illustrating alternative research paradigms
through which linkages between ecosystem services and land use may be used to inform
policy.
Moving forward, it is critical to acknowledge both the challenges and benefits of

research linking ecosystem services to land use. Recognizing Toman’s (1998) admo-
nition, a willingness to sacrifice scientific rigor in an effort to demonstrate that all of
nature is valuable risks the provision of useless or misleading information. In contrast,
well-conceptualized ecosystem services analysis—whether through valuation or other
perspectives—can help clarify tradeoffs necessary to enhance the benefits that humans
realize from ecosystems and motivate concomitant policies. These tradeoffs may involve
those aspects of nature whose degradation enhances human welfare, as well as those
for which the benefits of nature’s services warrant greater conservation. We recognize
that limitations of the scientific state-of-the-art for ecosystem services research may
leave some researchers desirous of a more complete set of revolutionary analytical tools.
Despite these temptations, we urge greater efforts in the careful evolutionary develop-
ment and evaluation of these tools, focusing on validity and provision of information
relevant to decisions at hand.
Acknowledgments
This chapter draws on work that was originally supported by the EPA STAR Decision
Making and Valuation for Environmental Policy grant (R829384) and STAR Graduate
Fellowship; USDA/CSREES/NRI Grant 2002-35401-11657 and USDA/NIFA/AFRI
grants 2009-55401-05038 and 2011-67023-30378, a USDA/NRCS Conservation
Innovation Grant, and the URI and UConn Agricultural Experiment Stations.
References
Balmford, A., B. Fisher, R. E. Green, R. Naidoo, B. Strassburg, R. K. Turner and A. S. L.
Rodrigues. 2011. Bringing ecosystem services into the real world: An operational frame-
work for assessing the economics consequences of losing wild nature. Environmental and
Resource Economics 48(2): 161–175.
Bateman, I. J., G. M. Mace, C. Fezzi, G. Atkinson, and K. Turner. 2011. Economic analysis for
ecosystem service assessments. Environmental and Resource Economics 48(2): 177–218.
Batie, S. S. 2003. The multifunctional attributes of northeastern agriculture: A research agenda.
Agricultural and Resource Economics Review 32(1): 1–8.
Bauer, D. M. and R. J. Johnston. 2013. The economics of rural and agricultural ecosystem ser-
vices: Purism versus practicality. Agricultural and Resource Economics Review 42 (1): iii–xv.
Bauer, D. M., P. W. C. Paton, and S. K. Swallow. 2010a. Are wetland regulations cost effective
for species protection? A case study of amphibian metapopulations. Ecological Applications
20 (3): 798–815.
Bauer, D. M., S. K. Swallow, and P. W. C. Paton. 2010b. Cost-effective species conservation in
exurban communities: A spatial analysis. Resource and Energy Economics 32:180–202.
Bergstrom, J. C., and R. C. Ready. 2009. What have we learned from over 20 years of farmland
amenity valuation research in North America? Review of Agricultural Economics 31(1): 21–49.
Bockstael, N. E., A. M. Freeman, R. J. Kopp, P. R. Portney, and V. K. Smith. 2000. On measuring
economic values for nature. Environmental Science and Technology 34(8): 1384–1389.
Bollinger, E. K., P. B. Bollinger, and T. A. Gavin. 1990. Effects of hay-cropping of eastern popula-
tions of the bobolink. Wildlife Society Bulletin 18: 143–150.
Boody, G., B. Vondracek, D. A. Andow, M. Krinke, J. Westra, J. Zimmerman, and P. Welle. 2005.
Multifunctional agriculture in the United States. BioScience 55: 27–38.
Boyd, J., and S. Banzhaf. 2007. What are ecosystem services? The need for standardized envi-
ronmental accounting units. Ecological Economics 63(2–3): 616–626.
Boyd, J., and A. Krupnick. 2013. Using ecological production theory to define and select envi-
ronmental commodities for nonmarket valuation. Agricultural and Resource Economics
Review 42(1): 1–32.
Boyer, T., and S. Polasky. 2004. Valuing urban wetlands: A review of nonmarket valuation stud-
ies. Wetlands 24(4): 744–755.
Brown, T. C., J. C. Bergstrom, and J. B. Loomis. 2007. Defining, valuing and providing ecosys-
tem goods and services. Natural Resources Journal 47(2): 329–376.
Carbone, J. C., and V. K. Smith. 2008. Evaluating policy interventions with general equilibrium
externalities. Journal of Public Economics 92(5–6): 1254–1274.
Carbone, J. C., and V. K. Smith. 2010. Valuing ecosystem services in general equilibrium. NBER
Working Paper No. w15844.
Champ, P. A., K. J. Boyle, and T. C. Brown, eds. 2003. A primer on nonmarket valuation.
Dordrecht, The Netherlands: Kluwer Academic.
Clark, C. W. 1976. Mathematical bioeconomics: The optimal management of renewable resources.
New York: John Wiley & Sons.
Claassen, R., A. Cattaneo, and R. Johansson. 2008. Cost-effective design of agri-environmental
payment programs: U. S. experience in theory and practice. Ecological Economics 65:
737–752.
Collins, S. L., S. R. Carpenter, S. M. Swinton, D. E. Orenstein, D. L. Childers, T. L. Gragson, N.
B. Grimm, J. M. Grove, S. L. Harlan, J. P. Kaye, A. K. Knapp, G. P. Kofina, J. J. Magnuson, W.
H. McDowell, J. M. Melack, L. A. Ogden, G. P. Robertson, M. D. Smith, and A. C. Whitmer.
2010. An integrated conceptual framework for long-term social-ecological research.
Frontiers in Ecology and the Environment doi:10.1890/100068.
Conrad, J. M. and C. W. Clark. 1987. Natural resource economics: Notes and problems. Cambridge,
UK: Cambridge University Press.
Costanza, R., R. d’Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S. Naeem,
R. V. O’Neill, J. Paruelo, R. G. Raskin, P. Sutton, and M. van den Belt. 1997. The value of the
world’s ecosystem services and natural capital. Nature 387 (May): 253–260.
Daily, G. C. 1997. Nature’s services. Covelo, CA: Island Press.
Daily, G. C., S. Polasky, J. Goldstein, P. M. Kareiva, H. A. Mooney, L. Pejchar, T. H. Ricketts,
J. Salzman, and R. Shallenberger. 2009. Ecosystem services in decision making: Time to
deliver. Frontiers in Ecology and Environment 7(1): 21–28.
Dale, V. H., and S. Polasky. 2007. Measures of the effects of agricultural practices on ecosystem
services. Ecological Economics 64: 286–296.
de Groot, R. S., M. A. Wilson, and R. M. J. Boumans. 2002. A typology for the classification,
description and valuation of ecosystem functions, goods and services. Ecological Economics
41(3): 393–408.
Duke, J. M., and R. J. Johnston. 2010. Nonmarket valuation of multifunctional farm and forest
preservation. In New perspectives on agri-environmental policies: A multidisciplinary and
transatlantic approach, eds. S. J. Goetz and F. Brouwer, 124–142. Oxford: Routledge.
Eagan, R. S., and P. W. C. Paton. 2004. Within-pond parameters affecting oviposition by wood
frogs and spotted salamanders. Wetlands 24:1–13.
Finnoff, D., A. Strong, and J. Tschirhart. 2008. Stocking regulations and the spread of invasive
species. American Journal of Agricultural Economics 90(4):1074–1084.
Finnoff, D., and J. Tschirhart. 2008. Linking dynamic ecological and economic general equilib-
rium models. Resource and Energy Economics 30(2):91–114.
Fisher, B., S. Polasky, and T. Sterner. 2011. Conservation and human welfare: Economic analysis
of ecosystem services. Environmental and Resource Economics 48(2): 151–159.
Fisher, B., K. Turner, M. Zylstra, R. Brouwer, R. de Groot, S. Farber, P. Ferraro, R. Green, D. Hadley,
J. Harlow, P. Jefferiss, C. Kirkby, P. Morling, S. Mowatt, R. Naidoo, J. Paavola, B. Strassburg, D.
Yu, and A. Balmford. 2008. Ecosystem services and economic theory: Integration for policy
relevant research. Ecological Applications 18(8): 2050–2067.
Fisher, B., R. K. Turner, and P. Morling. 2009. Defining and classifying ecosystem services for
decision making. Ecological Economics 68(3): 643–653.
Freeman, A. M. III. 2003. The measurement of environmental and resource values: Theory and
methods. Washington, DC: Resources for the Future.
Gardner, B. D. 1977. The economics of agricultural land preservation. American Journal of
Hanley, N., and E. B. Barbier. 2009. Pricing nature: Cost benefit analysis and environmental
policy. Cheltenham, UK: Edward Elgar.
Hanski, I. 1994. A practical model of metapopulation dynamics. Journal of Animal Ecology 63:
151–162.
Hanski, I., and O. Ovaskainen. 2000. The metapopulation capacity of a fragmented landscape.
Nature 404: 755–758.
Heal, G. M., and A. A. Small. 2002. Agriculture and ecosystem services. In Handbook of agricul-
tural economics, eds. B. L. Gardner and G. C. Rausser. Amsterdam: Elsevier.
Holland, D. S., J. Sanchirico, R. J. Johnston, and D. Joglekar. 2010. Economic analysis for eco-
system based management: Applications to marine and coastal environments. Washington,
DC: RFF Press.
Jiang, Y., S. K. Swallow, and P. W. C. Paton. 2007. Designing a spatially-explicit nature reserve
network based on ecological functions: An integer programming approach. Biological
Conservation 140: 236–249.
Johnston, R. J., D. M. Bauer, and S. K. Swallow. 2002c. Spatial factors and stated preference
values for public goods: Considerations for rural land use. Land Economics 78(4): 481–500.
Johnston, R. J., and J. M. Duke. 2007. Willingness to pay for agricultural land preservation
and policy process attributes: Does the method matter? American Journal of Agricultural
Economics 89(4): 1098–1115.
Johnston, R. J., T. A. Grigalunas, J. J. Opaluch, J. Diamantedes, and M. Mazzotta. 2002b. Valuing
estuarine resource services using economic and ecological models: The Peconic estuary sys-
tem study. Coastal Management 30(1): 47–66.
Johnston, R. J., J. J. Opaluch, T. A. Grigalunas, and M. J. Mazzotta. 2001. Estimating amenity
benefits of coastal farmland. Growth and Change 32(3): 305–325.
Johnston, R. J., G. Magnusson, M. Mazzotta, and J. J. Opaluch. 2002a. Combining economic

and ecological indicators to prioritize salt marsh restoration actions. American Journal of
Johnston, R. J., and R. S. Rosenberger. 2010. Methods, trends and controversies in contempo-
rary benefit transfer. Journal of Economic Surveys 24(3): 479–510.
Johnston, R. J., and M. Russell. 2011. An operational structure for clarity in ecosystem service
values. Ecological Economics 70(12): 2243–2249.
Johnston, R. J., Schultz, E. T., Segerson, K., Besedin, E. Y., and Ramachandran, M. 2012.
Enhancing the content validity of stated preference valuation: The structure and function of
ecological indicators. Land Economics 88(1): 102–120.
Johnston, R. J., K. Segerson, E. T. Schultz, E. Y. Besedin, and M. Ramachandran. 2011. Indices
of biotic integrity in stated preference valuation of aquatic ecosystem services. Ecological
Economics 70(11): 1946–1956.
Johnston, R. J., and S. K. Swallow, eds. 2006. Economics and contemporary land use pol-
icy: Development and conservation at the rural-urban fringe. Washington, DC: RFF Press.
Johnston, R. J., S. K. Swallow, C. W. Allen, and L. A. Smith. 2002d. Designing multidimensional
environmental programs: Assessing tradeoffs and substitution in watershed management
plans. Water Resources Research 38(7): IV 1–13.
Johnston, R. J., S. K. Swallow, D. M. Bauer, and C. M. Anderson. 2003. Preferences for residen-
tial development attributes and support for the policy process: Implications for manage-
ment and conservation of rural landscapes. Agricultural and Resource Economics Review
32(1): 65–82.
Johnston, R. J., S. K. Swallow, and T. F. Weaver. 1999. Estimating willingness to pay and resource
trade-offs with different payment mechanisms: An evaluation of a funding guarantee for
watershed management. Journal of Environmental Economics and Management 38(1):
97–120.
Josling, T. 2002. Competing paradigms in the OECD and their Impact on the WTO agricul-
ture talks. In Agricultural Policy for the 21st Century, eds. L. Tweeten and S. R. Thompson,
245–264. Ames: Iowa State University Press.
Just, R. E., D. L. Hueth, and A. Schmitz. 2004. The welfare economics of public policy: A practical
approach to project and policy evaluation. Cheltenham, UK: Edward Elgar.
Kawagoe, T., and T. Mori. 2001. Can the pivotal mechanism induce truth-telling? An
Experimental study. Public Choice 108: 331–354.
Kroeger, T., and F. Casey. 2007. An assessment of market-based approaches to providing eco-
system service on agricultural lands. Ecological Economics 64: 321–332.
Krutilla, J. V. 1967. Conservation reconsidered. American Economic Review 57(4): 777–786.
Marks, M., and R. Croson. 1998. Alternative rebate rules in the provision of a threshold public
good: An experimental investigation. Journal of Public Economics 67(2): 195–220.
Moilanen, A., and I. Hanski. 1998. Metapopulation dynamics: Effects of habitat quality and
landscape structure. Ecology 79: 2503–2515.
Millennium Ecosystem Assessment. 2005. Ecosystems and human well-being: Synthesis.
Washington, DC: Island Press.
Montgomery, C. A., G. M. Brown Jr., and D. M. Adams. 1994. The marginal cost of species pres-
ervation: The northern spotted owl. Journal of Environmental Economics and Management
26: 111–128.
Naidoo, R., and T. H. Ricketts. 2006. Mapping the economic costs and benefits of conservation.
PLoS Biology 4(11): E360. doi: 10.1371/journal.pbio.0040360.
Nalle, D. J., C. A. Montgomery, J. L. Arthur, N. H. Schumaker, and S. Polasky. 2004. Modeling

joint production of wildlife and timber in forests. Journal of Environmental Economics and
Management 48(3): 997–1017.
National Research Council. 2005. Valuing ecosystem services: Towards better environmental
decision-making. Washington, DC: National Academies Press.
National Research Council. 2010. Toward sustainable agricultural systems in the 21st century.
Washington, DC: National Academies Press.
Navrud, S., and G. J. Pruckner. 1997. Environmental valuation—to use or not to use? A com-
parative study of the United States and Europe. Environmental and Resource Economics
10(1): 1–26.
Nelson, E., G. Mendoza, J. Regetz, S. Polasky, H. Tallis, D. R. Cameron, K. M. Chan, G. C. Daily,
J. Goldstein, P. M. Kareiva, E. Lonsdorf, R. Naidoo, T. H. Ricketts, and M. R. Shaw. 2009.
Modeling multiple ecosystem services, biodiversity conservation, commodity production,
and tradeoffs at landscape scales. Frontiers in Ecology and Environment 7: 4–11.
Ovaskainen, O., and I. Hanski. 2001. Spatially structured metapopulation modes: Global and
local assessment of metapopulation capacity. Theoretical Population Biology 60: 281–302.
Poe, G. L., J. E. Clark, D. Rondeau, and W. D. Schulze. 2002. Provision point mechanisms and
field validity tests of contingent valuation. Environmental and Resource Economics 23(1):
105–131.
Polasky, S. E. Nelson, D. Pennington, and K. A. Johnson. 2011. The impact of land-use change
on ecosystem services: A case study in the state of minnesota. Environmental and Resource
Economics 48(2): 219–242.
Priess, J. A., M. Mimler, A.-M. Klein, S. Schwartze, T. Tscharntke, and I. Steffan-Dewenter.
2007. Linking deforestation scenarios to pollination services and economic returns in coffee
agroforestry systems. Ecological Applications 17(2): 407–417.
Ready, R. C., and C. Abdalla. 2005. The amenity and disamenity impacts of agriculture: Estimates
from a hedonic pricing model. American Journal of Agricultural Economics 87(2): 314–326.
Ribaudo, M., C. Greene, L. Hansen, and D. Hellerstein. 2010. Ecosystem services from agricul-
ture: Steps for expanding markets. Ecological Economics 69: 2085–2092.
Ricketts, T. H., G. C. Daily, P. R. Ehrlich, and C. D. Michener. 2004. Economic value of tropi-
cal forest to coffee production. Proceedings of the National Academy of Sciences of the USA
101: 12579–12582.
Rolfe, J., J. Bennett, R. Johnston and G. Kerr. 2011. Using benefit transfer to inform environmen-
tal policy making. Environmental Economics Research Hub Policy Brief. Crawford School
of Economics and Government, Australian National University.
Rondeau, D., G. L. Poe, and W. D. Schulze. 2005. VCM or PPM? A comparison of the perfor-
mance of two voluntary public good mechanisms. Journal of Public Economics 89(8):1581–
1592; doi: 10.1016/j.jpubeco.2004.06.014.
Rondeau, D., W. D. Schulze, and G. L. Poe. 1999. Voluntary revelation of the demand for public
goods using a provision point mechanism. Journal of Public Economics 72(3): 455–470.
Semlitsch, R. D. 1998. Biological delineation of terrestrial buffer zones for pond-breeding
amphibians. Conservation Biology 12: 1113–1119.
Semlitsch, R. D. 2007. Differentiating migration and dispersal processes for pond-breeding
amphibians. Journal of Wildlife Management 72: 260–267.
Semlitsch, R. D., and J. R. Bodie. 1998. Are small, isolated wetlands expendable? Conservation
Biology 12: 1129–1133.
Simpson, R. D. 1998. Economic analysis and ecosystems: Some concepts and issues. Ecological
Applications 8: 342–349.
Spencer, M. A., S. K. Swallow, J. F. Shogren, and J. A. List. 2009. Rebate rules in threshold public
good provision. Journal of Public Economics 93: 798–806; doi:10.1016/J.jpubeco.2009.01.005.
Swallow, S. K. 1994. Renewable and nonrenewable resource theory applied to coastal agricul-
tural, forest, wetland, and fishery linkages. Marine Resource Economics 9: 291–310.
Swallow, S. K. 1996a. Resource capital theory and ecosystem economics: Developing nonre-
newable habitats of heterogeneous quality. Southern Economic Journal 6(1): 106–123.
Swallow, S. K. 1996b. Economic issues in ecosystem management: An introduction and over-
view. Agricultural and Resource Economics Review 25(2): 83–100.
Swallow, S. K. 1997. Biodiversity loss: Economic and ecological issues, eds. C. Perrings, K. -G.
Mäler, C. Folke, C. S. Holling, and B.-O. Jansson. Cambridge: Cambridge University Press,
1995, ix, 332. Southern Economic Journal 64(1): 368–369.
Swallow, S. K., E. C. Smith, E. Uchida, and C. Anderson. 2008. Ecosystem services beyond valu-
ation, regulation and philanthropy: Integrating consumer values into the economy. Choices
23(2): 47–52.
Swinton, S. M., F. Lupi, G. P. Robertson, and S. K. Hamilton. 2007. Ecosystem services and
agriculture: Cultivating agricultural ecosystems for diverse benefits. Ecological Economics
64: 245–262.
Tallis, H., and S. Polasky. 2009. Mapping and valuing ecosystem services as an approach for
conservation and natural resource management. Annals of the N. Y. Academy of Sciences
1162: 265–283.
Toman, M. 1998. Why not to calculate the value of the world’s ecosystem services and natural
capital. Ecological Economics 25: 57–60.
Turner, R. K., and G. C. Daily. 2008. The ecosystem services framework and natural capital
conservation. Environmental and Resource Economics 39(1): 25–35.
U. S. Environmental Protection Agency (EPA). 2009. Valuing the protection of ecological systems
and services: A report of the EPA science advisory board. Washington, DC: Author.
Wainger, L. A., D. M. King, R. N. Mack, E. W. Price, and T. Maslin. 2010. Can the concept of eco-
system services be practically applied to improve natural resource management decisions?
Ecological Economics 69(5): 978–987.
Wainger, L., and M. Mazzotta. 2011. Realizing the potential of ecosystem services: A frame-
work for relating ecological changes to economic benefits. Environmental Management 48
(4): 710–733.
Wallace, K. J. 2007. Classification of ecosystem services: Problems and solutions. Biological
Conservation 139(3–4): 235–246.
Wilen, J. E. 1985. Bioeconomics of renewable resource use. In Handbook of natural
resource and energy economics, Vol. I, eds. A. V. Kneese and J. L. Sweeney, 61–124.
Amsterdam: North-Holland, 1985.
Zhang, W., T. H. Ricketts, C. Kremen, K. Carney, and S. M. Swinton. 2007. Ecosystem services
and disservices to agriculture. Ecological Economics 64: 253–260.
C HA P T E R 9
LAND USE AND

C L I M AT E C HA N G E
BRU C E A . MC C A R L , W I T S A N U AT TAVA N IC H ,

M A R K M U SUM BA , J IA N HONG E . M U, A N D
RU T H A I S A B OK HA E
The climate change issue raises a number of risks and decision-making opportunities.
The decision/risk space involves three dimensions: (1) societal vulnerability, in which
the effects of climate change influence current and future productivity; (2) societal
adaptation, in which adaptive actions are pursued to reduce the productivity effects of
climate change; these actions involve changes in operations accompanied by invest-
ments of resources; and (3) societal mitigation, in which actions are undertaken to
reduce the net emissions of greenhouse gases (GHGs) with the aim of reducing future
atmospheric concentrations of GHGs and their consequent effects on climate change.
This also involves modification in operations plus potential investments.
Land use is heavily involved with these climate change concerns. Land productivity,
land use, and land management (LPLULM) decisions are relevant. Land productivity
is affected by climate change, which, in turn, alters the returns to enterprises using land
(representing vulnerability to climate change). Also, LPLULM decisions can alter net
GHG emissions and contribute to reducing the future extent of climate change (miti-
gation). Finally, LPLULM provides possible mechanisms for altering management or
changing enterprise mix to enhance productivity in the face of climate change (pursuing
adaptation).
In this chapter, we discuss and review interrelationships among the vulnerability,
adaptation, and mitigation aspects of land use and climate change. We do this based on
the literature. A number of studies have addressed such issues and contained findings
that apply to vulnerability, mitigation, and adaptation. We review key research on cli-
mate change issues regarding LPLULM, identifying key findings, pointing out research
needs, and raising economic questions to ponder. In doing this, we go beyond previous
reviews and simultaneously treat the troika of vulnerability, mitigation, and adaptation
aspects of the issue. Hopefully, this will provide readers with a more comprehensive,
Land Use and Climate Change 227
multifaceted grasp of the spectrum of current issues regarding LPLULM and climate
change.
1. Land Use and Climate Change

Interrelationships
LPLULM is involved with all three aspects of the climate change issue. In terms of vulner-
ability, LPLULM productivity is sensitive to changes in climate. The Intergovernmental
Panel on Climate Change (IPCC) (2007a) documents that the climate is changing by
presenting data on increases in temperature, extreme events, heat waves, droughts, and
alterations in rainfall incidence intensity. In addition, hydrological cycles, incidence of
pests and diseases, and forest fires are also being affected.
In terms of mitigation, land cover change has been a major historical contributor to
atmospheric GHG accumulation and is potentially reversible. Houghton (2003) and
Golub et al. (2009) estimate that, since 1850, a third of the total anthropogenic emis-
sions of carbon have come from land use change. In contemporary terms, the IPCC
(2000) finds the current share of total anthropogenic emissions from LPLULM-related
sources to be 18% from forestry and 14% from agriculture. These emissions are mainly
from deforestation, in which forests are converted into cropland, pasture land, and
developed uses; and grassland conversion, in which land use is changed into cropland
from pasture or range. Furthermore, agriculture is estimated to account for 52% and
84% of global anthropogenic methane and nitrous oxide emissions, respectively (IPCC
2000; WRI 2005; Smith et al. 2007a). These emissions are mainly from land-based crop
and livestock production. In the face of this, authors such as Lal, Follett, and Kimble
(2003), Smith et al. (2007a, 2007b), and Fri et al. (2010) argue that LPLULM actions
may enhance sequestration or reduce emissions, thus reducing future atmospheric
concentrations.
LPLULM can also be used to adapt to a changing climate. Land use can be shifted
among enterprises by changing crops, tree, or livestock species and also by changing
uses between cropping, pasture, grazing, and forests to exploit relative changes in pro-
ductivity. One can also alter land management involving practices for crops, livestock,
and forest production to better accommodate a changed climate. Now, given this over-
view, we delve into the individual vulnerability, mitigation, and adaptation topics.
1.1 Climate Change Vulnerability and Land Use Change

Agriculture and forestry (AF) are decidedly vulnerable to climate change. Land use
economists have widely addressed this vulnerability by examining effects on direct AF
productivity, disturbances, land values, and water resources.
228 Bruce A. McCarl Et Al.
1.2 Climate Change and AF Productivity

The IPCC (2007b) indicates that we have observed increases in temperature, changes in
rainfall patterns, increased climate variability, and a greater frequency of extreme events.
These effects will differentially alter productivity across various types of crops, livestock,
and trees and across regions. In addition, atmospheric carbon dioxide (CO2) concentra-
tion increases enhance yields of some crops, grasses, and tree species. Findings on impli-
cations for major categories are reviewed later. Here, we limit coverage to findings using
observed data but not that IPCC (2007b) reviewed studies using simulation models.
1.2.1 Observed Crop Yields

A wide variety of studies have addressed climate change effects on crop yields. Deschenes
and Greenstone (2007) find that yields of corn and soybeans are negatively correlated
to growing degree days. Schlenker and Roberts (2009) and Huang and Khanna (2010)
find similar results and reveal a nonlinear effect of temperature on yields of corn and
soybeans. Attavanich and McCarl (2011) and McCarl, Villavicencio, and Wu (2008)
find that the effect of temperature on US state yields depends on location, with ben-
eficial consequences to colder (northern) areas and detrimental outcomes to the hot-
ter (southern) areas. Collectively, these and other studies show that climate change will
likely reduce yields in areas where heat stress is a factor and increase them in areas where
cold is a key factor.
Regarding the effect of precipitation, Chen, McCarl, and Schimmelpfennig (2004),
Isik and Devadoss (2006), and McCarl, Villavicencio, and Wu (2008) find that increased
precipitation enhances yields of corn, cotton, soybeans, wheat, and sorghum while
having a negative impact on wheat. An inverted U-shaped relationship between corn
and soybean yields and precipitation is found in Schlenker and Roberts (2009) and
Huang and Khanna (2010). Attavanich and McCarl (2011) show heterogeneity in pro-
jected climate change effects, identifying negative effects over the currently wetter US
Central and Northeast regions and positive effects for the drier North Plains regions,
where precipitation gains are projected.
Climate variability and extreme events are addressed in a number of studies. McCarl,
Villavicencio, and Wu (2008) find that increased temperature variation negatively
impacts yields of all crops, and Huang and Khanna (2010) also find this result for corn
and soybeans. McCarl, Villavicencio, and Wu (2008) and Attavanich and McCarl (2011)
find that an increase in precipitation intensity reduces crop yields, whereas an increase
in the Palmer drought index has a differential effect across crops. Chen and McCarl
(2009) examine the effects of hurricane incidence and find yield reductions ranging
from 0.20% to 12.90%, with the US Gulf Coast and the southern Atlantic coastal regions
being the most vulnerable areas.
Crop yields are also affected by atmospheric CO2 concentration. C3 crops are found
to be more responsive to CO2 than are C4 crops under the ample water conditions
(Ainsworth and Long 2005; Kimball 2006 Long et al. 2006; Attavanich and McCarl
2011).1 Leakey (2009) finds that C4 crops only benefit from elevated CO2 in times and
places of drought stress, as do Attavanich and McCarl (2011). Farmers in developing
countries have been found to be highly vulnerable to climate change. Butt et al. (2005)
combine biophysical and economic models to investigate implications of climate change
in Mali. They find that, under climate change, crop farmers are severely affected, and
overall food insecurity almost doubles.
1.2.2 Forests
Boisvenue and Running (2006) review previous literature related to climate change
impacts on forest productivity. They find that climatic change has a generally posi-
tive impact on forest productivity when water is not limiting. McMahon, Parker, and
Miller (2010) estimate that the Northeast US forest is growing at a much faster rate
than expected and attribute this to rising levels of atmospheric CO2, higher tempera-
tures, and a longer growing seasons. Foster et al. (2010) argue to the contrary that past
tree mortality could explain the difference in rates. Recent studies from the free-air
CO2 enrichment (FACE) experiments2 suggest that direct CO2 effects on tree growth
may need to be revised downward (Norby et al. 2005; Karnosky and Pregitzer 2006;
McCarthy et al. 2010). Allen et al. (2010) review studies and indicate that climate change
may enhance tree mortality due to drought and heat in forests worldwide. Sohngen et al.
(1999), in their global study on forest effects, find a market- and productivity-induced
shift to subtropical areas.
1.2.3 Grasslands
Changes in rainfall, temperature, and CO2 concentrations affect the productivity of
grasslands, an important fodder source for livestock production. The IPCC (2007a)
indicates projected declines in rainfall in some major grassland and rangeland areas
(e.g., South America, South and North Africa, western Asia, Australia, and southern
Europe). They state that grass production tends to increase in humid temperate regions
but that it would likely see decreases in arid and semiarid regions (IPCC 2007b). In
Australia, Cullen et al. (2009) predict an increase in grass production in subtropical and
subhumid regions of eastern Australia, whereas in southern Australia they predict slight
increases as of 2030 but decreases of up to 19% in 2070.
1
All plants must convert sunlight to energy by “fixing” carbon as part of photosynthesis. C3 crops are
crops in which the CO2 is first fixed into a compound containing three carbon atoms, whereas C4 crops
are crops in which the CO2 is first fixed into a compound containing four carbon atoms before entering
the Cavin cycle of photosynthesis. In brief, C4 crops are better adapted than C3 crops in an environment
with high daytime temperatures, intense sunlight, drought, or nitrogen or CO2 limitation. Examples of
C3 crops include soybeans, wheat, and cotton; examples of C4 crops are corn and sorghum.
2
In these experiments, air enriched with CO2 is blown into the rings where crops are grown in a real
field (not in a chamber). Then, a computer-control system uses the wind speed and CO2 concentration
information to adjust the CO2 flow rates to maintain the desired CO2 concentration. Finally, crop yield in
the elevated CO2 rings are compared to that in the control rings with nonelevated CO2 environment.
Wang et al. (2007a) project that the net primary productivity of grasslands3 in China
will increase 7–21% under 2.7–3.9°C increases in temperature and 10% increases in
precipitation coupled with doubled CO2. However, they predict a drop of 24% when
there are only increases in temperature. Mu et al. (2013) find in many regions land
use shifts from cropland use to grasslands under predicted climate change. The IPCC
(2007b) indicates that CO2 fertilization enhances grass growth, with C3 pasture grasses
and legumes positively responding and exhibiting about 10% and 20% productivity
increases, respectively (Nowak et al. 2004; Ainsworth and Long 2005). Shifts in forage
quality are also expected (Polley et al. 2012).
1.2.4 Livestock
Climate change affects livestock productivity. Warming climates can increase thermal
stress reducing livestock productivity, conception rates, and survival rates. Increased
climate variability and droughts may lead to livestock production reductions (IPCC
2007b; Thornton et al. 2009). Stocking rates may also decline as gross growth is reduced.
For example, Mu et al. (2013) find an inverted U shape between summer precipitation
and US cattle stocking rates and that cattle stocking rates decrease with increases in the
summer temperature and humidity index (THI). Mader et al. (2009) find that under
increased CO2 concentration scenarios, the west side of the US Corn Belt encounters
productivity losses for swine of as much as 22.4%, whereas on the east side, losses of over
70% occur. For beef, they find increasing temperature is beneficial to beef producers in
the western Corn Belt but not in the northwest and southeast regions. Finally, dairy pro-
duction is projected to decrease from 1.0 to 7.2, depending on location.
Livestock in developing countries are highly vulnerable. Sirohi and Michaelowa (2007)
state that the livestock impacts could be large and devastating for low-income rural areas.
Seo and Mendelsohn (2008a) find that net revenue for beef cattle is lower in warmer
places, but sheep net revenue is lower in wetter places. They also indicate the expected
profit from African livestock management will fall as early as 2020. Moreover, they show
that climate change as predicted would cause considerable reductions in the net incomes
of large livestock farms. Seo et al. (2009) find that a hot and dry climate results in a greater
incidence of livestock compared to crop production. Butt et al. (2005) indicate that under
climate change, livestock weights are projected to decrease by 14–16%.
1.3 Disturbances
Climate change can increase disturbances in the form of increased incidence of pest
and diseases and fires. Numerous studies find that increases in temperature affect pest
3 The Leymus chinensis meadow steppe is widely distributed in the east of the Eurasian region, and
more than half of the steppe is located in China, especially in the northeastern China Plain and Inner
Mongolian Plateau (Wang et al. 2007).
populations and migrations. Rising temperatures are also predicted to increase for-
est pests, crop pesticide usage costs, and wildfire risk (e.g., Chen and McCarl 2001;
Williams and Liebhold 2002; Gan 2004, 2005; Taylor et al. 2007; Hicke and Jenkins 2008;
Walther et al. 2009; Robinet and Roques 2010).
In a review of forestry studies, Taylor et al. (2007) find that the current outbreak
of the mountain pine beetle in British Columbia is an order of magnitude larger in
area and severity than all previous recorded outbreaks. Williams and Liebhold (2002)
project that outbreak areas for southern pine beetles increase with higher tempera-
tures and generally shift northward, whereas the projected outbreak areas for moun-
tain pine beetle shifts toward higher elevations. Hicke and Jenkins (2008) map climate
change effects on lodgepole pine stand susceptibility to mountain pine beetle attack,
concluding that forests in the southern Rocky Mountains have the highest level of
susceptibility.
Patriquin, Wellstead, and White (2007) find negative long-term economic implica-
tions of mountain pine beetle infestations in British Columbia. Schwab et al. (2009)
predict a significant medium-term timber supply shortage, reduced stumpage reve-
nues, and increased cost competition among primary wood products manufacturers.
Williams et al. (2010) estimate that about 7.6% of US southwestern forest and woodland
area experienced mortality associated with pine bark beetles between 1997 and 2008.
In terms of agriculture, crops are negatively affected by insect and disease pest out-
breaks. Chen and McCarl (2001) find that increases in rainfall raise pesticide usage costs
for corn, cotton, potatoes, soybeans, and wheat, whereas hotter weather increases pes-
ticide costs for all crops except wheat. Rosenzweig et al. (2001) review studies on agri-
cultural chemical use and conclude that, in a warmer climate, pests may become more
active and may expand their geographical range, resulting in increased use of pesticides
with accompanying health, ecological, and economic costs. Shakhramanyan, Schneider,
and McCarl (2013) find that climate change causes significant increases in pesticide use
and external costs.
For animal diseases, Purse et al. (2005) explore climate-induced shifts in bluetongue
virus incidence in Europe and find that strains have spread across 12 countries and 800
kilometers further north due to climate change since 1998. Saegerman, Berkvens, and
Mellor (2008) find similar results. Mu, McCarl, and Wu (2011) show that climate change
may have caused part of the current increase in avian influenza incidence and is likely to
further stimulate disease spread in the future.
Climate change also affects fire risk. Westerling et al. (2006) argue that climate change
has caused wildfire risk to increase particularly since the mid-1980s, with the greatest
increases occurred in mid-elevation Northern Rockies forests. Williams et al. (2010)
find that about 2.7% of US southwestern forest and woodland area experienced sub-
stantial mortality due to wildfires between 1984 and 2006. Moriondo et al. (2006) find
an increase in fire risk in the EU Mediterranean countries, especially in the Alps region
of Italy, the Pyrenees of Spain, and the Balkan mountains. Brown, Hall, and Westerling
(2004) argue that climate change will exacerbate forest fires and that new fire and fuels
management strategies may be needed. Chapter 11 by Montgomery in this handbook

provides additional material on fire and land use change.
1.4 Land Values
Climate change affects LPLULM, which in turn impacts land values. Overall, the effect
is mixed in developed countries, but negative in developing countries. Mendelsohn,
Nordhaus, and Shaw (1994) find that higher temperatures in all seasons except autumn
reduce average US farm values, whereas more precipitation outside of autumn increases
farm values. They estimate a climate change-induced loss in US farmland value rang-
ing from –$141 to $34.8 billion. Schlenker, Hanemann, and Fisher (2005) do a similar
study and find an annual loss in US farmland value in the range of $5–5.3 billion for dry-
land nonurban counties. Mendelsohn and Reinsborogh (2007) find that US farms are
much more sensitive to higher temperature than Canadian farms, but are less sensitive
to precipitation increases. Deschenes and Greenstone (2007) find that climate change
will lead to a long-run increase of $1.3 billion (in 2002 dollars) in agricultural land val-
ues. They indicate that land values in California, Nebraska, and North Carolina will be
lowered substantially by climate change, whereas South Dakota and Georgia will have
the biggest increases.
For developing countries, Seo and Mendelsohn (2008b) find that, in South America,
climate change will decrease farmland values except for irrigated farms. Moreover, they
find small farms are more vulnerable to the increase in temperature, whereas large farms
are more vulnerable to increases in precipitation. Mendelsohn, Arellano-Gonzalez, and
Christensen (2010) project that, on average, higher temperatures decrease Mexican land
values by 4,000–6,000 pesos per degree Celsius, amounting to cropland value reduc-
tions of 42–54% by 2100. Wang et al. (2009) find that, in China, an increase in tem-
perature is likely to harm rain-fed farms but benefit irrigated farms. A small value loss
is found in Southeast China farms, whereas the largest damage is discovered in farms in
the Northeast and Northwest (Wang et al. 2009).
1.5 Water Supply
Climate change has important consequences for the hydrological cycle and water avail-
ability (IPCC 2007b; Bates et al. 2008). Land use patterns are affected by this change via
the availability of irrigation water and the suitability of land for rain-fed production.
Regions where the majority of water supply comes from snow or glaciers are vulner-
able to climate change because higher temperatures cause a reduction in mountain stor-
age of water and seasonality of water availability (Gleick and Adams 2000; Barnett et al.
2005). Such regions include South American river basins along the Andes, the Greater
Himalayas, and much of the US West, including California (Coudrian et al. 2005; Xu
et al. 2009).
Climate change also poses water supply threats in Africa because much of the popula-
tion relies on local rivers. De Wit and Stankiewicz (2006) project that a 10% decrease in
precipitation in regions receiving about 1 meter of precipitation per year could reduce
runoff into rivers by 17%, whereas in regions receiving 0.5 meters, that runoff could
be reduced by 50%. Furthermore, they predict that, by the end of this century, surface
water access will be reduced across 25% of Africa. Paeth et al. (2009) find climate change
would cause a weakening of the hydrological cycle over most of tropical Africa, result-
ing in enhanced heat stress and extended dry spells. Additionally, on a global basis, the
Mediterranean Basin, Central America, and sub-tropical Australia are projected to
encounter declines in water availability (Bates et al. 2008) as is the Southwestern United
States (Seager et al. 2007).
In the United States, climate change is projected to reduce California snow accumula-
tion (Cayan et al. 2008). Barnett and Pierce (2009) find that climate change makes cur-
rent levels of Colorado River water deliveries unsustainable into the future. Reilly et al.
(2003) find that US irrigated agriculture needs for water are likely to decline approxi-
mately 5–10% and 30–40% for 2030 and 2090 due to increased precipitation and short-
ened crop-growing periods. McCarl (2008) finds that the US Pacific Southwest gains the
most under the climate change scenarios studied, whereas the US South encounters the
largest losses.
1.6 Vulnerability Research Needs

Although many research studies have focused on the vulnerability of AF land use to
climate change, there are a number of pressing research needs. First, most studies have
focused on developed countries. Thus, there is a need for future research in develop-
ing country settings, particularly those with the greatest projected climate change levels.
Second, a number of issues related to LPLULM require more thorough research, includ-
ing (1) increased incidence of extreme events including droughts, floods, and tropical
storms; (2) analysis of multiple drivers acting at once, including water supply/demand,
pests, diseases, fires, sea level rise, and extremes; (3) effects of shifts in risks in terms of,
for example, yield variability, pest outbreaks, droughts, and market prices; and (4) anal-
ysis of longer term decision making under uncertainty regarding long-term phenomena
like choice of tree species in the face of climate change uncertainty.
2. Mitigation and Land Use Change
LPLULM can alter net fluxes to the atmosphere through increases in sequestration or
reductions in emissions (McCarl and Schneider 2001). Sequestration in the ecosys-
tem can be increased through means like afforestation, forest management, grassland
expansion, biochar, and reduced tillage intensity. Emissions can be limited through
changes in land management and enterprise choice by means such as reducing fertiliza-
tion, altering livestock feeding and numbers, providing less intensive emitting products
like bioenergy, and reducing rice acreage. Here, we elaborate, discussing forestry and
agriculture separately.
2.1 Forestry-Based Mitigation

Forestry mitigation includes means such as (1) reduction in emissions from defores-
tation and degradation4 (REDD as discussed in Miles and Kapos 2008); (2) increasing
forest carbon density through management; (3) afforestation (increasing forested land
area); and (4) provision of substitutes for emission-intensive products, particularly
replacing fossil fuels, but also cement, steel, and other items (McCarl and Schneider
2000; Canadell and Raupach 2008).
Deforestation creates an estimated net emission of 1.5 billion tons of carbon per year
(Pg C year–1) when carbon in standing trees, understory, and soils is released upon har-
vest and subsequent land use change (Canadell et al. 2007). The IPCC estimates that
17% of emissions come from forestry sources, largely from deforestation (IPCC 2001).
Gullison et al. (2007) estimate that by reducing deforestation by 50% by 2050 and main-
taining this level until 2100; society can avoid emissions equivalent to 50 billion tons of
carbon.
Forest management offers another possibility for mitigation. Carbon density can be
increased by thinning, protecting against disturbances (fires, diseases, pests), changing
species mix, lengthening rotations, reducing harvest damage, accelerating replanting
rates, and lengthening use life of harvested projects (Nabuurs et al. 2007). Afforestation
can further reduce the net emissions. For example, in China, 24 million hectares of new
forest were afforested to offset 21% of China’s fossil fuel emissions in 2000 (Wang et al.
2007; Candell and Raupach 2008). Murray et al. (2005) show afforestation to be one of
the large possible strategies for AF to participate in mitigation.
Bioenergy is commonly discussed as a means to mitigate climate change. In forestry,
various trees species plus logging residues and forest byproducts have been proposed
for use as feedstocks for bioenergy to replace fossil fuels, and many of these use short
rotation trees (Cerri et al. 2004; Dias de Oliviera et al. 2005; Smith et al. 2008). Kaul et al.
(2010) show that significant carbon benefits can be obtained in the long run by using
land for short rotation energy crops and substituting biomass for fossil fuels.
Finally, in terms of forestry, there is a dynamic issue involved when temporary/imper-
manent carbon sequestration and permanent emissions reductions are considered. In
4
The Reducing Emissions from Deforestation and Forest Degradation (REDD) is an initiative
process to consider policy that reduces emission from deforestation and forest degradation initiated at
the Eleventh Session of Conference Parties (COP 11) to the United Nations Framework Convention on
Climate Change (UNFCC).
particular, the amount of carbon stock increase is limited by an approach to equilibrium.

That is, as the tree grows to maturity and is subject to harvest, it reaches a point at which
carbon quantity reaches an equilibrium state with no further meaningful gains possible
under a given management system (Birdsey 1996; Kim et al. 2008). Furthermore, the
sequestration in forestry can be reversed in the future by changing practices, forest or
other forces. In the face of this, suppliers often propose to lease carbon sequestration in
forests only for a limited time. This reduces the ultimate value of the carbon credits gen-
erated (Kim et al. 2008).
2.2 Agricultural-Based Emissions

Agriculture is a major emitter of GHGs but also has high potential to mitigate emis-
sions using current technologies, many of which can be implemented immediately.
Mitigation options in agriculture mainly include (1) enhancing carbon sequestration,
(2) intensification and extensification in agricultural production and livestock manage-
ment, and (3) substituting low-emission products for higher emission products (bio-
energy) and reducing emissions (McCarl and Schneider 2000; Clemens and Ahlgrimm
2001; Schneider and Kumar 2008; Smith et al. 2008).
Carbon sequestration enhancement involves increasing carbon stored in the ecosys-
tem (Cacho et al. 2003; Richard and Stokes 2004; Mendelsohn and Dinar 2009). This is
accomplished by some combination of increasing carbon inputs to the soil or reducing
carbon decomposition. Reductions in tillage intensity embody a reduction in soil dis-
turbance that, in turn, limits exposure of carbon to the atmosphere and the amount of its
oxidation and increases sequestration (Cerri et al. 2004; Smith et al. 2008). Conversion
of croplands to grasslands, forests, and perennials also reduces disturbance, which leads
to increased carbon in roots and in the soil. Global estimates have indicated that conver-
sion of all cropland to conservation tillage could sequester 25 billion tons of carbon over
the next 50 years (Pacala and Socolow 2004).
In terms of tillage or land use changes, one should note that the amount of carbon
increase is limited by an approach to equilibrium, as discussed in the forestry section.
Namely, as the carbon quantity increases, so does the decomposition rate and, at a point,
the soil becomes saturated, with no further meaningful gains possible under a given
management system (West and Six 2007; Kim et al. 2008). Furthermore, that can occur
in as few as 10 years (West and Post 2002).
Emissions can be reduced by lowering the use of inputs like fertilizer, pesticides,
and fossil fuels. In particular, reduced nitrogen fertilizer use limits N2O emissions
and also reduces the CO2 involved in nitrogen fertilizer manufacturing (Schlesinger
1999). Deintensification of tillage also reduces fossil fuel usage, as does changes in other
energy-intensive operations like drying and irrigation (McCarl and Schneider 2000).
Improved rice management can reduce methane emissions (Aulakh et al. 2001; Yan
et al. 2003; Smith and Cohen 2004; Smith et al. 2008).
Livestock are emissions sources through enteric fermentation and manure-related

emissions. Managing livestock using improved diets and feed additives aimed at sup-
pressing methanogens have been proposed to reduce emissions from enteric fermenta-
tion (Smith et al. 2007, 2008; Thornton and Geber 2010). Anaerobic digestion of animal
wastes reduces methane emissions while producing biogas (Monteny et al. 2006; Gerber
et al. 2008). Lowering the number of animals can also reduce emissions. For example,
a change in human diet from beef to plant based protein would likely reduce herd size
and total methane emissions, along with cropland needs and associated emissions
(Schneider and Kumar 2008). Management changes, feed additives, and animal breed-
ing that raises animal growth and spreads energy costs of maintenance across greater
feed intake may reduce the methane output per kilogram of animal product (Boadi et al.
2004; Smith et al. 2008).
Mitigation can be achieved by substituting products that replace fossil fuels. In agricul-
ture, production of bioenergy feedstocks may help achieve this. For example, an estimate
of the percentage reduction in net GHG emissions by using corn-based ethanol is 17%
relative to using gasoline (McCarl and Reilly 2007; McCarl 2008). However, one must be
careful of market effects that may simulate land use change elsewhere because this can
increase emissions (Murray et al. 2004; Fargione et al. 2008; Searchinger et al. 2008).
2.3 Role of Markets and Policies in Climate Change Mitigation

Cap-and-trade approaches have been implemented or contemplated as means of
increasing mitigation by providing economic incentives. The Kyoto Protocol suggests
such trading, and trading has been implemented in several forms and places such as
in Europe (Foxon 2010) and California. Theory indicates that market-based incentives
like carbon taxes or cap-and-trade are more economically efficient than are regulatory
approaches in controlling GHGs and are favored by most economists (Metcalf and
Reilly 2008; Raymond and Shively 2008).
There are implementation issues concerning the wide range of GHGs and the global
nature of climate change. Implementation issues mainly arise from differential char-
acteristics of additionality, permanence, uncertainty and leakage; transactions costs,
including measurement and monitoring costs; and property rights. Each is discussed in
the following sections.
2.3.1 Leakage
Leakage is a major mitigation concern in that practices may reduce net emissions in one
region but lead to increased emissions elsewhere due to reduced supplies and market
price signals. In particular, actions that divert production in the mitigating area may
well cause increases in production elsewhere, with accompanying emissions increases
(Murray et al. 2004). A number of authors have cautioned that this could well happen
with expansion of biofuels or afforestation because such activities compete with tradi-
tional cropland and forest land. This can result in reduced production and increased
market prices, thus stimulating other areas to expand production and, in turn, emis-
sions (Murray et al. 2004; Fargione et al. 2008; Searchinger et al. 2008; Mendelson and
Dinar 2009).
2.3.2 Additionality
Ideally, policy desires to only pay for “additional” GHG net emission avoidance, not
that which would have occurred under business as usual. This raises the issue of base-
line establishment, in which a without-policy baseline is compared to a with-policy alter-
native; ideally, only the additional contribution above the baseline would be eligible for
market trading (Smith et al. 2007). Baseline projection is difficult and also implies that
programs must be designed to anticipate future actions and not pay for actions that have
not occurred under current circumstances but that are projected to occur in the future,
in the absence of carbon markets. For example, consider deforestation: most studies use
the assumption that deforestation will continue (IPCC 2007b), but the extent of this is
uncertain, and there may be some changes in trends that portend less future deforesta-
tion (reductions in population growth and an increasingly renewable timber industry),
as argued in Sohngen et al. (1999) and Mendelsohn and Dinar (2009). Policy makers may
subsidize landholders to hold land in forests, but the question remains: would that forest
have been cut down in the absence of policy, and might we be paying for something that
never would have happened?
2.3.3 Permanence
Permanence is another major issue, particularly with carbon sequestration strategies,
in that carbon credits and offsets are not necessarily stored permanently or sold on a
forever basis (Murray et al. 2004; Sands and McCarl 2005; Smith et al. 2007). The prob-
lem is that carbon may not be stored permanently (permanence) due to such things as
possible future LPLULM changes, limited time of guaranteed storage (leasing), needs
for maintenance fees, approach to equilibrium, fires, or extreme events. In turn, this
can lead to the release of sequestered carbon and may merit significant price discounts
accounting for its nonpermanent nature (Kim et al. 2008).
2.3.4 Uncertainty
Uncertainty is a complex implementation issue. Agriculture and forestry, by their
very nature, are affected by climate; thus, both emissions reductions and sequestration
amounts will be affected and uncertain from year to year and over time. Uncertainty
in estimating the magnitude of GHG emissions and sequestration rates has inhibited
implementation of mitigation options in the AF sectors, causing some to argue against
inclusion of AF sequestration in trading schemes. There are also variations and corre-
lations among years, seasons, and locations that make estimation of the sequestration
volume difficult. Kim and McCarl (2009) present a discounting procedure for taking
this into account in trading, whereas Mooney et al. (2004, 2007) dimension the size of
the error and a sampling scheme.
2.3.5 Transactions Costs
Conveyance of carbon credits in markets will likely result in cost wedges between buyers
and sellers due to transaction costs. Kim (2011) separates such costs into a number of
components as discussed below.
2.3.5.1 Assembly Costs
Carbon market purchases would likely need large quantities of offsets (with, for exam-
ple, emissions of large power conglomerates in the hundreds of millions of tons) com-
pared to what a land user could produce. Typically, it would not be economical for an
offset purchaser in quest of 100,000 tons to deal with a single land user. An offset of
100,000 tons at an average sequestration rate of 0.25 tons of carbon per acre (and aver-
age rate from West and Post 2002) would require 400,000 acres. Considering a rough
average farm size of 400 acres (the average US farm was 418 acres in 2007), this offset
would involve nearly 1,000 farmers. Thus, there would be a role for brokers or aggrega-
tors to assemble groups to create marketable quantities. Costs arise in such a process.
Also, there will be costs involved in keeping the group of farmers together and dispers-
ing payments. The crop insurance case is one such scheme, and there transactions costs
are about 25% for brokers.
2.3.5.2 Measurement, Monitoring, and Certification

Market trading will also require measurement and monitoring to establish that offsets
are being produced and continue to be produced. This requires the development of
low-cost measurement and monitoring approaches based on sampling, with an inte-
gration of field level measurement, computer simulation, and remote sensing data
(Mooney et al. 2004, 2007).
There may also be a need for certain bodies to certify offset quantity estimates or the
effectiveness of practices and then monitor that the practice continues. For example,
under the Clean Development Mechanism (CDM), rules were established that indicate
the number of offset credits from various practices. Such certification again introduces
transactions costs.
2.3.5.3 Shortfalls, Enforcement, and Liability

Compliance with carbon contracts will not always happen, and an enforcement or lia-
bility mechanism may be needed. This may involve the setup and operation of shortfall
insurance, an enforcement entity, or a liability imposition mechanism. This again will
introduce transactions costs.
2.3.5.4 Additional Adoption Cost Incentives

Market participation may involve education and training of agricultural producers on
how to alter their practices so that they produce emission offsets most efficiently. Costs
may be borne by agencies, and this may also involve transactions costs.
2.3.6 Property Rights
A final issue involved in market design involves property rights. As argued in McCarl
and Schneider (2001), embarking on the road toward enhanced carbon sequestration
poses policy questions regarding private property rights. For example, if carbon pro-
grams involve land use conversion, there may be a need to ensure that these movements
are not offset by countervailing movements, and this may limit the property rights of a
number of land owners.
2.4 Mitigation Research Needs

A wide literature focuses on mitigation issues, but there are still pending unresolved
questions. Raymond and Shively (2008) pose these questions: Which methods are best
used, given transactions costs, regional variations, and uncertainties? Which strategies
should not be adopted by agriculture? In addition to these questions, we pose the follow-
ing: how does one design mitigation strategies to address leakage, additionality, uncer-
tainty, and permanence, all of which have been major obstacles? How do we expand
coverage to an international setting to avoid leakage and unnecessary shifts in compara-
tive advantage? What are the tradeoffs and synergies between sustainable development
and mitigation? How can one design incentives to practically harness the implementa-
tion of AF mitigation?
3. Adaptation and Land Use Change
Adaptation is the least explored economic area to date. Climate is expected to change
agricultural productivity and shift ecosystems over space (Zilberman et al. 2004;
Mendelsohn and Dinar 2009). Adaptation involves the purposeful manipulation of
LPLULM to increase productivity in the face of such shifts.
There are two types of adaptation: actions undertaken by private decision makers in
their own best interests (autonomous adaptation) and actions undertaken by the pub-
lic sectors in the name of society (IPCC 2007b). Prior authors have called the latter
“planned adaptation,” but we prefer “public adaptation” because it generally addresses
the public goods characteristics of underinvestment in certain adaptation actions.
3.1 A Conceptual View of Adaptation

Following Zilberman et al. (2004) and Mendelsohn and Dinar (2009), a theoretical view
of adaptation through AF land use change is illustrated in Figure 9.1. There, suppose L
depicts the total land available, which is assumed fixed. From right to left, the horizontal
PL PL
Farmland Forestland
L1 L2 L
FIGURE 9.1 Theoretical model of land use change under climate change.
axis shows the land allocated to agriculture; from left to right, it shows land used for for-
est. The two sloped lines are the marginal returns to land allocated to agriculture and
forestry. PL is the land price. L1 is current land allocated to agriculture. L−L1 is current
land allocated to forest. With climate change, the returns to agriculture and forest shift
and are represented as dashed lines. The revised land allocation is then L2, and L−L2.
This reflects substitution to adapt to climate change-induced increases in produc-
tivity. The framework shows that users will autonomously adapt to improve their sit-
uation in response to climate change (Mendelsohn and Dinar 2009). However, public
investments may be needed to either make alternative actions available, such as making
crop varieties more heat tolerant, adapting infrastructure (changing the stock of roads,
bridges, processing locations etc.), or providing information on heretofore unknown
adaptation possibilities.
3.2 LPLULM Adaptation Options
A number of potential LPLULM adaptation options are available. These are often vari-
ations of existing climate risk management strategies (Howden et al. 2007) including
changes in enterprise choice, crop, or livestock mix; moisture management; irriga-
tion, soil, and water conservation; and management of natural areas, among others
(McCarl 2007).
A number of authors have examined observed or potential adaptations in the AF
sector. In national studies, Adams et al. (1990), and Reilly et al. (2003) examine changes
in crop acreage and find northward shifts in crop mixes. Mu et al. (2013) examine the
ways climate change induced land use adaptation between crop and pasture in the US
and find that climate change causes shifts in land from crop to pasture and a lower
stocking rate. They estimate that projected climate change will decrease cropland by
6% and increase pasture land by 33% by the end of this century. Seo (2010) finds that,
in Africa, a hotter and wetter climate causes a shift from crops toward animals. In
addition, Reilly et al. (2003) examine how crops have shifted over time by constructing
the geographic centroid of production for corn and soybeans and find that, between
the early and later 1900s, both US corn and soybean production shifted northward by
about 120 miles. Attavanich et al. (2013) update this, finding that US corn and soy-
bean production has shifted northward, ranging from 100–150 miles between 1950
and 2010.
Studies also have shown that cropping system management adjustments can be used
to adapt (Adams et al. 2003; Easterling et al. 2003; Butt et al. 2005; Travasso et al. 2006;
Challinor et al. 2007). Reilly et al. (2003) show considerable potential to varietal adap-
tation, but Schlenker and Roberts (2009) suggest limited historical adaptation of seed
varieties or management practices to warmer temperatures. Jin et al. (1994) find that
using new rice cultivars and changing planting dates in southern China can substan-
tially adapt to climate change and increase rice yields. Kurukulasuriya and Mendelsohn
(2008a, 2008b) find that, in Africa, farmers adapt by shifting toward more heat-tolerant
crops as temperatures rises and farmers will also shift toward more heat-tolerant and
water-loving crops. In Greece, Kapetanaki and Rosenzweig (1997) find that changing
planting dates and varieties of corn can increase yields by 10%. In Spain, Iglesias et al.
(2000) find that hybrid seeds and altered sowing dates can allow for double cropping of
wheat and corn, thus increasing yields and reducing water use.
Within livestock systems, many adaptation options are connected with maintain-
ing the availability of fodder and feed and reducing heat stress from animal housing
(McCarl 2007; Parry et al. 2009). McCarl and Reilly (2008) estimate changes in the
size of the US livestock herd under 2030 climate scenarios and find increased sheep,
cow calf, dairy, turkey, hog, and broiler numbers with less feedlot beef animals.
In South America, Seo et al. (2010) discover that livestock increase with warming
climate but decrease when it becomes too wet. In Africa, Seo and Mendelsohn
(2008a) find that a warming climate is harmful to commercial livestock but is ben-
eficial to small landowners. Seo et al. (2009) find climate change will likely decrease
African dairy cattle but increase sheep and chickens, although adaptation measures
vary across agro-ecological zones.
Farmers can adapt to climate change by adjusting livestock numbers and species. Mu
et al. (2013) find that adaptation involves reductions in cattle stocking rates under pro-
jected climate change. Alternatively, farmers could switch breeds so that livestock can
adapt to a warmer temperature and changes in precipitation. Zhang et al. (2013) exam-
ine breed choices among cattle in Texas and find that heat-tolerant breeds like Brangus
cattle are used as an adaptation strategy in a hot and humid environment.
Climate change is projected to have far-reaching impacts on ecosystems and sup-
ported species (Chopra et al. 2005; Lemieux and Scott 2005). Adaptation of man-
aged forests could involve changes in tree species, harvesting patterns, pest control,
and location of managed woodland (McCarl 2007). Ecological models have predicted
that forests will expand globally and become somewhat more productive and also
that forest ecosystems would shift poleward and to higher elevations (Zilberman et al.
2004; Mendelsohn and Dinar 2009). Howden et al. (2007) argue that the forest sector
can plant better adapted tree species and can reduce disturbance losses by harvesting
high-risk stocks before they can be destroyed. Mendelsohn and Dinar (2009) indicate
that climate change will alter land allocation between forest and wild lands. Sohngen
and Mendelsohn (2003), Mendelsohn and Dinar (2009) and Sohngen et al. (2010) show
that forest adaptation is a dynamic process involving staged harvest decisions, thus lim-
iting the ability to change large forest stocks quickly.
Parry et al. (2009) argue that complementary relationships between adaptation and
mitigation can be exploited because adaptation actions can have positive or negative
mitigation effects and vice versa. In the forest sector, afforestation of degraded hill slopes
is an example of a mitigation action with a positive adaptation effect that would not only
sequester carbon, but also control soil erosion (IPCC 2007b).
3.3 Completeness of Adaptation

Adaptation has been found to improve welfare, so it is therefore very likely that peo-
ple will autonomously adapt (Butt et al. 2005; Mendelsohn and Dinar 2009). However,
most impacts due to climate change are projected to continue to increase for some time
(IPCC 2007b), implying a need for continuing adaptation. Furthermore, some adapta-
tion actions may not be practical due to costs or barriers. Therefore, it is likely that some
climate change impacts are unavoidable (Parry et al. 2009) and the resolution of who is
going to pay for adaptation is also a major issue.
3.4 Adaptation and Development

Social-economic development and adaptation are intimately linked (Parry et al. 2009).
Technological sophistication and progress are important determinants of farm produc-
tivity and adaptation potential and also influence adaptation demand. In particular, if
technological progress lags behind population growth, there will be increased compe-
tition among land uses, including those for adaptation and mitigation (Mendelsohn
and Dinar 2009; McCarl et al. 2012). Lobell et al. (2008) indicate that South Asia and
Southern Africa are regions that, without sufficient adaptation measures, will likely
suffer negative impacts on several crops, which are important to large, food-insecure
human populations.
It is not clear what level of adaptation investment is appropriate because we have
limited knowledge of climate change impacts, as well as limited studies on the effec-
tiveness and optimal level of adaptation (Parry et al. 2009). The United Nations
Framework Convention on Climate Change (UNFCCC ) has estimated that the
annual cost of AF adaptation ranged between $11.3 to $ 12.6 billion for 2030, with
developing contries needing $7 billion dollars (McCarl 2007). With such levels of
adaptation, about 80% of the costs of potential impacts might be avoided, but about
20% might not (Parry et al. 2009), and cost of adaptation may rise steeply after 2030
(IPCC 2007b).
3.5 Adaptation Research Needs

The choice of optimal adaptation levels is uncertain because information about future
potential impacts and adaptation effectiveness is scarce (Parry et al. 2009), and this
delays adaptation. Research efforts to narrow this uncertainty and identify robust adap-
tation are needed.
In addition, incorporating adaptation into integrated assessment models is
needed, as is the inclusion of information on both autonomous and public adapta-
tion cost. Although global estimates of adaptation cost have emerged (e.g., UNFCCC
2007; Nelson et al. 2009; Parry et al. 2009; and McCarl 2007 in a LPLULM sense),
costs of adapting to varying levels of climate change need analysis, to provide a
choice range for the level of adaptation investment. There is also a need to analyze
the unavoidable impacts and the resulting damage costs that we need to anticipate
(Parry et al. 2009).
Adaptation plans may suffer from maladaptation or leakage problems (Smith et al.
2000; Sathaye and Andrasko 2007; Sohngen and Brown 2008), in which actions in one
region cause lower net adaptation in other locations. Furthermore, the roles of many
nonclimatic factors, such as changes of commodity prices, population growth, econ-
omy development, or farm programs, and the like that interact with climatic stimuli in
influencing LPLULM adaptation decision making are rarely substantively examined
(Kandlikar and Risbey 2000; Schneider et al. 2000).
4. Interrelationships among Climate

Change Effects, Mitigation, and
Adaptation
The total burden of climate change consists of three elements: the costs of mitigation
(reducing the extent of climate change), the costs of adaptation (reducing the impact of
change), and the residual impacts that can be neither mitigated nor adapted to (Parry
et al. 2009). Mitigation and adaptation both avoid climate change but are fundamentally
different in timing, with adaptation providing an immediate avoidance and mitigation
a long-term reduction in extent. Some studies have attempted to understand the inter-
play among impacts, adaptation, and mitigation. Yet there are still many unanswered
questions.
Bosello et al. (2009) indicate that welfare is greater when adaptation and mitiga-
tion are implemented jointly and both contribute to better control of climate damages.
Estimations considering only single mitigation or adaptation actions are therefore likely
to yield biased results.
The major climate change policy question is: “What combination of emissions reduc-
tion and adaptation is appropriate in offsetting the impacts of climate change?” In
addressing this question, one must realize that adaptation and mitigation can be both
complementary and substitutes. The IPCC (2007b) reviews four major types of inter-
relationships between adaptation and mitigation: (1) adaptation actions that have con-
sequences for mitigation, (2) mitigation actions that have consequences for adaptation,
(3) decisions that include trade-offs or synergies between adaptation and mitigation,
and (4) processes that have consequences for both adaptation and mitigation.
Important implications arise from the interdependence between mitigation and
adaptation. Lecocq and Shalizi (2007) point out the need for mitigation and adap-
tation policies to be analyzed and implemented jointly, not separately. Mata and
Budhooram (2007) state that, in a hypothetical world where all net costs are borne
by a single global entity, choices would probably be driven by total cost minimization
rather than by aversion to “dangerous anthropogenic interference with the climate
system.” However, the complexities of costs and benefits and their widespread dis-
tribution make these assumptions implausible (Mata and Budhooram 2007). Rather,
society is saddled with the burden of optimally allocating resources subject to bud-
get constraints and uncertainties. In addition, action should incorporate learning and
irreversibility.
Some studies have tried to assess the optimal policy balance of mitigation and
adaptation using cost-benefit frameworks based on integrated assessment models
(IAMs) (IPCC 2007b). Temporal investment allocation results obtained using IAMs
in de Bruin et al. (2009) and in Wang and McCarl (2013) show that both adaptation
and mitigation are simultaneously employed, with adaptation prevailing initially then
mitigation investment taking over in the long run as the damages from GHG emissions
increase.
4.1 Research Needs
Most studies on climate change responses focus on single aspects of the
adaptation-mitigation nexus, without considering their interplay. Hence, substantial
research needs to address the optimal portfolio of adaptation and mitigation, along
with practical inquiries into the extent to which climate change vulnerability can be
addressed. The IPCC (2007c) indicates that the relationship between development paths
and adaptation-mitigation interrelationships requires further research. This is particu-
larly important in developing country settings.
5. Conclusion
LPLULM decision making is certainly affected by climate change and climate policy.
Actions can address adaptation or mitigation, and there will be climate change-induced
damages that are not mitigated or adapted to. In the future, we think substantial research
will need to be devoted to determine how land use decisions can facilitate adaptation
and mitigation, as well as the degree of vulnerability under alternative levels of action.
We only hope that this review will inform researchers about past efforts and potential
productive future ones.
Acknowledgments
We thank the editors for comments. Seniority of authorship is shared among authors.
References
Adams, R. M., C. Rosenzweig, R. M. Peart, J. T. Ritchie, B. A. McCarl, J. D. Glyer, R. B. Curry,
J. W. Jones, K. J. Boote, and L. H. Allen, Jr. 1990. Global climate change and U.S. agriculture
Nature 345: 219–224.
Adams R. M, B. A. McCarl, and L. O. Mearns. 2003. The effects of spatial scale of climate
scenarios on economic assessments: An example from U.S. agriculture. Climatic Change
60:131–148.
Ainsworth, E. A., and S. P. Long. 2005. What have we learned from 15 years of free-air CO2
enrichment (FACE)? A meta-analytic review of the responses of photosynthesis, canopy
properties and plant production to rising CO2. New Phytologist 165(2): 351–372.
Allen, C. D., A. K. Macalady, H. Chenchouni, D. Bachelet, N. McDowell, M. Vennetier, T.
Kitzberger, A. Rigling, D. D. Breshears, and E. H. Hogg. 2010. A global overview of drought
and heat-induced tree mortality reveals emerging climate change risks for forests. Forest
Ecology and Management 259(4): 660–684.
Attavanich, W., and B. A. McCarl. 2011. The effect of climate change, CO2 fertilization, and
crop production technology on crop yields and its economic implications on market out-
comes and welfare distribution. Paper presented at 2011 AAEA & NAREA Joint Annual
Meeting, July 24–26, Pittsburgh, Pennsylvania.
Attavanich, W., B. A. McCarl, Z. Ahmedov. S. W. Fuller, and D. V. Vedenov. 2013. Effects of
climate change on US grain transport. Nature Climate Change 3:638–643 doi:10.1038/
nclimate1892
Aulakh, M. S., R. Wassmann, C. Bueno, and H. Rennenberg. 2001. Impact of root exudates of
different cultivars and plant development stages of rice (Oryza sativa L.) on methane pro-
duction in a paddy soil. Plant Soil 230: 77–86.
Barnett, T. P., J. C. Adam, and D. P. Lettenmaier. 2005. Potential impacts of a warming climate
on water availability in snow-dominated regions. Nature 438(7066): 303–309.
Barnett, T. P., and D. W. Pierce. 2009. Sustainable water deliveries from the Colorado River in a
changing climate. Proceedings of the National Academy of Sciences of the USA 106(18): 7334.
Bates, B., Z. W. Kundzewicz, S. Wu, and J. Palutikof, eds. 2008. Climate change and water. Technical
Paper of the Intergovernmental Panel on Climate Change, IPCC Secretariat. Geneva.
Birdsey, R. A. 1996. Carbon storage for major forest types and regions in the conterminous
United States. Forests and Global Phange, 2, 1-25.
Boadi, D., C. Benchaar, J. Chiquette, and D. Masse. 2004. Mitigation strategies to reduce enteric meth-
ane emissions from dairy cows: Update review. Canadian Journal Animal Science 84: 319–335.
Boisvenue, C. É. L., and S. W. Running. 2006. Impacts of climate change on natural forest produc-
tivity–evidence since the middle of the 20th century. Global Change Biology 12(5): 862–882.
Bosello, F., C. Carraro and E. De Cian. 2009. An analysis of adaptation as a response to climate
change. University Ca’Foscari of Venice, Dept. of Economics Research Paper Series, (26_09).
Brown, T. J., B. L. Hall, and A. L. Westerling. 2004. The impact of twenty-first century climate
change on wildland fire danger in the western United States: An applications perspective.
Climatic Change 62(1): 365–388.
Butt, T. A., B. A. McCarl, J. Angerer, P. T. Dyke, and J. W. Stuth. 2005. The economic and food
security implications of climate change in Mali. Climatic Change 68(3): 355–378.
Cacho, O. J., R. L. Hean, and R. Wise. 2003. Carbon-accounting methods and reforestation
incentives. Australian Journal of Agricultural and Resource Economics 47: 153–179.
Canadell, J. G., C. L.Quéré, M. R. Raupach, C. B. Field, E. T. Buitenhuis, P. Ciais, T. J. Conway,
N. P. Gillett, R. A. Houghton, and G. Marlandi. 2007. Contributions to accelerating atmo-
spheric CO2 growth from economic activity, carbon intensity, and efficiency of natural
sinks. Proceedings of the National Academy of Sciences of the USA 104(47): 18866–18870.
Canadell, J. G., and M. R. Raupach. 2008. Managing forests for climate change mitigation.
Science 320: 1456–1457.
Cayan, D. R., E. P. Maurer, M. D. Dettinger, M. Tyree, and K. Hayhoe. 2008. Climate change
scenarios for the California region. Climatic Change 87: 21–42.
Cerri, C. C., M. Bernoux, C. E. P. Cerri, and C. Feller. 2004. Carbon cycling and sequestration
opportunities in South America: The case of Brazil. Soil Use Management 20: 248–254.
Challinor, A., T. Wheeler, P. Craufurd, C. Ferro, and D. Stephenson. 2007. Adaptation of
crops to climate change through genotypic responses to mean and extreme temperatures.
Agriculture, Ecosystems and Environment 119: 190–204.
Chen, C. C., and B. A. McCarl. 2001. An investigation of the relationship between pesticide
usage and climate change. Climatic Change 50(4): 475–487.
Chen, C. C., B. A. McCarl, and D. E. Schimmelpfennig. 2004. Yield variability as influenced by
climate: A statistical investigation. Climatic Change 66(1): 239–261.
Chen, C. C., and B. A. McCarl. 2009. Hurricanes and possible intensity increases: Effects
on and reactions from U.S. agriculture. Journal of Agricultural and Applied Economics
41(1): 125–144.
Chopra, K., R. Leemans, P. Kumar, and H. Simons. 2005. Ecosystems and human well-being: Policy
Responses. Washington, DC: Island Press.
Clemens, J., and H. J. Ahlgrimm. 2001. Greenhouse gases from animal husbandry: Mitigation
options. Nutrient Cycling in Agroecosystems 60: 287–300.
Coudrain, A., B. Francou, and Z. W. Kundzewicz. 2005. Glacier shrinkage in the Andes and
consequences for water resources. Hydrological Sciences Journal 50(6): 925–932.
Cullen, B. R., I. R. Johnson, R. J. Eckard, G. M. Lodge, R. G. Walker, R. P. Rawnsley, and M.
R. McCaskill. 2009. Climate change effects on pasture systems in south-eastern Australia.
Crop and Pasture Science 60(10): 933–942.
De Bruin, K. C., R. B. Dellink, and R. S. J. Tol. 2009. AD-DICE: An implementation of adapta-
tion in the DICE model. Climatic Change 95: 63–81.
De Wit, M., and J. Stankiewicz. 2006. Changes in surface water supply across Africa with pre-
dicted climate change. Science 311(5769): 1917–21.
Deschenes, O., and M. Greenstone. 2007. The economic impacts of climate change: Evidence
from agricultural output and random fluctuations in weather. American Economic Review
97(1): 354–385.
Dias de Oliveira, M. E., B. E. Vaughan, and E. J. Rykiel, Jr. 2005. Ethanol as fuel: Energy, carbon
dioxide balances, and ecological footprint. BioScience 55: 593–602.
Easterling, W. E., N. Chhetri, and X. Niu 2003. Improving the realism of modeling agronomic
adaptation to climate change: Simulating technological substitution. In Issues in the impacts
of climate variability and change on agriculture—Applications to the southeastern United
States, ed. L. O. Mearns, 149–173. Dordrecht: Kluwer Academic.
FAO. 2010. Managing forests for climate change. http://www.fao.org/docrep/013/i1960e/
i1960e00.pdf
Fargione, J., J. Hill, D. Tilman, S. Polasky, and P. Hawthorne. 2008. Land clearing and the biofuel
carbon debt. Science 319(5867): 1235–1238.
Foster, J. R., J. I. Burton, J. A. Forrester, F. Liu, J. D. Muss, F. M. Sabatini, R. M. Scheller, and
D. J. Mladenoff. 2010. Evidence for a recent increase in forest growth is questionable.
Proceedings of the National Academy of Sciences of the USA 107(21): E86.
Foxon, T. J. 2010. Climate change mitigation policy: An overview of opportunities and chal-
lenges. In Changing climates, earth systems and society, eds. E. F. J. Nulder, and J. Dowson,
231–241. Netherlands: Springer.
Gan, J. 2004. Risk and damage of southern pine beetle outbreaks under global climate change.
Forest Ecology and Management 191: 61–71.
Gan, J. 2005. Incorporating human and natural adaptations in assessing climate change
impacts on wildfire occurrence. In New research on forest ecosystems, ed. A. R. Burk, 61–73.
Hauppauge, NY: Nova Science.
Gerber P., N. Key, F. Portet, and H. Steinfeld. 2010. Policy options in addressing livestock’s con-
tribution to climate change. Animal 4: 393–406.
Gleick, P. H., and D. B. Adams. 2000. Water: The potential consequences of climate variability
and change. A report of the National Water Assessment Group, U.S. Global Change Research
Program, U.S. Geological Survey, U.S. Department of the Interior and the Pacific Institute
for Studies in Development, Environment, and Security, Oakland, California.
Golub, A., T. Hertel, H. Lee, S. Rose, and B. Sohngen. 2009. The opportunity cost of land use and
the global potential for greenhouse gas mitigation in agriculture and forestry. Resource and
Energy Economics 31: 299–319.
Gullison, R. E., P. C. Frumhoff, J. G. Canadell, C. B. Field, D. C. Nepstad, K. Hayhoe, R. Avissar,
L. M. Curran, P. Friedlingstein, C. D. Jones, and C. Nobre.2007. Tropical forests and climate
policy. Science 316: 985–986.
Hicke, J. A., and J. C. Jenkins. 2008. Mapping lodgepole pine stand structure susceptibil-
ity to mountain pine beetle attack across the western United States. Forest Ecology and
Management 255(5–6): 1536–1547.
Houghton, R. A. 2003. Revised estimates of the annual net flux of carbon to the atmosphere
from changes in land use and land management: 1850–2000. Tellus Series B: Chemical and
Physical Meteorology 55(2): 378–390.
Howden, S. M., J. F. Soussana, F. N. Tubiello, N. Chhetri, M. Dunlop, and H. Meinke. 2007.

Adapting agriculture to climate change. Proceedings of the National Academy of Sciences of
the USA 104(50): 19691–19696.
Huang, H., and M. Khanna. 2010. An econometric analysis of U.S. crop yield and cropland acre-
age: Implications for the impact of climate change. Paper presented at 2010 AAEA, CAES, &
WAEA Joint Annual Meeting, Denver, Colorado.
Iglesias, A., C. Rosenzweig, and D. Pereira. 2000. Agricultural impacts of climate change in
Spain: Developing tools for a spatial analysis. Global Environment Change 10: 69–80.
IPCC. 2000. IPCC Special Report: Land use, land-use change, and forestry, eds. T. W. Robert.,
I. R. Noble, B. Bolin, N. H. Ravindranath, D. J. Verardo, and D. J. Dokken. Cambridge,
IPCC. 2007a. Climate change 2007: The physical science basis. Contribution of working group
I to the fourth assessment report of the Intergovernmental Panel on Climate Change, eds.
S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, and H.
L. Miller. Cambridge, UK: Cambridge University Press.
IPCC. 2007b. Climate change 2007: Impacts, adaptation and vulnerability. Contribution of
working group II to the fourth assessment report of the Intergovernmental Panel on Climate
Change, eds. M. L. Parry, O. F. Canziani, P. J. Palutikof, P. J. van der Linden, and C. E. Hanson.
Cambridge, UK: Cambridge University Press.
IPCC. 2007c. Climate change 2007: Mitigation of climate change. Contribution of working group
III to the fourth assessment report of the Intergovernmental Panel on Climate Change, eds.
B. Metz, O. R. Davidson, P. R. Bosch, R. Dave, and L. A. Meyer. Cambridge, UK: Cambridge
University Press.
Isik, M., and S. Devadoss. 2006. An analysis of the impact of climate change on crop yields and
yield variability. Applied Economics 38(7): 835–844.
Jin, Z., D. Ge, H. Chen, and J. Fang. 1994. Effects of climate change on rice production and strat-
egies for adaptation in southern China. In Implications of climate change for international
agriculture: Crop modeling study, eds. C. Rosenzweig, and A. Iglesias, 1–24. Washington,
DC: U.S. Environmental Protection Agency.
Kandlikar, M., and J. Risbey. 2000. Agricultural impacts of climate change: If adaptation is the
answer, what is the question? Climatic Change 45: 529–539.
Kapetanaki, G., and C. Rosenzweig. 1997. Impact of climate change on maize yield in cen-
tral and northern Greece: A simulation study with CERES-Maize. Mitigation Adaptation
Strategies for Global Change 1: 251–271.
Karnosky, D. F., K. S. Pregitzer, J. Nosberger, S. P. Long, R. J. Norby, et al. 2006. Impacts of ele-
vated atmospheric [CO2] and [O3] on northern temperate forest ecosystems: results from
the Aspen FACE Experiment. In Managed ecosystems and CO2: Case studies, processes,and
perspectives, eds. J. Nosberger, S. P. Long, R. J. Norby, M. Stitt, G. R. Hendrey, and H. Blum,
213–229. Berlin, Germany.: Springer-Verlag.
Kaul, M., G. M. J. Mohren, and V. K. Dadhwal. 2010. Carbon storage versus fossil fuel substitu-
tion: A climate change mitigation option for two different land use categories based on short
and long rotation. Mitigation Adaptation Strategies for Global Change 15: 395–409.
Kelly, D. L., C. D. Kolstad, and G. T. Mitchell. 2005. Adjustment costs from environmental
change. Journal of Environmental Economics and Management 50: 468–495.
Kim, M-K., B. A. McCarl, and B. C. Murray. 2008. Permanence discounting for land-based
carbon sequestration. Ecological Economics 64(4): 763–769.
Kim, M-K., and B. A. McCarl. 2009. Uncertainty discounting for land-based carbon sequestra-
tion. Journal of Agricultural and Applied Economics 41(1): 1–11.
Kim, S. 2011. The effect of transactions costs on GHG emission mitigation for agriculture and
forestry. Ph.D. dissertation, Texas A&M University.
Kimball, B. A. 2006. The effects of free-air [CO2] enrichment of cotton, wheat, and sorghum. In
Managed ecosystems and CO2, 47–70. Berlin and Heidelberg: Springer.
Kurukulasuriya, P. and R. Mendelsohn. 2008a. A Ricardian analysis of the impact of climate
change on African cropland. African Journal Agriculture and Resource Economics 2:1–23.
Kurukulasuriya, P., and R. Mendelsohn. 2008b. Crop switching as an adaptation strategy to
climate change. African Journal Agriculture and Resource Economics 2: 105–126.
Lal, R., R. F. Follett, and J. M. Kimble. 2003. Achieving soil carbon sequestration in the United
States: A challenge to the policy makers. Soil Science 168: 827–845.
Leakey, A. D. B. 2009. Rising atmospheric carbon dioxide concentration and the future of C4
crops for food and fuel. Proceedings of the Royal Society B: Biological Sciences 276(1666): 2333.
Lecocq, F., and Z. Shalizi. 2007. Balancing expenditures on mitigation of and adaptation to
climate change: An exploration of Issues relevant to developing countries. Policy Research
Working Paper Series 4299. Washington, DC: The World Bank.
Lemieux, C. J., and D. J. Scott. 2005. Climate change, biodiversity conservation and protected
areas planning in Canada. The Canadian Geographer 49(4): 384–399.
Lobell, D. B., M. B. Burke, C. Tebaldi, M. D. Mastrandrea, W. P. Falcon., and R. L. Naylor. 2008.
Prioritizing climate change adaptation needs for food security in 2030. Science 319: 607–610.
Long, S. P., E. A. Ainsworth, A. D. B. Leakey, J. Nösberger, and D. R. Ort. 2006. Food for
thought: Lower-than-expected crop yield stimulation with rising CO2 concentrations.
Science 312(5782): 1918–21.
Mader, T. L., K. L. Frank, J. A. Harrington, G. L. Hahn, and J. A. Nienaber. 2009. Potential
climate change effects on warm-season livestock production in the Great Plains. Climatic
Change 97(3): 529–541.
Mata, L. J., and J. Budhooram. 2007. Complementarity between mitigation and adapta-
tion: the water sector. Mitigation and Adaptation Strategies for Global Change 12: 799–807,
doi:10.1007/s11027-007-9100-y.
McCarl, B. A., and U. A. Schneider. 2001. Greenhouse gas mitigation in U.S. agriculture and
forestry. Science 294: 2481–2482.
McCarl, B. A. 2007. Adaptation options for agriculture, forestry and fisheries. A report to the
UNFCCC Secretariat Financial and Technical Support Division. http://agecon2.tamu.edu/
people/faculty/mccarl-bruce/papers/1467mccarl.pdf
McCarl, B. A., and J. M. Reilly. 2007. Agriculture in the climate change and energy price
squeeze: Part 2: Mitigation opportunities. Department of Agricultural Economics, Texas
A&M University. http://agecon2.tamu.edu/people/faculty/mccarl-bruce/papers/1322agin
climate2mitigation2.doc.
McCarl, B. A. 2008. Bioenergy in a greenhouse mitigating world. Choices 23(1): 31–33.
McCarl, B. A., and J. M. Reilly. 2008. U.S. agriculture in the climate change squeeze: Part 1: Sectoral
sensitivity and vulnerability. Report to the National Environmental Trust. http://agecon2.
tamu.edu/people/faculty/mccarl-bruce/689cc/topic5b_Agriculture%20in%20the%20cli-
mate%20change%20squeez1.pdf
McCarl, B. A., X. Villavicencio, and X. Wu. 2008. Climate change and future analysis: Is station-
arity dying? American Journal of Agricultural Economics 90(5): 1241–1247.
McCarthy, H. R., R. Oren, K. H. Johnsen, A. Gallet-Budynek, S. G. Pritchard, C. W. Cook, S.

L. LaDeau, R. B. Jackson, and A. C. Finzi. 2010. Re-assessment of plant carbon dynamics at
the duke free-air CO2 enrichment site: Interactions of atmospheric CO2 with nitrogen and
water availability over stand development. New Phytologist 185(2): 514–528.
McCarl, B. A., X. Villavicencio, X. M. Wu, and W. E. Huffman. 2013. Climate change influences
on agricultural research productivity. Climatic Change 19: 815–824
McMahon, S. M., G. G. Parker, and D. R. Miller. 2010. Evidence for a recent increase in forest
growth. Proceedings of the National Academy of Sciences of the USA 107(8): 3611–3615.
Mendelsohn, R., W. D. Nordhaus, and D. Shaw. 1994. The impact of global warming on agricul-
ture: A Ricardian analysis. American Economic Review 84: 753–771.
Mendelsohn, R., and M. Reinsborough. 2007. A Ricardian analysis of U.S. and Canadian farm-
land. Climatic Change 81(1): 9–17.
Mendelsohn, R., and A. Dinar. 2009. Land use and climate change interactions. Annual Review
of Resource Economics 1: 309–332.
Mendelsohn, R., J. Arellano-Gonzalez, and P. Christensen. 2010. A Ricardian analysis of
Mexican farms. Environment and Development Economics 15(2): 153–171.
Metcalf, G. E., and J. M. Reilly. 2008. Policy options for controlling greenhouse gas emis-
sions: Implications for agriculture. Choices 23(1): 34–37.
Miles. L., and V. Kapos. 2008. Reducing greenhouse gas emissions from deforestation and forest
degradation: Global land-use implications. Science 13: 1454–1455.
Monteny, G. J., A. Bannink, and D. Chadwick. 2006. Greenhouse gas abatement strategies for
animal husbandry. Agriculture Ecosystems and Environment 112: 163–170.
Montgomery, C. A. 2014. Fire: An agent and a consequence of land use change. In The Oxford
handbook of land economics, eds. J. M. Duke and J. Wu, 281–301. New York: Oxford
University Press.
Mooney, S., J. M. Antle, S. M. Capalbo, and K. Paustian. 2004. Influence of project scale on the
costs of measuring soil sequestration. Environmental Management 33(S1): S252–263.
Mooney, S., K. Gerow, J. M. Antle, S. M. Capalbo, and K. Paustian. 2007. Reducing standard
errors by incorporating spatial autocorrelation into a measurement scheme for soil carbon
credits. Climatic Change 80:55–72.
Moriondo, M., P. Good, R. Durao, M. Bindi, C. Giannakopoulos, and J. Corte-Real. 2006.
Potential impact of climate change on fire risk in the Mediterranean area. Climate Research
31(1): 85–95.
Mu, J. E., B. A. McCarl, and X. M. Wu. 2011. Climate change influences on the risk of avian
influenza outbreaks and associated economic losses. Paper presented at the 2011 AAEA
AND NAREA Joint Annual Meeting, July 24–26, Pittsburgh, Pennsylvania.
Mu, J. E., B. A. McCarl, and A. Wein. 2013. Adaptation to climate change: Changes in farmland
use and stocking rate in the U.S. Mitigation and Adaptation Strategies for Global Change.
18: 713–730. doi: 10.1007/s11027-012-9384-4.
Murray, B. C., B. A. McCarl, and H. C. Lee. 2004. Estimating leakage from forest carbon seques-
tration programs. Land Economics 80: 109–124.
Murray, B. C., B. L. Sohngen, A. J. Sommer, B. M. Depro, K. M. Jones, B. A. McCarl, D. Gillig, B.
DeAngelo, and K. Andrasko. 2005. Greenhouse gas mitigation potential in U.S. forestry and
agriculture, EPA-R-05-006. Washington, DC: U. S. Environmental Protection Agency, Office
of Atmospheric Programs, Washington, D.C.
Nabuurs, G. J., O. Masera, K. Andrasko, P. Benitez-Ponce, R. Boer, M. Dutschke, E. Elsiddig,
J. Ford-Robertson, P. Frumhoff, T. Karjalainen, O. Krankina, W. A. Kurz, M. Matsumoto,
W. Oyhantcabal, N. H. Ravindranath, M. J. Sanz Sanchez, and X. Zhang, 2007: Forestry.

In Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change, eds. B. Metz,
O. R. Davidson, P. R. Bosch, R. Dave, L. A. Meyer. Cambridge, UK and New York, NY,
USA: Cambridge University Press.
Nelson, G. C., M. W. Rosegrant, J. Koo, R. Robertson, T. Sulser, T. Zhu, C. Ringler, S. Msangi, A.
Palazzo, M. Batka, M. Magalhaes, R. Valmonte-Santos, M. Ewing, and D. Lee. 2009. Impact
on agriculture and costs of adaptation. International Food Policy Research Institute (IFPRI).
http://ccsl.iccip.net/impact_agri.pdf
Norby, R. J., E. H. DeLucia, B. Gielen, C. Calfapietra, C. P. Giardina, J. S. King, J. Ledford, H.
R. McCarthy, D. J. P. Moore, and R. Ceulemans. 2005. Forest response to elevated CO2 is
conserved across a broad range of productivity. Proceedings of the National Academy of
Sciences of the USA 102(50): 18052–18056..
Nowak, R. S., D. S. Ellsworth, and S. D. Smith. 2004. Functional responses of plants to elevated
atmospheric CO2–Do photosynthetic and productivity data from FACE experiments sup-
port early predictions? New Phytologist 162: 253–280.
Pacala, S., and R. Socolow. 2004. Stabilization wedges: Solving the climate problem for the next
50 years with current technologies. Science 305: 968–972.
Paeth, H., K. Born, R. Girmes, R. Podzun, and D. Jacob. 2009. Regional climate change in tropi-
cal and northern Africa due to greenhouse forcing and land use changes. Journal of Climate
22(1): 114–132.
Parry, M., N. Arnell, P. Berry, D. Dodman, S. Fankhauser, C. Hope, S. Kovats, R. Nicolls, D.
Sattherwaite, R. Tiffen, and T. Wheeler. 2009. Assessing the costs of adaptation to climate
change: A critique of the UNFCCC estimates. London: International Institute for Environment
and Development and Grantham Institute for Climate Change.
Patriquin, M. N., A. M. Wellstead, and W. A. White. 2007. Beetles, trees, and peo-
ple: Regional economic impact sensitivity and policy considerations related to the
mountain pine beetle infestation in British Columbia, Canada. Forest Policy and
Economics 9(8): 938–946.
Polley, H. W., D. D. Briske, J. A. Morgan, K. Wolter, D. W. Bailey, and J. R. Brown. 2012. Climate
change and North American rangelands: Evidence, trends, and implications. Commissioned
paper under submission to Rangeland Ecology and Management.
Purse, B. V., P. S. Mellor, D. J. Rogers, A. R. Samuel, P. P. C. Mertens, and M. Baylis. 2005. Climate
change and the recent emergence of bluetongue in Europe. Nature Reviews Microbiology
3(2): 171–181.
Raymond, L., and G. Shively. 2008. Market-based approaches to CO2 emissions reductions.
Choices 23(1):38–40.
Reilly, J. M. (ed.). 2002. Agriculture: The potential consequences of climate variability and change
for the United States. New York: Cambridge University Press. http://www.usgcrp.gov/
usgcrp/Library/nationalassessment/Agriculture.pdf
Reilly, J., F. Tubiello, B. A. McCarl, D. Abler, R. Darwin, K. Fuglie, S. Hollinger, C. Izaurralde, S.
Jagtap, and J. Jones. 2003. U.S. agriculture and climate change: New results. Climatic Change
57: 43–67.
Richards, K., and C. Stokes. 2004. A review of forest carbon sequestration cost studies: A dozen
years of research. Climatic Change 63: 1–46.
Robinet, C., and A. Roques. 2010. Direct impacts of recent climate warming on insect popula-
tions. Integrative Zoology 5(2): 132–142.
Rosenzweig, C., A. Iglesias, X. B. Yang, P. R. Epstein, and E. Chivian. 2001. Climate change and
extreme weather events: Implications for food production, plant diseases, and pests. Global
Change and Human Health 2(2): 90–104.
Saegerman, C., D. Berkvens, and P. S. Mellor. 2008. Bluetongue epidemiology in the European
Union. Emerging Infectious Diseases 14(4): 539–544 539.
Sands, R. D. and B. A. McCarl, 2005: Competitiveness of terrestrial greenhouse gas offsets: are
they a bridge to the future? In Abstracts of USDA Symposium on Greenhouse Gases and
Carbon Sequestration in Agriculture and Forestry, March 22–24, Baltimore: USDA.
Sathaye, J. A., and K. Andrasko. 2007. Special issue on estimation of baselines and leakage in
carbon mitigation forestry projects. Mitigation Adaptation Strategies for Global Change
12: 963–970.
Schlenker, W., W. M. Hanemann, and A. C. Fisher. 2005. Will U.S. agriculture really benefit from
global warming? Accounting for irrigation in the hedonic approach. American Economic
Review 95(1): 395–406.
Schlenker, W., and M. J. Roberts. 2009. Nonlinear temperature effects indicate severe damages
to U.S. crop yields under climate change. Proceedings of the National Academy of Sciences of
the USA 106(37): 15594–15598.
Schlesinger, W. H. 1999. Carbon sequestration in soils. Science 286: 2095.
Schneider, S. H., W. E. Easterling, and L. O. Mearns. 2000. Adaptation: Sensitivity to natural
variability, agent assumptions and dynamic climate changes. Climatic Change 45: 203–221.
Schneider, U. A., and P. Kumar. 2008. Greenhouse gas mitigation through agriculture. Choices
23(1):19–23.
Schwab, O., T. Maness, G. Bull, and D. Roberts. 2009. Modeling the effect of changing market
conditions on mountain pine beetle salvage harvesting and structural changes in the British
Columbia forest products industry. Canadian Journal of Forest Research 39(10): 1806–1820.
Seager, R., M. F. Ting, I. M. Held, Y. Kushnir, J. Lu, G. Vecchi, H. P. Huang, N. Harnik, A.
Leetmaa, N. C. Lau, C. Li, J. Velez, and N. Naik. 2007. Model projections of an imminent
transition to a more arid climate in southwestern North America. Science 316: 1181–1184.
Searchinger T., R. Heimlich, R. A. Houghton, F. Dong, A. Elobeid, J. Fabiosa, S. Tokgoz, D.
Hayes, and T. -H. Yu. 2008. Use of U.S. croplands for biofuels increases greenhouse gases
through emissions from land-use change. Science 319: 1238–1240.
Seo, S. N., and R. Mendelsohn. 2008a. Measuring impacts and adaptations to climate change: a
structural Ricardian model of African livestock management. Agricultural Economics
38: 151–165.
Seo, S. N., and R. Mendelsohn. 2008b. An analysis of crop choice: Adapting to climate change
in Latin American farms. Ecological Economics 67: 109–116.
Seo, S. N., R. Mendelsohn, A. Dinar, R. Hassan, and P. Kurukulasuriya. 2009. A Ricardian analy-
sis of the distribution of climate change impacts on agriculture across agro-ecological zones.
Africa Environmental and Resource Economics 43: 313–332.
Seo, S. N. 2010. Managing forests, livestock, and crops under global warming: A microecono-
metric analysis of land use changes in Africa. Australian Journal of Agricultural and Resource
Seo, S. N., B. A. McCarl, and R. Mendelsohn. 2010. From beef cattle to sheep under global
warming? An analysis of adaptation by livestock species choice in South America. Ecological
Economics 69: 2486–2494.
Shakhramanyan, N. G., U. A. Schneider, and B. A. McCarl. 2013. U.S. agricultural sector analy-
sis on pesticide externalities—The impact of climate change and a Pigovian tax. Climatic
Change 1–13.
Sirohi, S., and A. Michaelowa. 2007. Sufferer and cause: Indian livestock and climate change.
Climatic Change 85(3): 285–298.
Smith, B., I. Burton, R. J. T. Klein, and J. Wandel. 2000. An anatomy of adaptation to climate
change and variability. Climatic Change 45(1): 223–251.
Smith, P. 2004b. Engineered biological sinks on land. In The global carbon cycle: Integrating
humans, climate, and the natural world, eds. Field, C. B., and M. R. Raupach, 479–491.
Washington, DC: Island Press.
Smith, P., D. Martino, Z. Cai, D. Gwary, H. Janzen, P. Kumar, B. A. McCarl, S. Ogle, F. O’Mara, C.
Rice, B. Scholes, and O. Sirotenko. 2007a. Agriculture. In Climate change 2007: Mitigation.
Contribution of working group III to the fourth assessment report of the Intergovernmental
Panel on Climate Change, eds. B. Metz., O. R. Davidson, P. R. Bosch, R. Dave, and L. A. Meyer.
Cambridge, UK and New York: Cambridge University Press.
Smith, P. D. Martino, Z. Cai, D. Gwary, H. Janzen, P. Kumar, B. A. McCarl, S. Ogle, F. O’Mara,
C. Rice, B. Scholes, O. Sirotenko, M. Howden, T. McAllister, G. Pan, V. Romanenkov, U.
Schneider, and S. Towprayoon. 2007b. Policy and technological constraints to implemen-
tation of greenhouse gas mitigation options in agriculture. Agriculture Ecosystems and
Environment 118: 6–28.
Smith, G. A., B. A. McCarl, C. S. Li, J. H. Reynolds, R. Hammerschlag, R. L. Sass, W. J. Parton, S.
M. Ogle, K. Paustian, J. A. Holtkamp, and W. Barbour. 2007c. Harnessing farms and forests in
the low-carbon economy: How to create, measure, and verify greenhouse gas offsets, eds. Zach
Willey and Bill Chameides, Durham, NC: Duke University Press.
Sohngen, B., R. Mendelsohn, and R. Sedjo. 1999. Forest management, conservation and global
timber markets. American Journal of Agricultural Economics 81:1–13.
Sohngen B., and R. Mendelsohn. 2003. An optimal control model of forest carbon sequestra-
tion. American Journal of Agricultural Economics 85:448–457.
Sohngen, B., and S. Brown. 2008. Extending timber rotations: Carbon and cost implications.
Climate Policy 8: 435–451.
Sohngen, B., R. Alig, and B. Solberg. 2010. The forest sector, climate change, and the global
carbon cycle–Environmental and economic implications. Technical Coordinator 37. http://
www.arlis.org/docs/vol1/C/693780362.pdf#page=45
Stern, N. 2007. The economics of climate change—The Stern review. Cambridge, UK: Cambridge
University Press.
Strassburg, B. B. N., A. S. L. Rodrigues, M. Gusti, A. Balmford, S. Fritz, M. Obersteiner, R.
K. Turner, and T. M. Brooks. 2012. Impacts of incentives to reduce emissions from deforesta-
tion on global species extinctions. Nature Climate Change 2: 350–355.
Taylor, S. W., A. L. Carroll, R.I. Alfaro, L. Safranyik, and B. Wilson, B. (2007). Forest, climate and
mountain pine beetle outbreak dynamics in Western Canada. The mountain pine beetle: a
synthesis of biology, management and impacts on lodgepole pine, 67-94.
Thomson, A. M., K. V. Calvin, L. P. Chini, G. Hurtt, J. A. Edmonds, B. Bond-Lamberty, S.
Frolking, M. A. Wise, and A. C. Janetos. 2010. Climate mitigation and the future of tropical
landscapes. Proceedings of National Academy of Science 107:19633–19638.
Thornton, P. K., J. Van de Steeg, A. Notenbaert, and M. Herrero. 2009. The impacts of climate
change on livestock and livestock systems in developing countries: A review of what we
know and what we need to know. Agricultural Systems 101(3): 113–127.
Thornton, P. K., and P. Gerber. 2010. Climate change and the growth of the livestock
sector in developing countries. Mitigation Adaptation Strategies for Global Change
15: 169–184.
Travasso, M., G. Magrin, W. Baethgen, J. Castaño, G. Rodriguez, J. Pires, A. Gimenez, G. Cunha,

and M. Fernandes. 2006. Adaptation measures for maize and soybean in South Eastern
South America. AIACC working paper No. 28. Washington, DC: AIACC.
Walther, G. R., A. Roques, P. E. Hulme, M. T. Sykes, P. Pysek, I. Kuhn, M. Zobel, S. Bacher, Z.
Botta-Dukát, and H. Bugmann. 2009. Alien species in a warmer world: Risks and opportu-
nities. Trends in Ecology and Evolution 24(12): 686–693.
Wang, W., and B. A. McCarl. 2013. Temporal investment in climate change adaptation and
mitigation. Climate Change Economics DOI: 10.1142/S2010007813500097
Wang, P., R. Sun, J. Hu, Q. Zhu, Y. Zhou, L. Li, and J. M. Chen. 2007a. Measurements and simu-
lation of forest leaf area index and net primary productivity in northern China. Journal of
Environmental Management 85: 607–615.
Wang, Y., G. Zhou, and Y. Wang. 2007b. Modeling responses of the meadow steppe dominated
by Leymus chinensis to climate change. Climatic Change 82(3): 437–452.
Wang, J., R. Mendelsohn, A. Dinar, J. Huang, S. Rozelle, and L. Zhang. 2009. The impact of cli-
mate change on China’s agriculture. Agricultural Economics 40(3): 323–337.
West, T. O., and W. M. Post. 2002. Soil organic carbon sequestration by tillage and crop rota-
tion: A global data analysis. Soil Science Society of America Journal 66: 1930–1946.
West, T. O., and J. Six. 2007. Considering the influence of sequestration duration and carbon
saturation on estimates of soil carbon capacity. Climatic Change 80: 25–41.
Westerling, A. L., H. G. Hidalgo, D. R. Cayan, and T. W. Swetnam. 2006. Warming and earlier
spring increase western U.S. forest wildfire activity. Science 313(5789): 940–943.
Williams, D. W., and A. M. Liebhold. 2002. Climate change and the outbreak ranges of two
North American bark beetles. Agricultural and Forest Entomology 4: 87–99.
Williams, A. P., C. D. Allen, C. I. Millar, T. W. Swetnam, J. Michaelsen, C. J. Still, and S. W. Leavitt.
2010. Forest responses to increasing aridity and warmth in the southwestern United States.
Proceedings of the National Academy of Sciences of the USA 107: 21289–21294
World Resources Institute. (WRI). 2005. Navigating the numbers, greenhouse gas data and
international climate policy. http://www.wri.org/publication/navigating-the-numbers/
Xu, J., R. E. Grumbine, A. Shrestha, M. Eriksson, X. Yang, Y. Wang, and A. Wilkes. 2009. The
melting Himalayas: Cascading effects of climate change on water, biodiversity, and liveli-
hoods. Conservation Biology 23(3): 520–530.
Yan, X., T. Ohara, and H. Akimoto. 2003. Development of region-specific emission factors and
estimation of methane emission from rice field in East, Southeast and South Asian coun-
tries. Global Change Biology 9: 237–254.
Zhang, Y. W., A. D. Hagerman, and B. A. McCarl. 2013. How climate factors influence the spa-
tial distribution of Texas cattle breeds. Climatic Change 118(2):183–195.
Zilberman, D., X. Liu, D. Roland-Holst, and D. Sunding. 2004. The economics of climate change
in agriculture. Mitigation and Adaptation Strategies for Global Change 9: 365–382.
C HA P T E R 10
L A N D U S E , C L I M AT E
C HA N G E , A N D E C O S YS T E M
S E RV I C E S
W I T S A N U AT TAVA N IC H , BE N JA M I N S . R ASH F OR D,

R IC HA R D M . A DA M S , A N D BRU C E A . MC C A R L
Recent studies, including those by the Intergovernmental Panel on Climate Change

(IPCC 2001a, 2001b, 2007a, 2007b), indicate that greenhouse gas (GHG) emissions
and resultant atmospheric concentrations have led to changes in the world’s climate,
including increases in temperatures, extreme temperatures, heat waves, droughts, and
rainfall intensity. Such changes are expected to continue, with substantial impacts on a
range of land uses. Agriculture is potentially the most sensitive economic sector to cli-
mate change, given that agricultural production is highly influenced by climatic condi-
tions. Changes in climate can have direct effects on crop yields and production costs, as
well as indirect effects on relative crop prices. Each effect can drive changes in cropping
patterns.
In view of its importance to economic well-being, effects of climate change on
agriculture have been well researched and documented, dating back at least 25 years
(see Adams et al. 1990; Zilberman et al. 2004; and various IPCC reports). A recent
review of climate change and agricultural effects and adaptations, including land
use, is found in Chapter 9 by McCarl et al. (this volume). This chapter differs from
Chapter 9 by McCarl et al. in that we discuss and provide an empirical application
demonstrating the linkages among agricultural land use, climate change, and ecosys-
tem services.
Adaptation, in the form of changes in crops and their locations, is the most likely
immediate reaction of agricultural producers to climate changes. Crop production,
for example, is expected to increase in high latitudes and decline in low latitudes
(see Adams et al. 1990; Zilberman et al. 2004; IPCC 2007b; 2007c; and Chapter 9 by
McCarl et al.). Research generally suggests that current zones where crops are suit-
able may shift more than 100 miles northward. In the US, northward shifts in the
256 Witsanu Attavanich et al.
crop production mix have already been observed. Southern sections of traditional
wheat-producing regions are now northern sections of corn-producing regions,
as is already being observed in North Dakota (Upper Great Plains Transportation
Institute 2011).
The combination of changes in rainfall, temperature, and carbon dioxide (CO2)
concentrations can also affect the productivity of pasture and rangelands, which are
an important input for livestock production and an important source of wildlife hab-
itat. Pasture production tends to increase in humid temperate grasslands, but is likely
to decrease in arid and semiarid regions (IPCC 2007b), although climate change may
decrease stocking rates. The combination of a northward shift in crop production
and decreasing productivity of pasture and rangeland could lead to substantial con-
version of land from low-intensity agricultural uses to intensive crop production.
Conversion of grassland systems (i.e., pasture and rangeland) to crop production is
associated with losses of grassland-dependent species (Green et al. 2005), releases of
sequestered carbon (Foley et al. 2005), decreases in water quality (Moss 2008), and
increases in soil erosion (Montgomery 2007). Shifts in crop production have been
hypothesized to have important environmental and ecological consequences. These
include increases in air and water pollution as land is converted to more intensive
cropping systems and the reduction of ecological diversity provided by these altered
landscapes. These various environmental and ecological effects are discussed in
IPCC (2007b).
The purpose of this chapter is to discuss the linkages between climate change,
changes in agricultural land use patterns, and the ecological performance of these
altered landscapes. The chapter first reviews the literature on the relationships among
these topics, including studies assessing farmers’ adaptations to a changing climate, and
possible changes in flora and fauna triggered by land use changes. This is followed by an
empirical study directed at one important consequence of such behavior—the effects
of changes in agriculture land use on the ecological performance of wetlands in the
Prairie Pothole Region of North America (PPR), as measured by wetland and waterfowl
abundance.
The PPR is a useful case study area because it is experiencing the effects of climate
change and rapid changes in cropping patterns. The PPR is characterized by highly pro-
ductive agricultural land, producing coarse and small grains, legumes, and livestock,
interspersed with millions of prairie pothole wetlands. Although many of the histori-
cal wetland-grassland complexes in the PPR have been previously altered by agriculture
(Tiner 1984; Kantrud et al. 1989), the region remains the most productive waterfowl
breeding area in North America (Batt et al. 1989). Climate change has the potential to
significantly alter the productivity of the PPR for waterfowl, both through direct effects
on wetlands (e.g., fewer wetland due to increased drought frequency) and through the
indirect effects of human response (i.e., land use change). Thus, this region offers an
excellent case study for understanding the interplay among climate change, human
response, and ecological outcomes.
Land Use, Climate Change, and Ecosystem Services 257
1. Literature Review
This section first reviews the existing literature on potential climate change impacts
on land use in US agriculture and associated adaptive response, with specific focus on
changes in crop production patterns. This is followed by a review of ecological effects
that may arise from the interplay of climate change and agricultural land use changes.
Finally, we review previous studies related to the response of waterfowl to climate
change and land use.
1.1 Change in US Crop Production Pattern as an Adaptive

Response to Climate Change
There are a number of ways that land use can be affected by climate change. For exam-
ple, climate change, through changes in temperature, precipitation, extreme events, and
snow cover, can induce changes in land values and land productivity through changes in
water supply; increased fire risks; productivity of crops, forests, pastures, and livestock;
and spatial and temporal distribution/proliferation of pests and diseases (see Chapter 9
by McCarl et al.).
Change in crop production patterns is one immediate adaptive response of agri-
cultural producers to changes in land value and land productivity. Crop production is
expected to increase in high latitudes and decline in low latitudes since increases in pre-
cipitation are likely in the high latitudes, whereas decreases in rainfall and increased
risk of drought are likely in most subtropical regions. Reilly et al. (2003) constructed
the geographic centroid of production for maize and soybeans and plotted its move-
ment from 1870 (1930 for soybeans) to 1990. They find that both US maize and soybean
production shift northward by about 120 miles. Similar results for corn and soybeans is
shown in Beach et al. (2009) and Attavanich et al. (2011). For example, Attavanich et al.
(2011) find that the production-weighted latitude and longitude of national production
trended northwest from 1950 to 2010 by approximately 100 and 138 miles for corn and
soybeans, respectively.
Most studies conclude that changes in crop yields and relative crop prices induced
by climate change will result in northward shifts in cultivated land (see, e.g., Adams
et al. 1990; Attavanich et al. 2011). The Lake states, Mountain states, and Pacific region
show gains in production; the Southeast, the Delta, the Southern Plains, and Appalachia
generally lose. Results in the Corn Belt are generally positive. Results in other regions
are mixed, depending on the climate scenario and time period. Attavanich and McCarl
(2011) find that percentage of planted acreage of corn, sorghum, soybeans, cotton,
and winter wheat increases the most in Appalachia, Corn Belt, Mountains, and Pacific
regions, respectively. Their results indicate that more cropland would shift to pasture/
grazing land under climate change.
1.2 Effects of Land Use Changes on Ecological Performance

The extant literature on agricultural effects and adaptations clearly demonstrates that
changes in the agricultural landscape are likely to occur as a result of climate change.
How these changes translate into changes in ecological performance requires infor-
mation from the natural sciences. A substantial literature exists on potential effects
of climate change on the environment (i.e., air and water), as well as ecological effects
on flora and fauna. Some of this literature deals only with the physical and biological
basis of such effects. Other studies tie these effects to economic outcomes, such as the
costs of mitigating climate effects on environmental quality or ecological services. Still
another set of studies include the relationship between climate and economic drivers
of land use (e.g., changes in forest or agricultural landscapes arising from changes in
temperature or precipitation or changes in crop prices) on these environmental and
ecological outcomes.
A comprehensive summary of these effects is beyond the scope of this chapter.
Various IPCC reports summarize possible environmental and ecological effects (IPCC
2007b). What is clear is that climate change is expected to adversely affect a range of
plant and animal species. For example, Hoegh-Guldberg et al. (2007) review previ-
ous studies and conclude that if atmospheric CO2 is stabilized at 380 part per million
(ppm), coral reefs will continue to change but will remain coral dominated and carbon-
ate accreting in most areas of their current distribution. However, if atmospheric CO2
is between 450 and 500 ppm, the density and diversity of corals on reefs are likely to
decline, which could lead to largely reduced habitat complexity and loss of biodiversity,
including losses of coral-associated fish and invertebrates.
Sekercioglu et al. (2008) assess risks of bird extinctions caused by climate change.
They reveal that for land birds, approximately 400–500 bird extinctions by 2100 are
projected under intermediate scenarios (surface warming 2.8°C by 2100 with 50% of
lowland bird species assumed to adjust their geographical and topographic distribu-
tions in response to warming), whereas up to 2,498 extinctions (30% of all land birds)
are forecasted under extreme scenarios (surface warming 6.4°C by 2100 with all spe-
cies assumed to adjust their distributions). In another study addressing avian species,
Jetz et al. (2007) estimate projected impacts of climate change and land use change on
the global diversity of birds. They predict that 11–21% of land bird species in the world
could be endangered by climate change and land conversion by 2100 under the four
Millennium Ecosystem Assessment (MA) global scenarios. They also suggest that land
conversion (e.g., deforestation and conversion of grasslands to croplands) could have a
much larger effect on species that inhabit the tropics.
Effects on mammalian species are also noted. For example, Welbergen et al. (2008)
study the effects of temperature extremes on behavior and demography of Australian
flying-foxes. They find that on January 12, 2002 in New South Wales, Australia tem-
peratures exceeding 42.8°C killed at least 3,679 individuals in nine mixed-species colo-
nies. The impacts of these temperatures had differential effects across subspecies, with
the tropical black flying fox experiencing a greater mortality rate than the temperate
grey-headed flying fox.
Reptiles and amphibians are also likely to be affected by climate change. During
field-level monitoring of nests at an alpine site in southern Australia for the period
1997–2006, Telemeco et al. (2009) found that lizards (Bassiana duperreyi, Scincidae)
responded to rising ambient temperatures by increasing their nest depth and increas-
ingly early oviposition; however, they were unable to adjust themselves entirely to
climate change. They reveal that rising ambient temperatures is likely to affect their
hatchling sex ratio.
Finally, numerous studies have documented a wide range of effects of climate
change on plants, both naturally occurring and managed, such as forest and agricul-
ture. For example, Feeley and Silman (2010) report the effects of land use and climate
change on population size and extinction risk of Andean plants. They find that plant
species from high Andean forests may benefit from climate change and expand their
population under a scenario that beneficial land use change practices are adapted
and deforestation is halted (best-case scenario). On the other hand, if the pace of
future climate change exceeds their abilities to migrate (worst-case scenario), all of
these Andean species are projected to experience large population losses and conse-
quently face risk of extinction. Moreover, all species are projected to experience large
population losses regardless of potential migration rates under a business-as-usual
land use scenario.
An example of a study explicitly linking landscapes to climate change and plant spe-
cies is by Lawson et al. (2010). This study links a spatially explicit stochastic population
model to dynamic bioclimate envelopes to investigate cumulative effects of land use,
changed fire regime, and climate change on persistence of a rare, fire-dependent plant
species (Ceanothus verrucosus) of southern California. They reveal that climate change
is the most serious factor determining the reduction of this plant species’ population.
Interactions of climate change with changes in fire regime and land use change could
increase risk to these species.
1.3 Integrated Assessments of Climate Change, Land Use, and

Ecological Performance
As noted, numerous studies over the past two decades have linked economic behav-
ior, changes in land use patterns and climate change. Most of these relate to agricultural
and forest landscapes. A subset of this literature has looked at the co-effects of land use
changes on ecological services and environmental quality, with climate change either
directly or implicitly assumed. These studies have examined the economic impacts of
land use changes or the cost of mitigating for these changes on the ecological or envi-
ronmental metrics of interest. Although a variety of methods are used to link land use
to ecological performance, two general approaches are most common. The first treats
land use as an exogenous input to ecological models that then predict the ecological
performance of alternative land use configurations (e.g., Lenzen et al. 2008; Polasky
et al. 2008). This approach cannot meaningfully inform consequences and policy impli-
cation of climate changes because the feedback among climate, land use response,
and ecological outcomes is incomplete. The second approach explicitly nests land use
change models within ecological models (e.g., Lewis and Plantinga 2006; Langpap and
Wu 2008). This approach typically maintains the modeling feedback loops necessary
to understand the joint effects of exogenous shocks (e.g., climate change) and land use
response on ecological performance. Some representative studies related to climate
change are discussed later.
Wu et al. (2004) explored the influence of cropping pattern changes in the Midwest
United States on regional water quality and, ultimately, on hypoxia potential in the Gulf
of Mexico. They found that changes in cropping patterns (e.g., more corn, less pasture)
and practices (e.g., minimum tillage) affected the run-off and erosion levels within the
region. Although climate change was not explicitly examined, the underlying model-
ing included the influence of differences in weather variables. A number of studies have
addressed the relationship among forest cover, riparian zone health, and water qual-
ity. For example, Watanabe et al. (2006) examined such relationships in the Pacific
Northwest. The water quality parameters of interest were stream temperatures that, if
elevated, can adversely affect cold water species, such as salmonids. The study noted
that even active management of the landscape, such as tree planting or riparian zone
protection, have limited potential to reduce water temperatures to desired levels. Other
studies, such those by as Langpap et al. (2011) and Seedang et al. (2008), also note the
difficulty (high costs) of obtaining reductions in water temperature through forest and
riparian mitigation activities when landscapes have been extensively altered by human
activities.
Pattanayak et al. (2005) performed an analysis of water quality co-effects associated
with greenhouse gas mitigation activities on agricultural lands in the United States. As
with other studies examining carbon sequestration on agricultural lands, they found
substantial carbon sequestration potential from use of alternative cropping practices on
agricultural lands. However, the study also found that such sequestration had an ancil-
lary effect on national water quality. Specifically, overall water quality increased by 2% as
a result of the sequestration practices. In another study of co-effects (co-benefits) of cli-
mate change mitigation policies, Plantinga and Wu (2003) assess the potential positive
externalities of afforestation to sequester carbon. The authors find substantial benefits
in terms of improved water quality (reduced soil erosion) and increased wildlife habitat
from an afforestation policy.
In discussing effects of land use changes on ecological or environmental services, it
is important to note that climate change is also expected to have impacts on both the
participation patterns of recreationists and their willingness to pay to experience recre-
ation activities. As climate change affects wetland resources and their productivity and
snowpack patterns, and redistributes wildlife habitat, the intensity and spatial distribu-
tion of associated recreation activities (e.g., fishing, skiing, wildlife watching and hunt-
ing) are also likely to change. In addition, it is expected that recreationists’ willingness to
pay for preservation of environmental services (use and nonuse values) will be affected.
Loomis and Crespi (1999) review the recreation literature regarding climate change and
conclude that climate change will increase both participation rates and willingness to
pay. Loomis and Richardson (2006) also confirm the effects of climate change on will-
ingness to pay for ecological services. In general, warmer temperatures, earlier springs,
and longer lasting summers are expected to increase the demand and willingness to pay
for a variety of recreation activities. Few studies, however, consider both the direct and
indirect (e.g., land use) effects of climate on recreation. Thus, although climate change
may increase the demand and willingness to pay for outdoor recreation and ecosystem
services, it remains to be seen whether climate-induced land use change will expand or
restrict the supply of recreation opportunities and ecosystem services.
1.4 Response of Waterfowl to Climate Change and Land Use

Waterfowl production in the PPR is highly dependent on the quantity and qual-
ity of wetlands and on the suitability of upland land cover for nesting. Thus, a robust
body of research has examined the relationship between wetland and grassland habi-
tats and waterfowl production (see, e.g., Batt et al. 1989). In general, waterfowl pop-
ulations are highly correlated with the number of wet basins, which generates the
historic boom-and-bust cycle in waterfowl populations (Baldassare and Bolen 1994).
Additionally, upland land cover, which provides critical waterfowl nesting habitat, can
mitigate or exacerbate the effects of pond numbers. Waterfowl nest success is generally
higher in large blocks of native grassland (see, e.g., Stephens et al. 2008) and lowest when
wetland complexes are surrounded by intensive crops (Cowardin et al. 1983). Although
waterfowl can adapt and persist in the margins of cropland (given sufficient wetlands),
population growth rates tend to decrease significantly in highly fragmented landscapes
(Klett et al. 1988).
Given the importance of both wetlands and upland land use, climate change has
the potential to substantially affect waterfowl productivity in the PPR. Some research
has explicitly considered the effect of climate change on wetland functions in the PPR
(Poiani et al. 1996; Johnson et al. 2004, 2005, 2010). In general, this research con-
cludes that the increases in temperature predicted for the PPR will result in shorter
hydro periods and less dynamic wetlands. With sufficient warming (e.g., + 4°C) much
of the PPR will lack wetland conditions necessary to support waterfowl nesting. The
effects of climate change on wetland productivity, however, are heterogeneously dis-
tributed across space, with optimal conditions shifting east as climate warms (Johnson
et al. 2010).
Other research has demonstrated that the effects of climate change on wetland pro-
ductivity depend on upland land use (Voldseth et al. 2007). Upland land uses affects
hydrological processes and vegetation dynamics and therefore influences downstream
prairie wetlands. Some wetland characteristics improve when uplands are in managed
cover (e.g., managed grassland or crops) because these covers increase water delivery
to wetlands. Voldseth et al. (2009) explicitly found that managed covers could partially
mitigate climate effects on wetland function; however, the authors note that although
the wetland may appear more dynamic when surrounded by managed covers, waterfowl
production would be limited due to a lack of adequate nesting habitat.
The research on climate impacts on wetlands and land use impacts on waterfowl sug-
gests that climate change could dramatically reduce waterfowl production in the PPR.
Research using historical climate and land use patterns indicates that conversion of
grassland to crops in the Canadian prairies exacerbated the effects of low water years
(Bethke and Nudds 1995). Additionally, Sorenson et al. (1998) found a strong correla-
tion between drought indices and waterfowl populations in the US PPR and predict that
climate change could reduce waterfowl population by as much as 70% compared to his-
torical levels. Their analysis, however, did not include the possible effects of changes in
upland land use. Although the past literature establishes the importance of both climate
and land use, none of the previously developed models is capable of predicting the joint
effect of climate change and the resulting land use response on waterfowl production in
the PPR.
2. Model Components, Data, and

Process Overview
Changes in temperature and precipitation can directly affect waterfowl production by

impacting the quantity and quality of wetlands and can also indirectly influence water-
fowl production through shifts in upland land use. We use three models to account for
the direct and indirect effects of climate change on waterfowl. In this section, we provide
a detailed description of the two component modeling systems, data used, and then dis-
cuss the model that links the two.
2.1 Model Components and Data

This section provides a detailed description of the two modeling systems (Agricultural
Sector Model [ASM] and Wetland and Waterfowl Model [WWM]) and data we use to
analyze the impact of climate change and land use on waterfowl production. Model
components and process overview are summarized in Figure 10.1.
2.1.1 Agriculture Sector Model

We use an ASM to analyze the complex market mechanism that would occur in the
agricultural sector as a result of climate change. The ASM has been developed on the
basis of past work by McCarl and colleagues (McCarl and Spreen 1980; McCarl 1982;
Schneider et al. 2007). It has been used in climate change-related studies for the IPCC,
Climate scenarios
Temp and precip

Temp and precip
Simulated changes
in crop yields
Spatial
Wetland and waterfowl model
Agricultural mapping of
Ponds = f(precip, temp, land use)
sector model simulated
land use Ducks = f(ponds, land use, harvest)
Waterfowl production given

projected changes in climate
and land use
FIGURE 10.1 Waterfowl survey strata in the US Prairie Pothole Region.
Environmental Protection Agency (EPA), and United States Department of Agriculture

(USDA).
In brief, the ASM model is a price endogenous, spatial equilibrium mathematical pro-
gramming model of the agricultural sector in the United States. It includes all states in
the conterminous United States, broken into 63 agricultural production subregions and
10 market regions (Table 10.1). It also captures land transfers and other resource alloca-
tions within the US agricultural sectors.
Simulated changes of crop yields under climate change scenarios are vital for this study
since climate change affects crop yields, which influences the relative profitability of alter-
native land uses. We obtain simulated changes of crop yields from Beach et al. (2009). They
use a modified version of the Environmental Policy Integrated Climate (EPIC) model,
which was first developed by Williams et al. (1984), to simulate yield changes of 14 crops.1
The authors use projected climate scenarios from four global circulation models (GCMs)2
used in the 2007 IPCC assessment report with the IPCC SRES scenario A1B, which is
characterized by a high rate of growth in CO2 emissions. The scenarios are derived from:
• GFDL-CM 2.0, GFDL-CM 2.1 models developed by the Geophysical Fluid

Dynamics Laboratory (GFDL), United States;
• Meteorological Research Institute Coupled Atmosphere-Ocean General
Circulation Model (MRI-CGCM 2.2) developed by the Meteorological Research
Institute and Meteorological Agency, Japan; and
1
Their studied crops are barley, corn, cotton, forage production, oats, peanuts, potatoes, rice, rye,
sorghum, soybeans, sugarbeets, tomatoes, and wheat.
2 It is common practice in climate change analysis to use several GCM projections to reflect the
uncertainty inherent in such projections.

Table 10.1 Agricultural Sector Model (ASM) regions and subregions

Market region Production region (states/subregions)
Northeast (NE) Connecticut, Delaware, Maine, Maryland, Massachusetts, New
Hampshire, New Jersey, New York, Pennsylvania, Rhode Island,
Vermont, West Virginia
Lake States (LS) Michigan, Minnesota, Wisconsin
Corn Belt (CB) All regions in Illinois, Indiana, Iowa, Missouri, Ohio (IllinoisN,
IllinoisS, IndianaN, IndianaS, IowaW, IowaCent, IowaNE, IowaS,
OhioNW, OhioS, OhioNE)
Great Plains (GP) Kansas, Nebraska, North Dakota, South Dakota
Southeast (SE) Virginia, North Carolina, South Carolina, Georgia, Florida
South Central (SC) Alabama, Arkansas, Kentucky, Louisiana, Mississippi, Tennessee,
Eastern Texas
Southwest (SW) Oklahoma, All of Texas but the Eastern Part (Texas High Plains,
Texas Rolling Plains, Texas Central Blacklands, Texas Edwards
Plateau, Texas Coastal Bend, Texas South, Texas Trans Pecos)
Rocky Mountains (RM) Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah,
Wyoming
Pacific Southwest (PSW) All regions in California (CaliforniaN, CaliforniaS)
Pacific Northwest (PNW) Oregon and Washington, east of the Cascade mountain range
Source: Attavanich and McCarl, 2011
• Coupled Global Climate Model (CGCM) 3.1 developed by the Canadian Centre
for Climate Modeling and Analysis, Canada.
We use these simulated yields results as an input in the ASM to simulate changes in land
use. We first estimate the base scenario (without climate change) and then compare base-
line results to results under climate change simulated from GCMs in 2050, which reflect the
change in crop yields and shifts of crop production patterns as a result of climate change. Due
to the uncertainty of factors in the future, we fix all supply-side factors to their current level in
the base year and only allow the effect of the northward shift of crop production patterns and
the change in crop yields. The introduction of change in crop yields and possibility of north-
ward migration of crops causes ASM to change its equilibrium allocation of land use, crop
mix, trade flows, commodity prices, production, and consumption. Changes in crop acreage
are then used to model the resulting response of wetlands and waterfowl in the PPR.
2.1.2 Wetland and Waterfowl Model

We use a simple regression approach to understand the potential effect of climate and
land use change on wetlands and waterfowl in the PPR. Our approach is similar in spirit
to past models, which have been successfully used to understand the relationship among
wetland numbers, weather characteristics, land use, and waterfowl populations (see
Johnson and Shaffer 1987; Bethke and Nudds 1995; Sorenson et al. 1998). Specifically,
we estimate two regression models using historical data. The first model relates pond
numbers to climate and land use characteristics:
Ponds = f (precipitation, temperature, land use) (1)

The number of waterfowl that settle in the PPR to breed is largely determined by the
availability of wetland habitat. Thus, climate or land use change that affects wet-
land availability is expected to influence breeding waterfowl populations in the PPR.
Previous research has demonstrated the important role of both land use and climate on
wetlands in the PPR (see, e.g., Voldseth et al. 2007; Johnson et al. 2010).
The second model relates waterfowl populations to pond numbers, land use, and
harvest:
Ducks = f (ponds, land use, harvest) (2)

Although ponds largely influence where waterfowl settle in the PPR, upland land use
can reallocate birds on the landscape because females also select landscapes based on the
availability of nesting cover. Harvest during the previous hunting season could also influ-
ence the number birds in the northward migration and thus the number of birds that set-
tle in the PPR. This simple set of regression models allows us to relate changes in climate
and land use to changes in waterfowl breeding populations. Estimates of breeding popula-
tion are the primary determinant of waterfowl hunting regulations and are thus one indi-
cator of the potential social impacts of climate-induced changes in waterfowl populations.
We use data from a variety of sources to estimate (1) and (2). Pond and waterfowl
numbers are from the US Fish and Wildlife Service (USFWS) Waterfowl Breeding
Population and Habitat Survey (USFWS 2009). The survey is one of the most extensive,
both in time and space, wildlife population and habitat surveys in the world. Since 1955,
the USFWS has used aerial surveys to estimate annual pond and waterfowl numbers
within temporally consistent survey strata. Six survey strata (41, 45–49) overlap the US
PPR (Figure 10.2). We therefore use pond and waterfowl estimates from these six strata
to estimate the regression models. For the waterfowl estimates, we use the total count
of dabbling ducks, which constitute the largest subgroup of waterfowl that breed in the
PPR and the bulk of the US harvest.
Historical land use data are from the National Agricultural Statistics Service (NASS
2010). We aggregate annual county-level estimates of area by crop to the strata level. To
be consistent with the ASM model, we focus on the primary field crops in the PPR (e.g.,
corn, soybeans, barley, oats, potatoes, sugar beets, and wheat). Additionally, since all
field crops have similar effects on wetlands and waterfowl nesting habitat, we convert
individual crop area to strata-level shares by dividing the total crop area (sum over indi-
vidual crops) by the total area in each survey strata.
41 45
ND 47
MT 42 43 46
MN
44 48
SD
ID 49
WY
IA
NE
FIGURE 10.2 Model components and process overview.
We collect historical precipitation and temperature data from the National Climate
Data Center (NOAA 2011). We use data from weather stations distributed across each
waterfowl survey strata to estimate average precipitation and temperature at the strata
level. Last, harvest data comes from the Flyways.us website (http://www.flyways.us/),
which is a collaborative effort between waterfowl management agencies to organize data
on North American waterfowl. Harvest data are reported annually at the flyway level for
the period 1961–2009; we therefore use the total harvest for the Central flyway to cap-
ture potential harvest impacts on waterfowl breeding populations.
2.2 Linking Changes in Cropland Use to Waterfowl

Recruitment
To link the effect of climate change on the agricultural sector to waterfowl response, we
use the ASM simulated changes in production of crops as inputs in the regression mod-
els described in Section 2.1. The change in crop production reflects agricultural reaction
to future climate conditions, given market mechanisms. This study compares “baseline”
scenario in 20073 (current condition) with four climate change scenarios in 2050, as dis-
cussed in Section 2.1.
We first use the ASM to predict regional shifts in cropping patterns due to climate
change using the yield effects simulated during 2045–2055 provided in Beach et al.
(2009) for 63 regions in the United States. Although this is a fairly fine level of spa-
tial detail for economic analysis, it is not sufficiently detailed for waterfowl response
3 We adjust the base year used in ASM from 2005 to 2007 to reflect empirical evidences from the latest
Agricultural Census.
modeling. Therefore, we used an auxiliary model to downscale ASM results for use
in the waterfowl model. Development of a county-level counterpart to the ASM crop
mix would not be necessary if we could use county as the ASM spatial specification.
However, not only would such a model be very large, but developing/maintaining
production budget, crop mix, and resource data for such a scale is daunting. Thus,
we run ASM at a more aggregate level and reduce the solution crop mixes to the
county level.
We disaggregate the ASM solution of crop acreage to the county level using a
county-level multiobjective mathematical programming model developed by Attwood
et al. (2000), and used in Pattanayak et al. (2005). The Attwood et al. (2000) model was
later modified by Attavanich (2011) to better reflect the possibility of crop expansion
into new production areas under climate change scenarios. The regionalizing downscal-
ing of Atwood et al. (2000) disaggregates the crop mixes and crop acreage solutions from
the sector model to the county-level by fixing the solutions close to the county-level
historical crop mix. This process cannot fully account for factors that fall significantly
outside the range of historical observation. The modified model uses the area of a par-
ticular crop allocated to an irrigation status in each county as the primary choice vari-
able. This choice variable is constrained so it matches the land area shift in the ASM
but minimally deviates from the Census of Agriculture, US Bureau of Census, USDA
National Resource Inventory (NRI), and USDA county crops data, after accounting for
crop migration due to climate change.
The ASM results provide county-level estimates of crop area, temperature, and pre-
cipitation. For projected climate data, we also obtains IPCC SRES scenario A1B4’s pro-
jected agricultural district level mean temperature and precipitation in the PPR from
four GCMs, as previously discussed. We then use estimated crop area, temperature, and
precipitation to simulate wetland and waterfowl numbers under each climate scenario
by (1) aggregating county crop area to waterfowl strata level and calculating crop shares,
(2) aggregating mean temperature and precipitation predictions under each climate
scenario to waterfowl strata using simple averages, and (3) using the land use and cli-
mate data in the estimated pond and duck equations (e.g., [1]‌and [2]). We use predicted
2007 pond and duck numbers as the baseline for comparison and assume that water-
fowl harvest remains constant on average. For the change in the land use share in the
baseline, we use change in average crop share between the 1900s and 2000s. Since we
do not know how yield levels are likely to change, and since the yield impact is relatively
small, we fix yields at the 2000–2009 average for all simulations. Also, since the climate
4
Scenario A1B most closely reproduces the actual emissions trajectories during the period since
the SRES scenarios were completed (2000–2008). It is reasonable to focus on A1B scenario group
versus those in the B1 and B2 scenario groups that have lower emissions projections because in recent
years actual emissions have been above the A1B scenario projections. At the same time, there has been
considerable interest and policy development to encourage nonfossil fuel energy, which is consistent
with the A1B scenario vs. A1F1 or A2 that assume a heavier future reliance on fossil fuels (Beach
et al. 2009).
predictions represent the decadal average predicted for 2045–2055, we use the same
predicted average temperature and precipitation for all lagged values (i.e., the two-year
lagged precipitation and the one-year lagged precipitation are both the predicted aver-
age precipitation for each climate scenario). Hence, our predicted changes in pond and
duck numbers should be interpreted as averages over the decade not values for any indi-
vidual future year.
3. Model Results
This section reports our empirical findings from the three models. We first provide the
ASM results of projected changes in cropland use. We then report the regression results
of the effect of changes in climate and cropland use on wetland and waterfowl produc-
tion obtained from the wetland and waterfowl model. Finally, we report the simulated
results of the responses of waterfowl populations to changes in climate and land use in
the PPR. Overall, we find that cropland in the PPR is likely to increase. Moreover, lower
pond numbers and higher crop shares are correlated with lower duck numbers. Thus,
ignoring land use change would lead to a significant underestimate of the impacts of
climate change on duck populations by as much as 10% or nearly 300,000 birds. Under
alternative climate scenarios, pond and wetland numbers decrease substantially, and
land use response to climate change generally exacerbates the negative effects of climate
change on duck populations.
3.1 Results from ASM and Its Spatial Mapping

Table 10.2 shows acreage of major crops in the PPR under climate change projected from
the IPCC scenarios compared to the base scenario. Overall, cropland in the PPR is likely
to increase. Considering major crop acreage, corn, soybeans, and hay are projected to
increase by 15%, 39%, and 19%, respectively, whereas acreage of other remaining major
crops tends to decrease, with wheat projected to have the largest acreage reduction.
Because climate-induced shifts in crop production patterns are expected to signifi-
cantly influence the productivity of the PPR for waterfowl, understanding changes in
the movements and distributions of cropland under climate change is important. Our
study provides such information. Figure 10.3 shows the estimated percent change of
county-level crop shares from the base scenario given climate change in 2050 from
four GCM scenarios. We calculate crop share by dividing the county-level acreage of all
major crops (i.e., barley, corn, oats, wheat, hay, silage, soybeans, and sugar beets, which
accounted for 95% of total crop acreage in the PPR in 2007) by the total land area in
that county. In all scenarios, small percent changes in crop share are found in almost
all Iowa counties. Across GCMs, results generally suggest that areas in the eastern sec-
tion of North Dakota, the western section of South Dakota, and the central to northern
Table 10.2 Acreage of major cropland use (1,000 acres) in the Prairie Pothole
Region (PPR) under climate change
Base MRI-CGCM 2.2 GFDL 2.0 GFDL 2.1 CGCM 3.1
Major croplanda (1000 acres)
Barley 2,216 1,438 1,557 1,544 1,510
Corn 19,085 19,961 22,040 21,904 20,614
Oats 513 372 438 371 383
Wheat 14,336 10,517 9,945 10,384 10,492
Hay 5,104 7,119 6,925 6,821 6,885
Silage 800 751 742 1,275 795
Soybeans 15,346 18,275 16,657 15,715 17,652
Sugar beets 710 398 437 404 436
Note: A Crop acreage in the PPR is calculated by breaking down results of ASM crop acreage into
the county level and reaggregating to the PPR level using spatial mapping approach discussed in
section 3.2.
section of Minnesota will have a large increase in crop share, which potentially reduces
waterfowl productivity. Increases in crop share are generally associated with conversion
of grassland to intensive cropland, which reduces quantity and quality of wetlands and
the suitability of upland land cover for waterfowl nesting. On the other hand, we predict
a large reduction of crop share in the southern section of Minnesota and the central to
southern part of South Dakota, which could benefit waterfowl if cropland is replaced by
land covers suitable to waterfowl production (e.g., grassland).
3.2 Results from the Wetland and Waterfowl Models

We use the data described in Section 2.1.2 to estimate (1) and (2). For each equation,
we use a log-linear specification and a one-way fixed effects model to capture unob-
served cross-sectional heterogeneity. We experimented with running the model as
a system and with correcting for autocorrelation and heteroscedasticity (i.e., Parks
method). Since our primary purpose is prediction and none of the alternative regression
approaches produced meaningfully different predictions, we report and use estimates
from the simple model.
In the pond equation (1), we include temperature (T), one- and two-year lags for
precipitation (P) because prairie wetlands are dependent on accumulated soil moisture
(Sorenson et al. 1998), and the change in the crop share (ΔCS) because changes in crop
area better capture potential wetland loss. In the duck equation (2), we include the cur-
rent year and one-year lag of ponds, the crop share, and the lagged harvest (H) since
birds are harvested in the fall and thus affect the following spring migration. The regres-
sions fit the data well with R2 of 0.75 and 0.83 for the pond and waterfowl models,
MRI-CGCM 2.2 relative to the base scenario GFDL 2.0 relative to the base scenario
GFDL 2.1 relative to the base scenario CGCM 3.1 relative to the base scenario
0-5 6 - 20 21 - 40 41 - 60 > 60
-35< -35 --21 -35 --11 -10 --6 -5 - 0
FIGURE 10.3 Estimated percent change of county-level crop share from the base scenario under climate change in 2050 from GCM scenarios
in the Prairie Pothole Region.
respectively, and highly significant F-statistics. The estimated equations, with fixed
effects omitted for simplicity and p-values in parentheses, are:
ln (Ponds ) = 13.76 − 0.04 × T + 0.01 × P + 0.09 × Pt −1 + 0.14 × Pt − 2 − 1.49 × ∆CS

(3)
(< 0.0001)(0.01) (0.85) (0.018) (0.0001) (0.06)
ln ( Ducks ) = 12.25 + 0.000003 × Ponds + 0.000001 × Ponds t −1 − 1.34 × CS + 0.00000008 × H

(< 0.0001) (< 0.0001) (0.013) (0.048) (0.073)
(4)
Parameter estimates generally have the expected sign and reasonable magnitudes.
Higher average temperatures, lower average precipitation, and higher shares of land in
crops decrease pond numbers. Similarly, higher pond numbers and lower crops shares
are correlated with high duck numbers. Harvest has a very small and positive effect on
duck numbers. This seemingly counterintuitive result is consistent with the theory that
harvest is compensatory (i.e., increased survival rates for birds not harvested compen-
sate for the loss of harvested birds—thus, the total population is no smaller than it would
have been in the absence of harvest). The estimate on harvest essentially implies that
every harvested duck is perfectly compensated for through increased production. The
estimated models allow us to predict impacts on waterfowl, given a climate scenario and
predicted land use from the ASM model.
3.3 Effects of Climate and Land Use Change on Waterfowl

Populations in the PPR
Our results suggest that climate change and its induced land use changes will have dra-
matic impacts on waterfowl in the PPR; however, the impacts vary substantially by cli-
mate scenario. To highlight the total impacts and variability across climate scenarios, we
calculate the change in pond and duck numbers (i.e., predicted ducks|baseline—pre-
dicted ducks|climate change), expressed as a percentage, using each GCM (Figure 10.4).
Under three of the four climate scenarios, pond and wetland numbers decrease sub-
stantially, with a worst-case scenario reduction in duck numbers of 25% from the 2007
baseline. For the GFDL 2.0 climate scenario, however, our results suggest an increase in
ponds (9%) and thus duck populations (4%).
Differences across scenarios are largely explained by differences in temperature and
precipitation predictions. The GFDL 2.0 scenario includes minor increases in tempera-
ture accompanied by significant increases in precipitation (by 1 inch, or approximately
25% on average across strata). The increases in precipitation are sufficient to offset any
negative impacts of temperature or land use change on pond numbers. The GFDL 2.1
scenario also has increased precipitation (0.2 inch on average); however, the increase
15%
Ponds Ducks
10%
5%
0%
–5%
MRI–CGCM 2.2
GFDL 2.0
GFDL 2.1
CGCM 3.1
–10%
–15%
–20%
–25%
–30%
FIGURE 10.4 Percent change in pond and duck numbers relative to 2007 baseline under four
alternative climate scenarios.
in temperature predicted in this scenario (2.3°F on average) is sufficient to cause large

decreases in pond and duck numbers. The most extreme scenarios, CGCM 3.1 and
MRI-CGCM 2.2, predict small decreases in precipitation combined with increases in
temperature (2–3°F on average). This combination results in substantial reduction in
pond and duck numbers.
The impacts of climate change on pond and duck numbers, however, are not com-
pletely explained by precipitation and temperature. Land use change plays an impor-
tant role. The ASM model generally predicts an increase in corn and soybeans acreage
and a decrease in wheat acreage in the PPR. The reduction in wheat and other minor
crops can be accompanied by increases in pasture or other major crops, especially corn
and soybeans. Although increases in pasture, which can be suitable waterfowl habitat,
should be positive for waterfowl, the net effect of land use change implies an increase in
the share of land in crops in most waterfowl strata under most climate scenarios. Land
use response to climate change therefore generally exacerbates the negative effects of
climate change on duck populations.
The extent of land use change impacts is best demonstrated by considering tempera-
ture and precipitation impacts absent of land use change. We therefore predict pond
and duck numbers assuming that crop shares remain at baseline levels. For the three
scenarios under which climate change reduces pond and duck numbers, ignoring land
use change would lead to a significant underestimate of climate change impacts (Table
10.3). Not accounting for land use response leads to underestimating ponds and ducks
by as much as 14% and 10%, respectively. This implies underestimating the effect of cli-
mate change on duck populations by nearly 300,000 birds, approximately 10% of the
average current harvest.
Table 10.3 Comparison of pond and duck prediction under climate change, with
and without land use response
Percent change from baseline
With land use change Without land use change
Ponds Ducks Ponds Ducks
MRI-CGCM 2.2 −28% −25% −17% −15%

GFDL 2.1 −19% −20% −5% −16%
CssGCM 3.1 −26% −24% −16% −22%
Although climate change and the associated land use response are likely to have sig-
nificant impacts on ducks in the PPR, the impacts are not uniformly distributed over
space. Predicted temperatures and precipitation under alternative climate scenarios dif-
fer by waterfowl strata. Thus, even with our highly aggregated strata-level data, land use
response and the ultimate impact on ponds and ducks have spatial variations that could
be important for targeting programs to mitigate climate impacts.
Regardless of climate scenario, the Montana portion of the PPR (strata 41) is pre-
dicted to gain ducks with climate change. This region has historically been a relatively
low duck production area because it receives less rainfall than regions to the east and
south. It also has the lowest crop share of any strata. With climate change, the region is
predicted to gain precipitation and have relatively little change in the share of land in
crops. Thus, the region could see increased pond numbers with little loss in waterfowl
nesting habitat (Figure 10.5).
The central portion of the PPR is predicted to see the largest negative impacts to duck
populations. In all climate scenarios, the strata in eastern North and South Dakota lose
significant portions of their current duck populations. These strata currently produce
the most ducks (78%) because they have relatively high pond numbers and, related, sig-
nificant land area not in crop production (>50%). With climate change, these strata are
predicted to experience small to no increase in precipitation, significant temperature
increases, and the largest relative increases in crop land area. As a result, this tradition-
ally productive waterfowl region will have fewer ponds, less nesting habitat, and, as a
result, significantly fewer ducks.
In contrast, the strata in eastern and southern North and South Dakota are predicted
to have very modest gains or losses in duck populations across climate scenarios. Here,
the explanation is largely unrelated to climate change factors. These regions are cur-
rently dominated by intensive crop production (>60%), and, as a result, have relatively
low pond numbers. They therefore have not attracted many breeding ducks in recent
history. The changes in temperature, precipitation, and land use predicted under alter-
native climate scenarios are not substantial enough to significantly change, in either
direction, the waterfowl potential of these regions.
% Change from Baseline
< –35% –34% to –25% –24% to –15% –14% to –0% > 0%
GFDL 2.0 GFDL 2.1
GFDL 3.1 MRI-CGCM2.2
FIGURE 10.5 Percent change in duck populations from baseline by waterfowl survey strata under alternative climate scenarios.
4. Conclusion
This application examines the joint effect of climate change and the resulting land use
response on waterfowl production in the PPR by linking a model of land use changes
induced by climate change with a wildlife habitat and productivity model. Our results
reveal that overall cropland in the PPR is likely to increase, but changes vary spatially
across the region. In all the climate scenarios, small percent changes in crop share are
found in almost all of counties in the Iowa part of the PPR. A majority of climate sce-
narios project that areas in the eastern section of North Dakota, the western section of
South Dakota, and the central to northern section of Minnesota are generally predicted
to have a large increase in crop share. On the other hand, a large reduction of crop share
is likely detected in the southern section of Minnesota and the central to southern part
of South Dakota.
Using the estimates from the climate, wetlands, and waterfowl productivity mod-
els, we also find that (1) higher average temperatures, lower average precipitation, and
higher shares of land in crops relative to pasture decrease pond numbers; (2) lower pond
numbers and higher crop shares are correlated with lower duck numbers; and (3) yield
increase have a very small and positive effect on duck numbers. In addition, when we
include alternative climate scenarios and their effects on crop mixes, we find that pond
and wetland numbers decrease substantially, with a worst-case scenario reduction in
duck numbers of 25% from the 2007 baseline. For the GFDL 2.0 climate scenario, how-
ever, our results suggest an increase in ponds (9%) and thus duck populations (4%). The
study also finds that land use response to climate change generally exacerbates the nega-
tive effects of climate change on duck populations.
The spatial heterogeneity in climate effects could pose serious challenges to waterfowl
conservation efforts targeted toward climate mitigation. Investments could, for exam-
ple, be targeted toward securing habitat in Montana. These investments could further
bolster the predicted increases in duck production, given climate change. The Montana
region, however, has historically produced a very small proportion of the region’s ducks.
Moreover, even with the predicted improvements with climate change, this region does
not produce sufficient additional ducks to offset those lost in other regions. In the three
climate scenarios that reduce duck populations, for example, removing all land from
crop production in the Montana portion of the PPR only offsets 5% of the duck losses in
the rest of the PPR.
This suggests that conservation investments will have to be focused in the central or
eastern portion of the PPR to have any chance of significantly mitigating climate effects.
These regions, however, have high historic shares of land in crops and/or are predicted
to gain significant crop shares under alternative climate scenarios. Land in these regions
is therefore likely to be more highly valued. Thus, conservation efforts will have to com-
pete with agriculture to secure wetland and nesting habitat. Given the duck deficits
predicted under several climate scenarios, limited conservation budgets will likely be
challenged to conserve the amount of area required to mitigate climate change impacts.
Conservation programs will therefore need to be strategically targeted to maximize cost
effectiveness (see, e.g., Rashford and Adams 2007).
The findings and conclusions reported in this section also have implications for the
general literature on land use and ecosystems services reviewed earlier in this chapter.
Specifically, deriving policy-relevant conclusions about complex ecological systems is
only possible by integrating models from multiple disciplines. Moreover, models must
contain the linkages between the economic forces that drive human processes (e.g., land
use decisions) and the ecological performance supported by those same processes (e.g.,
landscapes). The analysis developed here is only possible because of considerable invest-
ment in an ecosystem-based land use model that was developed with input from mul-
tiple disciplines. Although the need for an interdisciplinary approach is intuitive and
has been discussed in the environmental economics literature for decades, in practice,
one finds relatively few empirical applications that are sufficiently integrated to be useful
in assessing the efficacy of alternative policies.
One reason for the lack of empirical applications has been normal tension between
disciplines and the lack of incentives to perform such assessments. In our opinion,
this reluctance or hesitancy to pursue truly integrated analyses is diminishing, due to
enhanced funding opportunities and the broadened curricula of graduate programs in
resource economics and in the natural sciences, which encourages an interdisciplinary
“mindset” in new graduates. Studies of the type we report here demonstrate the poten-
tial utility of investments in interdisciplinary team and model building.
References
Adams, R. M., C. Rosenzweig, J. Ritchie, P. Peart, J. D. Glyer, B. A. McCarl, B. Curry, and J. Jones.
1990. Global climate change and agriculture. Nature 345: 219–224.
Attavanich, W. 2011. Essays on the effect of climate change on agriculture and agricultural
transportation. Ph.D. Dissertation. Texas A&M University.
Attavanich, W., and B. A. McCarl. 2011. The effect of climate change, CO2 fertilization, and crop
production technology on crop yields and its economic implications on market outcomes
and welfare distribution. Paper presented at 2011 AAEA & NAREA Joint Annual Meeting,
Pittsburgh, PA.
Attavanich, W., B. A. McCarl, S. W. Fuller, D. V. Vedenov, and Z. Ahmedov. 2011. The effect
of climate change on transportation flows and inland waterways due to climate-induced
shifts in crop production patterns. Paper presented at 2011 AAEA & NAREA Joint Annual
Meeting, July 24–26, 2011, Pittsburgh, PA.
Attwood, J. D., B. A. McCarl, C. C. Chen, B. R. Eddleman, B. Nayda, and R. Srinivasan. 2000.
Assessing regional impacts of change: Linking economic and environmental models.
Agricultural Systems 63(3): 147–159.
Baldassare, G. A., and E. G. Bolen. 1994. Waterfowl ecology and management. New York: John
Wiley & Sons.
Batt, B. D. J., M. G. Anderson, C. D. Anderson, and F. D. Caswell. 1989. The use of prairie pot-
holes by North American ducks. In Northern prairie wetlands, ed. A. van der Valk, 204–227.
Ames: Iowa State University Press.
Beach, R. H., C. Zhen, A. Thomson, R. M. Rejesus, P. Sinha, A. W. Lentz, D. V. Vedenov, and B.
A. McCarl. 2009. Climate change impacts on crop insurance. Contract AG-645S-C-08-0025.
Final Report. Research Triangle Park, NC: RTI International. Prepared for USDA Risk
Management Agency.
Bethke, R. W., and T. D. Nudds. 1995. Effects of climate change and land use on duck abundance
in Canadian Prairie-Parklands. Ecological Applications 5(3): 588–600.
Cowardin, L. M., A. B. Sargeant, and H. F. Duebbert 1983. Problems and potentials for prairie
ducks. Naturalist 34(4): 4–11.
Feeley, K. J., and M. R. Silman. 2010. Land-use and climate change effects on population size
and extinction risk of Andean plants. Global Change Biology 16(12): 3215–3222.
Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin, M. T. Coe,
G. C. Daily, and H. K. Gibbs. 2005. Global consequences of land use. Science 309: 570–574.
Green, R. E., S. J. Cornell, J. P. W. Scharlemann, and A. Balmford. 2005. Farming and the fate of
wild nature. Science 307: 550–555.
Hoegh-Guldberg, O., P. J. Mumby, A. J. Hooten, R. S. Steneck, P. Greenfield, E. Gomez, C.
D. Harvell, P. F. Sale, A. J. Edwards, and K. Caldeira. 2007. Coral reefs under rapid climate
change and ocean acidification. Science 318 (5857): 1737.
IPCC. 2001a. Climate change 2001: The scientific basis. Contribution of Working Group I to
the Third Assessment Report of the Intergovernmental Panel on Climate Change, eds. J.
T. Houghton, Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell and C.
A. Johnson. Cambridge, UK: Cambridge University Press.
IPCC. 2001b. Climate change 2001: Impacts, adaptation, and vulnerability. Contribution
of Working Group II to the Third Assessment Report of the Intergovernmental Panel on
Climate Change, eds. J. J. McCarthy, O. F. Canziani, N. A. Leary, D. J. Dokken and K. S. White.
Cambridge, UK: Cambridge University Press
IPCC. 2007a. Climate change 2007: The physical science basis. Contribution of Working Group
I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, eds.
S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor and H.
L. Miller. Cambridge, UK: Cambridge University Press.
IPCC. 2007b. Climate change 2007: Impacts, adaptation and vulnerability. Contribution of
Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on
Climate Change, eds. M. L. Parry, O. F. Canziani, P. J. Palutikof, P. J. van der Linden and C.
E. Hanson. Cambridge, UK: Cambridge University Press.
IPCC. 2007c. Climate change 2007: Mitigation of climate change. Contribution of Working
Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change, eds. B. Metz, O. R. Davidson, P. R. Bosch, R. Dave and L. A. Meyer. Cambridge,
Jetz, W, D. S. Wilcove, and A. P. Dobson 2007. Projected impacts of climate and land-use change
on the global diversity of birds. PLoS Biology 5(6): E157. doi:10.1371/journal.pbio.0050157
Johnson, D. H., and T. L. Shaffer. 1987. Are mallards declining in North America? Wildlife
Society Bulletin 15: 340–345.
Johnson, W. C., S. E. Boettcher, K. A. Poiani, and G. Guntenspergen. 2004. Influence of weather
extremes on the water levels of glaciated prairie wetlands. Wetlands 24: 385–398.
Johnson, W. C., B. V. Millett, T. Gilmonov, R. A. Voldseth, G. R. Guntenspergen, and D.
E. Naugle. 2005. Vulnerability of northern prairie wetlands to climate change. BioScience
55: 863–872.
Johnson, W. C., B. Werner, G. R. Guntenspergen, R. A. Voldseth, B. Millett, D. E. Naugle, M.
Tulbure, R. W. H. Carroll, J. Tracy, and C. Olawsky. 2010. Prairie wetland complexes as land-
scape functional units in a changing climate. BioScience 60: 128–140.
Kantrud, H. A., G. L. Krapu, and G. A. Swanson. 1989. Prairie basin wetlands of the
Dakotas: A community profile. Biological Report 85, U. S. Department of the Interior, U. S.
Fish and Wildlife Service, Jamestown, ND.
Klett, A. T., T. L. Shaffer, and D. H. Johnson. 1988. Duck nest success in the prairie pothole
region. Journal of Wildlife Management, 52(3): 431–440.
Langpap, C., and J. Wu. 2008. Predicting the effect of land-use policies on wildlife habitat abun-
dance. Canadian Journal of Agricultural Economics 56: 195–217.
Huang, B, C. Langpap, and R. M. Adams. 2012. The value of in-stream water temperature fore-
casts for fisheries management. Contemporary Economic Policy 30 (2): 247–261.
Lawson, D. M., H. M. Regan, P. H. Zedler, and J. Franklin. 2010. Cumulative effects of land
use, altered fire regime and climate change on persistence of Ceanothus verrucosus, a rare,
fire-dependent plant species. Global Change Biology 16 (9): 2518–2529.
Lenzen. M., A. Lane, A. Widmer-Cooper, and M. Williams. 2008. Effects of land use on threat-
ened species. Conservation Biology 23(2): 294–306.
Lewis, D. L., and A. J. Plantinga. 2006. Policies for habitat fragmentation: Combining econo-
metric models with GIS-based landscape simulations. Land Economics 83: 109–127.
Loomis, J., and J. Crespi. 1999. Estimated effects of climate change on selected outdoor rec-
reation activities in the United States. In The impact of climate change on the United States
economy, eds. R. Mendelsohn and J. Neumann, Chapter 11. Cambridge, UK: Cambridge
University Press.
Loomis, J., and R. Richardson. 2006. An external validity test of intended behavior: Comparing
revealed preference and intended visitation in response to climate change. Journal of
Environmental Planning and Management 49(4): 621–630.
McCarl, B. A., and T. H. Spreen. 1980. Price endogenous mathematical programming as a tool
for sector analysis. American Journal of Agricultural Economics 62(1): 87–102.
McCarl, B. A. 1982. Cropping activities in agricultural sector models: A methodological pro-
posal. American Journal of Agricultural Economics 64(4): 768–772.
Montgomery, D. R. 2007. Soil erosion and agricultural sustainability. Proceedings of the National
Academy of Sciences of the USA 104: 13268–13272.
Moss, B. 2008. Water pollution by agriculture. Philosophical Transactions of the Royal Society of
London, Series B 363: 659–666.
National Agricultural Statistics Service (NASS). 2010. County estimates. http://www.nass.
usda.gov/.
National Oceanographic and Atmospheric Administration (NOAA). 2011. National Climatic
Data Center. http://www.ncdc.noaa.gov/oa/ncdc.html.
Pattanayak, S. K., B. A. McCarl, A. J. Sommer, B. C. Murray, T. Bondelid, D. Gillig, and B.
DeAngelo. 2005. Water quality co-effects of greenhouse gas mitigation in US agriculture.
Climatic Change 71 (3): 341–372.
Plantinga, A. J., and J. Wu. 2003. Co-benefits from carbon sequestration in forests: Evaluating
reductions in agricultural externalities from an afforestation policy in Wisconsin. Land
Economics 79(1): 74–85.
Poiani, K. A., W. C. Johnson, G. A. Swanson, and T. C. Winter. 1996. Climate change and north-
ern prairie wetlands: Simulation of long-term dynamics. Limnology and Oceanography
41: 871–881.
Polasky, S. E., E. Nelson, J. Camm, B. Csuti, P. Fackler, E. Lonsdorf, C. Montgomery, D. White,
J. Arthur, B. Garber-Yonts, R. Haight, J. Kagan, A. Starfield, and C. Tobalske. 2008. Where
to put things? Spatial land management to sustain biodiversity and economic returns.
Biological Conservation 141:1505–1524.
Rashford, B. S., and R. M. Adams. 2007. Improving the cost-effectiveness of ecosystem manage-
ment: An application to waterfowl production. American Journal of Agricultural Economics
89(3): 755–768.
Reilly, J., F. Tubiello, B. A. McCarl, D. Abler, R. Darwin, K. Fuglie, S. Hollinger, C. Izaurralde, S.
Jagtap, and J. Jones. 2003. US agriculture and climate change: New results. Climatic Change
57 (1): 43–67.
Schneider, U. A., B. A. McCarl, and E. Schmid. 2007. Agricultural sector analysis on greenhouse
gas mitigation in US agriculture and forestry. Agricultural Systems 94(2): 128–140.
Seedang, S., S. Fernald, R. M. Adams, and D. L. Landers. 2008. Economic analysis of water
temperature reductions practices in a large river flood plain. River Research and Applications
24: 941–959.
Sekercioglu, C., S. Schneider, J. Fay, and S. Loarie. 2008. Climate change, elevational range
shifts, and bird extinctions. Conservation Biology 22 (1): 140–150.
Sorenson, L. G., R. Goldberg, T. L. Root, and M. G. Anderson. 1998. Potential effects of global
warming on waterfowl populations breeding in the Northern Great Plains. Climatic Change
40: 343–369.
Stephens, S. E., J. A. Walker, D. R. Blunck, A. Jayaraman, D. E. Naugle, J. K. Ringleman, and
A. J. Smith. 2008. Predicting risk of habitat conversion in native temperate grasslands.
Conservation Biology 22: 1320–1330.
Telemeco, R. S., M. J. Elphick, and R. Shine. 2009. Nesting lizards (Bassiana duperreyi) compen-
sate partly, but not completely, for climate change. Ecology 90(1): 17–22.
Tiner, R. W. 1984. Wetlands of the United States: Current status and recent trends. Cambridge,
Washington, DC: US Fish and Wildlife Service, US Government Printing Office.
Upper Great Plains Transportation Institute. 2011. Road investment needs to support agricul-
tural logistics and economic development in North Dakota. Agricultural Roads Study, North
Dakota State University.
USFWS. 2009. Waterfowl Breeding Population and Habitat Survey. U. S. Department of the
Interior, https://migbirdapps.fws.gov/mbdc/databases/db_selection.html.
Voldseth, R. A., W. C. Johnson, T. Gilmanov, G. R. Guntenspergen, and B. V. Millett. 2007.
Model estimation of land-use effects on water levels of northern prairie wetlands. Ecological
Voldseth, R. A., W. C. Johnson, G. R. Guntenspergen, T. Gilmanov, and B. V. Millett. 2009.
Adaptation of farming practices could buffer effects of climate change on northern prairie
wetlands. Wetlands 29: 635–647.
Watanabe, M., R. M. Adams, and J. Wu. 2006. The economics of environmental manage-
ment in a spatially heterogeneous river basin. American Journal of Agricultural Economics.
88(3): 617–631.
Welbergen, J. A., S. M. Klose, N. Markus, and P. Eby. 2008. Climate change and the effects of tem-
perature extremes on Australian flying-foxes. Proceedings of the Royal Society B: Biological
Sciences 275(1633): 419–425.
Williams, J. R., C. A. Jones, and P. T. Dyke. 1984. A modeling approach to determining the
relationship between erosion and soil productivity. Transactions of the American Society of
Agricultural Engineers 27(1): 129–144.
Wu, J., R. M. Adams, C. Kling, and K. Tanaka. 2004. From micro-level decisions to land-
scape changes: An assessment of agricultural conservation policies. American Journal of
Zilberman, D., X. Liu, D. Roland-Holst, and D. Sunding. 2004. The economics of climate change
in agriculture, mitigation and adaptation strategies. Global Change 9: 365–382.
C HA P T E R 1 1
FIRE
An Agent and a Consequence of
Land Use Change
C L A I R E A . MON TG OM E RY
As long as people and fire have coexisted on this planet, fire has been both purposefully
used as an agent and subsequently experienced as a consequence of land use change.
In fact, our ability to manipulate the landscape to our own purpose is fundamentally
dependent on our ability to use fire. This chapter begins with a history how people have
used fire as an agent of land use change over many millennia, how attitudes toward
fires have evolved over time, how fire policy has developed in the United States over
the last century, and what challenges for fire policy are emerging now. That is followed
by a description of three core themes that appear in the literature on the economics of
fire: spatial externalities, incentives, and risk-based decision analysis. The chapter closes
with a discussion of how future economics research might best contribute to the design
of efficient and effective fire policy for the future.
1. People and Fire
In the suite of books that comprise his Cycle of Fire and his subsequent synopsis,
“Fire: A Brief History” (Pyne 2001), Stephen Pyne, fire historian, describes three great
stages in the relationship between people, fire, and land use.
First, there was aboriginal fire. The colonization of vast areas by people was only
possible with fire as a tool. Prior to the capture of fire by humans, most places were
inhospitable for human habitation. People occupied only a small part of the land-
scape. However, once people could carry fire, they took it everywhere. As they col-
onized new places, they used fire to transform the landscape into, and maintain it
as, a place suitable for people to live in. Every major migration can be tracked in
the geologic record by a layer of charcoal that was left when the existing biota was
282 Claire A. Montgomery
burned and replaced with more fire-friendly ecosystems in which fuel was regener-
ated and made available for repeated burning. People brought more regularity to the
fire regime than had existed when fire occurrence depended on the coincidence of
lightning, burnable fuel, and dry weather, and most places burned more frequently
than before.
Regular use of fire altered ecosystems; it changed travel patterns of wildlife, moisture
regimes, and even entire microclimates. Vegetation and wildlife adapted to its regular
presence. It is usually this systematic and repeated use of fire by aboriginal people that
we really mean when we talk about “natural” fire regimes.
In this first stage, fire was used as a tool to convert large areas for humans to live
in, to hunt, to forage for food, and to protect villagers from predation, from warfare,
and, ironically, to protect from uncontrolled wildfire. Perhaps most importantly,
fire nurtured a sustainable supply of combustible fuel to be used for cooking, heat-
ing, and as the center of village life. Without anthropogenic fire, many areas would
revert to closed canopy forest, impenetrable by light and fire and hostile to human
presence.
Aboriginal fire was followed by agricultural fire. Large areas of existing vegeta-
tion had to be cleared for farming, and fire was an invaluable tool for that purpose.
Fire released nutrients from existing vegetation and, often, the period immediately
following the first burning was an enormously productive one. In some places, fire
was then used to beat back the encroaching vegetation and maintain the land in its
agricultural use. Once farming was established, fire could be part of a cycle of peri-
odic renewal that sustained continued productivity of a site; this included the burn-
ing of fields to stimulate sprouting of forage for livestock whose manure replenished
soil nutrients; it also included postharvest field burning to release nutrients from the
remaining vegetation, to prepare the field for replanting, and to purge the soil of dis-
ease and pests. In places where nutrients are so easily leached from soil that they are
mostly held in biomass, fire was part of the longer cycle of shifting agriculture. Fields
were cleared and nutrients released from the existing vegetation by fire. Farming con-
tinued until the soil was depleted. Then new sites were cleared by fire and old sites
were left for long periods of time in which the forest vegetation regrew and restocked
with nutrients.
Little remains of aboriginal fire; because it is free-ranging, it poses a threat in a densely
populated world. However, agricultural fire remains an important component of life.
For example, 1.2 million hectares of cropland is burned annually in the contiguous
United States alone (McCarty et al. 2009). Broadcast burning after clear-cut logging is a
common practice to prepare logged sites for forest regeneration (Van Lear and Waldrup
1991). Shifting agriculture is estimated to support up to 500 million people, mainly in
the tropics (Kleinman, Pimentel, and Bryant 1995). Biomass continues to be removed
from forests to provide fuel for the home; currently, fuelwood comprises nearly half of
global wood consumption (FAO 2011).
Now, most of the developed world is either in, or in transition to, the third stage in
which agricultural fire is supplanted by industrial fire. The nature of the relationship
Fire 283
between fire, people, and land use is, again, changing dramatically. Industrial fire is com-
bustion that is confined to engineered containers such as the engines of automobiles, the
boilers that produce steam heat, and the plants that burn coal to generate electricity to
power the appliances in urban homes. Industrial fire is ubiquitous in our lives and yet we
rarely see its flame. It is mostly fueled by fossilized biomass rather than living biomass. It
replaces agricultural fire with petrochemicals that enrich soil and kill pests—chemicals
that are produced elsewhere using industrial fire. Industrial fire doesn’t happen on the
landscape, but its implications for land use change are vast.
First, as living biomass is replaced with fossilized biomass to fuel combustion, people
are no longer altering only the landscape through fire; they are also altering the atmo-
sphere through the release of carbon from its geologic cache. The economics of carbon
and global climate change are addressed elsewhere in this book. Suffice it to say here that
climate change is a direct consequence of how the relationship between fire, people, and
land use is changing.
Second, the spread of industrial fire happened alongside urbanization. The combined
effect of these two trends is that people tend to see fire less as an integral part of their
homes and daily lives and more as a threat to both. People rarely see fire’s open flame
and, when they do, it is often in the form of wild and uncontrolled fire that destroys and
kills. Moreover, people who live primarily in the built environment tend to see the coun-
tryside less as a resource to be managed for use and more as a resource to be preserved to
provide refuge from the city and reserves for vegetation and wildlife. This urge to “pro-
tect nature” is having unintended consequences, at least with respect to wildfire. One
policy outcome has been the banishment of anthropogenic fire from most places. On
the other hand, policy for natural fire has been somewhat contradictory. In many places,
such as the western United States, fire is treated as a threat. It is aggressively suppressed
with the discipline and ardor of military action. In a few places where fire is recognized
as a natural ecological process, there have been experimental attempts to let wildfire
burn unhampered. For example, the National Park Service in the United States began
to reintroduce wildfire in the national parks in the 1950s and has largely held to that
policy in spite of a few spectacular fires, such as the fire that burned 320,000 hectares in
Yellowstone National Park in 1988, which have stirred public controversy (Carle 2002;
Omi 2005).
Indigenous people of the United States used fire for a variety of purposes, typically
burning in the spring and fall. Documented uses include driving game animals to places
where they were easier to hunt, encouraging sprouting of green forage for game, favor-
ing fire-adapted edible plants such as yucca, berries, camas, providing fire protection
by burning areas around settlements, controlling flies and mosquitoes, opening spaces
for easy travel, reducing fuels so that summer fires are less severe, and punishing and
harassing enemies (Williams 2000). Burning was not completely controlled, and there
were likely many escaped fires. Most areas appear to have been burned frequently—
sometimes every 1–3 years. The overall effect was that when European immigrants ven-
tured west, they found broad expanses of tall grass prairie, oak savanna, or chaparral,
particularly in river basins.
By the time European settlement of the west occurred, indigenous populations had
been decimated by disease and conquest; they and their fires were, for the most part,
removed from the landscape. Encroachment of forest on the valleys and plains was well
underway. For example, in the Willamette Valley of Oregon, early logging towns were
located high in the Coast Range or the foothills of the Cascades because that was where
the interface between oak savanna and Douglas fir forest was. Now the forest has crept
up to the Willamette River except where it has been blocked by agriculture and urban
development.
The elimination of frequent burning led to a build-up of fuels so that when fire did
occur, either from lightning strikes or escaped anthropogenic fire, it was more likely
to be large and catastrophic. For example, in 1871, the Peshtigo fire burned 500,000
hectares in Wisconsin and Michigan and killed at least 1,250 people. The Yacolt fire in
Washington burned about 400,000 hectares and killed 38 people in 1902. The Great Fire
of 1910 burned 1.2 million hectares in Washington, Idaho, and Montana killing over 85
people (Omi 2005).
This set the stage for federal fire policy in the 20th century. When the United States
National Forest Reserve system (later to become the USDA Forest Service) was estab-
lished in 1891, its first responsibility was to protect the forests from fire. Debate raged
about whether all wildfire should be suppressed or whether wildfire could be controlled
and used for beneficial purposes (Carle 2002). However, the great fires that were occur-
ring across the continent dampened the debate and a policy of aggressive fire suppres-
sion was adopted. The 1908 Forest Fires Emergency Act authorized unlimited spending
to fight wildfire (Omi 2005) effectively eliminating fiscal responsibility. In 1935, fed-
eral forest fire policy became formalized in the so-called “10:00 A.M. policy,” the goal of
which was to contain every wildfire by 10:00 A.M. the day after it was reported. Bambi
and Smokey the Bear brought a message to the public that forest fire is an enemy to be
vanquished.
The suppression policy was successful in reducing the extent of wildfire for a while,
but by the 1970s, it was becoming apparent to fire ecologists and forest managers that
a policy of aggressive suppression could not be sustained (Biswell 1980). Fire exclusion
was driving forest conditions well outside the range of variation that had prevailed in the
forests of the western United States for millennia. Forest fire fuels were accumulating in
the form of downed woody debris and dead standing trees. Without fire to cleanse the
forest of weak and diseased trees, whole forests were swept by insect infestations, add-
ing even more to fuel loads. Ingrowth of seedlings that would have been eliminated by a
light burning developed into ladder fuels capable of carrying fire into the forest canopy
where it is far more deadly. When wildfire occurred, it was becoming far more difficult
and costly to contain.
Recognizing the beneficial effects of light fire, federal land management agencies
began to revise fire policy to encourage preventative measures, such as mechanical fuel
removal, prescribed burning, and restoration thinning. The most recent guidelines allow
all fires, including human-caused or unplanned fires, to be used to achieve management
Fire 285
and resource goals (Lasko 2010). This means that both anthropogenic fire (prescribed
burning) and cautious use of wildfire are now considered viable management tools.
However, change has been slow to arrive. Although current policy allows for wildfire
use, less than 0.5% of wildfires originating on federal land were allowed to burn between
1998 and 2008 (NIFC 2011). Wildland fire activity in the western United States con-
tinues to increase in the 2000s, and fire suppression expenditures continue to escalate
(Calkin et al. 2005; NIFC 2011). Even though the 1908 Forest Fires Emergency Act was
repealed in 1978, Congress continues to reimburse the agencies for fire suppression
costs. The USDA Forest Service, which is responsible for approximately 70% of all wild-
land fire expenditures in the United States, tripled its annual expenditure on fire sup-
pression in the 2000’s over the levels of the previous three decades (Abt, Prestemon, and
Gebert, 2009).
At the same time, the aging of the baby-boom generation, increasing wealth, and
technological change that allows people to interact with one another at a distance
are fueling a wave of low density housing development along the edge of, and inter-
mixed with, wildland that has forest fire fuels. The extent of area that can be classified
as wildland-urban interface (WUI) has increased by over 50% since the 1970s and is
expected to increase by another 10% by 2030. Nearly 90% of the WUI in the 11 western
states of the United States can be classified as a high wildfire-hazard type (Theobald
and Romme 2007). People are drawn to the WUI because they want the amenities of a
forest or wildland setting. However, they also want the benefits of urban life, including
protection of lives and property from fire. This phenomenon of WUI development is
challenging to wildland managers because it imposes conflicting management man-
dates—to manipulate vegetation to block fire from destroying residences while, at
the same time, managing vegetation and wildlife in its natural state—of which fire is a
component.
Although the details of fire policy and forest ecology described here are specific to the
United States, the general pattern—ecosystems adapted to frequent burning by indig-
enous people, subsequent banishment of anthropogenic fire with industrialization and
urbanization, massive fuel accumulation, escalating wildfire severity and cost—is not
(Pyne 1995). In the Russian taiga, fire suppression was as aggressively military during
the Cold War as it was in the United States. Anthropogenic fire was totally banned dur-
ing the Brezhnev years. Now, with the resulting fuel accumulation, combined with the
collapse of the Soviet Union, the number of wildfires is fairly steady, but area burned and
expenditure to control fire are exploding. On the Iberian peninsula, fuel loading was
controlled more by intense grazing than by anthropogenic fire, but with the shift from
rural to urban life, grazing diminished, fuels built up, and wildfire became more fre-
quent and more damaging. The authoritarian regimes of Salazar in Portugal and Franco
in Spain took up the charge to suppress wildfire with the same military vigor as in the
United States. Similar patterns appear in Brazil, Sweden (a country thought to be named
after its long practice of slash and burn, or svedje, agriculture), Canada (Martell 2011),
and elsewhere.
2. Economics of Fire
Economists have only recently turned to these issues, but three themes appear to be emerg-
ing in the economics literature: inefficiency in the face of the spatial externalities associ-
ated with fire, the influence of institutional incentives (e.g., liability rules, insurance, and
regulations) on private landowners’ and public land managers’ decisions about fire risk
management, and the development of tools to guide fire and fuel management decisions.
2.1 Spatial Externalities

Spatial externalities arise because fire spreads. Any fuel management, timber harvest, or
fire that occurs on one unit of land affects the probability of fire reaching adjacent units
and, hence, fire risk on those units.
Konoshima et al. (2008, 2010) formulated the spatial problem for timberland as
a stochastic dynamic programming problem in which a single landowner chooses
the spatial configuration of fuel treatment and timber harvest in each time period
to maximize the expected net present value of timber production over two periods,
assuming Faustmann rotation age and no fire beyond the time horizon. The spa-
tial spread of fire was explicitly modeled using the equations that drive fire spread
in widely used fire simulation models, such as FARSITE (Rothermel 1972; Finney
1998) in order to estimate transition probabilities. Fuel treatment reduces fire spread
rate across the landscape; timber harvest generates revenue but leaves flammable
fuel on-the-ground and a new stand through which fire moves relatively quickly.
Ignoring the spatial externality, fire risk acts as an increase in the discount rate and
reduces optimal timber harvest age (Reed 1984). When the spatial externality is taken
into account, however, it acts to increase optimal harvest age, because timber harvest
speeds the spread of fire through the harvested unit to adjacent units. Hence, there is
a trade-off between harvesting earlier to protect on-site timber value and harvesting
later to protect adjacent timber value. Which effect dominates for a particular unit
depends on timber value on site, timber value on adjacent units, and the location of
the unit with respect to topography and prevailing wind. As formulated, this model
is relevant primarily for industrial timberland owners.
Crowley et al. (2009) extended the problem by modeling two adjacent timberland
owners, each choosing timber and fuel management to maximize the expected net
present value of timber production and amenities on their respective properties. Fire
suppression effort, when fire occurs, is determined (and its cost is born by) an external
government agency and depends on fuel and timber value. The problem was formulated
as a game and was solved for three cases:
1. Landowners understand the effect of fuel treatment on adjacent properties on

their own fire risk. They do timber and fuel management.
Fire 287
2. Landowners ignore the spatial fuel treatment externality. They do timber and
fuel management.
3. Landowners do timber management and no fuel treatment. This model is equiv-
alent to Reed (1984) in which an isolated timber landowner responds to fire risk
by harvesting earlier.
The solutions were compared to a “socially optimal” baseline in which fuel and timber
management decisions are made by a single agent to maximize expected net present
value of timber and amenities on both units. The results demonstrate the potential for
inefficient fuel management choices. Landowners do too little fuel treatment for two
reasons: they don’t bear fire suppression costs, which depend on fuel loads, and they try
to free-ride on their neighbors’ fuel treatment. In numerical simulations, it appears that
the suppression cost externality is a far greater problem than the spatial fuel treatment
externality. As formulated, this model is relevant primarily for small private timber
landowners such as those that dominate the landscape in the southern and southeastern
United States and in Scandinavia.
Busby, Albers, and Montgomery (2012) took the problem into the WUI where private
owners choose fuel management to protect their own buildings and the stream of ame-
nity values generated on their own and adjacent units that accrue to them. Public land
managers choose fuel management to protect public goods such as aesthetics, wildlife,
and ecosystem health. Because nontimber values dominate in the WUI, timber was
not included in this model. Fuel treatment reduces fire severity and, hence, extent of
damage on the treated unit and adjacent units. However, landowners ignore the effect
of their fuel treatment on fire spread. Fire suppression is exogenous and not part of the
landowners’ decision process. This is approximately true in the WUI where any wildfire
brings on a full (and expensive) fire-suppression effort.
The authors were particularly interested in the effect of the spatial pattern of owner-
ships (public and private) on the extent of inefficiency arising from spatial externalities
in both fire risk and amenity value. They formulated the problem as a game in which
one player represents a coordinated private landowner and the other player represents
the public-land-management agency. The game was solved for each of five spatial pat-
terns of ownership ranging from two adjacent blocks of public and private land to a
nine-square checkerboard of public and private ownership. The following cases were
modeled: spatial externality from fire spread only; spatial externalities from both fire
spread and amenity values; and three forms of response of fire severity to fuel treatment.
The results suggest that the spatial pattern of ownership in the WUI matters; increasing
fragmentation decreases efficiency as landowners free-ride on the fuel treatment effort
of adjacent landowners. That inefficiency is offset somewhat by the presence of off-site
amenities as landowners increase fuel treatment in order to provide protection for
amenities generated on adjacent units. The results also suggest that nonlinear response
functions for fire severity give rise to strategic behavior as landowners choose their
own fuel treatment levels to influence fuel treatment on adjacent units. Evidence from
an empirical analysis of homeowners’ fuel treatment choices in the WUI in Colorado
(Shafran 2008) indicates that landowners are indeed influenced by their neighbors’ fuel
treatment choices and that it matters whether adjacent property is publicly or privately
owned. These models are particularly relevant for understanding how optimal policy
might differ between the two main categories of WUI: the interface where residential
development presses up against wildland and the intermix where residential develop-
ment is dispersed throughout the wildland.
2.2 Incentives Matter
Fire management in and around the WUI is further complicated by the fact that both
private homeowners and agency fire managers face a complex mix of incentives for
managing fire risk.
Private homeowners may engage in risk averting activities and/or they may purchase
insurance. However, the efficiency of insurance and real estate markets depends on the
availability of accurate information about wildfire hazard. The state of California, where
structure values in the WUI are high, has implemented a natural hazards disclosure law
(AB 1195) that requires homeowners to inform potential buyers of hazard ratings for an
array of hazards, including wildfire, when they sell their homes. Troy and Romm (2007)
used hedonic pricing to explore how house prices were affected by implementation of
the law, but their results were mixed. Troy (2007) speculated that other fire insurance
laws may have actually promoted development in the WUI in California by subsidizing
insurance in high hazard zones for people who otherwise could not obtain it.
To the extent that fire reduces both market (e.g., structures) and nonmarket values
(e.g., amenities), insurance can only partially compensate a loss. Therefore, there is rea-
son for homeowners to avert risk even when full market insurance is available and risks
are accurately known. In fact, Talberth et al. (2006) analyzed experimental and survey
data and found that most households choose to purchase a mix of insurance and avert-
ing activities and that households that rated amenity values as high devote relatively
more of their budget to averting risk than those who do not.
Ehrlich and Becker (1972), in a theoretical analysis of the demand for insurance,
defined two types of risk-averting actions: “Self-insurance” reduces the size of a loss
when a hazardous event occurs and “self-protection” reduces the probability of a haz-
ardous event occurring. Fuel treatment by individual households is most effective as
self-insurance; homeowners treat fuel around their homes to reduce the intensity of any
fire that occurs and/or to create “defensible space” immediately around the structure.
Self-protection requires that fuel treatment be broad enough in scale to slow or block
the spread of fire across the landscape, increasing the likelihood that fire will be con-
tained before it reaches residential developments. Private landowners are unlikely to
undertake treatment at such a scale because it requires coordination among landowners
and because the large expanse of forest where fire often originates is mostly public in the
western United States. Hence, investment in self-protection is largely the responsibility
of public-land-management agencies.
Fire 289
Fire managers in public agencies face a bewildering mix of conflicting incentives.

Decisions regarding management for fire prevention and fire suppression are related and
they should be made simultaneously to maximize the expected net value of the landscape
or, equivalently, to minimize the expected value of all fire-related costs including the net
value change of the forest resulting from fire. The optimal fire-management problem
was first formulated as the “least-cost-plus-loss” model (Sparhawk 1925) in which fire
managers choose the optimal level of suppression effort to minimize suppression cost
plus loss to fire. The formulation was later modified to incorporate preventative measures
and beneficial effects of wildfire (Althaus and Mills 1982) so that fire managers choose
the optimal level of suppression and prevention to minimize treatment and suppression
cost plus net value change of the forest when the benefit of fire is accounted for as well as
the loss. However, under current policy, suppression and prevention decisions are dis-
connected. Because forest wildfire continues to be treated as an emergency, there is no
effective budget limit on suppression. Suppression cost savings that might result from
preventative fuel treatment do not accrue to the agencies making fuel-treatment deci-
sions. Donovan and Brown (2005) suggested adjusting the incentives faced by fire man-
agers by allotting them a joint budget for prevention and suppression, but in the absence
of such a measure, the current structure of incentives gives rise to an array of inefficien-
cies. Public agencies will spend too much on fire suppression and will be unlikely to allow
wildfire to burn even if it seems that it would be beneficial to do so.
There is substantial evidence that fire managers in public agencies have attitudes
toward risk that reinforce the problem. Wilson et al. (2011) used a web-based survey
instrument to explore whether risk biases are held by USDA Forest Service personnel
responsible for decisions regarding wildfire and, if so, how such biases affect the deci-
sions they make. Their survey results revealed that fire managers do exhibit risk-based
biases. They display an aversion to loss; they prefer long-term to short-term risk; and
they tend to rely on “status quo” rather than sophisticated risk-assessment tools when
faced with complexity and uncertainty. In another recent survey, fire managers indi-
cated that they tend to choose high-cost suppression strategies even though they say
they would prefer to choose more cost-effective strategies (Calkin et al. 2012). Berry
(2007) hypothesized that social and political pressures contribute to suboptimal sup-
pression decisions and, indeed, Donovan, Prestemon, and Gebert (2011) included an
array of variables representing newspaper coverage and political influence in a regres-
sion analysis of suppression costs for large fires, and they found empirical evidence to
support that notion.
With a virtually unlimited fire-suppression budget, not only is there little incen-
tive to control fire-suppression costs, but there is little incentive to invest efficiently in
fire-hazard reduction. The analysis reported in Crowley et al. (2009), described ear-
lier, demonstrates the potential for too little investment in fuel treatment when the
suppression cost savings that result are externalities. In an innovative approach to the
problem, Thompson et al. (2013) propose a method to provide feedback between sup-
pression expenditures and the general management budget at the national forest level
within the USDA Forest Service. They suggest establishing an insurance pool to finance
wildfire-suppression expenditures. Each national forest would pay a premium into the
pool based on past suppression expenditures, the level of fire risk, and management
to reduce that risk via preventative measures such as fuel treatment and beneficial use
of wildfire. Premiums would be adjusted regularly to reflect changes in risk status and
suppression cost containment. This structure would provide incentives to reduce risk
through fuel treatment, to place firebreaks around high-valued resources, and to use
cost-effective suppression strategies.
Furthermore, resources that are allocated to fuel treatment may not be applied
cost-effectively. Although the Healthy Forests Restoration Act directs that priority be
given to fuel treatment in and around the WUI (HFRA 2003, Sec. 103), that may not be
the strategy that most effectively protects private property values. Ager, Vaillant, and
Finney (2010) used repeated simulation of fire on a landscape in northeast Oregon to
generate burn probability profiles under two fuel-treatment strategies: one in which
areas with the greatest fuel accumulations were prioritized for treatment and one
in which areas in and near the WUI were prioritized for treatment. Their results sug-
gest that a strategy of treating relatively remote areas with the greatest fuel accumula-
tions could substantially reduce fire risk in the WUI. That is because fires that ignite
in remote places will spread more slowly and thus be less likely to reach the WUI (pro-
viding self-protection) than when treatment is concentrated in the WUI (providing
self-insurance).
The least costly method of fuel treatment, prescribed burning, is likely to be underuti-
lized due to concerns about liability. Prescribed burning reduces the risk associated with
wildfire by reducing fuel loads on the landscape, but it can escape prescription to wreak
havoc on nearby property. For example, the 2000 Cerro Grande fire was a prescribed fire
that escaped and burned 18,000 hectares and destroyed 235 homes in Los Alamos, New
Mexico.
Yoder et al. (2003) and Yoder (2004) developed an analytical model of prescribed
burning and ran simulations to explore the interaction of three related decisions under
different forms of liability rules. Landowners who benefit from burning choose when to
use prescribed burning and they choose the level of precaution to take against its escape.
Adjacent homeowners choose the level of self-insurance undertaken (fuel treatment on
their own property and creation of defensible space around their homes). Yoder’s model
indicates that there is a trade-off between risk from escaped prescribed fire while treat-
ing fuels and risk from wildfire if fuels are not treated. To the extent that liability rules
apply only to damage from escaped prescribed fire, prescribed burning for fuel treat-
ment purposes is underutilized. Strict liability (where the burner is liable for damage
regardless of precautionary measures taken) is likely to result in too little use of pre-
scribed burning for any purpose. However, most states have some form of negligence
rule that frees the burner from liability unless it can be shown that he or she was neg-
ligent in some way. Negligence rules may be designed to encourage the efficient level
of precaution on the part of the burner, but the decision about when and how much to
burn will still be distorted by the failure to impose liability for the wildfire hazard aris-
ing from the presence of untreated fuels. The homeowners will also make suboptimal
Fire 291
fire-risk-management decisions. To the extent that the burner is liable for damage from
escaped prescribed fire, homeowners will undertake too little self-insurance.
Although the federal government is not directly liable under state law for damage
from prescribed fire, it may be sued under the Federal Tort Claims Act if it or its employ-
ees can be shown to be negligent according to the laws of the place where the action
occurred (U.S.C. Title 28, 1346(b)). Hence, the incentives facing public fire managers
will also likely lead to overly cautious use of prescribed fire. There is a movement to bal-
ance liability for prescribed fire with liability for wildfire spreading from national forests
that have an abnormal accumulation of forest fire fuels. See, for example, the proposed
Enhanced Safety from Wildfire Act introduced in 2003 (USGPO 2003). These efforts, if
successful, could tilt the scale back toward fuel treatment.
2.3 Decision Support Tools

It may be possible to mitigate the effect of risk attitudes by developing risk-based deci-
sion support tools—at least for the agencies responsible for fire management on pub-
lic land. Because these tools process complex information systematically, they have the
potential to both inform risky decisions when they are being made and to provide a
vehicle for explaining these decisions after the fact. They also provide means for explor-
ing outcomes from policy alternatives via simulations. In this section, two approaches
to decision support tools are described: risk assessment and optimization under risk.
There are significant research efforts currently underway to develop tools under each
approach.
Risk assessments integrate information about risk (the likelihood of an event occur-
ring and the outcome if it does) in order to inform decision making. Thompson and
Calkin (2011) provide an excellent summary of current research and the challenges in
applying it to forest wildfire. They note that there are several sources of uncertainty that
must be accounted for in risk assessment. First, the location and timing of fire events
and the weather in which it occurs are inherently unpredictable. We do know some-
thing about probability distributions from historical frequencies, but we can predict
only patterns of events and not any particular event. There is also uncertainty due to
large gaps in our knowledge about how to model fire behavior, the response of fire to
various treatments and actions that might be taken, the effect of fire on resources, and
the relative values that people place on those resources. This uncertainty can be reduced
via research in the appropriate disciplinary fields. The latter, valuation, falls in the realm
of economics. Many, if not most, of the values that may be affected by fire are not easily
monetized because they are not conducive to market exchange (e.g., endangered species
habitat, visual amenities, air and water quality, carbon sequestration, cultural heritage,
and more). Nonmarket valuation techniques, including stated preference methods such
as contingent valuation and choice modeling, are developed and applied by environ-
mental economists in an array of resource contexts. Because fire-risk assessment may
pose some special challenges, this is a knowledge gap that is unlikely to be closed any
time soon. Venn and Calkin (2011) acknowledge these challenges and propose an eco-
nomics research agenda to address them.
Optimization takes risk assessment one step further by attempting to identify actions
and strategies that optimally achieve management objectives. All of the modeling chal-
lenges that must be addressed in fire-risk assessment must also be addressed in opti-
mization. The main challenge in moving beyond simulation-based risk assessment to
optimization is computational.
Optimization has a long history in forestry (Montgomery and Adams 1995). The
stand-level problem of how long to hold a tree before harvesting was first posed over
150 years ago by a German forester named Faustmann (1968) and confirmed just
35 years ago by Samuelson (1976). The problem is easy to solve numerically if the
timber-stand volume function is known. Economists have since extended the basic
analytical model to include the flow of ecosystem services dependent on attributes of
standing timber (Hartman 1976; Strang 1983). The Faustmann model was extended to
include fire by Reed (1984), who demonstrated that the optimal response to fire risk is to
harvest timber at a younger age.
The Faustmann model applies to a single isolated stand of trees. However, foresters
don’t manage stands; they manage forests composed of many stands of different ages.
Regulations on forest practices impose an array of spatial constraints, such as maximum
clear-cut size and limits on activities on adjacent stands because many of the benefits
from ecosystem services depend on the spatial pattern of vegetation, such as wildlife
habitat contiguity and connectivity.
When a problem is spatial, it is combinatorial and generally involves integer deci-
sion variables. For some problems, there may be no exact solution algorithm; for others,
the size of the problem quickly outgrows available computational resources (Bettinger,
Sessions, and Boston 2009) because the decision space grows exponentially with the
number of management units. Because of this so-called “curse of dimensionality,”
approximate methods such as genetic algorithms, simulated annealing, and Tabu search
(Reeves 1993) are used in forestry where they have found some degree of acceptance
(for example, Yoshimoto, Brodie, and Sessions 1994; Lichtenstein and Montgomery
2003; Nalle et al. 2004; and Hummel and Calkin 2005).
Adding a large-scale stochastic disturbance, such as fire, to the mix complicates the
problem enormously. Fire is spatial and temporal because its spread across the land-
scape depends on the spatial configuration of fuels, wildlife and other forest values at
risk from fire depend on the spatial configuration of vegetation, and both vegetation
and fuels evolve over time depending on the spatial configuration and timing of man-
agement activities. The fire problem is stochastic and dynamic because the optimality
of decisions made now depends on future fire events that cannot be predicted with any
certainty and for which possible outcomes are wildly diverse. Decisions in future peri-
ods will surely depend on the occurrence of fire events in the interim. Adaptation to new
information as it arrives must be accounted for in the current decision.
The problem can be formulated as a stochastic dynamic programming prob-
lem in which, in each period, the fire manager chooses a spatial vector of timber and
Fire 293
fuel management activities and a suppression response to fire, when it occurs, to

maximize the expected net present value of actions in the current period, assuming
value-maximizing decisions in all future periods once the outcomes from stochastic fire
events in the interim are known.
Most previous attempts to optimize fuel management on the forested landscape are
static. That is, they model the potential of the landscape to burn, given a post-treatment
configuration of fuel on the landscape. For example, Finney (2007) developed an algo-
rithm for identifying the pattern of fuel treatments that maximized the minimum time
it takes for a fire to spread across a particular landscape under typical high fire-severity
weather conditions (e.g.; wind direction and speed, relative humidity, temperature).
In Wei, Rideout, and Kirsch (2008) and Jones et al. (2010), the placement of fuel treat-
ments is optimized to minimize the expected value loss from the next fire. The burn
probability profile for the untreated landscape was estimated using repeated simulations
with random ignition points, assuming typical severe fire weather. Wei, Rideout, and
Kirsch (2008) simplified the computational aspect of the problem by decomposing the
burn probabilities for each unit into ignition and spread probabilities and assuming a
structure for the effect of fuel treatment on spread probability. They solved the prob-
lem using integer programming. Jones et al. tried to make the problem intertemporal
by using simulated annealing to schedule timber and fuel management over a 5-period
time horizon, assuming fire doesn’t actually occur. They used repeated fire simulations
to evaluate the effect of fuel treatments on the burn probabilities and, hence, expected
loss to the next fire. Vegetation is updated in each period using a forest vegetation simu-
lation model (Crookston and Dixon 2005) but it is not burned.
Konoshima et al. (2008, 2010) is the first attempt that this author is aware of to formu-
late and solve the fuel treatment and timber management problem simultaneously as a
stochastic dynamic programming problem. They solved the problem exactly for a very
stylized landscape of 7 management units over 2 time periods with 4 possible manage-
ment activities. They simulated the spread of fire from 7 possible fire-ignition locations
and 2 possible weather conditions, using fire-spread equations (Rothermel 1972) to esti-
mate transition probabilities for each possible configuration of post-treatment vegeta-
tion and fuels. They solved the problem using complete enumeration for each possible
initial configuration of vegetation and fuels on the landscape.
There are at least two problems with these analyses. First, in order for any bioeco-
nomic model of forest and fire management to provide useful guidance for resource
managers, it must capture important elements of the decision process itself and, at the
same time, it must represent the resource and management actions with sufficient real-
ism so that solutions can be translated into feasible actions on the ground. Fire ecolo-
gists and forest planners tend to favor simulation approaches that are relatively rich in
details of the landscape, fire behavior, and vegetation development. They simplify the
decision process, however, by ignoring its dynamic aspect. Economists favor analytical
approaches that represent the important elements of the decision process but, in order
to find solutions, the biological setting and processes are so stylized as to be practically
irrelevant.
The second problem is that once the spatial aspect of a resource-management prob-
lem is accounted for, it is very difficult to draw any general inference from the results.
Detailed models yield results that are specific to a particular landscape and cannot be
applied elsewhere, but even the more stylized models of economists can be hard to
interpret. In fact, general tendencies in the Konoshima results were identified by visual
examination of the solutions—of which there were many.
3. Fire Economics Research in the Future
Because fire is just now attracting the attention of economists, economic analysis of fire
is relatively rare in both the economics literature and the fire literature. However, the
consequences of ignoring fire or of prolonging the status quo of aggressive fire suppres-
sion and banishment of anthropogenic fire could be costly. Economists can make an
important contribution to the design of efficient and effective fire policy for the future.
In this concluding section, I suggest several paths that could advance existing research
in ways that may be especially relevant for pressing policy concerns. I close the section
by speculating about the role of economics in the broader context of the complex eco-
systems of which fire is an important component.
Existing studies of fire-risk management in the WUI take the current configuration
of land ownership and residential development as given. They focus on how households
choose to invest in risk-averting activities to protect private property values. The inter-
action between private landowners and public-land managers should continue to be
explored using game theory in order to learn more about the potential to reduce the
negative effects of spatial externalities via insurance, liability rules, cost-sharing, and
coordinating groups such as homeowners’ associations and cooperatives. However, the
larger question of whether there is too much residential development in fire-prone for-
ests has yet to be addressed. Does the fact that the federal government bears the cost
of forest-fire suppression, and some of the liability for the damage that wildfire causes,
create moral hazard? This question might be addressed by modeling land use change
and amenity migration. And, if the answer is yes, those models might be used to explore
means of internalizing wildfire risk in the decision to locate in the WUI.
The existing literature identifies several cases in which incentives that public-agency
fire managers face do not lead to socially optimal choices. The most important sources of
inefficiency appear to be (1) the lack of feedback between fire hazard reduction (e.g., fuel
treatment) and fire suppression decisions, and (2) the absence of any effective budget
limit for suppression. The incentives for individual managers within the agencies that
are responsible for fire suppression, and also for managing large expanses of public land
where fuel accumulation is most severe, are seriously misaligned with overall agency
objectives. This is an opportunity for the economics of public choice. There is much to be
gained by exploring what motivates the interactions between three sets of actors: indi-
viduals within the agencies, elected officials who determine policy and allocate funding,
Fire 295
and the voting individuals who elect them. Analysis could focus on realigning budgets
so that suppression and prevention are endogenous, developing performance measures
for fire managers that truly reflect desired outcomes, linking budgets to outcomes, and
lessening the effect of risk aversion on suppression decisions. Thompson et al.’s (2013)
proposal to establish an insurance pool to pay for fire suppression is one example of
what might come out of this line of research.
One limiting factor in extending risk assessment to the optimization of fire and
fuel management is the computational difficulty of the problem. The challenge is two-
fold: first, to solve the computational problem, and then to interpret the results once it
is solved. The appropriate decision framework for fire is stochastic dynamic program-
ming. Because fire is spatial, however, any specification of the problem that is realis-
tic enough to be informative to policy makers and land managers, and large enough in
scale to capture important fire behavior, will not be amenable to analytical methods and
will be too large to solve using exact methods. The approximate optimization methods
currently used in forestry are useful for large problems that involve spatial interactions,
but they are not dynamic and, hence, not adequate for modeling policy in the context of
large-scale stochastic disturbance, such as fire.
There is hope. New methods in operations research and computer science are being
developed and evaluated. One, in particular, approximate dynamic programming
(ADP), also known as reinforcement learning (Powell 2009), may prove useful for the
fire problem. The policy iteration version of ADP is intriguing because, instead of pro-
ducing a specific plan for a landscape, it produces a “rule” that recommends an action
based on attributes of the state (vegetation and fuels) and attributes of the stochas-
tic event (ignition and weather) that are known at the time of the decision. The basic
idea of ADP with policy iteration is to “learn” an optimal policy by iteratively solving
a deterministic problem for each of a large number of individual Monte Carlo simula-
tions of future time paths of random events (e.g., ignitions and weather). The solutions
provide data for estimating a policy rule using regression. The process is repeated until
the policy rule stabilizes. The policy rule, thus derived, would be applicable only to the
landscape on which it was developed. However, it may be possible to apply techniques
of machine learning to derive more general results by systematically “tweaking” the
attributes of the landscape in order “learn” a more general model for fire management.
The coefficients of the resulting policy rule are simple to interpret because they reveal
how optimal choices adjust to changing external conditions. In ongoing research, we
are attempting to apply ADP to the problem of when it is optimal to allow a wildfire
to burn, and we hope to extend it to the optimal placement of fuel treatments on a
landscape.
Finally (and in a necessarily speculative vein), the relationship between fire, people,
and land use is just one small aspect of the larger question of how we live on our planet
in a sustainable manner. Fire is an ecological process that plays an important role in
the functioning of complex ecosystems. Its exclusion has ramifications well beyond the
immediate impacts to people of increasing suppression costs and loss of property and
lives. Complex ecosystems often exhibit cycles of conservation and renewal that are
triggered by disturbances such as fires and hurricanes (Holling 1995). These cycles and
interactions occur at many scales in time and space.
In the Oregon Coast Range (for example), the natural disturbance regime appears to
be infrequent large-scale catastrophic fire events every 100–400 years. These fires play a
critical role in the renewal of the aquatic ecosystems that support salmon populations.
Without fire to kill large trees and trigger landslides, the stream systems become starved
of large woody debris and sediment. Over time, the complexity and quality of freshwater
salmon habitat declines and populations collapse (Reeves et al. 1995). Currently, sev-
eral evolutionarily significant units of Pacific salmon and distinct populations segments
of steelhead in the Pacific Northwest are listed as endangered under the Endangered
Species Act (NOAA 2011). Degradation of freshwater habitat that results from exclu-
sion of catastrophic disturbance from watersheds in coastal forests is one contributing
factor. Reeves and Duncan (2009) argue that disturbance is crucial to the maintenance
of salmon habitat and, since reintroduction of catastrophic fire is unlikely to be socially
acceptable, forested watersheds should be managed to mimic its effect. Instead, how-
ever, these watersheds are managed to maintain steady conditions over time and across
space so that habitat is moderately degraded everywhere and high-quality habitat exists
nowhere (Reeves, Burnett, and Gregory 2002).
C. S. Holling, one of the early contributors to the field of ecological economics,
describes case after case in which people who are uncomfortable with uncertainty
attempt to bring order to chaos, protect resources, and ensure a predictable supply of
ecosystem services by regulating ecosystems to uniform standards in order to dampen
disturbance cycles. There is no place for wildfire in these managed ecosystems. Natural
cycles of conservation, disruption, and renewal are interrupted so that, over time, eco-
systems become “brittle,” that is, less resilient and more vulnerable to collapse when dis-
turbance does occur (Holling 1995).
People are comforted by stability and it can be argued that one legitimate role of gov-
ernment is to reduce uncertainty, limit fluctuations, and maintain a stable economy.
However, as Holling notes, in ecological systems there is a trade-off between local sta-
bility and global stability. The challenge for economists is to inform the design of land
use policy so that it finds a balance between social acceptability and ecological resil-
ience, allowing for disturbance at local temporal and spatial scales in the interest of
sustaining stability at larger and longer scales. In the traditional regulatory approach to
resource policy, the standards that are imposed must be enforceable. For regulations to
be enforceable, they must be applied uniformly to outcomes that can be observed and
measured. Regulations that are too complicated and allow variability may be perceived
as arbitrary and unfair.
What can economics bring to this dilemma? In recent decades, there has been a grow-
ing interest in the application of the science of complexity to economic systems (Rosser
1999). One working definition of complexity from Durlauf (1998) states, “[A]‌system
is said to be complex when it exhibits some type of order as a result of the interactions
of many heterogeneous objects.” It has long been recognized that economic systems are
Fire 297
complex by that definition. Because the aggregate economy is the result of many agents,
all with different histories, endowments and objectives, interacting with one another in
markets, it is fundamentally complex (Colander 2009).
The emerging discipline of complexity economics is developing in many directions.
However, there do appear to be some common themes that are relevant for resource pol-
icy for complex ecosystems. One is that, when systems are complex, we can never know
enough about each of its elements to allow us to predict specific outcomes. We can only
predict general patterns that may occur (Hayek 1999). The idea that there exists a steady
state for the economy is replaced with the notion that economic systems evolve over
time and the path of that evolution cannot be predicted because it depends on a legacy
of past events (path dependency) and also on chance (Colander 2009). There appears to
be a sense that the most robust systems are those that are self-regulating. In other words,
if regulations that must be strictly defined and enforced can be replaced with an institu-
tional environment in which incentives are designed to lead individual agents to make
choices that are consistent with the overall objective, the outcome may be an environ-
ment more tolerant of local variation and, hopefully, more stable in the long run.
As promised, this section is mostly speculative. The discipline of complexity econom-
ics is in its infancy and it is not clear what it will yield. My thinking about its potential
application to land use and resource policy when disturbance is important for ecosys-
tem health is, likewise, in its infancy, and it is not clear what that will yield. However,
my hope is that this avenue of research will provide new and useful insights for future
resource management and policy—particularly with respect to fire.
References
Abt, K. L., J. P. Prestemon, and K. M. Gebert. 2009. Wildfire suppression cost forecasts for the
U.S. Forest Service. Journal of Forestry 107(4): 173–178.
Ager, A. A., N. M. Vaillant, and M. A. Finney. 2010. A comparison of landscape fuel treat-
ment strategies to mitigate wildland fire risk in the urban interface and preserve old forest
Structure. Forest Ecology and Management 259(8):1556–1570.
Althaus, I. A., and T. J. Mills. 1982. Resource values in analyzing fire management pro-
grams for economic efficiency. General Technical Report PSW-57. Berkeley, CA: U.S.
Department of Agriculture Forest Service, Pacific Southwest Forest and Range
Experiment Station.
Berry, A. 2007. Forest policy up in smoke: Fire suppression in the United States. Bozeman,
MT: Property and Environment Research Center. http://www.law1.northwestern.edu/sear-
lecenter/papers/Berry_forest_policy.pdf.
Bettinger, P., J. Sessions, and K. Boston. 2009. A review of the status and use of validation pro-
cedures for heuristics used in forest planning. International Journal of Mathematical and
Computational Forestry & Natural Resource Sciences 1(1): 26–37.
Biswell, H. 1980. Fire ecology: Past, present, and future. Keynote speech presented at annual
meeting of the American Association for the Advancement of Science, Davis, CA, June
22–27.
Busby, G. M., H. J. Albers, and C. A. Montgomery. 2012. Wildfire risk management in

a landscape with fragmented ownership and spatial interactions. Land Economics
88: 496–517.
Calkin, D. E., K. M. Gebert, J. G. Jones, and R. P. Neilson. 2005. Forest service large fire
area burned and suppression expenditure trends: 1970–2002. Journal of Forestry
103: 179–183.
Calkin, D. E., T. J. Venn, M. J. Wibbenmeyer, and M. P. Thompson. 2012. Estimating U.S. fed-
eral wildland fire managers’ preferences toward competing strategic suppression objectives.
International Journal of Wildfire Research 22: 212–222.
Carle, D. 2002. Burning questions: America’s fight with nature’s fire. Westport, CT: Praeger.
Colander, D. 2009. Complexity and the history of economic thought. In Handbook of research
on complexity, eds. J. Barkley Rosser Jr. and Kirby L. Cramer, 409–426. Northampton,
MA: Edward Elgar.
Crookston, N. L., and G. E. Dixon. 2005. The forest vegetation simulator: A review of
its structure, content, and applications. Computers and Electronics in Agriculture
49(1): 60–80.
Crowley, C. S. L., A. S. Malik, G. S. Amacher, and R. G. Haight. 2009. Adjacency externalities
and forest fire prevention. Land Economics 85(1): 162–185.
Donovan, G. H., and T. C. Brown. 2005. An alternative incentive structure for wildfire manage-
ment on national forest land. Forest Science 51(5): 387–395.
Donovan, G., J. P. Prestemon, and K. M. Gebert. 2011. The effect of newspaper cover-
age and political pressure on wildfire suppression costs. Society & Natural Resources
24(8): 785–798.
Durlauf, S. N. 1998. What should policymakers know about economic complexity? The
Washington Quarterly 21(1): 155–165.
Ehrlich, I., and G. S. Becker. 1972. Market insurance, self-insurance, and self-protection. The
Journal of Political Economy 80(4): 623–648.
Faustmann, M. 1968. Calculation of the value which forest land and immature stands possess
for forestry. In Martin Faustmann and the evolution of discounted cash flow, ed. M. Gane,
27–55. Oxford: University of Oxford Commonwealth Forestry Institute. (Originally pub-
lished in 1849.)
Finney, M. A. 1998. FARSITE: Fire area simulator—Model development and evaluation.
Research Paper RMRS-4. Missoula, MT: U. S. Department of Agriculture Forest Service,
Rocky Mountain Research Station.
Finney, M. A. 2007. A computational method for optimizing fuel treatment locations.
International Journal of Wildland Fire 16: 702–711.
FAO. 2011. State of the world’s forests, 2011. Rome: Food and Agriculture Organization.
Hartman, R. 1976. The harvesting decision when a standing forest has value. Economic Inquiry
14: 52–68.
Hayek, F. A. 1999. The theory of complex phenomena. In Critical approaches to science and
philosophy, ed. M. Bunge, 332–349. New Brunswick, NJ: Transaction Publishers. (Originally
published in 1964 by The Free Press.)
HFRA. 2003. Healthy Forests Restoration Act of 2003. http://ag.senate.gov/Legislation/
Compilations/Forests/healthy.pdf
Holling, C. S. 1995. What barriers? What bridges? In Barriers and bridges to the renewal of
ecosystems and institutions, eds. L. H. Gunderson, C. S. Holling, and S. S. Light, 3–37.
New York: Columbia University Press.
Fire 299
Hummel, S. S., and D. E. Calkin. 2005. Costs of landscape silviculture for fire and habitat man-
agement. Forest Ecology and Management 207(3): 385–404.
Jones, G. J., W. Chung, C. Seielstad, J. Sullivan, and K. Krueger. 2010. Optimizing spatial and
temporal treatments to maintain effective fire and non-fire fuels treatments at landscape scales.
Final Report FJSP Project 06-3-3-14. U.S. Department of Agriculture/U.S. Department of
the Interior, Joint Fire Science Program. http://www.fs.fed.us/rm/pubs_other/rmrs_2010_
jones_g002.pdf.
Kleinman, P. J. A., D. Pimentel, and R. B. Bryant. 1995. The ecological sustainability of
slash-and-burn agriculture. Agriculture Ecosystems & Environment 52: 235–249.
Konoshima, M., C. A. Montgomery, H. J. Albers, and J. L. Arthur. 2008. Spatial endog-
enous fire risk and efficient fuel management and timber harvest. Land Economics
84(3): 449–468.
Konoshima, M., H. J. Albers, C. A. Montgomery, and J. L. Arthur. 2010. Optimal spatial patterns
of fuel management and timber harvest with fire risk. Canadian Journal of Forest Research
40(1): 95–108.
Lasko, R. 2010. Implementing federal wildland fire policy—Responding to change. Fire
Management Today 70(1): 5–7.
Lichtenstein, M. E., and C. A. Montgomery. 2003. Biodiversity and timber in the coast range of
Oregon: Inside the production possibility frontier. Land Economics 79(1): 56–73.
Martell, D. 2011. The development and implementation of forest fire management decision
support systems in Ontario, Canada: Personal reflections on past practices and emerging
challenges. International Journal of Mathematical and Computational Forestry & Natural
Resource Sciences 3(1): 18–26.
McCarty, J. L., S. Korontzi, C. O. Justice, and T. Loboda. 2009. The spatial and temporal distribu-
tion of crop residue burning in the contiguous United States. Science of the Total Environment
407: 5701–5712.
Montgomery, C. A., and D. M. Adams. 1995. Optimal timber management. In Handbook of
environmental economics, ed. D. W. Bromley, 379–404. Oxford: Basil Blackwell.
Nalle, D. J., C. A. Montgomery, J. L. Arthur, N. H. Schumaker, and S. Polasky. 2004. Modeling
joint production of wildlife and timber in forests. Journal of Environmental Economics and
Management 48(3): 997–1017.
NIFC (National Interagency Fire Center). 2011. Wildland fire statistics. http://www.nifc.gov
NOAA (National Oceanic and Atmospheric Administration). 2011. ESA salmon listings.
http://www.nwr.noaa.gov/ESA-Salmon-Listings
Omi, P. N. 2005. Forest fires: A reference handbook. Santa Barbara, CA: ABC-CLIO.
Powell, W. B. 2009. What you should know about approximate dynamic programming. Naval
Research Logistics 56: 239–249.
Pyne, S. J. 1995. World fire: The culture of fire on earth. New York: Holt.
Pyne, S. J. 2001. Fire: A brief history. Seattle, WA: University of Washington Press.
Reed, W. J. 1984. The effects of the risk on the optimal rotation of a forest. Journal of Environmental
Reeves, C. R. 1993. Modern heuristic techniques for combinatorial problems. New York: John
Wiley & Sons.
Reeves, G. H., L. Benda, K. M. Burnett, P. A. Bisson, and J. R. Sedell. 1995. A disturbance-based
ecosystem approach to maintaining and restoring freshwater habitats of evolutionarily sig-
nificant units of anadromous salmonids in the Pacific Northwest. American Fisheries Society
Symposium 17: 334–349.
Reeves, G. H., K. M. Burnett, and S. V. Gregory. 2002. Fish and aquatic ecosystems of the Oregon
coast range. In Forest and stream management in the Oregon coast range, ed. S. D. Hobbs, J.
P. Hayes, R. L. Johnson, G. H. Reeves, T. A. Spies, J. C. Tappeiner II, and G. E. Wells, 68–98.
Corvallis: Oregon State University Press.
Reeves, G. H., and S. L. Duncan. 2009. Ecological history vs. social expectations: Managing
aquatic ecosystems. Ecology and Society 14(2): 8. www.ecologyandsociety.org/vol14/iss2/
art8/
Rosser J. B., Jr. 1999. On the complexities of complex economic dynamics. Journal of Economic
Perspectives 13(4): 169–192.
Rothermel, R. C. 1972. A mathematical model for predicting fire spread in wildland fuels. Research
Paper INT-115. Ogden, UT: U.S. Department of Agriculture Forest Service, Intermountain
Forest and Range Experiment Station.
Samuelson, P. A. 1976. Economics of forestry in an evolving society. Economic Inquiry
14: 466–492.
Shafran, A. P. 2008. Risk externalities and the problem of wildfire risk. Journal of Urban
Sparhawk, W. N. 1925. The use of liability ratings in planning forest fire protection. Journal of
Agricultural Research 30: 693–761.
Strang, W. J. 1983. On the optimal forest harvesting decision. Economic Inquiry 21: 576–583.
Talberth, J., R. P. Berrens, M. McKee, and M. Jones. 2006. Averting and insurance decisions In
the wildland-urban interface: Implications of survey and experimental data for wildfire risk
reduction policy. Contemporary Economic Policy 24(2): 203–223.
Theobald, D. M., and W. H. Romme. 2007. Expansion of the U.S. wildland–urban interface.
Landscape and Urban Planning 83: 340–354.
Thompson, M. P., and D. E. Calkin. 2011. Uncertainty and risk in wildland fire manage-
ment: A review. Journal of Environmental Management 92: 1895–1909.
Thompson, M. P., D. E. Calkin, M. A. Finney, and K. G. Gebert. 2013. A risk-based premium
approach to wildland fire finance and planning. Forest Science 59: 63–77.
Troy, A. 2007. A tale of two policies: California programs that unintentionally promote develop-
ment in wildland fire hazard zones. In Advances in the economics of environmental resources,
Vol. 6, eds. A. Troy and R. G. Kennedy, 127–140. Oxford: Elsevier JAI Press.
Troy, A. and J. Romm. 2007. The effects of wildfire disclosure and occurrence on property mar-
kets in California. In Advances in the economics of environmental resources, Vol. 6, eds. A.
Troy and R. G. Kennedy, 101–119. Oxford: Elsevier JAI Press.
USGPO (United States Government Printing Office). 2003. H.R. 2551 Summary. http://www.
gpo.gov/fdsys/pkg/BILLS-108hr2551ih/pdf/BILLS-108hr2551ih.pdf.
Van Lear, D. H., and T. A. Waldrop. 1991. Prescribed burning in regeneration. In Forest
regeneration manual, eds. M. L. Duryea and P. M. Dougherty, 235–250. Dordrecht,
Netherlands: Kluwer Academic.
Venn, T. J., and D. E. Calkin. 2011. Accommodating non-market values in evaluation of wildfire
management in the United States: Challenges and opportunities. International Journal of
Wildland Fire 20: 327–339.
Wei, Y., D. Rideout, and A. Kirsch. 2008. An optimization model for locating fuel treatments
across a landscape to reduce expected fire losses. Canadian Journal of Forest Research
38: 868–877.
Williams, G. W. 2000. Introduction to aboriginal fire use in North America. Fire Management
Today 60(3): 8–12.
Fire 301
Wilson, R. S., P. L. Winter, L. A. Maguire, and T. Ascher. 2011. Managing wildfire
events: Risk-based decision making among a group of federal fire managers. Risk Analysis
31(5): 805–818.
Yoder, J., M. Tilley, D. Engle, and S. Fuhlendorf. 2003. Economics and prescribed fire law in the
United States. Review of Agricultural Economics 25(1): 218–233.
Yoder, J. 2004. Playing with fire: Endogenous risk in resource management. American Journal
Yoshimoto, A., J. D. Brodie, and J. Sessions. 1994. A new heuristic to solve spatially constrained
long-term harvest scheduling problems. Forest Science 40(3): 365–396.
C HA P T E R 1 2
L A N D U S E A N D M U N I C I PA L
PROFILES
E DWA R D STON E A N D J U N J I E W U
Land is a fundamental resource, and the character of the landscape influences quality
of life in significant ways. “Land use” refers to more than simply the pattern of differ-
ent land covers (e.g., cropland, forests, urban) in space. Rather, land use is “the total of
arrangements, activities, and inputs that people undertake in a certain land cover type
to produce, change, or maintain it” (FAO/UNEP 1999). In other words, land use for a
particular parcel encompasses land cover, as well as management intensity and prac-
tices. Changing land use could mean changing land cover, or it could mean maintaining
the same land cover while altering management. Land use determines the availability
of primary inputs including food, fiber, building materials, and even developable land.
If the consequences of individual land use decisions were to fall entirely on individual
landowners, markets for these inputs should result in an efficient allocation of land uses.
However, individual land use decisions—and the resulting land use patterns—may gen-
erate environmental and social externalities.
First, the decision to extract ecosystem goods and services from the landscape may
generate an environmental externality. Removing ecosystem goods or managing land to
enhance the production of these goods may have indirect consequences on flows of sup-
porting ecosystem services, such as freshwater storage and release, soil formation and
fertility, biodiversity, and climate regulation (DeFries et al. 2004). Typically, land use
activities involve making natural resources available for human consumption at some
cost to environmental quality. This is the case for farming or forestry, as well as for urban
development (Foley et al. 2005).
Second, land use decisions and the resulting patterns may give rise to social externali-
ties, including the impacts on local public finance and school quality. Collective land
use decisions shape community well-being in ways beyond market and environmental
impacts. The link between land use and various indicators of environmental quality has
been the subject of much research. This chapter explores the social implications of land
use, which have been less well-studied. Collectively, the socially relevant features of a
Land Use and Municipal Profiles 303
city or a neighborhood may be termed its “municipal profile.” Relevant features include
local taxes and public services, public safety and health, open-space provision and natu-
ral amenities, income distribution, housing prices, development densities, demographic
composition and distribution, and transit and congestion.
Land use patterns affect municipal profiles. For example, urban sprawl has been
linked to obesity, congestion, and open-space loss (see, e.g., Nechyba and Walsh 2004;
Plantinga and Bernell 2007). Suburbanization is often associated with income stratifica-
tion and concentrated poverty, with their inherent fiscal and social implications (e.g.,
Mieszkowski and Mills 1993). Municipal profiles in turn affect land use patterns. The
expression “flight from blight” refers to falling incomes and deteriorating public safety
and services causing high-income households to relocate from city centers to suburbs
and thus contributing to suburbanization and sprawl. Conversely, high-income com-
munities may enact zoning and tax regimes that affect land use patterns by attracting
new residents and/or restricting the pattern of development.
In many cases—and in many economic models (e.g., Wu 2007)—the interaction
between land use and municipal profiles is self-reinforcing. “Flight from blight” further
diminishes central city incomes and tax revenues, leading to deteriorating public ser-
vices and safety and thus more flight. High-income suburbs with high tax revenue and
high levels of services attract more high-income households. Other urban development
phenomena are also self-reinforcing, including gentrification and urban revitalization.
Historically, urban expansion has been accompanied by the rise of the automobile
and patterns of suburbanization. This is markedly true in the United States but holds
internationally as well (Mieszkowski and Mills 1993). This much-lamented tendency
has been associated with social and environmental costs. Many households continue to
locate in the urban fringe. Despite social and environmental externalities, households
enjoy substantial private benefits from consuming more land and housing. Sprawl and
other “undesirable” development patterns leave many households better off. That is not
to say that such land use patterns are efficient. Indeed, they have given rise to regula-
tions and incentive-based policies aimed at correcting perceived inefficiencies associ-
ated with “excessive sprawl,” including urban growth boundaries, zoning protections for
open space, and impact fees for new development.
The aggregate location decisions of firms and households are the drivers behind both
land use and municipal profile change. Two primary bodies of economic literature
attempt to explain historical development patterns through the lens of household loca-
tional choice. The contemporary urban economics literature dates back to the mono-
centric city model, with early incarnations by Alonso (1964), Mills (1967), and Muth
(1969). This approach explains historical patterns of suburbanization in terms of rising
incomes, falling commuting costs, and newer housing on the periphery. In contrast, the
local public finance approach explains development patterns in terms of preferences for
alternative bundles of local taxes and public goods and services. This body of literature
expands on Tiebout’s (1956) household sorting model.
Although urban economics models capture the primary drivers of urban expan-
sion, they do not account for other factors that influence household locational choice
304 Edward Stone and JunJie Wu
within a metropolitan area, including amenities and public finances (Nechyba and
Walsh 2004). Local public finance models include these factors and better explain why
many households moving to the suburbs prefer to form homogeneous groups, but these
models are typically aspatial. Due to data limitations, the bulk of empirical research
in this area has focused on changes at the county or city level, although many relevant
decisions are made at a smaller scale. Recent developments in computing, especially
Geographic Information System (GIS) software, facilitate observation and analysis
within metropolitan areas.
The purpose of this chapter is twofold: first, to survey the most significant develop-
ments in theory and analyses that explore the interactions among household location
decisions, land use patterns, and municipal profiles; and, second, based on the survey, to
explore strategies to model these interactions using a case study from Portland, Oregon.
The remainder of this chapter is organized as follows. Section 1 reviews the literature
in urban economics and local public finance that focuses on various links between land
use and municipal profiles. Section 2 illustrates the role of emerging data and informa-
tion technologies in modeling household location choice within a metropolitan area,
as opposed to at the county or city level. Section 2 also discusses appropriate estimation
strategies. Section 3 provides a conclusion.
Household preferences and collective location decisions determine land use patterns
and neighborhood characteristics.1 This section first reviews the literature on household
location decisions and then focuses on the interactions between location decisions and
municipal profiles.
1.1 Household Location Decisions

Suburbanization has been a dominant trend in aggregate household location decisions
and urban spatial development in the modern era, particularly in the United States. As
previously mentioned, economists offer two main classes of theory explaining house-
hold residential location choices: urban economics and local public finance. The mono-
centric city model is a basic formalized model of the urban economics approach. In
1
Strictly speaking, land use and thus municipal profiles are determined by the collective location
decisions of both households and firms. Although an in-depth investigation of firm location decisions
is beyond the scope of this chapter, there is a rich literature in this area. One fundamental question is
whether jobs follow people or people follow jobs. For example, Muth (1971) finds that jobs primarily
follow people. Carlino and Mills (1987) find that people primarily follow jobs. By focusing on
households, the framework presented in this chapter implicitly assumes that jobs follow people.
early incarnations of the monocentric city model (e.g., Alonso 1964; Mills 1967; Muth
1969), all employment lies within the central business district (CBD), households are
differentiated by income, and the key difference between alternative household loca-
tions is distance to the CBD. Since housing closer to the employment center requires less
commuting, it is more desirable and therefore more expensive. Thus, households face a
tradeoff between commuting time and housing price. Those who choose to live farther
away incur higher commuting costs but face lower housing prices and can thus afford to
consume more housing. The primary driver behind suburbanization and modern urban
spatial expansion has been falling commuting cost due to the proliferation of the auto-
mobile and the development of highway systems. Simple CBD models account for this
driver and correctly predict expanding urban footprints in the face of decreasing trans-
portation costs. However, simple CBD models do not account for a number of other
relevant factors—including alternative transportation modes, locational amenities, and
age of the housing stock—nor do they predict multicentric metropolitan areas and vari-
ous observed historical development patterns. A number of researchers have relaxed
assumptions and generalized CBD models to address these concerns.
LeRoy and Sonstelie (1983) incorporate two alternative transport modes—one fast
and one slow—and demonstrate that when the rich are better able to afford the faster
mode, they will tend to suburbanize more rapidly than others. They argue that this was
the case early on with the automobile. However, as the cost of the faster mode falls (the
vast majority of American households can now afford car-commuting), the rich lose
this comparative advantage for suburbanizing. In fact, since wages—and thus opportu-
nity cost of time—are higher for high earners, LeRoy and Sonstelie predict gentrifica-
tion by the rich as commuting costs fall and the poor suburbanize. According to this
model, when the rich and poor use the same transport mode, the rich will tend to locate
in the city center.
Brueckner, Thisse, and Zenou (1999) add natural and historical amenities to explain
alternative income distributions across different cities. They observe the stark differ-
ence between most American cities, where high-income households tend to live in the
suburbs, and many European cities, where the wealthy occupy the central city.2 Their
model explains these differences in terms of differing levels of natural and historical
amenities across cities. As with classic CBD models, the rich are pulled to the suburbs
by their preference for more housing, which is available more cheaply on the periph-
ery; simultaneously, they are pulled into the center by their high time-cost of commut-
ing. However, this model also allows for heterogeneous levels of natural and historical
amenities between the center and the suburbs. When the central city, such as Paris, has
high levels of amenities, these constitute an additional attraction for the wealthy. On
balance, the time-cost effect and the amenity effect outweigh the housing price effect,
2 The simple CBD model is consistent both with the rich locating in the center (the ratio of
commuting cost per mile to housing consumption increases with income) and with the rich locating in
the suburbs (opposite). However, it seems implausible that the behavior of this ratio across countries
and the wealthy locate in the center. When the central city has low or even negative ame-
nity value, as in Detroit, the housing price effect dominates, and the wealthy locate in
the suburbs. A key assumption of this model is that preferences for amenities rise with
income.
Wu (2006) incorporates amenities in a different fashion. Distinguishing between
exogenous amenities (natural and historical features) and endogenous amenities (e.g.,
local public services), this study incorporates exogenous amenities in a modified CBD
model. Alternative locations within the city differ in terms of the distance to the employ-
ment center and the level of local amenities. In contrast to the Brueckner-Thisse-Zenou
model, spatially heterogeneous amenities in this model attract households to various
suburbs. With this spatial heterogeneity in amenities, households may be willing to pay
more for a nice location than for a short commute; thus, housing prices may not fall uni-
formly as distance from the center increases. At a given distance from the center, higher
income households will choose locations with better amenities. This model is consistent
with noncontiguous development patterns and non–distance-based patterns of income
segregation. Wu (2006) includes a model incorporating endogenous amenities as well,
discussed in Section 1.2, with local public finance models.
Brueckner and Rosenthal (2009) posit that age of the housing stock is an important
determinant of the location of high- and low-income households. The resulting model is
consistent with both suburbanization and gentrification. In addition to short commutes
and low housing prices, high-income households prefer newer housing. Commuting
concerns pull households inward; housing price concerns pull them outward. The
location of new housing determines the direction of the housing age effect. As a city
grows, new housing is always available on the periphery. Some new housing is also avail-
able in the interior—more so during periods of rapid redevelopment. If new housing
is abundant in the interior city, it exerts an additional pull, causing some high-income
households to locate in the center. Holding housing age constant, this model predicts a
negative relationship between income and distance—the rich prefer to live in the center.
This is in contrast with the traditional CBD model, in which suburbanization by the rich
implies a positive relationship between income and distance.
CBD models, including those just discussed, assume monocentricity—that is, all
firms (and thus all employment) locate in the CBD. Whereas household location is
determined endogenously within the model, firm location is exogenously given. Ogawa
and Fujita (1980) and Fujita and Ogawa (1982) relax this assumption and explore the
conditions under which a nonmonocentric city is the equilibrium urban spatial config-
uration. In addition to commuting cost, these models include a transaction cost param-
eter, which measures the benefits of spatial clustering for firms. When transaction costs
are high relative to marginal commuting costs, the incentive for firms to cluster out-
weighs the incentive for households to locate close to work. A monocentric city is the
differs enough to fully explain differences in spatial income distribution. See Brueckner, Thisse, and
Zenou (1999) for a more complete discussion.
equilibrium spatial arrangement. Higher relative marginal commuting costs give rise
to multiple dispersed employment centers because households have increasingly strong
incentives to minimize commuting distance. In the extreme case, in which firms do not
benefit from spatial proximity, the equilibrium spatial arrangement is a fully mixed city
with firms and residences dispersed throughout.
1.2 Interactions Between Residential Location Choices and

Municipal Profiles
Local public finance models offer an alternative lens through which to examine house-
hold locational choice. Even broadened to include amenities, housing age, and transit
considerations, CBD models do not fully capture the role of community characteristics.
Dating back to Tiebout (1956), local public finance models endogenize the provision
of public services. In other words, these models account for the interaction between
household locational choice and the levels of local taxes and public services. Households
choose a location based, in part, on their preferences for various bundles of local taxes
and public services at the community scale. They “vote with their feet.” Simultaneously,
households influence the level of taxes and services in a community through the repre-
sentative process and through peer and local public finance externalities. A brief dis-
cussion of the link between household locational choice and community characteristics
follows.
A household chooses a home based on income/wealth, own preferences, home char-
acteristics, and community characteristics. Based on their finances, families choose a
preferred option from available house-community combinations. The role of commu-
nity characteristics in this process is clear: families like nice, safe neighborhoods and
good school districts. The link between household location decisions and community
characteristics is more involved. Relevant community characteristics include tax rates
and the levels of amenities and public services. Some community characteristics are
exogenous; they are not affected by household location decisions. Consider natural fea-
tures, such as a hill, lake or river, or a well-established man-made attraction, for exam-
ple. These types of sites exist prior to any location decisions and will persist regardless of
those decisions. Other community characteristics are endogenous; they are affected by
household location decisions.
Collective location decisions—and the preferences and characteristics of the resulting
population—affect these endogenous community characteristics in three ways: voting,
local public finance externalities, and peer externalities. Spatial context also matters.
First, community residents vote for their preferred bundle of taxes and services. As the
voter base changes due to household relocations, the results of these votes may change.
Voting determines the local tax rate directly, but relocation decisions can alter the size of
the tax base, thus indirectly affecting local public service provision.
Second, if high-income households move out of central cities and into suburbs in clas-
sic “flight-from-blight,” this results in an erosion of the city tax base and a strengthening
of the suburban tax base, leading to deteriorating public services in the center and
enhanced services in the suburbs. This is the local public finance externality. Collective
location decisions that shift income distributions affect the ability of jurisdictions to
provide services.
Third, peer externalities also affect the level or quality of services independently of
finance. Consider public education, for example. Funding affects school quality, and
wealthier school districts tend to be better-funded—the local public finance external-
ity. Highly involved parents may also affect school quality. So, two comparably funded
districts with different levels of parental engagement might expect different results.
Peer externalities are present when the level of the public services provided depends
on the characteristics of the population being served, as well as on the level of funding.
Interestingly, peer externalities may preclude the possibility of leveling the playing field
by increasing funding to lagging communities. Depending on the scale of spillovers,
economists may alternatively term these effects family or neighborhood externalities.
A desire to take advantage of perceived peer externalities may influence location deci-
sion and has been put forth as an explanation for the formation of homogeneous sub-
urbs (Nechyba and Walsh 2004).
Of course, some community characteristics defy identification as purely endog-
enous or exogenous. The presence of a previously existing park or open space is
exogenous. However, the quality of experience in the park may be endogenous and
subject to change due to voting, local public finance externalities, peer externalities,
and spatial context. The community could vote to cut or boost maintenance fund-
ing. A weakening tax base could force maintenance reductions via local public finance
externalities. Citizen-use levels and participation in volunteer maintenance could
affect quality of experience, which are examples of peer externalities. Finally, spatial
context matters; a well-maintained park in a high-income neighborhood provides
amenities to local residents and increases values of nearby properties, whereas an
under-maintained park that serves as a focal point for criminal behavior is much less
valuable to local residents and could potentially be viewed as a disamenity (Anderson
and West 2006; Troy and Grove 2008). Home buyers value nearby shopping and tran-
sit access but may prefer not to live adjacent to a shopping center or highway. More
space is devoted in later sections to spatial context in the discussion of the home price
hedonics literature.
By incorporating interaction between community characteristics and household
location decisions, local public finance models go beyond their CBD counterparts.
Following Tiebout’s 1956 seminal paper, other researchers expand on Teibout’s general
equilibrium model. Ellickson (1971) derives the single-crossing property, a necessary
condition for equilibrium characterized by income stratification. Epple, Filimon, and
Romer (1984) incorporate housing markets. Epple and Sieg (1999) develop a general
method for estimating equilibrium models of local jurisdictions. Although these papers
generate strong predictions of characteristics of communities in equilibrium—includ-
ing income stratification across communities or, more generally, income stratification
across communities by preference—they ignore location.
Wu (2006) incorporates distance and exogenous amenities in a hybrid of CBD and

local public finance models. The first model from this paper, mentioned earlier, simply
adds exogenous amenities to a CBD model. A second model, however, includes both
exogenously determined amenities and endogenously determined taxes and public
services, not to mention location. This model predicts income stratification by amenity
level for a given distance from the city center.
In addition to urban economics and local public finance, papers from several eco-
nomic subgenres inform this investigation of the link between land use and municipal
profiles. Hedonic home pricing offers insight into the preferences driving household
location choice, which, in turn, drives land use change. Oates (1969) introduces hedonic
modeling to test Tiebout’s hypothesis, and a wide range of empirical studies use hedon-
ics to estimate the value of community characteristics (both positive and negative) as
capitalized in home sale prices (e.g., Bowes and Ihlanfeldt 2001; Irwin 2002; Anderson
and West 2006; Cohen and Coughlin 2008; Troy and Grove 2008).
One clear message emerges from this literature: when estimating the effect of ameni-
ties (or disamenities) on home prices, spatial context matters. For example, Cohen and
Coughlin (2008) find that the effect of proximity to the airport varies with distance. If
you are too close, airport noise drives down home prices; sufficiently far away to miti-
gate noise, proximity to the airport drives up home prices. There are many examples
from the literature on the amenity value of open space. Troy and Grove (2008), men-
tioned earlier, find that parks in high-crime areas may be disamenities. Other studies
have found that the amenity value of open space varies widely with distance from the
city center (Geoghegan et al. 1997), whether the site is permanently designated as open
space (Irwin and Bockstael 2001), type and proximity of open space (e.g., Smith et al
2002; Anderson and West 2006), and income and age structure of the neighborhood
(Anderson and West 2006), to name a few. Investigating how households value partic-
ular community characteristics—and how those values vary depending on context—
enhances understanding of household location decisions.
A number of related papers focus on urban sprawl. The term “sprawl” has negative
connotations and is often cited as an example of a land use pattern with negative social
implications. Nechyba and Walsh (2004) provide a comprehensive review of the litera-
ture on sprawl. They argue that, despite its negative reputation, sprawl occurs because
individual households are happier with the larger homes and lots that it offers. However,
they do identify four costs: road congestion, vehicle pollution, loss of open space, and
unequal service and public good provision across metro areas due to self-segregation
and associated pockets of affluence and poverty. Lopez (2004) and Plantinga and
Bernell (2007) investigate the link between obesity and urban sprawl. These papers and
most of the related literature focus on concrete sprawl impacts: weight, emissions, and
income distribution. Brueckner and Largey (2008) notably depart from this trend and
focus on sprawl and the reduction of social interaction. They investigate the premise
that low-density living reduces social interaction to the detriment of society as a whole.
Having reviewed the economic literature on the interaction between land use and
municipal profiles, an illustrative example follows. Specifically, in the next section, a
case study from the Northwest United States illustrates how researchers might go about
modeling these relationships. Whereas much of the economic literature investigates
household locational choice and land use change at a county or city level, emerging data
and information technology facilitate investigation within a metropolitan area.
2. Case Study
Household locational choice is a central driver of land use and municipal profile change.
Individually, households relocate based on a variety of factors. Family, career, or other
factors often determine the city, although municipal profile may play a role. Households
then choose a home within that city or metropolitan area based on the households’
characteristics (preferences, income, wealth), home characteristics (price, size, etc.),
and municipal profile (regulations, public goods, demographics, etc.). Relevant regula-
tions include taxes and land use regulations. Relevant public goods or amenities include
school quality, public safety, transit access, environmental quality, access to parks and
open space, and social amenities like shopping, dining, and culture. With these theoreti-
cal relationships in mind, how can researchers go about modeling this process? How can
researchers take advantage of emerging information technology and rapidly improv-
ing data availability? The following case study details the data collection and process-
ing used to model the link between household locational choice and municipal profiles
in the Portland, Oregon, and Vancouver, Washington, metropolitan area. A discussion
of alternative estimating strategies follows. Although the data collection and process-
ing described here are specific to this case study, the approach described is adaptable to
a variety of geographic areas and research questions, contingent on the availability of
appropriate GIS data.
2.1 Data Processing

GIS data differ from conventional data in that they are spatially explicit. Although con-
ventional data may contain variables describing spatial relationships, such as distance
from each observation to a park, they do not preserve the underlying spatial arrange-
ment of different features. A GIS dataset or layer includes the location (and shape) of
each feature and can therefore be represented as a map. Also associated with each layer
is a table containing one or more variables. Data layers differ in terms of their spatial
resolution; for the same geographic area, a high-resolution layer will contain more
observations than will a low-resolution layer. GIS software facilitates the use of multiple,
overlapping data layers.
The Portland, Oregon, and Vancouver, Washington, metropolitan area—specifically
Clackamas, Multnomah, and Washington Counties in Oregon and Clark County in
Washington—makes an excellent laboratory. Abundant GIS data are available for the
Clark Co.
Washington Co.
Portland
Multnomah Co.
Clackamas Co.
FIGURE 12.1 The study region.
area from a variety of sources, often at quite high spatial resolution. In addition to rich
data, the region consists of multiple jurisdictions—more than 40 incorporated towns
and cities in four counties and two states (see Figure 12.1) These jurisdictions differ sig-
nificantly in regulatory regimes, which should influence locational choice. Combined,
the four counties have an area of 3,727 square miles and a 2000 population of 1.79 mil-
lion, which grew to 2.07 million in 2010 (15.5% decennial growth).
Several issues must be addressed to move from a theoretical concept of household
locational choice to an empirical model of the effect of municipal profiles. First, a
dependent variable, some measure of household choice, is needed. Second, data quan-
tifying local municipal profiles are required. Third, the researcher must select a unit of
observation. Finally, GIS is used to process information from the underlying layers and
construct a dataset for estimating a model of the effects of municipal profiles on house-
hold location choice.
The data described here for the Portland, Oregon, and Vancouver, Washington, met-
ropolitan area do not include information on individual household location choices, but
they do include a measure of a direct consequence of these choices; that is, population
change at the US Census block level. The metropolitan area as a whole grew rapidly,
but certainly some areas grew more rapidly than others. Population change is a feasi-
ble dependent variable. The resulting model would aim to explain variations in popu-
lation change within the region using variables describing the local municipal profile.
Of course, some elements of municipal profile are exogenous, including natural ameni-
ties such as lakes, rivers, and other topographical features, whereas other elements of
2000 Land use: Parks and open space
FIGURE 12.2 The distribution of parks and designated open space.
municipal profile are endogenous to population change, including parks and designated
open space (see Figure 12.2) and public goods such as school quality (see Figure 12.3).
For the dependent variable, GIS-compatible population data are available at the
Census block level from both the 2000 and the 2010 US Census. Data relevant to local
municipal profiles are available from a variety of sources, most importantly the US
Census and local governments.
The 2000 US Census includes a number of potentially relevant demographic vari-
ables, including age and race/ethnicity at the block level. Data on education, income,
housing characteristics, and more are available at the block group level. To give an idea
of scale, in 2000, the study area contained 34,178 census blocks and 1,160 block groups.
For the three counties in Oregon, the elected regional government, METRO, main-
tains the Regional Land Information System (RLIS), a high-quality GIS database with a
variety of data layers. Potentially relevant layers for quantifying local municipal profiles
include zoning, water features, parks and designated open space, mass transit, and tax-
lots, which includes parcel-level data on land use and home characteristics (for residen-
tial properties). The Clark County, Washington, Assessor also offers similarly detailed
GIS data. Although similar, these datasets are not identical, and substantial care is nec-
essary to ensure consistency when merging data across states. The end result is a single
map covering the entire study area for each relevant layer.
For some layers, including school districts, these GIS data contained maps but no rel-
evant variables. If available, the relevant data are easily incorporated into GIS. In the
case of school districts, an index of school quality is constructed using test score data
2000 School districts
Legend
School districts
sch_qual
< 0.85
0.85 – 0.95
0.95 – 1.05
1.05 – 1.15
>1.15
FIGURE 12.3 School quality measured by an index of test scores relative to state averages.
available from Oregon and Washington state departments of education. In each case,
reading and math test scores are reported for multiple grade levels. A composite score
is a viable option to compare districts across multiple grades and subjects. However,
because data and testing vary across the two states, composite scores from Oregon and
Washington are not comparable. This is resolved by normalizing using state averages.
For Washington, 2000 district-level reading and math scores for the 4th, 7th, and
10th grades are reported as percentages. Each score is divided by the corresponding
state-level score, converting scores from percentages to shares of the state average. These
shares are then averaged across subjects and grade levels with equal weight. The result
is an index measuring district-level test score performance relative to the state average.
For Oregon, reading and math test results are reported at the 3rd-, 5th-, 8th-, and
10th-grade levels. Instead of percentages, the share of students who do not meet, meet,
or exceed performance standards is reported. In addition, data are at the individual
school level, and the number of students taking each test is known. First, a single score
for each subject and grade level at each school is constructed. This score is the share of
students who meet the standard plus two times the share of students who exceed the
standard. At this point, each score is normalized using the corresponding state score,
then averaged across subjects, grade levels, and schools. Additionally, for Oregon, these
averages are weighted based on the number of students taking each test. The result is a
single value, which again measures district test performance relative to the state average,
making it comparable to the Washington index. However, normalizing by state averages
implicitly assumes little systematic difference between Oregon and Washington. Any
such systematic difference would be captured by the intercept in estimation.
With underlying data in place, the next step is choosing units of observation. Existing
geographies tend to be problematic. Using counties would provide only four observa-
tions and ignore variation in population change and local municipal profiles within the
counties. Using cities drops unincorporated areas and again ignores variation, particu-
larly in the largest city, Portland. US Census geographies, including census blocks and
census block groups, are much smaller than counties and cities and so can capture vari-
ation within cities and counties. However, a considerable proportion of census geog-
raphies shift boundaries over time, leading to consistency problems when measuring
population change. Furthermore, the size of census geographies varies widely, as census
blocks and block groups are drawn to have roughly equal populations. Thus, rural cen-
sus blocks with low population density are much larger than densely populated urban
census blocks. Finally, and perhaps more problematically, census geographic boundar-
ies are not random; they tend to follow evident development patterns and form homo-
geneous units. Although these make sense as cohesive units within a city or county,
nonrandom boundaries can lead to endogeneity issues and biased estimates (Banzhaf
and Walsh 2008).
Researchers can avoid problems associated with existing geographies by construct-
ing new units of observation in GIS. For this case study, a grid of two-mile diameter
circles is overlaid on the study area, and those circles not completely within the study
area are dropped. These circles do not represent cohesive communities in any tradi-
tional sense.3 Rather, this method constitutes an effective sampling methodology that
allows us to take advantage of high-resolution spatial data. Of course, a grid of two-mile
circles is not the only option. Alternative diameters, shifting the grid incrementally, and
random locations as opposed to a grid are possible. Indeed, comparing alternative units
is a good strategy for testing the sensitivity of coefficient estimates. This approach has
been used by Banzhaf and Walsh (2008) to test the Tiebout hypothesis that “people vote
with their feet.”
GIS software and data are used to quantify variables measuring population change
(see Figure 12.4) and municipal profiles for each circular “community” or observation.
This procedure varies depending on the data in question. Some GIS layers cover the
entire study area, such as census geographies, tax lots, zoning, and school districts. These
layers each contain one or more potential explanatory variables. For each layer, a GIS
script aggregates the variables of interest from the underlying geometry to the circular
observations. A number of variables relevant to municipal profile are constructed in this
fashion, including median household income (Figure 12.5) and home value (Figure 12.6).
3 Indeed, because the four counties in the study area include rural areas devoted to forestry (including
State and National Forests) and agriculture, there are 178 out of 844 (21%) observations with zero
population, as noted in Table 12.1. Clearly, these observations are not communities.
2000–2010 Population change
Legend
Pop. Change
chPop_bl
< −100
[−100, 0)
0
(0. 250]
> 250
FIGURE 12.4 Population change between 2000 and 2010.
For other GIS layers that do not fully cover the study area, such as bus stops and parks,
the appropriate measure is less clear. The bus stop layer contains only points, so an
appropriate measure might be the number of bus stops in a circular observation. One
can measure access to parks and open space in a variety of ways: park acreage within
the observation, distance to the nearest park, and number or acreage of parks within
some distance, among others. One can differentiate parks by type from the data as well.
Also, one recalls that the hedonics literature reveals variations in amenity values of open
space depending on many factors including income, proximity, type of open space, age,
urban density, and crime. Interaction terms allow models to capture differential effects.
For example, a community park located in a low-income, high-crime neighborhood
may not be valued as much as a park located in a high-income, low-crime neighbor-
hood. A model specification including interactions between crime or income and park
proximity variables might pick up this effect whereas an alternative specification would
not. Anderson and West (2006) provide a good discussion of interaction terms in this
context, in addition to a hedonic model with multiple types of open space and multiple
interaction terms. Table 12.1 provides summary statistics of some of the constructed
variables.
2000 Median household income
Legend
Median HH income
medhhi_fix
< $35k
$35k – $50k
$50k – $65k
$65k – $80k
> $80k
FIGURE 12.5 Median household income in 2000.
2.2 Model Specification and Estimation
This case study explored underlying data collection, construction of units of observa-
tion, and quantification of variables. Although specific, the process described is appli-
cable to other regions and research questions. The data clearly indicate correlations
between land use and municipal profiles (Table 12.2). The challenge is to specify an
appropriate model to identify the causal relationships among the dependent variable,
population change, and the independent variables, which quantify various aspects of
local municipal profile. Several estimation strategies are available to the researcher
investigating household locational choice and municipal profiles. The appropriate
estimation strategy depends on the precise research question. In all cases, an appropri-
ate estimation strategy must account for the fact that some variables measuring local
municipal profiles are exogenous, whereas others are endogenous. Examples of exog-
enous variables include natural features and historical development patterns. Examples
of endogenous variables include median household income, school quality, and prop-
erty tax rate. In the absence of endogeneity, the researcher could simply regress popula-
tion change on variables quantifying municipal profile. Due to endogeneity, this simple
approach would yield biased estimates. Because some of the explanatory variables are
2000 Mean value of single-family homes
Legend
Mean home value
Value
0 Signal-family homes
< $150k
$150k – $200k
$200k – $250k
> $250K
FIGURE 12.6 The mean value of single-family homes in 2000.
affected by the dependent variable and thus correlated with the error term, the model
becomes a system of simultaneous equations. In structural form, the model looks like:
Y = Xnβn + X s βs + εy , (1)

X si = Y γ iy + Xn γ ni + Z i γ z + εix , i = 1, 2, …, n (2)

where Y is the population change vector, Xn is exogenous municipal profile variables,

X s = ( X 1s , X s2 , …, X sn ) is the endogenous municipal profile variables, Zi is a vector of
variables that affect endogenous profile variable i, but do not affect household location
choices directly, the β’s and γ’s are the respective coefficients, and the ε’s are the error
terms. The structural model can be estimated in different ways, depending on the avail-
ability of appropriate instrumental variables and data.
If variables Zi can be identified for each endogenous profile variable, and data on Zi are
available, then (Z1 , Z 2 , …, Z n ) can serve as a set of instrumental variables because they
are correlated with the endogenous explanatory variable and uncorrelated with the error
term (i.e., causally unrelated to the dependent variable). In this case, the structural model
Table 12.1 Summary statistics

Circular observations Study area
Total change
Variable Mean Std. Err. Min Max (%)
Change in population 267.25 800.80 −809.25 9,206.78 276,942

(15.5%)
Mean/Study
Area
Distance to city center 25.45 12.62 0.00 56.36
Share of forest land 0.49 0.41 0.00 1.00 0.50
Share of agricultural land 0.12 0.23 0.00 0.97 0.12
Share of single-family home 0.12 0.19 0.00 0.76 0.12
Share of multi-family home 0.00 0.01 0.00 0.10 0.00
Share of commercial land 0.01 0.03 0.00 0.23 0.01
Share of industrial land 0.01 0.05 0.00 0.74 0.01
Share of national forests 0.26 0.43 0.00 1.00 0.46
Share of water 0.02 0.06 0.00 0.59 0.02
Share of park and designated 0.03 0.07 0.00 0.80 0.03
open space
Mean home value ($) 197,135 92,901 0.00 1,351,806 182,131
Mean lot size 1.81 8.30 0.00 170.17 0.63
School quality 1.04 0.15 0.75 1.53 1.05
2000 property tax levy 11.14 2.05 8.08 19.92 14.02
Population share in poverty 0.09 0.06 0.00 0.28 0.09
Urban population share 0.24 0.38 0.00 1.00 0.90
* Observations with zero population (n = 178) dropped from mean and standard deviation
calculations.
can be estimated using two or three-stage least squares or partial or full information
maximum likelihood estimation methods. For example, using two-stage least squares,
first regress each endogenous variable in Xs on all exogenous variables in the model
s . Then replace endogenous variables
(Z1 , Z 2 , …, Z n and Xn) and obtain fitted values, X
with fitted values in (1) in the second-stage regression.
s βs + ε y
Y = X n βn + X (3)

Estimates derived from instrumental variables and two-stage least squares are only as
reliable as the instruments. If the chosen instruments are correlated with the error term,
the bias problems encountered in the structural form remain unresolved. If the chosen
instruments are poor (weakly correlated with the endogenous variables they are replac-
ing), the result is poorly fitted values with little variation generated in the first stage. For
this case study, appropriate instruments would need to be correlated with endogenous
Table 12.2 Correlation coefficients

Medium household
Mean home value income School quality Property tax levy
Share of forest −0.47* −0.14* 0.13* −0.52*

land
Share of 0.15* 0.10* −0.01 0.00
agricultural land
Share of 0.33* 0.11* −0.07 0.45*
single-family home
Share of 0.16* −0.15* −0.15* 0.28*
multi-family home
Share of 0.22* −0.15* −0.08* 0.26*
commercial land
Share of industrial 0.18* −0.17* −0.06 0.31*
land
Share of park and 0.30* 0.15* −0.13* 0.36*
designated open
space
* Significant at the 5% level

Observations with no population in 2000 and in 2010 are dropped (n = 666).
amenities, uncorrelated with the error term, and not included in the set of explanatory
exogenous amenities. It can be challenging to identify such variables.
When appropriate instruments cannot found, researchers may resort to estimating
the reduced form of the structural model to uncover useful information about the effect
of amenities on location choices. Solving for Y and X s = ( X1s , X s2 , …, X sn ), one can derive
each of these endogenous variables as a function of Xn and perhaps (Z1 , Z 2 , …, Z n ).
These reduced-form equations can then be estimated using an appropriate method. The
related literature strongly suggests that exogenous natural amenities influence develop-
ment patterns, and these development patterns in turn affect the level of endogenous
social amenities (see, e.g., Wu 2006). Thus, exogenous amenity variables can be used to
explain endogenous amenity variables. The reduced form approach has a major draw-
back: estimation does not identify the structural coefficients found in (1). So, although
estimating the reduced form in this case sheds light on how natural amenities affect the
level of social amenities, it does not reveal the effects of various elements of municipal
profile on population change.
The model specification in (1) potentially includes multiple endogenous covari-
ates, for example, tax rate, school quality, and park access. This case study has a broad
research question. How do elements of municipal profile affect population change?
By narrowing the research question to focus on a single endogenous covariate, asking
instead how property tax rate affects population change, a number of other estimation
strategies from the treatment effects literature become available. Ordinary least squares
(OLS) estimation of treatment effects is typically biased because treatment effective-

ness depends on factors that determine whether an observation gets treated. In a medi-
cal context, the effectiveness of medical intervention depends on the characteristics of
the patient. At the same time, the characteristics of the patient determine whether the
patient receives the treatment. Although a full discussion of treatment effects is beyond
the scope of this chapter, a number of measures developed for nonexperimental settings
in the medical field are increasingly being adopted by economists.
Two relevant estimation strategies from this literature are propensity score matching
and difference-in-difference estimators, both described later in the context of property
taxes. Propensity score matching could be used to evaluate the effect of differential tax
rates on population growth. Propensity score matching estimates the treatment effect
by systematically comparing pairs of observations from the treatment and nontreat-
ment groups that are otherwise alike. This method first estimates a model predicting
likelihood of treatment and pairs observations for comparison based on the result-
ing fitted values. For property taxes, this would involve identifying low- and high-tax
observations, regressing tax rate on variables thought to influence tax rate (e.g., income
distribution, demographics), and calculating predicted tax rates using the estimated
coefficients. Each low-tax observation is paired with the high-tax observation with the
closest predicted tax rate. By controlling across a number of relevant covariates, pro-
pensity score matching improves the likelihood that observed differences in population
change are, in fact, the result of different tax rates.
Difference-in-difference methods, conversely, measure the effect of a treatment at a
given point in time. The idea behind this method is to compare the treated group to
itself before treatment, as well as to some other untreated control group. Simply evaluat-
ing treated observations relative to themselves before treatment does not account for
events or trends that occur during treatment and affect the entire treatment group. If
the researcher fails to include a nontreatment control group, then changes attributable
to trends affecting the general population will be attributed inappropriately to treat-
ment. In the property tax context, local population changes should not be attributed
to changes in local tax rates without first accounting for the population change trends
in the region. If the region as a whole is growing, then it is likely misleading to attribute
local population growth entirely to local changes in tax rate. The researcher can net out
the regional trend by comparing the treated group to an untreated control group.
The preceding case study details the data collection and processing used to construct
a model of household locational choice (population change) in the Portland, Oregon
metropolitan area. GIS software facilitate creative solutions for data at different resolu-
tions. Still, substantial care is necessary in model specification and estimation to avoid
the pitfalls associated with interactions and endogeneity.
In fact, the organization of the case study, particularly the model estimation sec-
tion, closely follows the authors’ efforts to take advantage of rich data for the study area
while avoiding the aforementioned pitfalls. Although detailed presentation of results is
beyond the scope of this chapter, a brief discussion of the empirical work that provides
the basis for the case study follows.
The initial focus of this research was causal links between municipal profile and
land use change, broadly, and, specifically, the effect of natural and social amenities on
household location choice. The inclusion of a broad set of municipal attributes, some of
which are undoubtedly endogenous, precludes unbiased OLS estimation. Broad con-
trols also render instrumental variable estimation infeasible in practice due to the dif-
ficulties of identifying appropriate instruments. Without sacrificing broad controls, a
reduced form model explaining endogenous municipal attributes in terms of exogenous
attributes remains a feasible option. In this case, reduced-form estimation reveals that
exogenous natural and historical amenities do indeed influence the level of endogenous
municipal characteristics, including population change and density, median income,
school quality, property taxes, and home values. Results indicate how natural charac-
teristics (e.g., slope, elevation) and proximity to different natural amenities (e.g., water
bodies, parks by type) influence endogenous characteristics. Of course, reduced-form
estimation does not shed light on the underlying relationships between location choice
and endogenous municipal characteristics. Furthermore, although they illustrate pref-
erences, reduced-form results may have little policy relevance since natural features are
difficult to change.
To quantify the underlying relationships in the absence of appropriate instruments,
one alternative approach is to abandon broad controls and focus on a single municipal
feature. In this case, although biased, preliminary OLS estimates highlight the impact of
race/ethnicity on population change. Neighborhoods with high concentrations of black
residents tended to shrink. Other minority neighborhoods grew fast, especially Asian
neighborhoods, whereas majority neighborhoods grew modestly. These observations
gave rise to a more focused research question: how do minority concentrations affect
local municipal profile or neighborhood quality?
Simple correlations reveal that high minority concentrations are associated with
lower school quality and higher crime. However, this approach ignores systematic dif-
ferences between minority and majority groups, for example, in terms of income and
educational attainment. To isolate the effect of minority concentration from the effect of
these systematic differences, a more sophisticated method is required. In this case, treat-
ment effects methods, specifically propensity score matching, are appropriate. Under
propensity score matching, pairs of observations that differ in terms of minority con-
centration but that are similar in other dimensions of municipal profile are compared.
In this context, that means a higher minority concentration community compared to a
lower minority concentration community with the most similar other characteristics.
Controlling for other dimensions of municipal profile can yield results that differ strik-
ingly from simple correlations. For example, once other relevant municipal attributes
were controlled for, communities with higher concentrations of black residents exhib-
ited significantly lower crime rates than communities with black resident concentra-
tions closer to the study area mean.
This case study provides a fairly specific example of data collection and process-
ing. It also provides a general guide to estimation procedures and several descriptions
of empirical applications. The fundamental challenge of modeling the relationships

between land use and municipal profiles is the interconnected nature of individual
location decisions and outcomes at the neighborhood, city, or regional scale. Quality
data do not preclude the fundamental challenges of identification in the presence of
endogeneity.
3. Conclusion
Land use and quality of life are inextricably linked. Collective household location deci-
sions affect local municipal profiles, the character of cities, and landscapes. Shifting
populations affect municipal profiles through voting, local public finance externalities,
and peer externalities. At the same time, local municipal profiles affect location deci-
sions because households choose the bundle of regulations and public goods they pre-
fer. Understanding these effects and interactions is central to managing development
and land use change in the future.
Several bodies of work within the economics literature shed light on household loca-
tion decisions. Urban economics models identify the primary drivers behind observed
suburbanization trends: rising incomes and falling commuting costs. However, these
models tend to ignore many regulations, public goods, and amenities that affect house-
hold location choice. Local public finance models include preferences for these alterna-
tive bundles, but many such models are aspatial.
Emerging GIS software and rapidly improving data availability facilitate analy-
sis within metropolitan areas, as opposed to at the county or city level. Although this
approach is promising, the sound judgment of the researcher remains necessary. In
particular, the researcher must construct relevant measures of local municipal profiles,
often from a profusion of underlying GIS data. Some choice between alternative empiri-
cal measures of the same theoretical variable may be necessary. A significant effect may
only appear with properly specified interaction terms. An inappropriate estimation
strategy can bias results.
One additional obstacle facing researchers in this area is measuring endogenous
social amenities. How do researchers measure the intangible desirability of neighbor-
hoods and districts? For example, shopping is an amenity, but it is difficult to quantify.
Even data on the location of retail stores are insufficient because big-box suburban
shopping centers are qualitatively different from walkable urban shopping districts.
Although measures of some social amenities are hard to come by, spatial analysis and
the increasing profusion of GIS data present a wider array of potential measures than
has previously existed. Although interactions between land use and municipal profiles
are complex and may be difficult to measure, the tools available to address this issue
have never been better.
Acknowledgments
This material is based on work supported by the US Forest Service Pacific Northwest
Research Station under JVA No. 11-JV–11261985–073. Any opinions, findings, and
conclusions or recommendations expressed in this material are those of the author(s)
and do not necessarily reflect the views of the US Forest Service.
References
Alonso, W. 1964. Location and land use. Cambridge: Harvard University Press.
Anderson, S. T., and S. E. West. 2006. Open space, residential property values, and spatial con-
text. Regional Science and Urban Economics 36(6): 773–789.
Banzhaf, H. S., and R. P. Walsh. 2008. Do people vote with their feet? An empirical test of
Tiebout. American Economic Review 98(3), 843–863.
Bowes, D. R., and K. R. Ihlanfeldt. 2001. Identifying the impacts of rail transit stations on resi-
dential property values. Journal of Urban Economics 50(1): 1–25.
Brueckner, J. K., J. F. Thisse, and Y. Zenou. 1999. Why is central Paris rich and downtown Detroit
poor? European Economic Review 43(1): 91–107.
Brueckner, J. K., and A. G. Largey. 2008. Social interaction and urban sprawl. Journal of Urban
Economics 64(1): 18–34.
Brueckner, J. K., and S. S. Rosenthal. 2009. Gentrification and neighborhood housing cycles: Will
America’s future downtowns be rich? Review of Economics and Statistics 91(4): 725–743.
Carlino, G. A., and E. S. Mills. 1987. The determinants of county growth. Journal of Regional
Science 27: 29–54.
Cohen, J. P., and C. C. Coughlin. 2008. Spatial hedonic models of airport noise, proximity, and
housing price. Journal of Regional Science 48(5): 859–878.
DeFries, R. S., J. A. Foley, and G. P. Asner. 2004. Land-use choices: Balancing human needs and
ecosystem function. Frontiers in Ecology and the Environment 2(5): 249–257.
Ellickson, B. 1971. Jurisdictional fragmentation and residential choice. American Economic
Review 61(2): 334–339.
Epple, D., R. Filimon, and T. Romer. 1984. Equilibrium among jurisdictions: Toward an inte-
grated treatment of voting and residential choice. Journal of Public Economics 24(3): 281–308.
Epple, D., and H. Sieg. 1999. Estimating equilibrium models of local jurisdictions. Journal of
FAO/UNEP. 1999. Terminology for integrated resources planning and management. Rome, Italy
and Nairobi, Kenya: Food and Agriculture Organization/United Nations Environmental
Programme.
Foley, J. A. et al. 2005. Global consequences of land use. Science 309(5734): 570–574.
Fujita, M., and H. Ogawa. 1982. Multiple equilibria and structural transition of non-monocentric
urban configurations. Regional Science and Urban Economics 12(2): 161–196.
Geoghegan, J., L. A. Wainger, and N. E. Bockstael. 1997. Spatial landscape indices in a hedonic
framework. Ecological Economics 23(3): 251–264.
Irwin, E. G., and N. E. Bockstael. 2001. The problem of identifying land use spillovers: Measuring
the effects of open space on residential property values. American Journal of Agricultural
Economics 83(3): 698–704.
Irwin, E. G. 2002. The effects of open space on residential property values. Land Economics
78(4): 465–480.
LeRoy, S. F., and J. Sonstelie. 1983. Paradise lost and regained: Transportation innovation,
income, and residential location. Journal of Urban Economics 13(1): 67–89.
Lopez, R. 2004. Urban sprawl and the risk for being overweight or obese. American Journal of
Public Health 94: 1574–1579.
Mieszkowski, P., and E. S. Mills. 1993. The causes of metropolitan suburbanization. Journal of
Economic Perspectives 7(3): 135–147.
Mills, E. S. 1967. An aggregative model of resource allocation in a metropolitan area. American
Economic Review 57: 197–210.
Muth, R. F. 1969. Cities and housing. Chicago: University of Chicago Press.
Muth, R. F. 1971. Migration: Chicken or egg? Southern Economic Journal 37(3): 295–306.
Nechyba, T. J., and R. P. Walsh. 2004. Urban sprawl. Journal of Economic Perspectives
18(4): 177–200.
Oates, W. E. 1969. The effects of property taxes and local public spending on property val-
ues: An empirical study of tax capitalization and the Tiebout hypothesis. Journal of Political
Economy 77(6): 957–971.
Ogawa, H., and M. Fujita. 1980. Equilibrium land use patterns in a non-monocentric city.
Journal of Regional Science 20(4): 455–475.
Plantinga, A. J., and S. Bernell. 2007. The association between urban sprawl and obesity: Is it a
two-way street? Journal of Regional Science 47(5): 857–879.
Smith, V. K., C. Poulos, and H. Kim. 2002. Treating open space as an urban amenity. Resource
and Energy Economics 24(1): 107–129.
Tiebout, C. M. 1956. A pure theory of local expenditure. Journal of Political Economy
64(5): 416–424.
Troy, A., and J. M. Grove. 2008. Property values, parks, and crime. Landscape and Urban
Planning 87(3): 233–245
Environmental Economics and Management 52(2): 527–547.
Wu, J. 2007. How does suburbanization affect local public finance and communities? Review of
Agricultural Economics 29(Fall 2007): 564–571.
PA R T I I I
M E T HOD OL O G IC A L
DE V E L OP M E N T S
C HA P T E R 13
AN ASSESSMENT OF
EMPIRICAL METHODS FOR
M O D E L I N G L A N D U S E
E L E NA G . I RW I N A N D D OU G L AS H . W R E N N
Many of society’s most pressing socioeconomic and environmental issues relate in some
way to land use and land use change. Environmental problems such as carbon cycling
(Post et al. 1982; Schimel 1995), terrestrial water cycles (Vorosmarty and Sahagian
2000), loss of biodiversity (Sala et al. 2000), and climate change (Vitousek et al. 1997)
are directly or indirectly impacted by anthropogenic land use decisions. From a socio-
economic perspective, issues such as sprawl and suburbanization, congestion, public
service provisioning, and segregation are fundamentally related to land use change and
land use policies (Anas et al. 1998; Glaeser anf Kahn 2004; Nechyba and Walsh 2004).
The wide-ranging issues surrounding land use have led policy makers and research-
ers alike to develop land use models as a means of better understanding policy and other
effects. A variety of empirical land use modeling approaches is evident across multiple
academic disciplines. These approaches often have been distinguished by a key differ-
ence in research focus: identification of specific parameters of the underlying process
versus spatial prediction of land use patterns. Economists typically have focused on
causal identification of the underlying economic processes that generate land use out-
comes and patterns using reduced-form models to identify one or more key parameter
values. For example, hedonic models of land values are common in which the research
question is to identify the influence of a specific landscape feature or spatially articulated
policy on equilibrium land prices. The advantage of this approach is that consistent and
unbiased parameter estimates can be recovered to infer something about the effect of a
marginal change on the equilibrium. This approach is limited for spatial prediction or
counterfactual policy simulation, however, since the focus is on explaining the observed
price equilibrium, and the underlying structural parameters of demand and supply are
not recovered. Geographers and others outside of economics typically have focused on
empirical prediction of land use patterns and changes, for example, by calibrating model
328 Elena G. Irwin and Douglas Wrenn
parameters to derive transition rules that describe the evolution of land use or land cover
over space and time. Although these models are useful for description and perhaps for
the very short-run prediction of patterns, they provide little insight into the underlying
economic and other processes that generate these patterns. Thus, they also cannot be
used for counterfactual policy simulation or any spatial prediction with nonmarginal
changes. More recently, these disciplinary distinctions have blurred. Geographers have
sought to develop agent-based models that provide a process-based approach to land
use modeling, and economists have pursued structural econometric models that can be
used to predict large-scale changes over time and space.
The purpose of this chapter is to provide an overview and assessment of the main
methods used to model land use and land use change, with a focus on newer methods.
We focus on empirical models, which we define broadly as models that use data on land
use and the underlying demand and supply processes to specify model parameters in
some way. We array these models along two dimensions: first, models that are struc-
tural versus reduced-form and second, econometric models versus other empirical
approaches that are used to specify parameter values. Rather than providing an in-depth
primer on these modeling techniques, our goal is to present a general overview and a
targeted assessment. The key questions we seek to address are (1) what are the advan-
tages and disadvantages of these various empirical approaches to modeling land use and
land use change, (2) which questions are best suited to be answered using one versus the
other approach, and (3) where are the gaps in the current literature?
The remainder of the chapter is organized as follows. Following a general discussion of
modeling approaches, we turn to the particular case of modeling land use and land use
change. We ask what makes modeling land use and land use change special and how the
various modeling approaches stack up with respect to these key considerations. We pro-
vide a discussion of each of the main modeling approaches, highlighting strengths and
weaknesses and illustrating with a few recent examples from the literature. We conclude
with some thoughts about the existing gaps in the literature and future research needs.
1. An Overview of Modeling Methods
Any economic model begins with a structural model of the underlying economic pro-
cesses, for example, supply and demand equations, indirect utility function (household
side), or cost function (firm side), that are hypothesized to generate an observed mar-
ket equilibrium. Most economists would consider a structural econometric model to be
the gold standard for empirically modeling this process. Put simply, this approach uses
econometric methods to recover the full set of parameters of the underlying structural
model by making explicit assumptions about what is and is not observed (Timmins and
Schlenker 2009). For example, in a model of land development, the underlying struc-
tural parameters of a land developers’ profit or cost functions would be recovered using
econometric methods given a series of assumptions about functional form, choice sets,
An Assessment of Empirical Methods for Modeling Land Use 329
equilibrium in the market, and the distribution of the unobservables. The advantage of
a structural econometric approach is that by modeling these processes explicitly, it is
possible to account for the endogeneity of prices and other market-level or nonmarket
feedbacks that determine the equilibrium. A structural modeling approach is neces-
sary for counterfactual policy simulation, in which the goal is to evaluate the impacts
of a nonmarginal policy change on land use outcomes.1 This is particularly important
when modeling complex processes such as land use in which nonmarginal feedbacks
can arise from interactions within and between the socioeconomic and biophysical
systems.
Although we consider structural econometric modeling to be the benchmark, the
type of model implemented is ultimately determined by the particular research ques-
tion, limitations of theory and data, and the willingness of the researcher to make certain
assumptions. One of the main disadvantages of structural econometric modeling is that
the researcher must be willing to make certain modeling assumptions regarding the dis-
tributions of unobserved variables, the choice sets, the equilibrium relationship in the
market, and the functional forms that represent the behavioral equations. Theoretical
models often provide key insights into the processes at work—for example, utility maxi-
mization or cost minimization—but they rarely provide an explicit functional form for
the objective functions, a specific distribution for the error structure, or guidance on
specification of choice sets or how to best define equilibrium in the market. In addi-
tion, even if a robust empirical specification can be established, it still may be difficult to
gather data on all of the processes deemed important in answering a particular research
question. This is especially true in the area of land use modeling, given the complexity of
interactions across multiple spatial and temporal scales.
Given the challenges involved in fully estimating structural models, alterna-
tive approaches are often pursued. One alternative is to retain a structural model-
ing approach, but to take a less rigorous approach to parameter specification. Rather
than estimating the structural parameters in a manner that is fully consistent with the
observed data, a more ad hoc econometric strategy may be pursued or a combination
of empirical approaches used. For example, utility or cost parameters may be estimated
using multiple datasets taken from different sources or settings, or key parameter values
may simply be taken from the results of other studies reported in the literature (e.g.,
such as a demand or supply elasticity). A limitation of this approach is that the param-
eters may reflect different underlying conditions of demand and supply that are not
1
Although it is not always obvious what constitutes a marginal versus nonmarginal change, the intent
is to distinguish marginal changes as small changes that do not shift the underlying equilibrium and
nonmarginal changes as those that are large enough that they could. For example, a major downzoning,
in which the maximum allowable number of lots decreases from one house per acre to one house per
50 acres would be likely to shift residential land supply and induce a demand response and therefore
is nonmarginal. Conversely, an incremental change in a local jurisdiction’s budget for farmland
preservation is unlikely to generate a large shift in the demand or supply of land preservation and is
therefore marginal.
consistent with any single equilibrium. The advantage is that it may be easier to param-
eterize a structural model using some combination of empirical methods and data or, in
cases in which econometric estimation of the structural model is infeasible, this may be
the only possible means to parameterize a structural model.
In other cases, when model intractability or data limitations prevent structural
econometric estimation of the model, a reduced-form model is estimated instead.
Reduced-form models are models of an equilibrium outcome (e.g., land use, land
use change, or land or housing prices) derived from an underlying structural model
of demand and supply and expressed in terms of the simultaneous equilibrium rela-
tionship. Unless further structure is imposed, the explanatory variables included in a
reduced-form equation cannot be attributed to a specific underlying structural process
but instead reflect the net effects of these variables on the equilibrium outcome. In most
cases, this implies that parameter estimates cannot be used to simulate the impacts of a
nonmarginal change on land use outcomes.
In many reduced-form models of land use and land use change, the model may not be
fully reduced to only exogenous variables—that is, the model may include one or more
endogenous explanatory variables that are determined by the same equilibrium process as
the dependent variable. This is particularly true for land use models in which local inter-
actions imply that many of the variables are jointly determined by the same equilibrium
process. For example, in the case of open space spillovers that influence the amenity value
of a location, the spatial distribution of open space is usually endogenous to the land mar-
ket, implying that the spatial patterns of residential and open space are jointly determined.
If the endogeneity is properly dealt with, then the estimation will yield a consistent esti-
mate of the reduced-form parameters, which can then be used in hypothesis testing or for
simulation of marginal changes. The results can only be interpreted as representing the
effect of a marginal change and are conditional on the assumption that the equilibrium is
unchanged.
In some cases, a reduced-form model is preferred when the research is focused on the
identification of one or several key reduced-form parameters rather than on recovering
the underlying structure. As discussed in depth in Chapter 18 by Towe, Lewis, and Lynch
in this handbook, this is the case with reduced-form models used in quasi-experimental
designs in which the structural system has been fully solved so that the dependent vari-
able is defined only in terms of exogenous variables. This approach is touted as being
free of functional form assumptions and, assuming that the covariates are indeed exog-
enous, provides a transparent identification strategy. It should be noted, however, that
reduced-form models are not free from assumptions, and it is best to begin any modeling
exercise with a review of the relevant theoretical literature. Keane (2010) points out that
even in the case of the most clever instrument or quasi-experimental setting, if the initial
modeling exercise is not based on some economic model, then it is impossible to assign
meaning to the output.
2. What Is Special About Modeling

Land Use?
We do not consider the approach to modeling land use or land markets to be any dif-
ferent from the basic approach used in economics to model any behavioral pro-
cess. Land development outcomes, for example, are determined by the constrained
utility-maximizing location decisions of households based on their preferences over
parcel, neighborhood, and regional characteristics, and the supply decisions of develop-
ers and landowners are based on expectations of profits and costs subject to technologi-
cal and regulatory constraints. However, there are several features of land and related
markets (e.g., housing) that deserve special attention because of their importance in
determining market equilibrium and the spatial distribution of land use and land use
change.
First, land is an extremely heterogeneous good over which individuals have hetero-
geneous preferences and heterogeneous expectations about the future. Each of these
sources of heterogeneity is hypothesized to influence land use outcomes. The evolution
of leapfrog development, for example, has been alternatively explained as the result of
heterogeneous land quality (Harvey and Clark 1965), heterogeneous preferences over
large exurban lots (Newburn and Berck 2011), or heterogeneous expectations by land
developers (Mills 1981).
Second, land use and land use changes are the outcomes of market interactions
among many heterogeneous individuals, which generate market and nonmarket feed-
backs that also influence these land use outcomes. These feedbacks may exist at mul-
tiple spatial and temporal scales. For example, individual households may buy and sell
houses depending on their own agreed-upon terms of trade, which are influenced by the
number of other buyers and sellers of similar houses on the market and by nonmarket
feedbacks, such as congestion externalities or agglomeration benefits. The implication
for modeling is that many so-called explanatory variables are likely to be endogenous in
a model of land use or land use change and that relevant feedbacks may exist at multiple
scales, which introduces additional challenges for empirical modeling.
Third, both the causes and impacts of land use change are cumulative and often irre-
versible, implying that dynamics over time are important. Economists typically use the
word “dynamics” to mean forward-looking behavior, whereas most noneconomists
use it to mean changes in state variables over time without regard to expectations. Both
types of dynamics are important in the land use modeling context. Forward-looking
behavior is a critical element of many economic land use models—for example, farmers
who make current planting decisions based on anticipated future prices of agricultural
commodities and forest managers who make optimal harvesting decisions based on
expected growth rates and future prices. Accounting for changes in land use over time
is an equally important modeling goal and one that is particularly important for policy
evaluation and scenarios.
3. A Comparison of Approaches
The complexities involved in modeling land use and land use change prevent any single
model from accounting for all the aspects of land use and land use change. Here we sum-
marize three main approaches to modeling land use and land use change: reduced-form
econometric models, structural econometric models and, for lack of a better term,
other spatial simulation models. We provide a synthesis of their strengths and weak-
nesses particularly as they relate to the ability of the modeling approach to capture the
critical aspects of land use modeling discussed above: heterogeneity, interactions, and
dynamics. Our interest is in models that are able to account for spatial heterogeneity and
interactions and therefore we focus on models that are able to capture heterogeneity and
interactions at either a parcel or neighborhood scales.
3.1 Reduced-Form Econometric Models of Land Use Change

Because land use is most often characterized as a categorical variable, estimation of land
use and land use change models using spatially disaggregated data requires a discrete
choice framework. The earliest reduced-form models of land use and land use change
focused on binary or multinomial discrete choice models of discrete land use or land
cover categories (e.g., Bockstael 1996; Nelson and Hellerstein 1997; Bockstael 1996).
Recognizing the importance of intertemporal decision making by landowners, the next
generation of spatially disaggregate models of land use focused on the optimal timing
decision of conversion at the parcel level (Irwin and Bockstael 2002; Newburn and Berck
2006; Towe et al. 2008). More recently, researchers have incorporated both discrete and
continuous change aspects of land use (Lewis et al. 2009; Lewis 2010; Wrenn 2012). For
example, conditional on the decision to convert a parcel to a residential subdivision, the
parcel owner must make a decision about the optimal number of buildable lots to create.
The outcome of both the discrete land use change and the density of development, which
is a continuous variable, have important impacts on many issues directly related to land
use and its spatial configuration, including ecosystem fragmentation and loss of farm-
land, urban sprawl, and optimal zoning and regulatory policy.
There are a number of reasons why researchers are only beginning to model these
joint decisions. First, modeling timing and intensity require a modeling technique that
can jointly model the two decisions and account for the necessary correlations between
them. Although multistage econometric models have existed for some time, land use
researchers are just beginning to adapt these to the particular behavioral processes
unique to land use change.
Second, much of the theory in the real options investment literature that has been
used to motivate econometric land use change models (Capozza and Helsley 1990; Dixit
and Pindyck 1994) is focused on the timing decision and does not account for variation
in parcel sizes, regulatory structures, or the potential for intensity differences between
parcels. Versions of the real options model that do account for timing, intensity, and
space provide theoretical predictions and testable hypotheses for both the timing
and intensity decisions, but those models have only recently been adapted to land use
questions.
And finally, a lack of data on the original decision-making parcel has made the
second-stage intensity decision irrelevant. Although geographic information system
(GIS) parcel-level data are used extensively in reduced-form models, researchers usu-
ally only have data on individual lots (children) and not on the original parcel (parent)
from which they were formed. In order to model this parent-child process, it is nec-
essary to combine individual parcels into their original parent parcel; this is particu-
larly important in the case of residential subdivision development where hundreds of
small lots can be produced from a single agriculture parcel. As a solution to this prob-
lem, researchers have combined plat maps with GIS parcel data as a means of determin-
ing the parent-child process. Using these new parcel-level panel datasets, models of the
optimal timing and intensity decision can be estimated.
Two recent papers provide good examples of how these new datasets can be com-
bined with joint econometric models to evaluate important land use policy questions.
In a model of land use and ecosystem change, Lewis (2010) combines a spatial-temporal
dataset of land use change with a Probit-Poisson model to estimate an econometric
model of timing and intensity. Using the coefficient values from the model, Lewis (2010)
generates land use change simulations that are then combined with an ecological model
to determine how current land use configurations and policies are impacting the likeli-
hood of extinction of green frogs. Because the survival rate of green frogs is determined
not just by the timing and location of development, but also by the intensity of develop-
ment, it is imperative that the second-stage intensity decision be modeled in the analysis
of this question. In his simulation model, Lewis shows that intensity indeed plays an
important role in determining the future probability of extinction for green frogs (see
also Chapter 15 by Plantinga and Lewis in this handbook).
In a Probit-Poisson model of residential subdivision development, Wrenn (2012) com-
bines historical subdivision data with data on regulatory delay over space and time to eval-
uate the role of spatiotemporal heterogeneity in explaining fragmentation and sprawl in an
urbanizing Maryland county. Real options theory predicts that delay in costs can impact
both the timing and intensity of an investment decision if the owner of the real option has
control over both the timing and the size of the project. In the land use context, this implies
that it is important to account for the effect of cost or regulatory delay on both the timing
and intensity decision. He finds that delay indeed impacts both decisions, reducing both
the probability of development and the optimal number of lots created. The simulations
also provide evidence that policy-induced differences in regulatory cost are contributing
to increased sprawl and fragmentation of the landscape.
These models provide improvements over previous reduced-form models, but still
have a number of limitations. First, although they do a good job of accounting for parcel
heterogeneity, they do not explicitly account for agent heterogeneity because the agent’s
characteristics are not modeled separately from those of the parcel. In the case of panel
data, individual random effects are used to account for some of this unobserved hetero-
geneity, but this technique does not explicitly separate out the individual characteristics.
Second, because the models are reduced-form in nature, they are not able to capture
general equilibrium feedbacks. As a result, endogeneity issues are dealt with using tra-
ditional econometric techniques, and analysis and simulation studies are confined to
marginal changes. Third, the models can account for out-of-sample predictions of land
use change by using the coefficient values from the models to simulate alternate land-
scapes over time. However, these predictions are limited to marginal changes because
the model does not account for feedbacks or forward-looking expectations during the
estimation process.2
To account for feedback effects and agent heterogeneity as well as model nonmar-
ginal changes in policy instruments, researchers have begun to apply structural models
to model the underlying demand and supply processes of land use.
3.2 Structural Econometric Models of Land and

Housing Markets
Structural models of land use include models of the demand or supply of land itself or of
output markets in which the derived demand for land as an input is modeled. For exam-
ple, models of agricultural production that consider land as a derived input generate
predictions of agricultural land use as the result of farmers’ optimal cropping decisions.
Empirical structural models of land use are not new; structural agricultural production
models that posited land use as the result of optimal decision making by a representative
farmer and are estimated using county-level data have a long tradition in agricultural
economics (e.g., Lichtenberg 1989; Wu and Segerson 1995; Plantinga 1996). However,
when richer models were made possible by the availability of spatially disaggregated
microdata (e.g., at the parcel level), new identification challenges were also introduced
(e.g., due to heterogeneity and selection effects) that made structural econometric mod-
eling more challenging. This led to a focus on reduced-form models with clear iden-
tification strategies. Nonetheless, some progress in estimating structural econometric
models with microdata has been made, most notably with models of household demand
for residential housing3 and, to a lesser extent, models of land and housing supply. These
2
Although reduced-form models do not model dynamics in the traditional sense, with
forward-looking agents forming expectations and maximizing discounted present values of uncertain
future payoffs (e.g., Rust 1987), some researchers have built individual predictions of future economic
variables into their models using time-series techniques to capture the dynamics (Cunningham 2007;
Bulan et al. 2009).
3 For a full review of equilibrium sorting models and their connection to hedonic models and models
of differentiated products, see Chapter 14 in this handbook by Klaiber and Kuminoff and reviews by
Klaiber and Smith (2011) and Kuminoff, Smith, and Timmins (2012).
models have been usefully applied to simulate policies or the effect of other hypothetical
nonmarginal changes that may push the system to a new equilibrium. Here we focus our
discussion on structural modeling of urban land and housing markets with an emphasis
on how these models account for endogenous price and other market and equilibrium
feedbacks.
The basic assumption of all structural models is that the land and housing markets are
in a spatial equilibrium. Spatial equilibrium conditions differ depending on the model
assumptions. For example, in a residential land use model with homogeneous house-
holds, such as the canonical monocentric model, spatial equilibrium is characterized by
equal utility across space since any advantage or disadvantage of a location is capitalized
into its price. Given heterogeneity in preferences or incomes, as is the case with house-
hold sorting models, spatial equilibrium in the residential land market is characterized
by each household having made an optimal decision, given the location and supply deci-
sions of all other agents. In other words, it is not possible for any agent to make himself
better off by making a different choice. As with the homogeneous case, the equilibrat-
ing element of this process is price. The resulting spatial equilibrium is often character-
ized by a hedonic function that is comprised of the market-clearing bid and offer curves
of heterogeneous agents (Rosen 1974). Although Rosen (1974) provides a thorough
description of how the hedonic equilibrium is achieved, he does not make explicit how
these bids and offers lead to a spatial equilibrium. Tiebout (1956) was the first to observe
that it is the process of heterogeneous agents sorting themselves across differentiated
neighborhoods that determines the spatial equilibrium. In this model, the larger region
is subdivided into many heterogeneous or differentiated locations; as agents sort, they
bid on different neighborhoods, and prices are determined by the intersection of the
bid and offer curves of the agents moving to each neighborhood. Households choose a
neighborhood, conditional on their budgets, to maximize utility, and developers pro-
vide housing, given technological and regulatory constraints, to maximize profits. The
sorting process continues until no one has the incentive to move and the region reaches
a spatial equilibrium.
It is the application of the Tiebout (1956) theory to housing markets that allows
researchers to specify and estimate equilibrium sorting models. By combining the infor-
mation provided by the hedonic equilibrium (Rosen 1974; Ekeland et al. 2004) with a
description of the choice process that leads heterogeneous agents to sort (McFadden
1974; Berry et al. 1995), researchers are able to estimate the structural parameters of the
model; that is, it is the information revealed by households about their preferences for
different neighborhoods that allow researchers to specify and estimate the structural
parameters of the equilibrium sorting model and to characterize the heterogeneity of
preferences for local public goods and amenities. The equilibrium sorting framework
combines a mixture of discrete (neighborhood) choices with continuous choices (char-
acteristics of the houses). The approach makes explicit how preference-induced sorting
can lead to endogenous feedbacks that make it very challenging to use estimates from
reduced-form land use models to determine amenity values (Epple and Sieg 1999; Bayer
and Timmins 2005).
In some cases, the supply of amenities can be taken as exogenous, and households
sort to take advantage of these amenities. Often, however, the amenities in a particular
area are either completely determined by the sorting process or significantly affected
by it. As an example, one neighborhood may have a particularly nice open space area.
As households move to this area, they bid up the price. If higher house prices result in
greater tax revenues for schools, then the attraction of the open space area can lead to
increased school quality, which leads to additional feedbacks. As more people move in,
it can also lead to a reduction in the quality of the open space areas as they become con-
gested or as the actions of residents degrade the quality of them. Thus, even if the origi-
nal amenity was exogenous, its future value and the values of other public goods may be
determined by the sorting process.
Because many environmental amenities are not explicitly traded, researchers have
used the theory and econometrics of hedonic and discrete-choice models of differenti-
ated product markets to uncover willingness to pay (WTP) for various amenities from
local housing values. The theory of capitalization (Oates 1969) provides evidence that
house prices will reflect the value of both the structural characteristics of the house as
well as the local public goods and amenities in the house’s neighborhood. By observing
the outcome of the sorting process and the resulting prices of houses sold across differ-
ent neighborhoods, researchers can estimate values of households’ WTP for the public
goods and environmental amenities in those neighborhoods.4 Because the models are
structural, they can also be used to simulate counterfactual policy analysis to analyze the
costs and benefits of alternative policies.
Given the clear advantages of this equilibrium sorting framework, it is important to
highlight both what these models do well and what they do not do so well in terms of
land use modeling. Unlike reduced-form models, structural models capture agent het-
erogeneity (separate from neighborhood heterogeneity) by including information on
income, household composition, education, and other demographic factors that are
hypothesized to affect preferences. These data are interacted with neighborhood-level
amenity values, which allows the value placed on local public goods and amenities to
vary with the characteristics of the individual household. One disadvantage, however, is
that because structural models must be estimated at the neighborhood level, they miss
much of the spatial heterogeneity at the parcel level that is captured in reduced-form
models.5
4
This outcome relies on a number of assumptions that may not hold in every context, such as being
able to capture and measure all the relevant amenities as well as find suitable instruments that can control
for the unobservable amenities that cannot be measured.
5
To achieve consistency and efficiency in equilibrium sorting models, the number of households
or decisions makers must rise faster than the number of choices (Berry et al. 2004). Thus, the choice
set in these models includes neighborhoods, which are geographically aggregated areas that have been
constructed to be internally homogeneous but differentiated relative to other locations. Even assuming
consistent segregation of the study area into neighborhoods, it is still not possible, for technical reasons,
to estimate these models at the parcel level.
Another advantage of these models is their ability to account for market and non-
market feedbacks. Researchers first specify a utility or indirect utility function that
reflects the fundamental preference relationships from microeconomic theory (i.e.,
monotonic and convex preferences). Then, using data on the individual characteristics
of the agents and their respective choices, they estimate the structural parameters of
the function. Because these estimates reflect the underlying preference structure of the
agents and how they respond to changes in market fundamentals, the model can be used
to examine counterfactual policy scenarios in which large-scale nonmarginal changes
occur, which cause a resorting of agents and a shift the market equilibrium. Although
this is a clear advantage from the policy analysis perspective, these feedbacks and their
predictions for land use change are only conducted at the neighborhood scale or higher.
In most cases, this is sufficient, but smaller scale analysis of feedbacks and land use
change may be necessary for certain research questions, such as those related to land
fragmentation, loss of biodiversity, and the impact of land use change on ecosystems.
In addition to their lack of spatial disaggregation, another disadvantage of current
structural models is that they are largely static. By static, we mean that they do not
account for the formation of expectations by the agents about future house price values
and local neighborhood amenities and how individual decisions affect those values.
A variety of papers apply equilibrium sorting models to policy questions, but only
a few look explicitly at issues of land use. We highlight two of the most recent exam-
ples: Walsh (2007) and Klaiber and Phaneuf (2010). Both of these papers deal with the
issue of open space allocation and valuation but use different modeling techniques. We
explain the key differences between the two and highlight their key findings.
Walsh (2007) applies a vertical or “pure characteristics” sorting model (Epple and Sieg
1999) to the question of the equilibrium impacts of open space protection and growth
control policies. As is the case with all equilibrium sorting models, the larger region is
delineated into exhaustive and mutually exclusive neighborhoods. The vertical sorting
model assumes that all households place equal weight on the value rank of the neighbor-
hoods but have heterogeneous preferences for those neighborhoods as a result of dif-
ferences in personal characteristics. The main implication of this vertical specification
is that households, when they substitute, only consider those neighborhoods that are
directly adjacent to them in preference and income space.
Using this model, Walsh (2007) evaluates open space policy with an approach that
incorporates the endogenous formation of private open space and residential land.
Each neighborhood has two types of open space—public open space, which he assumes
is exogenously provided, and private open space, which is provided as a result of the
development decisions of households and land developers. Households value their rela-
tive location to the public open space and trade off this type of open space with private
open space relative to their distance from it. The key finding of this paper is that if one
allows for the endogenous adjustment of private open space, then an increase in public
open space can actually lead to a reduction in overall open space in the region. This
results stems from the fact that as public open space in a neighborhood increases, it both
increases the demand for the location and reduces the amount of private open space
needed as households have better access to the public amenity. The result is an overall
reduction in land that is preserved by a policy that was designed to do just the opposite.
Klaiber and Phaneuf (2010) apply a random-utility, horizontal sorting model (Bayer
et al. 2004) to a similar research question related to open space valuation and protec-
tion but allow the preference structure and types of open space to be more flexible.
Unlike the vertical sorting model, horizontal models allow preferences for alterna-
tives, as well as the weights of the individual rankings of those alternatives, to vary. For
example, in the vertical model, all households would assign the same neighborhood
ranking to good school districts regardless of the characteristics of the household, and
household-specific preferences would determine their actual choice. In the horizontal
model, the relative ranking of each choice is not the same for all households and instead
depends on the characteristics of the household. For example, households with young
children are likely to rank neighborhoods with good schools higher than would senior
citizens.
The innovation of the Klaiber and Phaneuf model, in terms of land use and open space
nonmarket valuation and allocation, is to allow not only the preferences and ranking of
neighborhood alternatives to vary by household type, but also to allow open space ame-
nities to be disaggregated into multiple categories. The paper also uses data that allows
for higher spatial resolution than past work. This paper provides several important find-
ings. First, they find that distinguishing among different types of open space is critical in
determining the tradeoffs people make during the sorting process. They find that differ-
ent types of open space amenities are valued differently depending on the characteristics
of the household. Second, they find that as the scale of the open space policy increases,
the welfare estimates between partial and general equilibrium models diverge.6 Finally,
they show that localized and targeted open space policies are more likely to be efficient
than region-wide policies that do not account for household heterogeneity. This result
follows from the fact that the interactions between household characteristics and differ-
ent types of open space provide drastically different WTP values.
Like the reduced-form models reviewed in the previous sections, most equilibrium
sorting models to-date are static in the sense that they do not take into account the for-
mation of expectations by the agents in terms of prices, costs, and other amenity vari-
ables. However, given that location choice and land use conversion are either irreversible
or difficult to reverse, this implies that agents, when choosing to relocate (demand) or
develop (supply), consider not just current market conditions, but also take into account
the current and expected future value of all state variables. For example, conventional
static sorting models assume that agents move freely between locations without regard
for technological, institutional, or social constraints. In the real world, however, agents
6
This finding is particularly important in the analysis of land use policy. For marginal land use policy
changes, it indicates that partial equilibrium or reduced-form analysis may be sufficient to provide
insight into marginal willingness to pay values of the policy. However, when policy changes are such
that they induce a resorting of households, then it is likely that reduced-form analysis will not provide
accurate welfare measure.
consider moving costs, wealth constraints resulting from changes in home equity values,
and social ties to current locations when deciding to move from one place to another.
These constraints influence the sorting equilibrium and the estimation of household
demand for locational attributes. Bayer et al. (2009), for example, show the importance
of accounting for moving costs in estimating households’ demand for air quality.
Unlike static choice models, dynamic discrete choice models take account of both
the evolution of the model’s state variables and how agents form expectations about
future values of these variables. They may also account explicitly for the interactions
between agents. The parameters are estimated in these dynamic models by nesting a
Bellman-style optimization equation in an empirical optimization technique. The
results of these models describe agent’s preferences and beliefs about technological
and intuitional constraints as they evolve over time. One of the main issues with these
models is the curse of dimensionality and the fact that the complexity of the estimation
techniques required to solve these models increases rapidly with the number of state
variables or interactions among agents (Aguirregabiria and Mira 2010). Each state vari-
able is endogenously determined within the dynamic optimization framework, and
agents are assumed to form expectations over each of these variables in each period.
Thus, as the size of the state space increases so does the complexity of the empirical
estimation process, making large state space models over many periods empirically
infeasible.
Two recent papers have attempted to estimate structural models with forward-looking
expectations. Bayer et al. (2011) extend the structural household sorting model to
include the formation of expectations about prices and local amenities and show that
including these in the model has a substantial effect on the marginal WTP values relative
to those computed using the static model.7 They find a significant divergence between
the predictions of the dynamic model from those of the static sorting model. The main
factor driving the wedge between the values in the two models is the introduction of
moving costs and of wealth formation from capital gains in house values, which impact
the budget constraint apart from income.
Murphy (2013) estimates a dynamic model of housing supply that takes into account
both variable and fixed costs, as well as uncertainty over future land use regulations.
It is has been shown by a number of authors (Mayer and Somerville 2000; Glaeser
et al. 2005) that land use regulations, when they translate into a reduction in the elas-
ticity of housing supply, can lead to significant increases in home prices. In this paper,
Murphy captures local cost uncertainties in the development process using a dynamic
fixed effects technique and finds that the majority of the rise in home prices in the San
Francisco Bay area over his study period could be attributed to regulatory effects that
reduce the ability of the supply side of the market to respond to demand. This result has
7 They give the example of a neighborhood in which the change in environmental quality or crime
is expected to be drastic in the future. As a result, households may be more willing to purchase in these
neighborhoods at a lower price if the future expected improvements mean larger expected capital gains.
important implications for policy in the area of housing supply and provides further
evidence of the impact of land use regulations on home prices and supply responses to
demand.
3.3 Other Spatial Simulation Models of Land Use Change

Because of data demands, necessary assumptions, and complexity of the modeling
techniques that are typically needed, it is not always be possible to estimate a structural
econometric model. This is particularly true when the interest is in accounting for the
sequencing of land use changes over time, for example, due to growing population in a
region. In such cases, a static spatial equilibrium assumption is problematic since the
long-run spatial equilibrium land use pattern is assumed to be instantaneously reached,
and the sequencing over time or space of individual agents’ decisions is not determined
(Chen et al. 2011).
Spatial simulation models have emerged as a means of representing land markets in
a spatially explicit framework that can better capture changes in land use patterns over
time. Although a variety of approaches to spatial simulation exist, these models have in
common the approach of a spatially explicit framework in which changes in land use
patterns over time are simulated as the result of individual-level decisions regarding
land use or location. Many different types of land use models may be adapted to a spatial
simulation framework. For example, as discussed in Chapter 15 by Plantinga and Lewis
in this handbook, coefficients from land use models estimated with aspatial plot-level
data on land use, land characteristics and net returns have been used to simulate pre-
dicted spatial patterns of land use under alternative policy scenarios by applying these
estimates to spatially disaggregated land use data. And, as we have already discussed and
is further detailed in Chapter 14 by Klaiber and Kuminoff in this handbook, structural
econometric models can be used to simulate changes in residential location and land
use at a neighborhood scale. Because these empirical simulation methods are discussed
in detail in these other chapters, here we focus on two other types of spatial models that
are used in simulation: (1) spatial equilibrium models that account for dynamics by
incorporating some exogenous change over time, such as population or income growth
and (2) agent-based models that focus on individual market trades in the absence of an
overall market equilibrium assumption.
Although the strength of these types of spatial simulation models is in representing
spatial heterogeneity and capturing some forms of dynamics, their weakness is in model
specification. In the absence of a fully structural econometric estimation approach,
empirical specification of a spatial simulation model may proceed in various ways. For
example, reduced-form parameters may be estimated using datasets taken from different
sources or settings. Secondary data on observable outcomes, including land use, prices,
and other variables that influence land or housing markets, or primary data collected
from surveys or lab or field experiments may be used. In other cases, estimates reported
in the literature may be used to specify key parameters or reasonable ranges of these
parameters, for example, demand or supply elasticities. Once specified, simulation of

these spatial simulation models can proceed in an analogous fashion to that of structural
econometric models by using the empirically specified structural equations to predict
land use outcomes, given a policy shift or other changes. A limitation of this approach to
empirical specification is that the parameters may reflect different underlying conditions
of demand and supply that are inconsistent with any single equilibrium (or, more gener-
ally, with any single observable outcome). Unobserved differences in macroeconomic
variables or other constraints that differ across different regions, scales, or time periods,
for example, can cause differences in the underlying demand and supply processes.
Spatial equilibrium simulation models are based on an assumption of instanta-
neous adjustment of prices to a spatial equilibrium. Spatial equilibrium prices evolve
over time, given exogenous changes in population, income, or some other variable.
Agents may be myopic, in which case the equilibrium is conditional on current levels,
or forward-looking, in which case spatial equilibrium is determined by agents’ expecta-
tions over future growth. The model may account for market frictions, such as informa-
tional asymmetry, credit constraints, construction lags, search costs, which may prevent
instantaneous adjustment to the long-run spatial equilibrium. Thus, it is possible to
draw a distinction between a short-run equilibrium, in which market frictions create
a binding constraint, and an unconstrained long-run equilibrium. Given an analytical
expression for price as a function of heterogeneous space and specified parameter val-
ues, simulation methods can be used to predict how growth over time generates changes
in spatial equilibrium land use patterns.
Although there are very good examples of structural spatial simulation mod-
els that consider how spatially heterogeneous costs or amenities influence land use
patterns, many are static and describe only the long-run spatial equilibrium in the
absence of growth (e.g., Wu and Plantinga 2003; Tajibaeva et al. 2008). The canonical
urban economic model with growth is Capozza and Helsley (1989), in which the influ-
ence of deterministic population growth on the land value gradient is considered in a
one-dimensional monocentric model. However, because their interest is in characteriz-
ing how growth influences land values, the authors do not consider changes in the land
development pattern over time. Instead they solve for an analytical expression of long
run spatial equilibrium land prices to show how the market value of developed and agri-
cultural land depends on the growth rate. Newburn and Berck (2011) extend this model
by simulating it across time for a stylized exurban region with population growth. Their
model provides a good example of how a spatial equilibrium model, which is dynamic
in terms of how population growth influences land rents and development decisions,
can be simulated and made to “step through time” to generate spatial predictions of
land use change over time. Household behavior is specified by a Cobb-Douglas util-
ity function in which households are assumed to trade-off a composite good and lot
size. Lots may be either small (suburban) or large (exurban) and household preferences
over these lot sizes are heterogeneous. Deterministic population growth is capitalized
into land rents and leads to a contiguous expansion of suburban development over
time. However, given differences in the development costs of suburban versus exurban
lots and preference heterogeneity in which some households have a greater preference
for large lots, leapfrog development can emerge in the exurban region located beyond
the suburban fringe. The resulting spatial equilibrium land rents and other model out-
puts, which include each household type’s optimal location choice, the rate of suburban
expansion, and the conditions for exurban leapfrog development, are expressed analyti-
cally. The model is then simulated to explore the magnitude of the effects of sewer and
commuting costs on suburban development and how city size influences the extent of
exurban leapfrog development. The simulations are performed in discrete time, with an
exogenously determined population growth rate for each household type. An iterative
approach is used to solve for equilibrium in each time period in both the suburban and
exurban land markets. In each period, for a given level of population of each household
type, initial values of the proportion of suburban and exurban land at each distance are
assumed. The model equations that characterize the spatial equilibrium are then used to
iterate between prices and these relative proportions until the predicted boundary of the
suburban area converges to a constant value.
The clear strength of this approach—simulation of spatial equilibrium models with
some source of exogenous growth—is that these are structural models of the land mar-
ket in which prices are modeled in a theoretically consistent manner that reflects both
individual-level preferences and constraints and market-level conditions. A primary
limitation of this approach is that the degree of spatial dynamics or agent and spatial
heterogeneity that can be considered is limited since an analytical expression for spatial
equilibrium prices is usually necessary to close the model.
The second approach, agent-based modeling,8 does not impose spatial equilibrium
and instead focuses on individual trades in the absence of this assumption. The essential
features of these models are typically heterogeneous agents, defined by a set of behav-
ioral rules and their interactions that evolve the system over time. Given a set of detailed
initial conditions (e.g., that fully specify the institutional arrangements, initial number
and types of consumers and firms, endowments, decision-making and trading rules,
geography), agents carry out production, pricing, and trade activities that generate feed-
backs (e.g., profits, utility, learning) that determine future decisions. A key departure
is the lack of an equilibrium constraint: given the initial specifications of the economic
system, the dynamics are driven solely by agent trading that is typically not subject to a
market-level equilibrium condition.
The primary advantage of agent-based modeling is that many more details can be
incorporated into the model, such as greater spatial disaggregation and heterogeneity
at an individual agent and parcel scale. However, because the equilibrium assumption
is dropped, these advantages come at the expense of additional model complexity that
8
See the Chapter 16 by Parker in this handbook for a much more in-depth discussion and for
many examples of this modeling approach. See Parker and Filatova (2008) and Chen et al. (2011) for
further discussion of agent-based models of land markets and Irwin (2010) for further comparison of
agent-based and other economic modeling approaches.
is necessary to specify the bidding and market interaction processes. More importantly,
this approach is not consistent with the basic spatial equilibrium theory of land markets
and as a result, represent market feedbacks in an ad hoc manner.
Filatova, Parker, and van der Veen (2009a) and Magliocca et al. (2012) provide good
examples of how agent interactions can be modeled in an agent-based framework by
specifying household offer bids, seller ask bids, and the interactions between individual
buyers and sellers. Filatova et al. (2009a) specify the household’s WTP as a function of
income net of transportation costs and expenditures on a composite good, household
utility, and the price of the composite good. The functional form used to specify the
WTP function is ad hoc but reflects standard demand relationships, such as increasing
WTP with income. The landowner’s willingness to accept (WTA) is given by the reser-
vation rent, assumed to be equal to agricultural land rents. Buyers and sellers interact via
a specified sequence of events that includes sellers announcing their WTA bids and buy-
ers searching for the location that generates the largest surplus. To account for market
feedbacks, they adapt an approach used in agent-based finance models (LeBaron 2006),
in which the individual WTP and WTA bids are adjusted by a multiplicative factor
(1 + ε), where ε = (NB − NS)/(NB + NS), NB = number of buyers and NS = number
of sellers. This is admittedly ad hoc, but allows bidding to be adjusted based on agent
perceptions of market conditions. Given positive gains from trade, then the transaction
price is set assuming that the buyer and seller divide these gains equally.
Although this model and related work by Filatova and Parker (e.g., Parker and Filatova
2008; Filatova et al. 2009b) is innovative, a limitation of this approach is that the bids are
not explicitly derived from a specific utility maximization model and, as a result, assump-
tions about the microeconomic foundations of the model, including the substitutability
between location and the composite good, are not made explicit. Magliocca et al. (2012)
improve on this by deriving the household’s WTP function from a Cobb-Douglas utility
function and the developer’s derived demand for land from an expected profit function.
In addition, they separately model land and housing markets, an innovation that is not
usually done, and consider other sources of heterogeneity, including multiple housing
types, variable minimum lot zoning, and heterogeneous expectations among landown-
ers. However, they use the same ad hoc adjustment procedure as Filatova et al. (2009a) to
account for market competition in which the WTP and WTA bids are adjusted based on
the relative number of buyers and sellers.
Several papers have combined agent-based modeling with an equilibrium assump-
tion at each point in time as a means of incorporating spatial equilibrium into an
agent-based framework. For example, Caruso et al. (2007) consider the emergence of
different forms of residential sprawl as the result of endogenous neighborhood ameni-
ties in a two-dimensional urban economic model. Two types of neighborhood ameni-
ties are considered, both of which are locally defined as a function of the neighborhood
land use pattern at a given location: open space amenities, which decrease with the
amount of nearby development, and social amenities, which increase with nearby devel-
opment. Households make optimal location decisions by trading off these competing
neighborhood amenities with the travel costs associated with a given location. When
the benefits of surrounding open space outweigh the costs of travel and lower social
amenities, households will find it optimal to locate away from the urban fringe and in
so doing generate sprawl. Given homogeneous households that are myopic (i.e., they
do not anticipate population growth), short run equilibrium land rents are determined
by the equalization of utility across space in a given time period and are conditional on
the population level in each time period. Land rents are bid up over time as additional
households enter the region and reach a long run spatial equilibrium when utility inside
and outside the region are equalized.
Chen et al. (2013) follow a similar approach to incorporating population growth into
a model of leapfrog development, but use a different approach to modeling household
bidding. They begin with a Cobb-Douglas model of utility with households that are het-
erogeneous in income and a highly stylized landscape that is distinguished by distance
from the urban center. The main innovation of this paper is that the household’s opti-
mal bidding function is derived as a function of preferences, income, and the expected
number of competing bidders relative to supply using an auction model. Specifically, a
first-price, sealed auction model is used to derive an analytical expression for the agent’s
optimal bid for location, which maximizes the household’s expected surplus associ-
ated with a given location (defined as the difference between its maximum WTP and
its actual bid) multiplied by the probability of winning. This approach incorporates key
market conditions, namely the number of households in the region, the distribution of
income and supply of land at each location, that, along with travel costs, determine the
competitive level of bidding for each land parcel. Landowners determine market land
rents by selecting the highest bid and development occurs if this bid exceeds the land-
owners’ opportunity cost. Given heterogeneous households that are myopic, the short
run equilibrium is defined by the optimal choice of each household and landowner, the
lack of incentive for agents to renegotiate, and the current population and income dis-
tribution. Market conditions change over space due to income differences and over time
as additional households enter the region, which increases competition and bids away
the surplus that a household can attain at any given location. Changes in surplus permit
the model to “step through time” by providing a temporal and spatial ordering of the
location choices of heterogeneous households based on their utility-maximizing loca-
tion decisions. The model is implemented using simulation methods that allow each
household type and location to be explicitly considered in the bidding process, and
therefore agent and spatial heterogeneity can readily be incorporated. Given this specifi-
cation of the bidding process, Chen et al. (2013) hypothesize that leapfrog development
can emerge if households are able to retain a larger surplus at more remote locations
due to fewer bidders and a greater supply of substitutable land at these distances. The
main result of the paper confirms this hypothesis. However, leapfrog development is
ultimately a short-run property of the model since, over time, as more households move
into the exurban region, the spatial differences in surplus are gradually bid away.
Like the assumption of a spatial equilibrium, the assumptions about price forma-
tion and market interactions that are necessary for agent-based modeling of land and
housing markets are maintained assumptions. Unlike spatial equilibrium models,
however, they usually omit any kind of equilibrium constraint and thus are incon-
sistent with a standard microeconomic model of behavior. In addition, although
assumptions may be motivated by stylized facts about the trading process, they are
often difficult to test empirically. The perceived lack of a stronger theoretical or empir-
ical basis for modeling price formation has generated skepticism among some econo-
mists regarding the efficacy of agent-based modeling for modeling markets. However,
as the advances in modeling just noted illustrate, it is possible to develop agent-based
models that are derived from microeconomic foundations and that incorporate some
notion of a spatial equilibrium. Although unsolved challenges regarding questions
of dynamics and spatial equilibrium remain, we believe that recent work in both
agent-based modeling and in other simulation models that are more prevalent in eco-
nomics are pushing these modeling methods closer together. This is a point that we
discuss in the Section 4.
4. Conclusion
In this chapter, we have sought to provide an overview and assessment of empirical

methods for modeling land use by addressing the following questions: (1) what are the
advantages and disadvantages of these various empirical approaches to modeling land
use and land use change; (2) which questions are best suited to be answered using one
versus the other approach, and (3) where are the gaps in the current literature? In draw-
ing conclusions, we reiterate that there is no single “right” modeling method and that,
instead, the appropriate method depends on the research question, modeling goals, and
available data. Although in many cases a specific modeling method is appropriate and
will suffice, we are also encouraged by the complementarities that may be possible by
using a combination of methods. Here, we see at least three compelling opportunities.
First, it may be possible to usefully combine reduced-form parameter estimates with a
structural econometric model. For example, Chetty (2009) uses a structural model to
derive expressions for the welfare consequences of various policies that are functions
only of high-level elasticities, which can be estimated with reduced-form models. He
then combines these reduced-form parameter estimates with a structural model to con-
duct counterfactual policy simulations. Whereas Chetty is focused on the welfare effects
of policies, it may also be possible to use an analogous approach to modeling land mar-
kets if it is possible to use the structural model to identify opportunities for simplify-
ing the number of structural parameters that must be estimated and instead estimate
one or more reduced-form parameters. This approach could be useful for estimating
the parameters of aggregate-level demand and supply relationships, for example, at the
neighborhood, county, or metropolitan scales.
Second, we see complementarities between the structural spatial equilibrium mod-
els, including the structural econometric models and other types of spatial equilib-
rium simulation models, and agent-based models. In some ways, these methods are
converging as more microeconomic foundations are included in agent-based models

and as more heterogeneity, disaggregation, and dynamics are introduced into spatial
equilibrium models. There are clear gains to be had from continuing to work toward
narrowing the gap between these two approaches. An eventual goal is a modeling
framework that includes the disaggregation and dynamics of the agent-based models
and a theoretically consistent specification of spatial equilibrium and market feedbacks
from the economic structural models. In addition, because of their added flexibility,
agent-based models may be useful in testing the maintained assumptions of economic
structural models, such as comparing model predictions from long-run spatial equilib-
rium with short-run equilibrium.
Third, we see complementarities between the spatial modeling approaches that we
have focused on here and general equilibrium models in which other sectors of the
economy are considered. Most models of housing or land markets ignore other input or
output markets and thus are limited to considering partial equilibrium effects on land
or housing rents or land use outcomes. However, recent work illustrates the importance
of accounting for feedbacks from related markets. Kuminoff (2008) accounts for the
jointness of households’ residential location and employment decisions in an integrated
model of housing and labor markets. Desmet and Rossi-Hansberg (2010) illustrate how
feedbacks across multiple input and output markets matter for regional spatial dynam-
ics and growth. Combes et al. (2005) account for feedbacks across multiple input and
output markets and demonstrate the importance of accounting for these multiple chan-
nels of adjustments for policy analysis.
Although we are encouraged by this work that integrates multiple models or mod-
eling approaches to address current research gaps, other critical gaps in modeling
also need attention. First, there is a pressing need for empirical testing of the many
maintained assumptions that must be made to make models tractable. As data and
computational power increase, so does the scope, complexity, and number of main-
tained assumptions of our models. Kuminoff (2009) provides an example of testing
the implications of maintained assumptions of functional form, preference distribu-
tions, and neighborhood delineation for identification and welfare measurement in
structural models of household locational choice. He also shows how uncertainty
regarding functional form and distributional assumptions can be measured and
assessed. More work along these lines is needed. This need applies equally to struc-
tural econometric and other structural empirical models, as well as to agent-based
models that rely on a number of maintained assumptions about the agent bidding
and market interactions processes.
Second, incorporating dynamics into location and land use choice models remains
a substantial challenge. The curse of dimensionality presents a key challenge since
the complexity of the model increases as the number of state variables or interactions
among agents increases, and estimation of large state space models over many peri-
ods is empirically infeasible. Spatial simulation techniques can fill this gap. Because
spatial simulation models are not limited by the same estimation constraints, they
can allow researchers to build land use models that are based on theory and that
allow for more complex interactions and dynamic feedbacks by accounting for more
state variables. If this translates into a more realistic model of real-world outcomes,
then there is a clear advantage in using these models in modeling dynamic land use
outcomes provided that the parameter and functional form assumptions can be
empirically verified.
Finally, one of the ultimate tests of any modeling approach is how useful it is for
policy. As applied researchers, we often conclude with the policy implications of
our findings. However, the number of times that academic articles actually have an
impact on the policy process is limited at best. Policy makers operate under sub-
stantial time and resource constraints and are usually looking for a quick approxi-
mate answer rather than a precise estimate that takes time to produce. Even when
we design our models to run policy scenarios, we may not capture the realities of the
political or policy process that determines the feasibility of a particular policy or pol-
icy change. The criteria by which the usefulness of our models are judged for real-time
policy decisions are different from the criteria we would choose for their academic
evaluation, but yet the optimal tradeoff between academic rigor and real-world prac-
ticality is uncertain. Certainly, rigorous policy analysis requires more than mod-
els that just generate pattern predictions, which implies that some representation
of agents’ demand and supply decisions is needed. However, the tradeoffs among
models that are fully structural, incorporate dynamics and multiple feedbacks, and
account for interactions across multiple scales remain uncertain. Much more work
is needed to assess the relative costs and benefits of different modeling approaches
for policy analysis and the relative gains from added model complexity versus the
time and opportunity costs that these innovations require. Unfortunately, such ques-
tions have received scant attention in the literature—in large part due to publishing
incentives that academic economists face that reward novel results and methods far
more than replication or model comparison. Replication and cross-site comparisons
are emphasized much more in other scientific disciplines. We see value in encourag-
ing this kind of research in land economics that could synthesize and assess model
results and modeling methods in a way that is useful for non-Ph.D. economists and
policy makers.
Acknowledgments
We would like to thank Josh Duke for very helpful feedback on an earlier draft of
this chapter. Support from the National Science Foundation (DEB LTER-1027188,
WSC-1058059, GSS-1127044), the U.S. Forest Service’s Northern Research Station, and
the James S. McDonnell Foundation is gratefully acknowledged.
References
Aguirregabiria, V., and P. Mira. 2010. Dynamic discrete choice structural models: A survey.
Journal of Econometrics 156: 38–67.
Anas, A., A. R. Arnott, and Small, K. 1998. Urban spatial structure. Journal of Economic
Literature 36: 1426–1464.
Bayer, P., R. McMillan, and K. Reuben. 2004. What drives spatial segregation? New evidence
using census microdata. Journal of Urban Economics 56: 514–535.
Bayer, P., N. Keohane, and C. Timmins. 2009. Migration and hedonic valuation: The case of air
quality. Journal of Environmental Economics and Management 58(1): 1–14.
Bayer, P., R. McMillan, A. Murphy, and C. Timmins. 2011. A dynamic model of demand for
houses and neighborhoods. Unpublished manuscript available upon request from authors.
Bayer, P., and C. Timmins. 2005. On the equilibrium properties of locational sorting models.
Journal of Urban Economics 57, 462–477.
Berry, S., J. Levinsohn, and A. Pakes. 1995. Automobile prices in market equilibrium.
Econometrica 63: 280–296.
Berry, S. T., O. Linton, and A. Pakes. 2004. Limit theorems for estimating the parameters of dif-
ferentiated product demand systems. Review of Economic Studies 71: 613–654.
Bockstael, N. 1996. Modeling economics and ecology: The importance of a spatial perspective.
American Journal of Agricultural Economics, 785: 1168–1180.
Bulan, L., C. Mayer, and C. Somerville. 2009. Irreversible investment, real options, and com-
petition: Evidence from real estate development. Journal of Urban Economics 65: 237–251.
Capozza, D., and R. Helsley. 1989. The fundamentals of land prices and urban growth. Journal
Capozza, D., and R. Helsley. 1990. The stochastic city. Journal of Urban Economics 28: 187–203.
Caruso, G., D. Peeters, J. Cavailhes, and M. Rounsevell. 2007. Spatial configurations in a peri-
urban city: A cellular Automata-based microeconomic model. Regional Science and Urban
Economics 37(5): 542–567.
Chen, Y., E. G. Irwin, and C. Jayaprakash. 2011. Incorporating spatial complexity into eco-
nomic models of land markets and land use change. Agricultural and Resource Economics
Review 40(3): 321–340.
Chen, Y., E. G. Irwin, and C. Jayaprakash. 2013. Thin markets and leapfrog development.
Unpublished manuscript available upon request from authors.
Chetty, R. 2009. Sufficient statistics for welfare analysis: A bridge between structural and
reduced-form methods. Annual Review of Economics 1: 451–488.
Combes, P. P., G. Duranton, and H. G. Overman. 2005. Agglomeration and the adjustment of
the spatial economy. Papers in Regional Science 84(3): 311–349.
Cunningham, C. 2007. Growth controls, real options, and land development. The Review of
Economics and Statistics 89: 343–358.
Desmet, K., and E. Rossi-Hansberg. 2010. On spatial dynamics. Journal of Regional Science
50(1): 43–63.
Dixit, A., and R. Pindyck. 1994. Investment under uncertainty. Princeton, NJ: Princeton
University Press.
Ekeland, I., J. Heckman, and L. Nesheim. 2004. Identification and estimation of hedonic mod-
els. Journal of Political Economy 112: S60–S109.
Epple, D., and H. Sieg 1999. Estimating equilibrium models of local jurisdictions. Journal of
Political Economy 107: 645–681.
Filatova, T., D. P. Parker, and A. van der Veen. (2009a). Land market interactions between het-
erogeneous agents in a heterogeneous landscape: Tracing the macro-scale effects of indi-
vidual trade-offs between environmental amenities and disamenities. Canadian Journal of
Filatova, T., Parker, D., and A. van der Veen. (2009b). Agent-based urban land markets: Agent’s
pricing behavior, land prices and urban land use change. Journal of Artificial Societies and
Social Simulation, 12(1) http://jasss.soc.surrey.ac.uk/12/1/3.html
Glaeser, E., J. Gyourko, and R. Saks. 2005. Why is Manhattan so expensive? Journal of Law and
Glaeser, E., and M. Kahn. 2004. Sprawl and urban growth. In Handbook of regional and urban
economics, Vol. 4, eds. J. V. Henderson and J.-F. Thisse, 2481–2527. Amsterdam: Elsevier.
Harvey, E. O., and W. Clark 1965. The nature and economics of urban sprawl. Land Economics
41: 1–9.
Irwin, E. G. 2010. New directions for urban economic models of land use change: Incorporating
spatial dynamics and heterogeneity. Journal of Regional Science 50: 65–91.
Irwin, E., and N. Bockstael. 2002. Interacting agents, spatial externalities and the endogenous
evolution of residential land use patterns. Journal of Economic Geography 2: 31–54.
Keane, M.P. 2010. Structural vs. atheoretic approaches to econometrics. Journal of Econometrics
156:3–20.
Klaiber, H. A., and D. Phaneuf. 2010. Valuing open space in a residential sorting model of the
twin cities. Journal of Environmental Economics and Management 60: 57–77.
Klaiber, H. A., and N. V. Kuminoff. 2014. Equilibrium sorting models of land use and residen-
tial location. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 352–379
Klaiber, H. A., and V. K. Smith. 2011. Preference heterogeneity and non-market benefits: the
roles of structural hedonic and sorting models. In International handbook on non-market
environmental valuation, ed. J. Bennett, 222–253. Edward Elgar. Massachusetts.
Kuminoff, N. 2008. Partial identification of preferences in a dual-market locational equilibrium.
Working Paper.
Kuminoff, N. 2009. Decomposing the structural identification of non-market values. Journal of
Kuminoff, N. V., V. K. Smith, and C. Timmins. 2012. The new economics of equilibrium sorting
and policy evaluation using housing markets. Forthcoming: Journal of Economic Literature.
LeBaron, B. 2006. Agent-Based Computational Finance. In K. L. Judd & L. Tesfatsion (eds.),
Handbook of computational economics, Vol. 2 Agent-based computational economics (pp.
1187–1233): Elsevier B.V.
Lewis, D. 2010. An economic framework for forecasting land-use and ecosystem change.
Resource and Energy Economics 32: 98–116.
Lewis, D., B. Provencher, and V. Butsic. 2009. The dynamic effects of open-space conserva-
tion policies on residential development density. Journal of Environmental Economics and
Management, 57: 239–252.
Lichtenberg, E. 1989. Land quality, irrigation development, and cropping patterns in the north-
ern high plains. American Journal of Agricultural Economics 71: 187–194.
Magliocca, N., V. McConnell, M. Walls, and E. Safirova. 2012. Zoning on the urban fringe: Results
from a new approach to modeling land and housing markets. Regional Science and Urban
Economics 42(1–2): 198–210.
Mayer, C., and C. Somerville. 2000. Residential construction: Using the urban growth model to
estimate housing supply. Journal of Urban Economics 48: 85–109.
McFadden, D. 1974. Conditional logit analysis of qualitative choice behavior. In Frontiers in
Econometrics. ed. P. Zarembka, 105–142. New York: Academic Press.
Mills, D. 1981. Growth, speculation, and sprawl in a monocentric city. Journal of Urban
Murphy, A. 2013. A dynamic model of housing supply. Unpublished manuscript available
upon request from author.
Nechyba, T., and R. Walsh. 2004. Urban sprawl. Journal of Economic Perspectives 18: 177–200.
Nelson, G., and D. Hellerstein. 1997. Do roads cause deforestation? Using satellite images in
econometric analysis of land use. American Journal of Agricultural Economics 79: 80–88.
Newburn, D., and P. Berck. 2006. Modeling suburban and rural-residential development
beyond the urban fringe. Land Economics 82: 481–499.
Newburn, D., and P. Berck. 2011. Exurban development. Journal of Environmental Economics
Oates, W. 1969. The effects of property taxes and local public spending on property values: An
empirical study of tax capitalization and the Tiebout hypothesis. The Journal of Political
Economy 77: 957–971.
Parker, D. 2013. An economic perspective on agent-based models of land-use and land-cover
change. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 402–429.
Parker, D., and T. Filatova. 2008. A conceptual design for a bilateral agent-based land mar-
ket with heterogeneous economic agents. Computers, Environment and Urban Systems
32: 454–463.
Plantinga, A. 1996. The effect of agricultural policies on land use and environmental quality.
Plantinga, A., and D. Lewis. 2014. Landscape simulations with econometric-based land use
models. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 380–401.
Post, W., W. Emauel, P. Zinke, and A. Stangenberger. 1982. Soil carbon pools and world life
zones. Nature 298: 156–159.
tion. Journal of Political Economy 82: 34–55.
Rust, J. 1987. Optimal replacement of GMC bus engines: An empirical model of Harold Zurcher.
Econometrica 55: 999–1033.
Sala, O., F. S. Chapin, III, J. Armesto, E. Berlow, J. Bloomfield, R. Dirzo, E. Huber-Sanwals, L.
F. Huenneke, R. B. Jackson, A. Kinzig, R. Leemans, D. Lodge, H. A. Mooney, M. Oesterheld.,
N. L. Poff, M. T. Sykes, B. H. Walker, and D. H. Wall. 2000. Global biodiversity scenarios in
the year 2100. Science 287: 1770–1774.
Schimel, D. 1995. Terrestrial ecosystems and the carbon cycle. Global Change Biology 1: 77–91.
Tajibaeva, L., R. Haight, and S. Polasky. 2008. A discrete-space urban model with environmen-
tal amenities. Resource and Energy Economics 30(2): 170–196.
Tiebout, C. 1956. A pure theory of local expenditures. The Journal of Political Economy
64: 416–424.
Timmins, C., and W. Schlenker. 2009. Reduced-form versus structural modeling in environ-
mental and resource economics. Annual Review of Resource Economics 1: 351–380.
Towe, C., C. Nickerson, and N. Bockstael. 2008. An empirical examination of the timing of
land conversions in the presence of farmland preservation programs. American Journal of
Towe, C., R. Lewis, and L. Lynch. 2014. Using quasi-experimental methods to evaluate
land policies: Application to Maryland’s priority funding legislation. In The Oxford
handbook of land economics, eds. J. M. Duke and J. Wu, 452–480. New York: Oxford
University Press.
Vitousek, P., H. Mooney, J. Lubchenco, and J. Melillo, 1997. Human dominance of earth’s eco-
systems. Science 277: 494–499.
Vorosmarty, C., and D. Sahagian. 2000. Anthropogenic disturbance of the terrestrial water
cycle. BioScience 50: 753–765.
Walsh, R. 2007. Endogenous open space amenities in a locational equilibrium. Journal of Urban
Wrenn, D. 2012. Time is money: An empirical examination of the dynamic effects of regula-
tory uncertainty on residential subdivision development. PhD Dissertation. The Ohio State
University.
Wu, J., and K. Segerson. 1995. The impact of policies and land characteristics on poten-
tial groundwater pollution in Wisconsin. American Journal of Agricultural Economics
77(4): 1033–1047.
Wu, J., and A. Plantinga. 2003. The influence of public open space on urban spatial structure.
C HA P T E R 1 4
E Q U I L I B R I U M S O RT I N G
MODELS OF LAND USE AND
R E S I D E N T IA L C H O I C E
H . A L L E N K L A I BE R A N D N IC OL A I V. K UM I NOF F
Far better an approximate answer to the right question, which is often

vague, than an exact answer to the wrong question, which can always be
made precise.
—John Tukey (1962)
Americans are remarkably mobile. Since World War II, 18% of the US population has
moved to a new residence every year, on average. As Charles Tiebout (1956) famously
observed, these movers face a public goods counterpart to the private market shopping
trip. They choose among residential communities that differ in their housing prices and
in their provision of amenities, such as local public goods, urban attractions, and envi-
ronmental services. The location choices that they make reveal features of their prefer-
ences. As heterogeneous households sort themselves across the urban landscape, their
collective location choices will influence housing prices as well as the supply of ameni-
ties through a combination of voting, social interaction, and feedback effects. To better
understand this two-way interaction between people and their surrounding environ-
ment, economists have developed equilibrium models of the sorting process.
Equilibrium sorting models begin with a formal description for the spatial landscape
and the structure of household preferences. Utility-maximizing location choices are
expressed as a function of the observable characteristics of households, houses, and
communities, as well as of structural parameters describing latent preferences. This
functional relationship is then inverted, using the logic of revealed preferences to char-
acterize the distribution of preferences in the population of households. Estimation
results are used to calculate the willingness to pay for large-scale changes in landscape
amenities. Sorting models can also be used to simulate how people and markets will
adjust to unexpected events and make predictions for “general equilibrium” benefit
Equilibrium Sorting Models of Land Use and Residential Choice 353
measures and future land use trends. This is a new and exciting framework for policy
evaluation that offers the potential to improve our understanding of land economics.
Compared to the standard quasi-experimental framework for describing how land-
scape changes affect housing prices, the development and estimation of a structural
sorting model can seem intimidating. The analyst must be willing to collect additional
data and think deeply about the economic forces that underlie market equilibria.
Econometric identification may seem less transparent. It may be necessary to code the
estimator from scratch, and the results may be viewed with skepticism by critics who
dislike structural modeling. Despite these challenges, the potential insights from for-
mulating, estimating, and interpreting an equilibrium sorting model far outweigh the
learning costs. Put simply, the equilibrium sorting methodology allows us to provide
approximate answers to the right questions about the relationships among land use, res-
idential choice, and public policy.
This chapter summarizes the equilibrium sorting methodology. We have two main
objectives. First, we intend to make the empirical models accessible to economists
who are new to the literature. Thus, we provide more detail about datasets and estima-
tors than one finds in the typical journal article. Our second objective is to clarify the
relationship between the newer structural models of the sorting process and the older
reduced-form models of hedonic equilibria that have long served as workhorses for eco-
nomic analysis of land use and household location choice. We argue that the two frame-
works are inseparable. Hedonic price functions describe sorting equilibria, and what
we learn about the sorting process influences how we interpret hedonic price functions.
We intend this chapter to be more pragmatic than previous efforts to characterize
the literature. Considerable space is devoted to (1) empirical descriptions for the spa-
tial landscape and household preferences, (2) econometric procedures for estimat-
ing structural preference parameters, and (3) procedures for simulating how markets
adjust to unexpected events. This leaves us with less space to cover historical back-
ground, systematically catalog empirical results, or recommend directions for future
research. Readers interested in these topics are directed to Palmquist (2005); Klaiber
and Smith (2011); Epple, Gordon, and Sieg (2010); and Kuminoff, Smith, and Timmins
(2013).
The chapter proceeds as follows. Section 1 begins with a general description for the
spatial landscape that nests empirical hedonic and sorting models. Then we define the
household’s location choice problem, characterize a sorting equilibrium, and briefly
summarize results on existence and uniqueness. In Section 2, we move from theory to
practice. Focusing on the two predominant microeconometric frameworks—the “pure
characteristics model” (Epple and Sieg 1999) and the “random utility model” (Bayer
et al. 2004)—we explain how to build an empirical sorting model and estimate struc-
tural parameters. Datasets, modeling assumptions, and econometric procedures are
covered. Section 3 explains how the estimation results can be used to simulate how peo-
ple and markets would adjust to an unexpected event. Many of the insights gleaned from
the estimation and simulation of sorting models also have important implications for
hedonic estimation. Section 4 summarizes insights on the causes and consequences of
354 H. Allen Klaiber and Nicolai V. Kuminoff
omitted variable bias, benefit measurement, and the interpretation of land value capital-
ization effects. Finally, Section 5 concludes.
1. Conceptual Framework
1.1 The Spatial Landscape

Consider a metropolitan region comprised of j = 1, …, J housing communities, each of
which contains N j houses.1 The region is assumed to be sufficiently small that most work-
ing households could relocate anywhere in the region without having to move to a dif-
ferent job. At the same time, the region is assumed to be self-contained in the sense that
few households would consider living outside the region. Some regions that meet these
criteria may be small and isolated, such as the Grand Junction metro area in western
Colorado. Others may be large and integrated, such as the San Francisco-Oakland-San
Jose consolidated metropolitan statistical area, containing more than 4 million people
spread out over several interconnected cities and suburbs.
Within the region, each housing community provides a unique bundle of amenities,
g j . “Amenities” are defined broadly to include any nonmarketed goods and services that
matter to households. Examples include local public goods produced from property
tax revenue (public education, police and fire protection), environmental services (air
quality, microclimate), proximity to urban attractions (central business district, shop-
ping, dining), and the demographic composition of the community (race, age, wealth).
Within a community, individual houses differ in their structural characteristics. The
vector hn j will be used to describe the physical attributes of a particular house, n, located
in community j. Examples include the square footage of the house, the number of bed-
rooms, and the quality of building materials.
Households are heterogeneous. They differ in terms of their incomes (y), preferences
(α), and demographic characteristics (d). Each household will maximize its utility by
choosing a specific house in its preferred community. We use n j to denote the house-
hold’s simultaneous choice of a community and a house within that community:
max U i ( g j , hn j , b, α i ) subject to yi = b + Pn j . (1)

nj

In the budget constraint, the price of the numeraire commodity (b) is normalized to 1 and
Pn j represents the annualized after-tax price of housing.
The collective location choices made by the population of households may influence
the spatial distribution of amenities. For example, as open space gets converted to urban
development, new opportunities for dining and nightlife may emerge, along with increased
1 The terms “community” and “neighborhood” are used interchangeably in the literature.
traffic and air pollution. Homeowners may be asked to vote on assessments to fund the
preservation of remaining open space or to support public schools. The academic perfor-
mance of students in those schools may depend on the incomes and education levels of
their parents. Although we do not model these mechanisms formally, it is important to
keep them in mind because they create a need for instruments in econometric estimation.
Finally, three assumptions are typically maintained to reduce the amount of friction in
the market. First, everyone is assumed to have perfect information about the spatial land-
scape. Second, everyone is assumed to face the same schedule of prices. Finally, households
are assumed to be freely mobile.
1.2 Characterizing a Sorting Equilibrium

In a sorting equilibrium, prices, physical housing characteristics, amenities, and location
choices are all defined such that no household could improve its utility by moving, and each
household occupies exactly one house. Equation (2) provides a formal statement of this
condition.
U i ( g j , hn j , yi − Pn j , α i ) ≥ U i ( g k , hmk , yi − Pmk , α i ), ∀ i, m, k : Pmk < yi ,
and
d ∑A
nj
i ,n j = 1, ∀ i, j , (2)

where Ai ,n j is an indicator variable that equals 1 if and only if household i occupies house n
in community j. Although we suppress temporal subscripts, equation (2) is best viewed as
a single-period snapshot of market outcomes. It may or may not be a long-run steady state.
Current incomes and preferences may reflect temporary factors. Credit may be unusually
easy (or difficult) to obtain. The average household may be unusually optimistic (or pessi-
mistic) about the future asset value of housing. Budget constraints may reflect other transi-
tory macroeconomic or microeconomic shocks. As these factors change over time, so will
the features of the sorting equilibrium.
With a few mild restrictions on preferences, the market outcomes from a sorting
equilibrium can be described by a hedonic price function. If U i ( g j , hn j , b, α i ) is con-
tinuously differentiable, monotonically increasing in the numeraire, and Lipschitz con-
tinuous, then theorem 1 from Bajari and Benkard (2005) can be invoked to prove that
equilibrium prices must be functionally related to housing characteristics and ameni-
ties, Pn j = P ( g j , hn j ).2 This result places less discipline on the price function than Rosen’s
(1974) hedonic model. Households are not assumed to be free to choose continuous
quantities of each amenity. Nor is the market assumed to be perfectly competitive. In
2 Although Bajari and Benkard (2005) treat nonprice attributes as exogenous, it is straightforward to
extend their result to the case of endogenous amenities by assuming that households ignore their own
contributions to each amenity.
fact, Bajari and Benkard demonstrate that no assumptions about the supply side of the
market are needed to prove that equilibrium can be described by a price function.
Relaxing the assumptions of Rosen’s model has costs and benefits. The main benefit
is a more realistic description of the spatial landscape. Although households may be
able to purchase approximately continuous quantities of physical housing characteris-
tics, the same is not true for landscape amenities. Air quality changes discretely from
air basin to air basin; test scores change discretely from school district to school district;
and some communities are adjacent to open space, whereas others are not. The cost of
relaxing Rosen’s continuity assumption is that we lose the ability to translate the price
function gradient into measures of the marginal willingness to pay (MWTP) for ame-
nities. Nevertheless, we shall see that the price function still plays an important role in
estimation.
A second, stronger, restriction that has proven useful in characterizing sorting equi-
libria is the single-crossing condition. Single crossing helps to characterize the ways in
which households sort themselves across locations according to their heterogeneous
incomes and preferences. To see the intuition, consider the simplest form of preference
heterogeneity—vertical differentiation. In a “vertical” model, households differ only in
their preferences for housing “quality” relative to the numeraire. They are assumed to
agree on a ranking of locations by overall quality, q = f ( g , h). Given this assumption,
equation (3) defines the slope of an indirect indifference curve in (q, p) space.
dp
M (q, p, α , y ) = |V = V . (3)
dq

If M is monotonically increasing in ( y | α ) and (α | y ), then indifference curves in the

(q, p) plane will satisfy single crossing in y and α. Under this condition, any sorting equi-
librium must satisfy three properties: boundary indifference, increasing bundles, and
stratification.3
To interpret the three properties, it is useful to first order locations by quality. Without
loss of generality, let the ordering be defined such that q1 < < qR. The increasing bun-
dles property requires that households must pay for the amenities provided by higher
ranked locations through higher housing prices: P1 < < PR . Stratification requires that
households are stratified across the R locations by (α | y ) and ( y | α) . In other words, all
else constant, households in higher ranked locations will have higher income and stron-
ger preferences for amenities. Finally, boundary indifference defines the set of values for
(α, y) that would make a household exactly indifferent between locations r and r + 1.
Figure 14.1 provides a simple illustration of a sorting equilibrium that satisfies the
three properties. Consistent with increasing bundles, the price ranking of communities
3
For additional background on the role of single-crossing conditions in equilibrium sorting models
see Epple and Romer (1991); Epple and Sieg (1999); and Kuminoff, Smith, and Timmins (2013).
p1 < p2 < p3
j=3
j=2
j=1
FIGURE 14.1 Partition of households into communities by preferences and income.
matches the ranking by overall amenity provision. The figure partitions (α, y) space
into three cells corresponding to (α, y) combinations that rationalize the choice of each
community. For example, community 1 would maximize utility for any household with
values for income and preferences in the lower left cell of the partition. The boundar-
ies between adjacent cells define the (α, y) combinations that would make a household
exactly indifferent between the corresponding communities. To see how households are
stratified across communities notice that, conditional on preferences, wealthier house-
holds choose communities with more public goods. Likewise, conditional on income,
households with stronger preferences choose communities with more public goods. This
two-dimensional stratification is consistent with Tiebout’s (1956) reasoning and helps
to explain why we sometimes observe low-income households living in high-amenity
communities and high-income households living in low-amenity communities.
Stratification, increasing bundles, and boundary indifference are particularly help-
ful in estimating the class of pure characteristics models covered in Section 2.2. The
single-crossing condition is sufficient, but not necessary, to guarantee that a sorting equi-
librium will satisfy these properties. In addition to providing a simple characterization
of equilibrium, the single-crossing condition can help to guarantee that equilibria exist.
1.3 Existence and Uniqueness

General proofs of existence and uniqueness require fairly strong restrictions on prefer-
ences and amenities. One strategy is to assume that households have identical prefer-
ences (αi = α, ∀i), so they differ only in their incomes. In this case, the single-crossing
condition makes it possible to prove existence in the presence of an endogenously deter-
mined amenity (Ellickson 1971; Westoff 1977). Another strategy is to allow preference
heterogeneity but rule out social interactions (Nechyba 1997). Bayer and Timmins
(2005) develop a third approach. They smooth the preference function by adding an
idiosyncratic iid shock to utility. This allows them to prove existence in a setting where
households with heterogeneous preferences for exogenous amenities share a common
marginal utility for a single endogenous amenity. Whether the equilibrium is unique
depends on whether marginal utility is positive or negative.
In the presence of more complex preference structures, analysts have used numerical
simulations to demonstrate that equilibria may exist (Epple and Platt 1998; Sieg et al.
2004; Walsh 2007; Klaiber and Phaneuf 2010; Kuminoff and Jarrah 2010; Kuminoff
2011). Despite the lack of general proofs for existence and uniqueness, the empirical
literature has moved forward with preference structures that allow considerable het-
erogeneity and acknowledge the potential endogeneity of amenities. Analysts simply
assume that the available data reflect an equilibrium. Then they write down a utility
function and estimate values for the structural parameters that justify those data as an
equilibrium.
2. Estimation
Moving from theory to estimation requires three sources of information: (1) a defini-

tion for the choice set, (2) a parametric representation of the preference function, and
(3) assumptions for the statistical distributions used to characterize sources of unob-
served heterogeneity. Although specific modeling choices differ from study to study,
most applications can be grouped into two broad frameworks: random utility models
(RUM) based on Bayer, McMillan, and Reuben (2004) and pure characteristics models
(PCM) based on Epple and Sieg (1999).4
The RUM and PCM frameworks provide alternative characterizations of the same
sorting equilibrium. They require data on the same core variables: prices, housing
characteristics, household demographics, and spatially delineated amenities. Data
sources vary. Housing prices and structural characteristics are typically drawn from
the same sources as the hedonic literature—assessor databases or the US Census of
Housing. Data on consumer demographics are typically drawn from the Census of
Population. Data on amenities have been drawn from a variety of federal and state
government agencies. Although it is possible to calibrate empirical models using
aggregate data, the rule of thumb is to use the highest resolution microdata that are
available.5
4
Ferreyra (2007) proposed a third “general equilibrium” approach building on earlier work by
Necheyba (1997, 1999, 2000).
5 Sieg et al. (2004); Bayer, Ferreira, and McMillan (2007); and Klaiber and Phaneuf (2010) provide
particularly detailed descriptions of how their datasets were assembled.

2.1 The Random Utility Framework

The random utility framework builds on McFadden’s (1974) seminal discrete choice
model. Bayer, McMillan, and Reuben (2004) developed the first application to residen-
tial sorting, using data from the San Francisco area. A key feature of their application
is the recognition that both housing prices and amenities may be endogenous in the
estimation process. Consider housing prices. Unobserved attributes of communities
that make them more desirable also increase the demand to locate there. Ceteris pari-
bus, equilibrium prices must be higher in more desirable communities. The implica-
tion of this logic is the need to use instrumental variables to disentangle the correlation
between equilibrium housing prices and unobserved amenities. A similar argument
applies to amenities that are endogenously determined through the sorting process.
Bayer, McMillan, and Reuben (2004) show that the structure of the sorting model itself
can help to overcome these econometric challenges.
Subsequent applications refined the RUM framework and used it to estimate prefer-
ences for school quality (Bayer et al. 2007), land use (Klaiber and Phaneuf 2010), and
air quality (Tra 2010). A distinguishing characteristic of these applications is the way
they define locations as particular “types” of housing. For example, Klaiber and Phaneuf
define a housing type as a unique (house size, time period, community) rather than
an individual house. This aggregation follows from Berry, Linton, and Pakes (2004)
who demonstrate that consistent estimation for this class of RUM model requires the
number of consumers to exceed the number of choice alternatives. Bayer, Ferreira,
and McMillan (2007) use micro-census data on individual houses, whereas Tra uses
information on sampled houses contained within Census public use microdata areas
(PUMAs) grouped by common housing characteristics. Each of these approaches either
implicitly or explicitly aggregates individual houses into “housing types within commu-
nities” that form the choice set.
2.1.1 Parameterization of the Model

Parameterization of the model begins by dividing utility into observed and unob-
served components. A location-specific unobservable, ξ, is used to represent hous-
ing characteristics and amenities that are observed by households, but not the analyst.
Additionally, an “error” term, ε, is added, recognizing that households may have idio-
syncratic preferences for each location.
The utility a household receives from choosing a particular housing type, t, in com-
munity j, is usually expressed as a linear function of its attributes,
Vtij = α hi ht j + α ig g j + α ip pt j + ξt j + ε iy j . (4)

The way that communities are subdivided into housing types varies from study to
study. At one extreme, a type could be defined as precisely as an individual house. At the
opposite extreme, a type could be defined as coarsely as the mean or median house in a
particular community. Most studies use definitions between these extremes for reasons
that we discuss in the context of the mechanics of the estimator. Meanwhile, communi-
ties are often defined using Census aggregates, such as PUMAs, tracts, or block groups.
Three features of (4) are worth noting. First, the i superscripts on α allow households
to differ in their relative preferences for different attributes. This generalizes the “ver-
tical” preference structure introduced earlier and is often referred to as “horizontal”
differentiation.6 Second, the marginal utility of income is implicitly assumed to be con-
stant. It is suppressed in (4), as is the custom in random utility models.7 Last, although
the choice of a location is deterministic from the perspective of each household, assum-
ing a statistical distribution for the idiosyncratic term, εti j, makes it possible to derive a
closed-form expression for the share of households who choose each housing type.
Assuming εti j is distributed according to an iid type I extreme value distribution
produces a familiar logit expression for the probability that household i chooses each
housing type,
exp(α ih ht j + α ig g j + α ip pt j + ξt j )
Prtij = . (5)
∑ exp(α h i
h sk + α ig g k + α ip psk + ξ sk )
s ,k

Aggregating (5) over i = 1, . . . I households generates the expected share of households

choosing a particular housing type,
1
σt j =
I ∑pr . i
tj (6)
i
This share forms the foundation for market clearing in the model. There is no direct
assignment of individual households to specific housing types. Instead, equilibrium is char-
acterized using the predicted share of households selecting each housing type.
Ensuring market clearing requires that the predicted share of households choosing each
housing type must be identical to the observed share for that type. In other words, housing
supply must equal housing demand. This condition is satisfied by the inclusion of the alter-
native specific unobservables, ξt j . Given a distributional assumption for εti j , Berry (1994)
demonstrates that including a complete set of alternative specific unobservables results in
predicted and observed market shares coinciding as a necessary condition for maximum
likelihood estimation.8
6
The “vertical” and “horizontal” terminology is adapted from Lancaster (1979).
Tra (2010) includes a nonlinear income term of ln( yi − ph ), which preserves the budget constraint
7
but presents difficulties for welfare measurement (Herriges and Kling 1999; McFadden 1999).
8 This property holds for the linear exponential family of models that includes conditional logit.
2.1.2 Estimation Procedures
Recall that the specification for utility in (4) allows for horizontal preference heterogeneity.
Past applications have taken advantage of this flexibility by decomposing each preference
parameter, αi, into the sum of a constant component and a component that varies along
observable demographic characteristics of households:
αi = α 0 + α1d i . (7)

Using this decomposition, the utility function can be expanded as (8).
Vtij = α h0 ht j + α1h d i ht j + α 0g g j + α1g d i g j + α 0p pt j + α1p d i pt j + ξt j + εti j. (8)

All of the structural parameters in (8) can be recovered using a two-stage approach to
estimation.
The first stage recovers parameters that vary with household demographic characteristics
(α1h , α1g , α1p ), as well as the mean indirect utility for each alternative (θt j ). The second stage
uses the first-stage estimate for mean indirect utility to recover preference parameters com-
mon to all households (α 0h , α 0g , α 0p ). This partitioning is shown in equations (9a) and (9b)
Vtij = α1h d i ht j + α1g d1 g j + α1p d i pt j + θt j + εti j (9a)

θ t j = α + α h0 ht j + α 0g g j + α 0p pt j + ξt j , (9b)

using the script-free α term in (9b) to represent an intercept.9

In principle, the parameters in (9a) could be estimated using a standard conditional
logit model. For many applications, however, the number of housing types is large, mak-
ing gradient-based maximum likelihood estimation burdensome due to the propaga-
tion of mean indirect utility parameters. To reduce this computational burden, past
studies have relied on the results from Berry (1994). Specifically, a contraction mapping

algorithm enables recovery of estimates for each mean indirect utility parameter (θt j )
without using gradient-based searches. This computational “trick” speeds model con-
vergence significantly.10
9
An intercept is included to account for the normalization that occurs in first-stage estimation.
Evaluating differences in utility prevents recovery of the full j = 1 . . . J mean indirect utility parameters.
In practice, researchers often normalize by setting the first mean indirect utility parameter equal to zero
and estimate the remaining J - 1 parameters.
 prtij 
10 The standard contraction mapping routine is: θ s +1 = θ s − ln 
iteration of the contraction mapping routine.

tj tj

i
∑  , where s indexes the
σ t j 
Second stage estimation of (9b) raises several econometric issues. First, because the
dependent variable consists of estimated mean indirect utilities from (9a), some addi-
tional criteria must be satisfied to establish consistency and asymptotic normality
(Berry et al. 2004). Let T = ∑t j represent the total number of distinct housing types.
t, j
Consistency and asymptotic normality are defined as T → ∞. To guarantee consistency,

the number of households must grow large relative to the number of types: T log T → 0 .
I
T2
Asymptotic normality requires the additional restriction that is bounded. These
I
two requirements help motivate the characterization of housing types.
Consistency cannot be established if t is defined as an individual house because this
results in T = I. At the same time, it seems important to recognize that the prices and
structural characteristics of houses vary within the Census aggregates used to define
housing communities. Empirical studies have sought a middle ground that addresses
both issues. For example, Klaiber and Phaneuf (2010) use square footage to divide the
houses in each Census block group into “small,” “medium,” and “large” terciles. Then
they define housing types using the median values of structural housing characteristics,
amenities, and prices for the houses in each block group and size category.
Another econometric issue is that prices, and potentially amenities, are likely corre-
lated with the error term in (9b), confounding ordinary least squares (OLS) estimation.
A popular instrumentation strategy is to exploit the logic of the sorting process to form
an “optimal” instrument (Bayer and Timmins 2007). The insight behind the IV strat-
egy explained by Bayer and Timmins is to utilize the variation in prices that reflects
exogenous characteristics of distant locations. Such instruments are relevant because
the equilibrium levels of endogenous attributes at each location are influenced by the
attributes of all other locations through the sorting equilibrium. Their validity relies on
the assumption that the analyst can identify “exogenous” attributes at distant locations
that are uncorrelated with ξ t j .
If we treat amenities as exogenous and employ the Bayer-Timmins instrument for
price, Klaiber and Phaneuf (2010) demonstrate that the two-step estimator can proceed
as follows:
S tep 1. Estimate (9a) to obtain parameter estimates.

0*
Step 2. Make a guess for the coefficient on price, α ρ .
Step 3. Move the term α 0* p pt j in (9b) to the left hand side and add additional control
variables (denoted by tildes) formed within rings around each choice, alterna-
tive to the right hand side of the modified (9b).
Step 4. Estimate the modified (9b) from step 2 via OLS and set the residual to zero.
Step 5. Calculate the mean indirect utility implied by step 3, denoting it by θ t j .
Step 6. Use the first stage estimates from step 0 along with the initial guess for the coefficient
on price and the estimates obtained in step 4 to solve for the set of prices, ρhiv such
that aggregate predicted shares exactly equal observed shares of each alternative tj.
Step 7. Perform IV estimation of (9b) using ptIVj as an instrument.
Step 8. Use the estimate of α 0p from step 7 to iterate starting at step 1 until the estimate
of α 0p converges.
2.2 The Pure Characteristics Framework

Most of the recent empirical models developed within the pure characteristics frame-
work build on earlier work by Dennis Epple and his co-authors (e.g., Epple et al. 1984,
1993; Epple and Romer 1991). These studies introduced a constant elasticity of substitu-
tion (CES) specification for preferences as an example. Epple and Platt (1998) calibrated
the CES function to data on housing market outcomes, and Epple and Sieg (1999) devel-
oped a structural estimator. Their approach to estimation was refined in subsequent
work by Sieg et al. (2002, 2004). The PCM framework has since been used to investigate
the benefits of numerous amenities, including landscape attributes in Portland, Oregon
(Wu and Cho 2003), air quality in Southern California (Smith et al. 2004) and Northern
California (Kuminoff 2009), open space in the Raleigh-Durham area of North Carolina
(Walsh 2007), and school quality in Phoenix, Arizona (Klaiber and Smith 2012).
2.2.1 Parameterization of the Model

One of the distinguishing features of the PCM framework is a mixed discrete-continuous
depiction of the choice set. Households are assumed to be free to choose continuous
quantities of physical housing characteristics in each of a discrete number of residen-
tial communities. Under this assumption, the location choice process can be character-
ized by the choice of a community. Conditional on that choice, a household will select a
house with the optimal combination of physical characteristics.
Sieg et al. (2002) illustrate how the discrete-continuous representation for the choice
set influences how we define the “price of housing” in an indirect utility function. They
demonstrate that as long as hn j enters utility through a separable subfunction that is
homogeneous of degree 1, housing expenditures can be expressed as the product of a
price index and a quantity index, Pn j = q(hn j ) ⋅ p( g j ). In this case, p j = p( g j ) replaces Pn j
in the indirect utility function. Equation (10) steps through this logic.
U[ g j , hn j ( g j , Pn j , αi , yi ), yi − Pn j ( g j , hn j ), αi ]
= U[ g j , hn j ( g j , q(hn j ) ⋅ p j ( g j ), α i , yi ), yi − q(hn j ) ⋅ p j ( g j ), α i ]
= V ( g j , p j , α i , yi ). (10)

The first equality follows from Sieg et al. (2002). The second equality simply rewrites
utility in indirect terms. The “price of housing” in each community, pj, represents the
implicit price (per unit of q) to consume the bundle of nonmarket amenities provided by
that community.
The same assumptions that allow Sieg et al. (2002) to factor housing expenditures into
price and quantity indices also support a strategy to estimate p1 ,…, pJ from a hedonic
regression. Taking logs of the expenditure function yields a general expression for an
estimable hedonic model,
ln(Pn j ) = ln[q(hn j )] + ln[ p j ( g j )].

(11)
Given an assumption for the functional form of the quantity index, microdata on
housing sales can be used to recover p1 ,…, p J as community-specific fixed effects in a
regression of log sale prices on housing characteristics.11 Normalizing the smallest fixed
effect to equal one produces the prices that enter the discrete-choice model of commu-
nity selection.
Equation (12) illustrates the CES specification for preferences. It describes the utility
that household i obtains from living in community j.
1
ρ
  ( yi )1− ν − 1  βPjη+1 − 1  ρ
ρ 

( )
Vi , j = α i G j + exp 
 1− ν 
 exp  −
 1+ η  


 , (12)
    
 
where G j = γ 1 g 1, j + + γ R −1 g R −1, j + ξ j , and F(α, y) ~ lognormal.

This indirect utility function does not correspond to any closed-form expression for
direct utility, but it has several useful properties. It recognizes that physical housing
characteristics may not be perfect substitutes for amenities. It also generates a conve-
nient Cobb-Douglas specification for the demand for housing.12 Finally, the CES speci-
fication maps directly into the underlying theory from Section 1. It yields parametric
expressions for boundary indifference, stratification, and increasing bundles that serve
as the basis for the estimation algorithm.13
The first term inside the CES nest represents utility from amenities. Households obtain
utility from a linear index of amenities provided by each community. They are assumed
to agree on a common set of weights for the amenities in the index (γ 1 ,.…, γ R −1 ), but
11
For example, if the quantity index is assumed to be multiplicative, then the regression is a simple
linear-in-logs specification with fixed effects for communities. Data on the transaction prices of actual
housing sales are converted to annualized values by adapting the formula from Poterba (1992).
12
This follows from the exponential form of the term in square brackets.
13 For additional discussion of the properties of CES specifications for utility in sorting model see
Epple, Filimon, and Romer (1984, 1993); Epple and Romer (1991); Epple and Platt (1998); Epple and Sieg
(1999); and Sieg et al. (2002).
they differ in their overall preferences for amenities relative to the private good compo-
nent of housing and the numeraire (αi). Of the R amenities in the index, R–1 are observ-
able. g R , j = ξ j represents the composite effect of community-specific attributes that
are observed by households but not the analyst. As in the RUM model, ξ varies across
choices but is restricted to be the same for every household.14 This is an example of what
Berry and Pakes (2007) label the “pure characteristics” approach to modeling choice
among differentiated objects. Utility is defined purely over the characteristics of com-
munities; there is no idiosyncratic location-household-specific εij shock.
The second term inside the CES nest represents utility from the private good com-
ponent of housing. Households are assumed to share the same elasticity of substitution
between amenities and private goods (ρ) and the same demand parameters for the pri-
vate good component of housing: price elasticity (η), income elasticity (ν), and demand
intercept (β). Applying Roy’s Identity to (12) yields a simple expression for the demand
for housing,
qi = β pηj yiν . (13)

Although households share a common set of demand parameters, notice that indi-
vidual demand varies with income.
A key feature of the CES specification in (12) is that preferences are vertical. Since
households have identical relative preferences for g1, . . . , gR, they agree on the ranking of
communities by the G index. Given the expected signs for the housing demand parame-
ters (β > 0, η < 0, ν > 0), preferences satisfy single crossing if ρ < 0.15 This makes it possi-
ble to describe how households sort themselves across communities in equilibrium. To
see this, first order communities by price: p1 < p2 < < pJ . Increasing bundles implies
G1 < G2 < < GJ . Equation (14) uses boundary indifference to implicitly define the (α, y)
combinations that make a household exactly indifferent between j and j + 1.
 y 1− ν − 1   Q j +1 − Q j 
ln (αi ) − ρ  i
 1− ν 
 = ln  ρ
 G j − G j +1 
ρ
 ρ
 = B j , j +1 , where Q j = exp  −
 1+ η
( )
β p ηj +1 − 1  .

(14)
Notice that all of the heterogeneity in income and preferences appears to the left of the
equality. The stratification property implies that any household with income and pref-
erence such that: ln(αi ) − ρ[( yi1− ν − 1) / 1 − ν] < B j , j +1 will prefer community j to every
higher ranked community: j + 1, j + 2, . . ., J. Therefore, the left side of (14) can be used to
14
Although the RUM and PCM frameworks both use ξ to represent choice-specific unobserved
attributes, they will generally recover different estimates for ξ due to the different spatial scales at which
they define the variable and due to their different specifications for preferences.
15 Empirical studies have been unanimous in confirming the expected signs of these four parameters.
characterize the sorting of households into communities. This result plays an important
role in the mechanics of the estimator.
2.2.2 Estimation Procedures

Estimation procedures vary slightly from study to study. Here we describe the simu-
lated GMM approach developed by Sieg et al. (2004). Treating the first-stage esti-
mates for housing prices as known constants, the GMM estimator can be used to
recover all of the structural parameters. Let θ represent a vector of these parameters,
θ = [β, η, ν, ρ, µ α , µ y , σ α , σ y , λ, G1 , γ 1 ,…, γ R −1 ] . Equation (15) defines the GMM objec-
tive function, where is a set of instruments, represents the moment conditions, and A is
the covariance matrix of moments.
′
 1 J   1 J 
θ = arg minθ ∈Θ 
J ∑ z j m j (θ ) A−1 
J ∑ z m (θ )
j j (15)
 j =1   j =1 
Sieg et al. demonstrate that the seven moment conditions in (16) can be used to iden-
tify all the parameters in θ.16
 G j − γ 1 g 1,1 − − γ R −1 ⋅ g R −1, j 
 
 y 25
j −y 25
j 
 50 50 
 y j − y j 
  (16)
m j (θ ) =  75 75
y j − y j .
 25 
 ln Pn∈j − ln β − (η + 1) ln p j − ν ln y j 
25
 ln P 50 − ln β − (η + 1) ln p − ν ln y 50 
 n ∈j j j 
 ln Pn75∈j − ln β − (η + 1) ln p j − ν ln y 75 
 j 

The first moment condition is based on the level of amenity provision. Given a value
for overall provision of amenities in the cheapest community, G1, the sorting behavior
implied by vertical differentiation allows G2 , … , GJ to be defined recursively. The pre-
dictions for G2, . . . , GJ are then used to identify the (constant) weights in the amenity
index. The residual to the moment condition defines the composite unobserved ame-
nity in each community (ξ1 , …, ξJ ).
16 The particular moment conditions selected by Sieg et al. are somewhat arbitrary. In principle,
one could use fewer moment conditions and additional instruments. Alternatively, one could
develop moment conditions based on different quantiles of the distributions of income and housing
expenditures.
The next three moment conditions are based on the model’s prediction for the dis-
tribution of income. Under the maintained assumptions on preferences, the informa-
tion in θ can be used to simulate community-specific income distributions. Three of the
moment conditions match the 25th, 50th, and 75th quantiles from the simulated dis-
tributions of income in each community ( y 25 50
j ,y 75
j ,y j ) to their empirical counterparts
25 50 75
( y j , y j , y j ).
The last three moment conditions use the simulated income distributions to match pre-
dicted and observed quantiles from the distribution of housing expenditures in each com-
munity. The expenditure moments are obtained by multiplying (13) by price and taking logs.
Instruments are required to address endogeneity in the moment condition based on
provision of amenities. The problem is that observed and unobserved amenities may
be correlated. If households sort themselves across communities according to their
income and preferences for a seemingly exogenous amenity—air quality, for example—
their location choices may influence the levels of other endogenous amenities, such as
public school quality, inducing correlation between them. PCM applications have fol-
lowed Epple and Sieg (1999) in developing instruments from monotonic functions
of each community’s rank in the price index. These instruments will be valid as long
as unobserved amenities are of second-order importance; that is, if they affect house-
holds’ location choices without affecting the price rank of a community. The relevance
of the instruments stems from the expectation that communities with higher levels of
observed amenities will tend to be higher in the price ranking.
The mechanics of the simulated GMM estimator are straightforward. It can be imple-
mented using a Nelder-Mead algorithm that iterates over the following steps.
Step 1. Select a starting value for θ = β , η, ν , ρ, µα , µ y , σ α , σ y , λ , G1 , γ 1 , … , γ R −1 
Step 2. Draw I “households” from F(α, y) ~ lognormal. In some applications, I is set to

the actual population of the study region. In other cases, it is scaled down by
an order of magnitude to reduce computational demands.
 yi1−ν − 1
Step 3. Calculate K i = ln (α i ) − ρ  for all i = 1, … , I and use it to sort house-
 1 − ν 
holds in ascending order. Epple and Sieg (1999) demonstrate that the vertical
model implies that, in any equilibrium, households will sort themselves across
communities according to Ki, such that households with higher values for Ki
will always locate in higher ranked communities.
Step 4. Sort households across communities. Let S1 , … , SJ represent the observed
population counts of each community such that ∑S j = I . Starting with the
j
lowest Ki, assign the first S1 households to community 1. Then assign the next
S2 households to community 2, and so on.
Step 5. Given G1, solve for G2 to make the boundary person between communities 1
and 2 indifferent between them. Then given G2, solve for G3, and so on. . . .
Step 6. Calculate y 25 50 75 for each community.

j , yj , yj
Step 7. Use y 25 50 75

j , y j , y j , and G2 (θ ),… , G J (θ ) and θ to evaluate the GMM objective
function (15). If the minimization criteria of the numerical algorithm are sat-
isfied, stop. If not, update θ and return to step 2.
2.3 Comparing the RUM and PCM Frameworks

The RUM and PCM frameworks are each capable of explaining a given dataset as a sort-
ing equilibrium. This makes it difficult to compare the two models based on in-sample
performance. In our opinion, neither model is strictly preferred to the other. Each has
some features that seem flexible and others that seem restrictive.
PCM models provide a relatively flexible preference function, recognizing that pub-
lic and private goods are not perfect substitutes. They also embed a budget constraint.
The identifying assumption is that each household is able to afford a subset of houses
in the community where it actually locates and in the communities that are adjacent in
the price ranking. In contrast, the PCM maintains a relatively strong assumption about
the importance of unobserved amenities. Unobserved amenities that influence the price
ranking of communities threaten the validity of the rank-based instruments.
Advantages of the RUM model include its relatively flexible characterization of the
choice set. It recognizes that zoning regulations may prevent home buyers from choos-
ing continuous quantities of housing characteristics. Moreover, the instruments pro-
posed by Bayer and Timmins (2007) are robust to the presence of unobserved amenities
that influence the price ranking of communities. Yet, the RUM model also makes strong
assumptions. The linear specification for utility assumes amenities and structural hous-
ing characteristics are perfect substitutes. Likewise, every household is assumed to be
capable of purchasing every house.
Both frameworks maintain strong assumptions about preference heterogeneity. The
PCM’s vertical characterization fails to recognize that households are likely to differ in
their relative preferences for landscape amenities. Households with young children may
be primarily concerned about public school quality, for example, whereas retirees may
place more weight on proximity to golf courses. RUM models are capable of recognizing
these tradeoffs. However, that flexibility comes at a cost. The RUM model’s flexible treat-
ment of preference heterogeneity is enabled by its strong assumption that every house-
hold’s preferences for the unobserved attributes of every house happen to be drawn
from the same iid type I extreme value distribution. Kuminoff (2009) illustrates how the
two frameworks present a bias-variance tradeoff. By restricting the extent of preference
heterogeneity, the PCM introduces some bias. The RUM framework relaxes the restric-
tion that causes the bias, but it does so in a way that increases the scope for distributional
assumptions to influence the results.
It is important to keep in mind that the “flexible” and “restrictive” assumptions of
RUM and PCM models are not inexorably linked to either framework. They reflect
modeling decisions embedded in the original estimators developed by Epple and Sieg
(1999) and Bayer, McMillen, and Reuben (2004). A clever econometrician could mix,
match, and alter the features of the two models to develop new estimators. That said,
no amount of econometric cleverness can ever identify the true behavioral model with
absolute certainty. Perhaps the best way to evaluate the validity of a sorting model is to
test its out-of-sample predictions for how people and markets will adjust to unexpected
changes in the spatial landscape.
3. Evaluating the Benefits of Large-Scale

Changes in the Spatial Landscape
Estimates for the structural parameters of a RUM or PCM model can be used to develop
theoretically consistent predictions for the distribution of benefits from large-scale
changes in the spatial distribution of prices or amenities. One can easily calculate partial
equilibrium measures of willingness to pay (WTP) for a prospective policy change. The
model can also be used to simulate the transition to the new equilibrium that would fol-
low the introduction of the policy. Comparing the ex ante and ex post equilibria makes
it possible to predict migration patterns, capitalization effects, changes in the levels of
endogenous amenities, and the corresponding “general equilibrium” measures of WTP.
In this section, we define “partial” and “general” equilibrium benefit measures and then
discuss how to close the model and solve for a new equilibrium.
3.1 Benefit Measurement

Consider a policy that changes the supply of a single amenity in community j from g j1 to
g *j1. A partial equilibrium measure of the willingness to pay for this change, WTPPE , holds
constant all other features of the equilibrium.17 In contrast, a general equilibrium measure,
WTPGE, accounts for potential changes in housing prices, location choices, and the levels
of other endogenous amenities. Equations (17a) and (17b) formalize this distinction,

( )
V α i , yi − WTPPE , g *j1 , g j~1 , hn j , Pn j = V (α i , yi , g j1 , g j~1 , hnj , Pnj )

(17a)

( )
V α i , yi − WTPGE , g k*1 , g k* ~1 , hmk , Pm* k = V (α i , yi , g j1 , g j~1 , hn j , Pn j ),

(17b)
17 Calculation of partial equilibrium benefit measures differs between the PCM and RUM
frameworks. In the PCM (17a) is inverted to calculate WTP directly. In the RUM, the idiosyncratic error
term means that WTP must be defined as an expected value using a version of the usual log-sum rule.
where g j = [ g j1 , g j ∼1 ]. In (17b) the change in subscripts from nj to mk recognizes that

households may respond to the change by moving to a new location. The asterisk super-
*
scripts on g k∼1 and Pm* k recognize that, as people resort, their behavior may affect the
levels of other endogenous amenities, and prices may need to adjust to clear the market.
As Δgj = gj – gj grows or impacts a larger number of households, it becomes increasingly
important to model general equilibrium feedback effects. Overall, the richness in this
characterization for how people interact with their surrounding environment makes the
general equilibrium sorting model a powerful framework for policy evaluation.
3.2 Closing the Model

The RUM and PCM estimators essentially characterize housing demand, treating the
supply of housing as fixed. However, solving for a new equilibrium requires character-
izing both supply and demand, as well as any sources of friction in the market. Thus, to
close the model, the analyst must define the supply of housing, formalize their assump-
tions about moving costs, write down production functions for endogenous amenities,
and clarify whether households are treated as owners or renters. The way that each of
these issues is treated varies from application to application. However, three general
trends are worth discussing.
First, land use policies often play dual roles. They simultaneously enhance open space
amenities and they restrict urban development. As such, a new land use policy target-
ing the current supply of an amenity may also influence the future supply of housing.
Although equilibrium sorting models are capable of modeling this connection, few
applications have done so. Instead, the supply of housing is usually treated as fixed or
defined by a constant-elasticity assumption (e.g., Sieg et al. 2004; Smith et al. 2004;
Klaiber and Phaneuf 2010; Kuminoff 2011). This approach simplifies computation
of the new equilibrium, but risks overlooking important policy implications. Future
research that models the impacts of land use policies on both amenities and housing
supply would be a welcome addition to the literature. Walsh (2007) provides an initial
example of how this can be done.
Second, the initial general equilibrium applications have mostly treated households
as being freely mobile. In our experience, this assumption tends to produce a good
deal of consternation among seminar audiences. Anyone who has gone through the
process of moving to a new house is all too familiar with the costs involved: physical
costs, search costs, time costs, borrowing costs, and the psychological cost of adjusting
to a new environment. The good news is that the structure of a sorting model makes it
straightforward to utilize prior information about moving costs (Kuminoff 2009). For
example, Kuminoff (2011) models the changes in commuting costs and wage rates that
occur when working households alter their job and/or house locations. Likewise, Bayer,
Keohane, and Timmins (2009) demonstrate that some moving costs can be estimated
using related information, such as the location of an individual’s hometown.
Finally, all of the applications we discuss in this chapter treat households as renters.
Capital gains from housing sales are assumed to be captured by absentee landowners.
This approach simplifies computation of the new equilibrium, but abstracts from issues
that matter to policy makers. Many policies are effectively enacted on the owners of cap-
ital, especially policies influencing individual tax treatment. With this in mind, future
research that builds changes in assets into the budget constraint would be another useful
addition to the literature.
3.3 Solving for a New Equilibrium in a Random Utility Model

Solving for a new equilibrium in the RUM framework requires calculating housing
prices, location choices, and the levels of endogenous amenities such that housing sup-
ply and housing demand equate in all locations. Klaiber and Phaneuf (2010) describe
the solution process for the special case where amenities are exogenous. The basic idea
is to iterate over price changes until the predicted market shares for each housing type
equal the supply of housing for that type. The steps are as follows:
Step 1. Given the new spatial distribution of amenities, use the estimated preference
parameters to calculate the aggregate demand for each housing type, σ tdj,0,
where d stands for “demand” and 0 indicates that this is the initial iteration of
the algorithm.
d ,0 s
Step 2. Determine whether excess demand (σ t j > σ t j ) or excess supply (σ tdj,0 < σ tsj )
exists for each housing type.
Step 3. For types with excess demand, increase prices by a small percentage. Decrease
prices by a small percentage for types with excess supply.18
Step 4. Using the new prices, recalculate the aggregate housing demand for each type,
σtdj ,1.
d s
Step 5. Continue iterating over steps 2–4 until σt j = σt j for every type.
3.4 Solving for a New Equilibrium in the Pure

Characteristics Model
As in the RUM framework, it is straightforward to solve for a new PCM equilibrium
in the special case where amenities are exogenous. The “vertical” restriction on pref-
erence heterogeneity allows the problem to be formulated as a one-dimensional
root-finding problem. To see this, first recall that communities will always be ordered by
18 A weighted average of previous and new prices can help to prevent oscillation in convergence. The
magnitude of price changes can be weighted to be proportional to the difference in observed shares to
speed convergence.
their equilibrium housing prices and provisions of public goods: p1 < p2 < < pJ and
G1 < G2 < < GJ . Following a shock to public goods, the new equilibrium price ranking
must be identical to the new ranking by G. Using this fact, the solution algorithm pro-
ceeds as follows:
*
S tep 1. Make a guess for the new price of housing in the cheapest community, p1 .
Step 2. Use the left side of (14) to sort households into community 1 until total hous-
ing demand equals supply, aggregating over (13) to calculate demand.
*
Step 3. Use the last household sorted into community 1 to solve for the value of p2 that
satisfies (14).
Step 4. Repeat steps 2–3 for communities 2 through J, or until all households are
assigned to communities.
*
Step 5. If there is excess housing supply in community J, increase p1 and return to step
*
2. If there is excess demand, decrease p1 and return to step 2.
This recursive structure effectively reduces the simulation to a one-dimensional

problem where the new equilibrium price of housing in community 1 is adjusted until
the market clears in community J.
3.5 Endogenous Amenities

RUM and PCM solution algorithms can be modified to recognize that, as households
resort, their behavior can affect the supply of endogenous amenities. The way this is
modeled is context-specific. We briefly describe three examples, each of which finds that
endogenous adjustment of amenities is important for characterizing the impacts of a
prospective policy.
Klaiber and Smith (2012) use a PCM to evaluate the general equilibrium implications
of reductions in teaching staff in Maricopa County (Arizona) school districts. School
quality is measured using the student/teacher ratio. Mandated reductions in teaching
staff reduce school quality, inducing some households to move. As households with
school-aged children move, the number of students in each school district changes,
which feeds back into the student/teacher ratio, inducing additional households to
move. . . . and so on until prices, location choices, and the student/teacher ratio all con-
verge in equilibrium.
Walsh (2007) uses a PCM to investigate the impact of public open space preservation
on households and urbanization in Wake County, North Carolina. He endogenizes the
supply of housing by recognizing that privately owned farmland will tend to be devel-
oped as the demand for housing increases. As a result, land preservation polices can
have unintended consequences. Suppose that public funds are used to purchase a small
amount of scenic open space near a residential neighborhood. If the amenities associ-
ated with the preserved parcels increase the demand for housing in the neighborhood,
it may actually accelerate the rate at which the remaining privately owned open space is
developed.
Finally, Bayer and McMillan (2005) use a RUM to assess the role of households’ pref-
erences for several amenities, including the demographics of their neighbors. Measures
of demographic composition, such as average income, average education, and neigh-
borhood population shares by race, are directly determined by the sorting process. As
a result, a public policy that influences an exogenous amenity is shown to be capable of
altering neighborhood demographic composition.
4. Implications for Hedonic Estimation
Since hedonic and sorting models describe the same underlying equilibrium, advances
in the sorting literature also improve our understanding of the challenges associated
with using reduced-form hedonic regressions to evaluate the benefits of prospective
changes in the spatial landscape. We briefly summarize three ways in which the theory,
estimation, and simulation of sorting models has clarified the challenges with hedonic
estimation.
4.1 The Economics of Omitted Variable Bias

Omitted variables systematically confound the identification of conventional
hedonic regressions. This stylized fact has motivated an entire subliterature on
quasi-experimental approaches to estimation (Parmeter and Pope 2013). The experi-
mentalist perspective is that the analyst never observes all of the landscape ameni-
ties that are correlated with the amenity of interest. Breaking the correlation requires
instruments that effectively randomize the amenity “treatment.” The equilibrium
sorting literature complements the experimentalist perspective by providing an
explanation for omitted variable bias and suggesting further implications for benefit
measurement.
If people choose where to live based, in part, on their heterogeneous incomes and
preferences for amenities, then their location choices will influence the long run lev-
els of endogenous amenities (Ferreyra 2007; Walsh 2007; Epple and Ferreyra 2008;
Bayer and McMillan 2010). Under single-crossing restrictions on preferences, it is
natural to expect multiple amenities to be spatially correlated. As wealthier house-
holds move to areas with nice microclimates and low crime rates, for example, they may
vote to pass special assessments that enhance local public education. If data on micro-
climates and crime rates are unavailable, then conventional hedonic estimates of the
MWTP for school quality will tend to be biased upward. This logic helps to explain why
quasi-experimental estimates of the MWTP for school quality are typically less than
half the size of estimates from conventional hedonic regressions (Black 1999; Bayer et al.
2007; Kuminoff and Pope 2014).
Endogenous amenities present an additional challenge for benefit measurement.
A public policy that alters the spatial distribution of one amenity may influence the
long-run levels of other endogenous amenities. In this case, hedonic price func-
tions do not provide enough information to evaluate the welfare implications of the
policy.
4.2 Benefit Measurement and Policy Evaluation

The empirical hedonic literature is mostly limited to estimating the willingness to pay
for marginal changes in amenities.19 However, estimates for average MWTP are often
used to approximate the benefits from prospective policies that would produce nonmar-
ginal changes. Sorting models underscore the limitations of this strategy and provide a
means to address them.
Hedonic and sorting models tend to generate similar estimates for average MWTP.
For example, Sieg et al. (2004) find that the average MWTP for reduced ozone concen-
trations is approximately $67 (1990 dollars) per household in the Los Angeles metro
area. This figure is well within the range of estimates from comparable hedonic studies
($8–$181).20 Bayer, Ferreira, and McMillan (2007) provide a more refined comparison.
Using the same data and the same quasi-experimental identification strategy, they find
that hedonic and RUM estimates of the average MWTP for school quality differ by less
than 14%. However, average MWTP is rarely a sufficient statistic for policy evaluation.
Policy makers care about distributional implications. Moreover, developing credible
benefit measures requires recognizing the demand is less than perfectly elastic and that
people may react to the policy by adjusting their behavior.
Heterogeneity in preferences and the supply of amenities can lead to wide benefit dis-
tributions. For example, Sieg et al. (2004) find that the average marginal WTP for air
quality in Los Angeles County is twice as large as in neighboring Ventura County. When
they evaluate the nonmarginal ozone reductions that actually occurred between 1990
and 1995, the difference in WTP between Los Angeles and Ventura increases to 800%!
This difference arises from a combination of lower baseline levels of ozone in Ventura, a
smaller reduction in Ventura between 1990 and 1995, and heterogeneity in preferences
and income. Predicted adjustments to housing prices and location choices also have
19
Rosen’s (1974) original vision for hedonic demand estimation remains unfulfilled due to the
difficulty with identifying demand curves (Bartik 1987; Epple 1987).
20
Klaiber and Phaneuf (2010) provide a more detailed comparison. Using the same dataset (but
different controls for omitted variables), they find that hedonic and sorting models produce very similar
estimates of MWTP for some types of open space ($30 vs. $28 for a 1% increase in local parks) and
very different estimates for other types of open space (–$277 vs. $618 for a 1% increase in agricultural
preserves).
significant welfare implications. Partial and general equilibrium benefit measures differ
by over 100% for the average Ventura household.
4.3 The Wedge Between Capitalization Effects and Benefit

Measures
Public policies or unexpected events that shock the spatial distribution of an amenity
can also be used to identify the rate at which that amenity is capitalized into property val-
ues. The quasi-experimental branch of the hedonic literature has focused on developing
clever research designs for identifying these “capitalization effects” (see Parmeter and
Pope [2013] for examples). These studies typically reformulate the price function within
a panel data framework, using first differences, fixed effects, or difference-in-difference
estimators. The resulting estimates for capitalization effects are interesting, but they can-
not be interpreted as benefit measures unless we are prepared to make a series of heroic
assumptions about people and markets.
One of the key maintained assumptions that make it possible to interpret marginal
capitalization effects as measures of MWTP is that the gradient of the hedonic price
function is constant over the duration of the study. This assumption effectively requires
demand curves for the amenity to be perfectly elastic. If demand is downward sloping,
the adjustment to a new sorting equilibrium will generally produce a wedge between the
marginal capitalization effect and the MWTP. The size of the wedge will depend on the
distribution of income and preferences, the supply response, and concomitant changes
to the landscape over the duration of the study.
The wedge between capitalization and willingness to pay can be very large. Kuminoff
and Pope (2014) find that capitalization effects for reported changes in public school
quality tendto differ from quasi-experimental measures of ex ante and ex post MWTP
by more than 100%. Likewise, Klaiber and Smith (2013) find it difficult to predict the
size or the direction of the bias in using capitalization effects to approximate the benefits
of nonmarginal changes. These findings reinforce the earlier theoretical results of Lind
(1973) and Starrett (1981), as well as simulation results from Sieg et al. (2004) and Smith
et al. (2004), where predicted changes in housing prices bear little resemblance to pre-
dicted changes in benefits. Thus, the collective evidence from the sorting literature sug-
gests that capitalization effects for amenities are best interpreted literally, as a statistical
description of changes in housing asset values.
5. Conclusion
Equilibrium sorting models provide a powerful framework for modeling the two-way
interaction between people and their surrounding environment. They have tremen-
dous potential for policy evaluation. The Clean Air Act, the Clean Water Act, and
the Superfund program are examples of major public policies designed to produce
large-scale changes in the spatial distribution of nonmarket amenities. We would like
to understand their distributional implications and be able to predict how new policies
will affect consumer welfare and market outcomes. Equilibrium sorting models are the
first revealed preference framework capable of meeting this task while recognizing that
people adapt to changes in their surrounding environment.
Like every revealed preference framework, sorting models rely on maintained
assumptions about the structure of consumer preferences. This means their predictions
for benefit measures, housing market outcomes, and the evolution of the surrounding
landscape are best viewed as approximations. How accurate are these approximations?
The ability to answer this question is one of the novelties of the literature. Sorting mod-
els make testable predictions for market and nonmarket outcomes! Thus, the same types
of natural experiments and policy discontinuities that have been used to develop instru-
ments for reduced-form hedonic models could also be used to test a sorting model’s
predictions for property value capitalization effects and migration patterns. Future evi-
dence on external validity would help to refine the current generation of estimators and
continue to advance the literature.
Finally, our objective has been to provide an introductory guide to sorting models for
empirical analysts. We have tried to be clear about the subtleties of the microeconomet-
ric models and the mechanics of estimation and simulation procedures. Nevertheless,
our own experience has been that the most effective way to learn a sorting model is to
“get your hands dirty.” Readers who are up to the challenge can find examples of data
and code on our webpages.
Acknowledgments
Our research on equilibrium sorting has benefited from collaborations and conver-
sations with several colleagues. We thank without implicating Pat Bayer, Spencer
Banzhaf, Antonio Bento, Amy Binner, Kelly Bishop, Keith Evans, Paul Fackler, Michael
Hanemann, Abdul Jarrah, Alvin Murphy, Ray Palmquist, Chris Parmeter, Dan Phaneuf,
Jaren Pope, V. Kerry Smith, Chris Timmins, Roger von Haefen, Randy Walsh, and Kent
Zhao. We also thank JunJie Wu for helpful comments on an earlier draft of this chapter.
References
Bajari, P., and C. L. Benkard. 2005. Demand estimation with heterogeneous consumers and
unobserved product characteristics: A hedonic approach. Journal of Political Economy
113(6): 1239–1276.
Bartik, T. J. 1987. The estimation of demand parameters in hedonic price models. Journal of
Bayer, P., F. Ferreira, and R. McMillan. 2007. A unified framework for measuring preferences
for schools and neighborhoods. Journal of Political Economy 115(4): 588–638.
Bayer, P., N. Keohane, and C. Timmins. 2009. Migration and hedonic valuation: The case of air
quality. Journal of Environmental Economics and Management 58(1): 1–14.
Bayer, P., and R. McMillan. 2005. Racial sorting and neighborhood quality. NBER Working
Paper No. 11813.
Bayer, P., and R. McMillan. 2010. Tiebout sorting and neighborhood stratification. ERID
Working Paper 49.
Bayer, P., R. McMillan, and K. Reuben. 2004. An equilibrium model of sorting in an urban
housing market. NBER Working Paper No. 10865.
Bayer, P., and C. Timmins. 2005. On the equilibrium properties of locational sorting models.
Journal of Urban Economics 57(3): 462–477.
Bayer, P., and C. Timmins. 2007. Estimating equilibrium models of sorting across locations. The
Economic Journal 117(518): 353–374.
Berry, S. 1994. Estimating discrete-choice models of product differentiation. The RAND Journal
of Economics 25(2): 242–262.
Berry, S., O. B. Linton, and A. Pakes. 2004. Limit theorems for estimating the parameters of dif-
ferentiated product demand systems. Review of Economic Studies 71(3): 613–654.
Berry, S., and A. Pakes. 2007. The pure characteristics demand model. International Economic
Review 48(4): 1193–1225.
Black, S. E. 1999. Do better schools matter? Parental valuation of elementary education.
Quarterly Journal of Economics 114(2): 577–599.
Ellickson, B. 1971. Jurisdictional fragmentation and residential choice. American Economic
Review 61(2): 334–339.
Epple, D. 1987. Hedonic prices and implicit markets: Estimating demand and supply functions
for differentiated products. Journal of Political Economy 95(1): 59–80.
Epple, D., and M. M. Ferreyra. 2008. School finance reform: Assessing general equilibrium
effects. Journal of Public Economics 92(5–6): 1328–1351.
Epple, D., R. Filimon, and T. Romer. 1984. Equilibrium among local jurisdictions: Toward
an integrated treatment of voting and residential choice. Journal of Public Economics
24(3): 281–308.
Epple, D., R. Filimon, and T. Romer. 1993. Existence of voting and housing equilibria in
a system of communities with property taxes. Regional Science and Urban Economics
23(5): 585–610.
Epple, D., and G. J. Platt. 1998. Equilibrium and local redistribution in an urban economy when
households differ in both preferences and incomes. Journal of Urban Economics 43(1): 23–51.
Epple, D., B. Gordon, and H. Sieg. 2010. Drs. Muth and Mills meet Dr. Tiebout: Integrating
location-specific amenities into multi-community equilibrium models. Journal of Regional
Science 50(1): 381–400.
Epple, D., and T. Romer. 1991. Mobility and redistribution. Journal of Political Economy
99(4): 828–858.
Epple, D., and H. Sieg. 1999. Estimating equilibrium models of local jurisdiction. Journal of
Ferreyra, M. M. 2007. Estimating the effects of private school vouchers in multi-district econo-
mies. American Economic Review 97(3): 789–817.
Herriges, J., and C. Kling. 1999. Nonlinear income effects in random utility models. Review of
Klaiber, H. A., and D. J. Phaneuf. 2010. Valuing open space in a residential sorting model of the
Twin Cities. Journal of Environmental Economics and Management 60(2): 57–77.
Klaiber, H. A., and V. K. Smith. 2011. Preference heterogeneity and non-market benefits: The
roles of structural hedonics and sorting models. International handbook on non-market
environmental valuation, ed. J. Bennet, 222–253. Northampton, MA: Edward Elgar.
Klaiber, H. A., and V. K. Smith. 2012. Developing general equilibrium benefit analyses for social
programs: An introduction and example. Journal of Benefit-Cost Analysis 3(2).
Klaiber, H. A., and V. K. Smith. 2013. Quasi experiments, hedonic models, and estimating trad-
eoffs for local amenities. Land Economics 89: 413–431.
Kuminoff, N. V. 2009. Decomposing the structural identification of nonmarket values. Journal
of Environmental Economics and Management 57(2): 123–139.
Kuminoff, N. V. 2011. An intraregional model of housing and labor markets for estimating
the general equilibrium benefits of large changes in public goods. AERE 2011 Summer
Conference Sponsored Session Paper. http://www.webmeets.com/aere/2011/Prog/viewpa-
per.asp?pid=487.
Kuminoff, N. V., and A. S. Jarrah. 2010. A new approach to computing hedonic equilibria
and investigating the properties of locational sorting models. Journal of Urban Economics
67(3): 322–335.
Kuminoff, N. V., and J. C. Pope. 2014. Do ‘capitalization effects’ for public goods reveal the pub-
lic’s willingness to pay? International Economic Review, in press.
Kuminoff, N. V., V. K. Smith, and C. Timmins. In press. The new economics of equilibrium
sorting and its transformational role for policy evaluation. Journal of Economic Literature
51(4): 1007–1062.
Lancaster, K. J. 1979. Variety, equity, and efficiency. New York: Columbia University Press.
Lind, R. C. 1973. Spatial equilibrium, the theory of rents, and the measurement of benefits from
public programs. Quarterly Journal of Economics 87(2): 188–207.
McFadden, D. 1974. Conditional logit analysis of qualitative choice behavior. In Frontiers in
econometrics, ed. Paul Zarembka, 105–142. New York: Academic Press:
McFadden, D. 1999. Computing willingness-to-pay in random utility models. In Trade, theory
and econometrics: Essays in honor of John S. Chipman, eds. J. Moore, R. Riezman, and J.
Melvin, 253–274. London: Routledge.
Nechyba, T. J. 1997. Existence of equilibrium and stratification in local and hierarchical Tiebout
economies with property taxes and voting. Economic Theory 10(2): 277–304.
Palmquist, R. B. 2005. Property value models. In Handbook of environmental economics, Vol.
2, eds. Karl-Göran Mäler and Jeffery Vincent, 763–820. Amsterdam: North Holland Press:
Parmeter, C. F., and J. C. Pope. 2013. Quasi-experiments and hedonic property value methods.
In Handbook on experimental economics and the environment, eds. John List and Michael
Price. Northampton, MA: Edward Elgar.
Poterba, J. M. 1992. Housing and taxation: Old questions, new answers. American Economic
Review 82: 237–242.
tion. Journal of Political Economy 82(1): 34–55.
Sieg, H., V. K. Smith, H. S. Banzhaf, and R. Walsh. 2002. Interjurisdictional housing prices in
location equilibrium. Journal of Urban Economics 52(1): 131–153.
Sieg, H., V. K. Smith, H. S. Banzhaf, and R. Walsh. 2004. Estimating the general equilibrium
benefits of large changes in spatially delineated public goods. International Economic Review
45(4): 1047–1077.
Smith, V. K., H. Sieg, H. S. Banzhaf, and R. Walsh. 2004. General equilibrium benefits for
environmental improvements: Projected ozone reductions under EPA’s prospective anal-
ysis for the Los Angeles air basin. Journal of Environmental Economics and Management
47(3): 559–584.
Starrett, D. A. 1981. Land value capitalization in local public finance. Journal of Political
Economy 89(2): 306–327.
Tiebout, C. M. 1956. A pure theory of local expenditures. Journal of Political Economy
64(5): 416–424.
Tra, C. I. 2010. A discrete choice equilibrium approach to valuing large environmental changes.
Journal of Public Economics 94 (1–2): 183–196.
Tukey, J. W. 1962. The future of data analysis. Annals of Mathematical Statistics 33(1): 1–67.
Walsh, R. L. 2007. Endogenous open space amenities in a locational equilibrium. Journal of
Urban Economics 61(2): 319–344.
Westoff, F. 1977. Existence of equilibria in economies with a local public good. Journal of
Economic Theory 14(1): 84–112.
Wu, J., and S-H. Cho. 2003. Estimating households’ preferences for environmental ameni-
ties using equilibrium models of local jurisdictions. Scottish Journal of Political Economy
50(2): 198–206.
C HA P T E R 15
L A N D S C A P E S I M U L AT I O N S
W I T H E C ONOM ET R IC - BASE D
L A N D U S E M O D E L S
A N DR EW J. PL A N T I NG A A N D DAV I D J. L EW I S
The spatial configuration of land use and land cover has important influences on popu-
lations of birds (Askins 2002; Faaborg 2002) and amphibians (Kolozsvary and Swihart
1999; deMaynadier and Hunter 2000), the health of riverine systems (Gergel et al. 2002)
and estuaries (Hale et al. 2004), human perceptions of scenic quality (Palmer 2004), and
the extent of urban sprawl (Carrion-Flores and Irwin 2004). Land use change results in
changes in the spatial pattern of land use, often in ways that diminish environmental
quality. For example, habitat fragmentation can occur when changes in land use trans-
form a contiguous habitat patch into disjunct patches. Many species of conservation
interest are sensitive to habitat fragmentation, including birds (Askins 2002; Faaborg
2002), amphibians (Kolozsvary and Swihart 1999; Lehtinen et al. 2003), and large mam-
mals (Costa et al. 2005; Noss et al. 2006). Land use change is the leading driver of biodi-
versity loss in terrestrial ecosystems and is expected to remain so in the future (Sala et al.
2000; Wilcove et al. 2000; Millennium Ecosystem Assessment 2005).
Much of the habitat important for biodiversity conservation occurs on privately
owned land. One study found that 70% of species listed under the U.S. Endangered
Species Act (ESA) depend on nonfederal land, most of which is privately owned, for the
majority of their habitat (Natural Heritage Data Center Network 1993). In landscapes
dominated by private ownership, landowners lack the incentive to coordinate decisions
to influence the spatial land use pattern and the environmental outcomes that depend
on it. Econometric-based landscape simulation models have been developed to under-
stand the nature and extent of this market failure problem and to identify and quantify
the effects of corrective land use policies. A landscape simulation begins with a spatial
representation of the landscape, such as a land use map in which the unit of analysis is
a land parcel, and simulates changes in the landscape through the use of rules applied
at the unit scale. An econometric-based simulation model uses rules derived from
Landscape Simulations with Econometric-Based Land Use Models 381
econometric estimation. For example, Lewis and Plantinga (2007) estimate an econo-
metric model that relates observed land use changes to economic returns to alternatives
uses. The econometric results are then incorporated into a landscape simulation model
used to study how forest fragmentation is affected by incentive-based policies that
modify the relative returns to different uses. Lewis, Plantinga, and Wu (2009) analyze
the spatial targeting of incentives to increase contiguous forest habitat, and Lewis et al.
(2011) consider the relative efficiency of voluntary incentive-based policies in achieving
biodiversity conservation objectives. The latter analysis combines an econometric land
use model, landscape simulations, and a biological model of biodiversity that depends
on the spatial pattern of land use.
The development of econometric-based simulations for landscapes dominated by
private ownership presents four basic challenges. The first is to represent variation in
the private economic returns to land at the same scale at which land use varies. Hedonic
price studies reveal that returns to urban land uses vary considerably at fine spatial
scales. Housing prices, for example, are affected by proximity to the central business dis-
trict, roads, and amenities (Wu et al. 2004), as well as by spatial interactions with neigh-
boring parcels (Irwin and Bockstael 2002). Returns to rural land uses, such as cropland
and forests, typically exhibit little variation at this scale because output and input prices
for land-based commodities are relatively constant over space. Factors that can cause
varation in rural land returns include soil quality, which affects crop and timber yields,
and access to markets. Land use regulations, such as zoning restrictions, can also have
important effects on economic returns (Grout et al. 2011).
The second challenge is to model the private information that landowners possess
about the returns to their land. Researchers have incomplete information about private
returns because of unobservable parcel attributes, landowner characteristics such as
managerial expertise, and private nonmarket benefits (e.g., recreation) associated with
particular uses of the land. The random utility framework is a common way to accom-
modate the incomplete information. The returns to land are represented by a determin-
istic component and a random error observed only by the landowner. This gives rise to
a probabilistic model of land use change, as in Lubowski, Plantinga, and Stavins (2006).
Lewis et al. (2011) estimate a mixed logit model that includes random parameters to
account for spatial and temporal correlation in land use decisions. Their results indicate
a significant degree of unobserved heterogeneity in returns to land.
The third challenge is how to best account for land use intensity. In addition to choos-
ing the use of their land, landowners must decide on the intensity of use. For exam-
ple, once the landowner has chosen to develop her land, she must also decide on how
many housing lots to build per acre or how many floors to add to a commercial building.
Likewise, a farmer who allocates his land to crops must decide which crops to produce
and how intensively to cultivate them. Finally, the forest owner must choose species
and rotation length, among other management decisions. Land use intensity is, thus,
the set of secondary choices faced by a landowner once the land use decision has been
made. Land use intensity has important implications for econometric land use models
because it affects the economic return to the chosen use. In many previous studies, land
382 Andrew J. Plantinga and David J. Lewis
use intensity is implicitly assumed in the measurement of net returns to each use (e.g.,
Stavins and Jaffe 1990; Plantinga 1996; Lubowski et al. 2006).1 Lewis, Provencher, and
Butsic (2009) and Lewis (2010), however, model land intensity as a joint decision with
land use. Explicit representation of land use intensity may be warranted if differences
in intensity are important for the landscape-level processes of interest. For example, in
the application presented here, the intensity of development—measured as the number
of shoreline housing lots—has important effects on the green frog population we study.
The fourth challenge arises from the probabilistic nature of the land use transition
rules derived from econometric analysis (Bockstael 1996). The researcher can determine
whether a particular parcel is more likely to convert than another parcel but not that any
particular parcel will convert with certainty. Some analysts present maps showing the
spatial distribution of the estimated probabilities (Bockstael 1996; Cropper et al. 2001),
whereas others form deterministic rules from probabilistic ones (e.g., Chomitz and Gray
1996; Irwin and Bockstael 2002). A problem with the latter approach is that a given deter-
ministic rule is only one of many possible rules. Thus, the simulation produces a single
landscape that represents only one of what is typically a very large number of potential
landscapes. An alternative is to generate a large number of different landscapes conform-
ing to the underlying probabilistic rules. However, one must then summarize this infor-
mation in a way that effectively conveys the range of potential outcomes.
This chapter discusses landscape simulations based on econometric land use models,
emphasizing ways to overcome the four challenges just mentioned. Section 1 reviews
the related literature. Section 2 presents the basic methodology for econometric model-
ing of private land use decisions, and Section 3 describes the use of these models in land-
scape simulations. An application of the methods is provided in section 4, and a final
section considers directions for future research.
1. Previous Literature
Numerous studies in the economics literature seek to explain observed land use
decisions in terms of profit-maximizing behavior. Early studies employed aggregate
(typically county-level) data on land use (Stavins and Jaffe 1990; Plantinga 1996;
Hardie and Parks 1997), whereas more recent analyses have used plot-level data
(Lubowski et al. 2006; Lewis et al. 2011) and spatially explicit land use or land cover
data (Bockstael 1996; Cropper et al. 2001; Nelson et al. 2001; Irwin and Bockstael
2002, 2004; Carrion-Flores and Irwin 2004). An advantage of spatial data is that they
allow spatial processes to be modeled explicitly. For example, Bockstael (1996) uses a
1 Lubowski, Plantinga, and Stavins (2006), for example, construct net returns to forest by assuming
that landowners choose the existing mix of forest species in their county and follow the Faustmann rule
in determining the rotation length.
hedonic function of residential development value to predict the potential developed

value of agricultural parcels. The hedonic function includes measures of distances
to cities, water access, and neighborhood characteristics. The potential development
values are used, along with other controls, in a probit model of land conversion esti-
mated with spatially explicit data.
The results of econometric estimation provide a set of rules governing
parcel-level changes in land use. By combining econometric results with a geo-
graphic information system (GIS)-based landscape representation, Lewis and
Plantinga (2007), Lewis, Plantinga, and Wu (2009), Nelson et al. (2008), and Lewis
et al. (2011) simulate future land use patterns under alternative biodiversity con-
servation policy scenarios. Their simulations account for land conversion into
urban use, as well as exchanges between rural uses (e.g., cropland to forest), which
have important implications for species habitat. Many earlier landscape simulation
studies focused on urbanization (e.g., Bockstael 1996; Carrion-Flores and Irwin
2004) or deforestation (e.g., Nelson et al. 2001). Researchers in other disciplines,
notably geography, have also made important contributions to the development of
spatial models of landscape change (Clarke and Gaydos 1998; Wu 1998, 2002; Li
and Gar-On Yeh 2000 Allen and Lu 2003; Guzy et al. 2008). A criticism of the mod-
els in the geography literature is that the transition rules represent human decisions
yet typically are not based on well-specified and empirically validated models of
human behavior (Wu and Webster 2000).
The literature on systematic conservation planning (SCP) is also concerned with
characterizing future landscapes (Margules and Pressey 2000). In contrast to the
simulation approach just discussed, SCP uses optimization methods to identify sites
for conservation. In the basic formulation of the problem, sites are chosen to maxi-
mize species conservation, subject to a constraint on the total area conserved or total
conservation budget allotted (e.g., Kirkpatrick 1983; Vane-Wright et al. 1991; Camm
et al. 1996; Church et al. 1996; Csuti et al. 1997). Extensions of the basic optimiza-
tion problem incorporate land costs (e.g., Ando et al. 1998; Balmford et al. 2000;
Polasky et al. 2001), considerations of compactness or contiguity (e.g., Fischer and
Church 2003; Onal and Briers 2003), and dynamics (e.g., Costello and Polasky 2004;
Meir et al. 2004; Newburn et al. 2006; Strange et al. 2006). More recent studies in the
SCP literature have analyzed complex spatial patterns that affect species persistence,
including habitat fragmentation and dispersal ability (e.g., Cabeza and Moilanen
2003; Nalle et al. 2004; Moilanen et al. 2005; Polasky et al. 2005, 2008; Nicholson et al.
2006; Jiang et al. 2007).
2. Econometric Models
The results of econometric estimation provide a set of rules governing land use change
in landscape simulations. In this section, we present a general framework for specifying
and estimating econometric land use models with parcel-scale data. We also discuss
strategies to address the first three challenges described: how to represent variation in
the private economic returns to land at the same scale at which land use varies, how to
model the private information that landowners possess about the returns to their land,
and how to account for land use intensity.
2.1 Model Specification

Landowners are assumed to allocate a land parcel of uniform quality to the use that max-
imizes the present discounted value of expected net revenues minus conversion costs.
It is convenient to assume that landowners form static expectations. That is, landown-
ers consider currently available information and form an expectation about the future
annual return to their land that is constant (although can be updated as new informa-
tion becomes available). The assumption of static expectations yields a simple decision
rule under which the use generating the greatest annualized net revenues net of conver-
sion costs is chosen (Plantinga 1996). The problem with relaxing this assumption is that
the land use decision then depends on the future sequence of net returns, and one must
apply Bellman’s equation to find the optimal solution. This complicates the estimation
problem considerably, although previous authors have estimated structural models of
dynamic decision making (e.g., Rust 1989; Provencher 1995). An interesting application
of these methods would be to the land use decision problem.
Assuming static expectations, the landowner compares the annualized net return to
alternative uses and allocates her parcel to the use providing the greatest return. The net
return (NRikt) equals the annual net revenues generated from parcel (Rikt) less annual-
ized conversion costs (Cijkt), where i indexes the parcel, k the chosen use, j the initial use,
and t the time. In general, the researcher cannot observe all of the factors that determine
net returns to the landowner, motivating a specification of net returns that includes
deterministic and random components, such as:
NRikt = β0 jk + β1 jk Rikt + β2 jk Cijkt + µijkt , (1)

where (β 0 jk , β 1 jk , β 2 jk ) are parameters specific to the j-to-k transition and μijkt is a ran-
dom error term.
Lewis et al. (2011) adopt the following specification of net returns, which is a special
case of (1):
NRikt = β0 jk + β1 jk Rc (i )kt + β2 jk LCCi Rc (i )kt + µijkt , (2)

where Rc (i )kt is the average net revenue from use k at time t in county c(i) where parcel i
is located and LCCi indicates the productivity, as measured by the Land Capability Class
(LCC) rating, of parcel i. The LCC system assigns a rating of I through VIII to a land par-
cel, where lower numbers indicate higher productivity for agricultural crops. The inter-
action of Rc (i )kt and LCCi allows the net revenue for parcel i to deviate from the county
average net revenue due to observable land quality. The effects of annualized conversion
costs are assumed to be constant across parcels and time and are measured implicitly by
β0 jk . Similar specifications are used in Lubowski et al. (2006) and Lewis and Plantinga
(2007).
The three studies mentioned earlier meet the first modeling challenge—representing
variation in the private economic returns to land at the same scale at which land use var-
ies—using the interaction of the county average net return and parcel-level land quality.
An alternative approach is to use parcel-level data to estimate separate hedonic price
models for the net returns to each use. These models can incorporate spatial variables,
such as distances to urban centers and features of the surrounding landscape, and be
used to predict net returns for the unselected land uses in the choice set. In this fash-
ion, Bockstael (1996) estimates a hedonic price model of the value of land in residential
housing. Parcel-level predictions of the value of land in residential use are then incorpo-
rated into a probit model to explain development of agricultural land. Carrion-Flores
and Irwin (2004) and Newburn, Berck, and Merenlender (2006) enter the determinants
of net returns (e.g., slope, elevation, soil characteristics, distances to cities, zoning, and
neighboring land uses) directly into the land use change model. The disadvantage of this
reduced-form approach is that one loses economic information—specifically, the rela-
tionship between land use decisions and net returns—that can be used in simulations of
incentive-based policies, such as subsidies for land conversion.
The random utility framework, adopted for all of the econometric land use models
discussed in this section, addresses the second challenge—modeling the private infor-
mation that landowners possess about the returns to their land. In (1), the determin-
istic component of net returns, β 0 jk + β 1 jk Rikt + β 2 jk Cijkt , is assumed to be common
knowledge, whereas the random error μijkt is observed by the landowner, but not by the
researcher. The average net revenues from crop production in a county, Rijkt, are typ-
ically observable, but the researcher is unlikely to observe deviations from the mean
return due to landowner-specific skills, knowledge, and other individual attributes.
These deviations are captured in μijkt and represent a landowner’s private information
about her returns. In all but one of the studies mentioned, researchers impose assump-
tions on the distribution of μijkt that yield probit or multinomial logit models. Lewis et al.
(2011) have panel data on land use change and thus can use a more flexible random
parameters specification:
µijkt = σ1 jk ϖ1c(i ) jk + σ2 jk ϖ2ijk + εijkt , (3)

where (εijkt , ϖ1c(i ) jk , ϖ2ijk ) are random variables and (σ1 jk , σ2 jk ) are parameters. The
random parameters allow for spatial correlation (σ1 jkϖ1c(i ) jk takes the same value for all
parcels within a county) and temporal correlation (σ1 jkϖ2ijk takes the same value for a
given parcel in all time periods).2
A useful property of random utility models is that they define a distribution over—in
the land use context—the maximum net return from each land parcel. Given the start-
ing use j, and K possible land use choices, the maximum net return derived from parcel
i in time t is:
Rijt* = max{β0 jk + β1 jk Rikt + β2 jk Cijkt + µijkt }kK=1 . (4)

The assumption that μijkt is distributed type I extreme value allows (4) to be
rewritten:
1    
Rijt* =  ln 
ξ j   ∑ exp(β
k
0 jk + β1 jk Rikt + β2 jk Cijkt ) − γ  + νijt ,
 
(5)

where γ is Euler’s constant and νijt is distributed type I extreme value with location
parameter equal to zero and scale parameter ξj (Ben-Akiva and Lerman 1985). Equation
(5) can be used in landscape simulations to introduce land uses other than those in the
original choice set, as long as one can assume that landowners will accept the maxi-
mum net return from their parcel as compensation for adopting the new use. Lewis et al.
(2011) use this approach to model habitat conservation on private land. In this context,
equation (5) defines a distribution over landowners’ willingness to accept conservation
payments.
This discussion assumes that the net returns in (1) are fixed. In most cases, however,
the net return is chosen by the landowner when she selects the land use intensity. The
appropriate way to model this is the third challenge discussed earlier. Formally, for par-
cel i, use k, and time t, the landowner chooses the intensity m to solve:
{
NRikt = max NRikt ,m }mM=1 ,
k
(6)
m
where Mk is the number of intensity choices associated with use k. The simplest approach
is for the researcher to assume she knows the solution to (6) (or at least the determinis-
tic component of the solution). Provided that one can observe the choice of intensity, a
2 Many econometric challenges arise with the estimation of econometric land use models,
particularly when spatial processes are an important feature of land use decisions. See Brady and Irwin
(2011) for discussion of these issues.
more flexible approach is to model the intensity decision explicitly. A natural extension
of the random utility models discussed here is to model intensity as a nested choice con-
ditional on land use. In this formulation, the landowner is assumed to simultaneously
choose land use and intensity, conditional on the net returns to each use-intensity com-
bination. Lewis, Provencher, and Butsic (2009) estimate a probit model of the binary
development decision jointly with a count model of the number of housing lots. They
address the sample selection problem inherent to their data; namely, that the number
of housing lots are observed only for developed parcels. Landscape simulations are nor-
mally concerned with the population of land parcels, which argues for the use of econo-
metric methods that can mitigate sample selection bias.
3. Landscape Simulations
The results from the estimation of econometric land use models translate into a set of
rules governing land use change. In the case of random utility models, specifically, the
researcher obtains a K × K matrix of land use transition probabilities for each parcel:
Pijkt = F (X it′ β jk ), (7)

where Xit is a vector of explanatory variables for parcel i in time t (e.g., the net returns
to each of the alternative uses) and β jk is the vector of estimated parameters associated
with the use j-to-k transition. The transition matrices are then matched to parcels in
a GIS using the variables in Xit. Figure 15.1 illustrates how this is done for the model
specification in (2). One obtains GIS layers on land ownership, political boundaries, soil
quality, and initial land cover and overlays them to define distinct parcels on the land-
scape. Each parcel corresponds to a set of transition probabilities defined in (7). The
land ownership layer is needed to eliminate public land parcels since the econometric
model applies only to private lands. The county and soil quality layers indicate the values
of Rc (i )kt and LCCi to use in applying (7), and the land cover layer indicates the initial set
of estimated parameters to use. If j is the initial land use of the parcel, then the relevant
parameter set is β jk , k = 1, . . . , K. In a similar fashion, one can associate a maximum net
return distribution with each parcel in the GIS.
Once this matching exercise is complete, Monte Carlo methods are used to simulate
future changes in the landscape. To begin, suppose that parcel i is initially in crop use and,
according to the matched set of transition probabilities for the parcel, will remain in crops
with a 70% probability and change to pasture, forest, and urban use, each with a 10% prob-
ability. A random draw from a specified distribution, such as a U(0,1), determines whether
the parcel remain in crops (e.g., if the random draw is between 0 and 0.70), changes to pas-
ture (between 0.70 and 0.80), and so on. This procedure is repeated for every parcel and
results in a period t+1 landscape. The transition probabilities are then updated for each
A
ty ty
B Political boundaries
oun un
C o layer
C
Land ownership
layer
Soil quality
layer
Land cover
layer
FIGURE 15.1 Matching land use transition matrices to parcels in a geographic information
system (GIS).
parcel on the t+1 landscape. For example, if parcel i changed to pasture use, then transition
probabilities for this parcel must be computed with a different parameter set. Or, net returns
may be different in period t+1 due, for example, to endogenous price feedbacks. This pro-
cess is repeated until a landscape representation is obtained for the future period of interest.
Of course, the simulated landscape is only one of many possible landscapes consis-
tent with the underlying transition rules. Some earlier authors have used the transition
probabilities to form a deterministic rule for land use change (Chomitz and Gray 1996;
Irwin and Bockstael 2002). For example, Nelson and Hellerstein (1997) and Nelson
et al. (2001) assume that each parcel will be put to the use with the highest estimated
transition probability. This practice, however, is at odds with the random utility frame-
work underlying the econometric model. Because of the unobserved component of net
returns, the researcher does not have the information needed to predict changes in land
use with certainty. Only probabilistic statements about land use changes can be made.
To characterize the range of potential outcomes, one can repeat the process described
here many times to generate a large number of future landscapes, each of which is con-
sistent with the probabilistic transition rules. This, however, raises our fourth modeling
challenge: how does one summarize this information in a way that effectively conveys
the range of potential outcomes?
Lewis and Plantinga (2007) solve this informational challenge with landscape metrics
that summarize the spatial pattern of land use. The focus of their study is forest fragmen-
tation, and so, for each landscape, they compute the average forest patch size and the
area of core forest (forest parcels that are completely surrounded by other forest parcels).
This defines a distribution over the landscape metrics. In a similar fashion, Lewis et al.
(2011) convert each landscape into a biodiversity score using a biological model that
combines simulated landscapes with information on species and habitats. They summa-
rize the results by computing the mean biodiversity score.
An important question that arises in Monte Carlo analysis is how many repetitions
are enough? In the context of our problem, how many landscapes need to be simu-
lated? In most cases, the number of possible landscapes will be astronomically large.
For instance, there are 5 × 1047 possible ways to arrange three land uses on a 100-parcel
landscape. Fortunately, the researcher is interested not in describing all possible land-
scapes, but rather with characterizing the distribution over the outcome of interest,
such as a fragmentation metric or a biodiversity score. Stability in the outcome distri-
bution is likely to be achieved after a relatively small number of simulations. The ideal
approach would be to implement a convergence rule that would end the simulations
when additional landscapes change the outcome distribution in a sufficiently small way,
although this may not be feasible if multiple computer programs are in use. In his study
of forest fragmentation, Lewis (2005) found that the first three moments of the distribu-
tions defined over five fragmentation indices changed very little once 500 landscapes
had been simulated. As a further test, Lewis generated two samples of 500 landscapes
and tested for differences in the sample moments across the two samples. Of course,
these tests need to be done for each application to ensure the stability of the outcome
distributions.
The power of econometric-based landscape simulations lies with their use for inves-
tigating effects of land use policies on landscape-scale environmental outcomes. If Xit
includes measures of net returns, then one can simulate the effects of incentive-based
policies, such as subsidies for afforestation or habitat conservation. Lewis et al. (2011)
evaluate a suite of conservation policies, ranging from a simple per-acre subsidy applied
uniformly across the landscape to targeted policies that account for biological charac-
teristics of land parcels. The authors generate landscapes for each policy scenario, com-
paring the mean biodiversity score in each case to the mean score obtained under a
reference scenario with no policy.
4. Application
In this section, we present an application of an econometric-based landscape simulation

model based on Lewis et al. (2009) and Lewis (2010). A model of shoreline development
along 140 lakes in northern Wisconsin is described. The model represents both the deci-
sion to develop and the development intensity, where the unit of observation is a par-
cel of land. The model is used in a landscape simulation and coupled with a previously
published regression model of green frog populations expressed as a function of a lake’s
development density (Woodford and Meyer 2003).
4.1 Econometric Specification

A landowner is assumed to make a binary decision to develop shoreline parcel i or leave
it undeveloped. Denoting development by k = 1 and the current undeveloped use by
k = 0, conversion is optimal if the net value of conversion (NVC) is positive:
NRi1t − NRi 0t = NVCit = U (X it ) + µ l > 0, (8)

where NVC is measured as a reduced-form function of observable parcel attributes Xit

(e.g., soil quality, distance to town centers) and an unobservable μl specific to lake l (e.g.,
the scenic beauty of the lake). NVC is an indirect function of the land use intensity deci-
sion upon conversion to the developed use. Formally, the value of choosing density m
(i.e., m housing lots) is given by:
V1m (X it ) + ϕi 1m + εi1t , (9)

where Vim is a density-specific function of observable variables Xit, ϕi1m is a

time-invariant density-specific unobservable for development, and ε i1t is an unobserv-
able for developed use in time t that is independent of density. The optimally chosen
density is, then:
mit* ( Xit , ωi ) = argmax{V1m (X it ) + ϕi1m }mM=1 1 (10)

m
and the net return to developed use is given by:
NRi1t = V1m* (X it ) + ϕi1m* + εi1t . (11)

The net return to developed use is a random variable because it is derived by maximiz-
ing over a set of random variables.
A logical modeling approach would be to estimate an econometric model of expected
land use intensity Emit* as a function of a set of observable variables Xit. However, since
both Emit* and the net value of conversion NVCit are derived from operations on the
same set of random variables φilm, there necessarily exists a sample selection problem
*
in estimation of Emit : the analyst only observes the intensity decision for those parcels
converted to the developed use. We assume that we can represent (10) as a Poisson pro-
*
cess, where Emit depends on Xit and the random variable ωi, where ωi reinforces that
the optimal density choice in (10) is a random variable generated by an operation on
the set of random variables φilm. The probability that mit* = m, m = 1, 2, … , M1 follows a
zero-truncated Poisson distribution:
exp[− exp(θ X it + σ2 γ i )][exp(θ X it + σ2 γ i )]m

Pr[mit* = m | X it , γ i ] = , (12)
m !(1 − exp[− exp(θ X it + σ2 γ i )])
where ω i = σ2 γ i is a normally distributed random variable with standard deviation σ2,

implying that λi is a standard normal random variable, and θ is a parameter vector.
To account for the sample selection problem discussed earlier, we assume that the
net value of conversion depends on the unobservable εi1t that is correlated with γi,
specifically:
NVCit = U ( X it ) + µ l = δ X it + µ l + εi1t . (13)
In sum, the binary decision to develop is determined by (13), which features, first, spa-
tial correlation in the unobservables induced by the presence of a common unobservable
(μl) for all parcels on lake l, and, second, an unobservable (εi1t) that is correlated with the
unobservables in the land use intensity decision (γi). If we make the assumption that εi1t
is a standard normal, then the conditional probability of development (dit = 1) is given by,
Pr(dit = 1 | X it , µ l ) = Φ(δ X it + µ l ). (14)

And, if we assume that εi1t and γi are joint standard normal with correlation coefficient
ρ, then using the properties of the joint normal distribution (Greene 2012), we obtain:

(
Pr(dit = 1 | X it , µ l , γ i ) = Φ [δ X it + µ l + ργ i ]/ 1 − ρ2 . ) (15)
Conditioning the probability in (15) only on observables Xit requires integrating out
μl and γi:
∫∫ Pr[m
*
Pr(dit = 1 | X it ) = it = m | X it , γ i ][Pr(dit = 1 | X it , µ l , γ i )]φ(γ i )φ(µ l )d γ i dµ l ,

(16)
where ϕ is the standard normal density function. Thus, the probability of the observed
behavior (dit, mit) on parcel i at time t is,
Pr(dit , mit | X it ) = ∫∫[(1 − d it ) + dit Pr[mit* = mit | X it , γ i ]]
× [Φ((2dit − 1)(δ X it + µ l + ργ i ]/ 1 − ρ2 )]φ(γ i )φ(µ l )d γ i dµ l . (17)

Of particular importance in this model is the lack of statistical independence across

parcel decisions, as γi captures parcel-specific and time-invariant unobservables
whereas μl captures lake-specific and time-invariant unobservables. Thus, this specifica-
tion includes both temporal and spatial correlation in the unobservables.
Lewis et al. (2009) estimate (17) by maximum simulated likelihood, where Dl denotes
the full set of development and intensity decisions on lake l. Conditional on a draw of γi
and μl, the probability of Dl is,
Pr(Dl ) = ∏∏[(1 − d
i t
it ) + dit Pr[mit* = mit | X it , γ i ]]
× [Φ((2dit − 1)(δ X it + µ l + ργ i ]/ 1 − ρ2 )]. (18)

Taking R sets of draws of γi and μl, the simulated approximation to the likelihood
function is,
R
Pr Sim (Dl ) = R−1
∑ Pr(D ).
r
l (19)
r =1
The econometric model is applied to legally subdividable lakeshore parcels across

140 lakes in Vilas County, a popular vacation destination in northern Wisconsin. The
panel data were derived from a number of sources, including a GIS parcel database, the
Wisconsin Department of Natural Resources (WI DNR), US Department of Agriculture
(USDA) soil surveys, and town governments in Vilas County. The GIS parcel database
was constructed from county tax parcel data and historic plat maps and consists of com-
plete spatial coverage of all parcel boundaries in 4-year intervals from 1974 through
1998. The set of independent variables used to estimate (19) consists of parcel character-
istics (size, soil restrictions, distance to town, zoning), lake characteristics (water clarity,
lake size and depth, shoreline open-space), and time-period dummy variables. There
were 335 individual subdivisions that occurred between 1974 and 1998 on a landscape
that began with 1,310 legally subdividable shoreline parcels. The lakeshore development
process was dominated by fairly small developments because 82% of recorded subdivi-
sions generated fewer than six new parcels each. More details on the model, in addition
to estimation results and treatment of potentially endogenous variables, are found in
Lewis et al. (2009).
4.2 Simulation Model and Results

Here, we illustrate two important simulation issues. First, we show how to include
both categorical land use change and land use intensity measures within a landscape
simulation. Second, we demonstrate how an econometric land use model can be cou-
pled with an ecological model as a solution to the problem of summarizing information
from a large number of simulated landscapes. We draw on Lewis’s (2010) simulation
study of shoreline land development in northern Wisconsin. The econometric model
in this chapter treats the net returns to land as a reduced-form expression of a set of
soil characteristics, lake characteristics (water clarity, lake size, etc.), and a zoning policy
variable indicating the minimum shoreline frontage for new residential lots. Output
from the econometric model includes parcel-specific estimates of the probability of
subdivision (a Probit model) and the expected number of new lots upon subdivision (a
Poisson model). Importantly, the Poisson model of the expected number of lots can also
be used to estimate the probability of each possible choice of density (one new lot, two
new lots, etc.).
The development and intensity probabilities are functions of the set of indepen-
dent variables, enabling Lewis (2010) to use the model to simulate the landscape effects
of changes to shoreline zoning policies. The use of a joint model of categorical land use
change and land use intensity raises the challenge of using two probability models (with
correlation across the models) for the simulation. The following steps were used in the
simulation:
1. Following the Krinsky and Robb method (1986), draw a parameter vector from
the econometrically estimated distribution to calculate the estimated Probit and
Poisson probabilities for each parcel.3
2. Standard normal random draws are multiplied by the corresponding standard
deviations from step 1 to generate a draw from the estimated random parameter
distributions.
3. A complete time path (t = 1 . . . T) of development is estimated for each lake.
• Draw a U ~ [0,1] random number r1 for each parcel, where development
occurs if r1 is less than or equal to the estimated subdivision probability; oth-
erwise, the parcel is assumed to remain in its current state.
• If developed, use the estimated Poisson probability, Pr[m* = m], of the num-
ber of new lots m as follows: Draw a U ~ [0,1] random number r2; one new
lot is created if r2 ≤ Pr[m* = 1], two new lots are created if Pr[m* = 1] < r2 ≤
Pr[m* = 2], and so forth until m* is equal to the maximum number of lots
allowable under zoning.
• Repeat these two steps until t = T.
4. Steps 1–3 are repeated to produce a large number of simulated landscapes.
s K is the estimated parameter

3 A simulated parameter vector is equal to ψ = ψ + C ′ λ , where ψ
vector, C is the K×K Cholesky decomposition of the estimated variance-covariance matrix, and λK is a
K-dimensional vector of draws from a standard normal distribution.
This simulation procedure accounts for variation in the estimated model parame-
ters and the random error terms. Furthermore, since step 1 uses the covariance matrix
of parameters from the joint estimation of the Probit and Poisson models, the simula-
tion accounts for the estimated unobserved correlation between land development
and land intensity and implicitly addresses the sample selection problem discussed
earlier.
Each simulated landscape is evaluated in terms of habitat for green frog populations.
The coupled economic-ecological model exploits the predictions from the econometric
model of landscape pattern, which are then used as input to the ecological model. Lewis
(2010) predicts shoreline development across each of 140 lakes, and shoreline develop-
ment density is used to predict the population of green frogs. The green frog population
model is a regression model developed by Woodford and Meyer (2003) that includes
shoreline development density as an independent variable.4 Notably, the Woodford and
Meyer (2003) model was estimated with green frog data from lakes in our study region
in northern Wisconsin. The spatial scale of the model is a lake (i.e., each lake is a habitat
patch), which nicely fits the scale of Lewis’s (2010) simulations, which provide lake-level
estimates of development density. Since development density is the driver of green frog
populations, this model also illustrates the value of modeling a land use intensity choice
rather than just land use categories.
Figure 15.2 illustrates a 20-year forecast from the econometric model as an empirical
distribution of the number of lots on a select lake in northern Wisconsin. As expected,
elimination of the zoning policy increases the likelihood of a larger number of lots being
built. The coupling of the econometric model with the ecological model is performed by
using the predicted shoreline development density for each simulation as an input into
the ecological model to generate a predicted probability of extinction for green frogs.5
Figure 15.2 illustrates how relaxing the zoning constraint along the lakes translates into
a greater probability of extinction for green frog populations. The results in Figure 15.2
draw on a large number of probabilistic landscape simulations, each of which is consis-
tent with the underlying econometric models. Thus, the results are represented in terms
of empirical distributions of development densities and extinction probabilities. By
modifying an independent variable in the econometric model, we see how these empiri-
cal distributions change as a function of a policy change.
4
The regression model from Woodford and Meyer (2003) is very simple and is estimated as E(Frogs
| Lots) = 2.537 − 1.189 × Lots, where Frogs is the number of Frogs per 100 m shoreline and Lots is
the number of developed lots per 100 m shoreline. This function is slightly revised from the original
published version that was sent directly to us by James Woodford. See the original paper, Woodford and
Meyer (2003), for additional information. The Lots variable is generated during each iteration of the
simulation and plugged into this function to get a Frogs measure. More complex ecological models can
be coupled to the economic model provided that the ecological outcomes of interest can be related to the
predicted landscapes. See Lewis et al. (2011) for an example involving a larger set of species.
5 Rather than use the expected number of green frogs, we use the properties of the simple regression
function to generate extinction probabilities. The regression function from Woodford and Meyer has an
Baseline
No
zoning
change
scenario
Number of lots on lake Probability of extinction
FIGURE 15.2 Coupling a landscape simulation with an ecological model for a select lake—
two frequency plots.
5. Future Research
In landscapes dominated by private ownership, landowners lack the incentive to coordi-

nate decisions to influence the spatial land use pattern and the environmental outcomes
that depend on it. Econometric-based landscape simulation models have been devel-
oped to understand the nature and extent of this market failure problem and to identify
and quantify the effects of corrective land use policies. In this chapter, we have dis-
cussed—and suggested solutions to—four challenges that arise with econometric-based
landscape simulations: (1) representing variation in the private economic returns to
land at the same scale at which land use varies, (2) modeling the private information that
landowners possess about the returns to their land, (3) accounting for land use intensity
as part of the land use decision process, and (4) recognizing the probabilistic nature of
the land use transition rules derived from econometric analysis.
Further challenges remain, including what we term the “salt-and-pepper” effect.
To illustrate this, we present a simulated future landscape for the area surround-
ing Madison, Wisconsin (Figure 15.3). The simulation was done using land use tran-
sition probabilities of the form in (7) applied to 30-meter pixels.6 The existing urban
areas are the large black shapes, and most of the small black dots are projected future
urban land. Clearly, the degree to which future urban land is scattered across the land-
scape is unrealistic. One would expect most future urban land to be added near existing
estimated mean number of green frogs per 100 m of E(Frogs | Lots) = 2.537 − 1.189 × Lots. Also, using
the sum of squared residuals, the model has an estimate of σ = 1.48. The probability of extinction is
calculated at each simulation iteration by plugging the predicted Lots into the regression function and
using the cumulative normal distribution function with mean E(Frogs | Lots) and σ = 1.48 to find the
probability that fewer than zero frogs occur on each lake.
6 In particular, each K × K set of transition probabilities varies only by soil quality and county.
FIGURE 15.3 The salt-and-pepper effect (urban land is shown in black).
urban areas and along transportation corridors. One of the reasons for this result is the
decision-making scale. We assumed in this simulation that a land use decision is made at
the scale of each pixel on the landscape, which produces implausibly small areas (dots)
of urban land. But, what is the right decision-making scale? This is a question critical to
land use modeling,7 but not one that can be easily answered in practice. One approach
would be to assume that ownership determines the scale at which land use decisions are
made. That is, each landowner could be assumed to allocate her parcel to a single use.
But, clearly, there are many exceptions to this, as in the case of a diversified farm with
land in crops, pasture, and forests. In the case of rented land, the use—and, particularly,
the intensity—decision may be made by somebody other than the owner. And, finally,
ownership can involve complicated legal arrangements that make it difficult to establish
the actual owner of a particular parcel of land. In their simulation analysis, Lewis and
Plantinga (2007) defined decision-making units in terms of contiguous blocks of land
allocated to single uses. This mitigated the salt-and-pepper effect, but likely had other
shortcomings.
7 In addition to affecting simulated landscape patterns, as demonstrated in Figure 15.3, the
decision-making scale can have important influences on land use decisions if scale economies are
present.
The salt-and-pepper effect also can occur if the econometric land use model fails
to account for important spatial processes. For example, urban development is often
more likely to occur near roads. Ignoring this dependency in the econometric model
will carry through to landscape simulations and likely produce a scattered pattern of
future urbanization. The remedy is to estimate spatially explicit econometric mod-
els of land use, which we regard to be the most important next step in the develop-
ment of econometric-based landscape simulations. The earlier work in this chapter
focused on the linkage between the spatial pattern of land use at the landscape scale
and ecological outcomes but did not emphasize the spatial relationships that affect
land use decisions. To represent these spatial processes, one needs high-quality spa-
tial data to use in the estimation of econometric land use models. These data are
increasingly available, but their use gives rise to a number of additional econometric
challenges. We conclude this chapter by emphasizing the importance of economic
theory in guiding the development of spatial econometric models to be used for
landscape simulations. Readers are referred to Brady and Irwin (2011) for a more
complete discussion.
There are surely important spatial processes that affect land use decisions, but what
are they exactly? Why is urban development more likely to occur near to existing urban
land (one can imagine negative externalities pushing development farther away)? If a
person’s land borders a farm, are they more likely to choose an agricultural use and, if so,
why? These are examples of theoretical questions that should motivate the specification
of spatial econometric models. One finds theoretically grounded spatial land use mod-
els in Irwin and Bockstael (2002) and Lewis et al. (2011). Irwin and Bockstael (2002)
conjecture that residential development creates a negative spatial externality that affects
land use decisions on neighboring land parcels. Their empirical analysis is motivated
by and finds support for the underlying theory. Lewis et al. (2011) model the growth
in organic dairy farms in Wisconsin, accounting for a positive spatial externality that
reduces the fixed costs of learning. In these studies, the underlying theory makes clear
that neighboring land uses are determined endogenously, requiring the use of instru-
mental variables, as in Irwin (2002), or of panel data methods, as in Lewis et al. (2011).
The combination of economic theory and appropriate econometric procedures is criti-
cal if the intent is to use the econometric results in a landscape simulation. In this case,
the underlying spatial process is identified explicitly and thus can be reproduced in the
simulation.
References
Allen, J., and K. Lu. 2003. Modeling and prediction of future urban growth in the Charleston
Region of South Carolina: A GIS-based integrated approach. Conservation Ecology 8(2): 2.
Ando, A., J. Camm, S. Polasky, and A. Solow. 1998. Species distributions, land values, and effi-
cient conservation. Science 279: 2126–2128.
Askins, R. A. 2002. Restoring North America’s birds: Lessons from landscape ecology, 2nd ed. New
Haven, CT: Yale University Press.
Balmford, A., K. J. Gaston, A. S. L. Rodrigues, and A. James. 2000. Integrating conservation
costs into international priority setting. Conservation Biology 11: 597–605.
Ben-Akiva, M., and S. Lerman. 1985. Discrete choice analysis. Cambridge, MA: MIT Press.
Bockstael, N. E. 1996. Modeling economics and ecology: The importance of a spatial perspec-
tive. American Journal of Agricultural Economics 78(5): 1168–1180.
Brady, M., and E. Irwin. 2011. Accounting for spatial effect in economic models of land
use: Recent developments and challenges ahead. Environmental and Resource Economics
48(3): 487–509.
Cabeza, M., and A. Moilanen. 2003. Site-selection algorithms and habitat loss. Conservation
Biology 17: 1402–1413.
Camm, J. D., S. Polasky, A. Solow, and B. Csuti. 1996. A note on optimal algorithms for reserve
site selection. Biological Conservation 78: 353–355.
Carrion-Flores, C., and E. G. Irwin. 2004. Determinants of residential land-use conversion and
sprawl at the rural-urban fringe. American Journal of Agricultural Economics 86(4): 889–904.
Chomitz, K. M., and D. A. Gray. 1996. Roads, land use, and deforestation: A spatial model
applied to Belize. The World Bank Economic Review 10(3): 487–512.
Church, R. L., D. M. Stoms, and F. W. Davis. 1996. Reserve selection as a maximal coverage
problem. Biological Conservation 76: 105–112.
Clarke, K. C., and L. J. Gaydos. 1998. Loose-coupling a cellular automaton model and
GIS: Long-term urban growth prediction for San Francisco and Washington/Baltimore.
International Journal of Geographic Information Science 12(7): 699–714.
Costa, L. P., Leite, Y. L. R., Mendes, S. L., and A. D. Ditchfield. 2005. Mammal conservation in
Brazil. Conservation Biology 19(3): 672–679.
Costello, C., and S. Polasky. 2004. Dynamic reserve site selection. Resource and Energy
Cropper, M., J. Puri, and C. Griffiths. 2001. Predicting the location of deforestation: The role of
roads and protected areas in North Thailand. Land Economics 77(2): 172–186.
Csuti, B., S. Polasky, P. H. Williams, R. L. Pressey, J. D. Camm, M. Kershaw, A. R. Kiester, B.
Downs, R. Hamilton, M. Huso, and K. Sahr. 1997. A comparison of reserve selection algo-
rithms using data on terrestrial vertebrates in Oregon. Biological Conservation 80: 83–97.
deMaynadier, P. G., and M. L. Hunter. 2000. Road effects on amphibian movements in a forested
landscape. Natural Areas Journal 20: 56–65.
Faaborg, J. 2002. Saving migrant birds: Developing strategies for the future. Austin, TX: University
of Texas Press.
Fischer, D. T., and R. L. Church. 2003. Clustering and compactness in reserve site selection: An
extension of the biodiversity management area selection model. Forest Science 49(4): 555–565.
Gergel, S. E., E. H. Stanley, M. G. Turner, J. R. Miller, and J. M. Melack. 2002. Landscape indica-
tors of human impacts to riverine systems. Aquatic Sciences 64(2): 118–128.
Greene, W. H. 2012. Econometric analysis, 7th ed. Upper Saddle River, NJ: Prentice Hall.
Grout, C. A., W. K. Jaeger, and A. J. Plantinga. 2011. Land-use regulations and property val-
ues in Portland, Oregon: A regression discontinuity design approach. Regional Science and
Guzy, M. R., C. L. Smith, J. P. Bolte, D. W. Hulse, and S. V. Gregory. 2008. Policy research using
agent-based modeling to assess future impacts of urban expansion into farmlands and for-
ests. Ecology and Society 13(1): 37.
Hale, S. S., J. F. Paul, and J. F. Heltshe. 2004. Watershed landscape indicators of estuarine benthic
condition. Estuaries 27(2): 283–295.
Hardie, I. W., and P. J. Parks. 1997. Land use with heterogeneous land quality: An application of
an area base model. American Journal of Agricultural Economics 79: 299–310.
Irwin, E., and N. E. Bockstael. 2002. Interacting agents, spatial externalities and the evolution of
residential land use patterns. Journal of Economic Geography 2: 331–354.
Irwin, E. G., and N. E. Bockstael. 2004. Land use externalities, open space preservation, and
urban sprawl. Regional Science and Urban Economics 34: 705–725.
Jiang, Y., S. K. Swallow, and P. W. C. Paton. 2007. Designing a spatially-explicit nature reserve
network based on ecological functions: An integer programming approach. Biological
Conservation 140: 236–249.
Kirkpatrick, J. B. 1983. An iterative method for establishing priorities for the selection of nature
reserves: An example from Tasmania. Biological Conservation 25(2): 127–134.
Kolozsvary, M. B., and R. K. Swihart. 1999. Habitat fragmentation and the distribution of
amphibians: Patch and landscape correlates in farmland. Canadian Journal of Zoology
77(8): 1288–1299.
Krinsky, I., and A. Robb. 1986. On approximating the statistical properties of elasticities. The
Lehtinen, R. M., J. B. Ramanamanjato, and J. G. Raveloarison. 2003. Edge effects and extinc-
tion proneness in a herpetofauna from Madagascar. Biodiversity and Conservation
12(7): 1357–1370.
Lewis, D. J. 2005. Managing the spatial configuration of land: The economics of land use and
habitat fragmentation. Unpublished PhD dissertation, Department of Agricultural and
Resource Economics, Oregon State University.
Lewis, D. J. 2010. An economic framework for forecasting land use and ecosystem change.
Resource and Energy Economics 32(2): 98–116.
Lewis, D. J., and A. J. Plantinga. 2007. Policies for habitat fragmentation: Combining economet-
rics with GIS-based landscape simulations. Land Economics 83(2): 109–127.
tion. American Journal of Agricultural Economics 91(4): 1080–1096.
Lewis, D. J., B. Provencher, and V. Butsic. 2009. The dynamic effects of open-space conserva-
Management 57(3): 239–252.
Lewis, D. J., A. J. Plantinga, E. Nelson, and S. Polasky. 2011. The efficiency of voluntary incentive
policies for preventing biodiversity loss. Resource and Energy Economics 33(1): 192–211.
Li, X., and A. Gar-On Yeh. 2000. Modeling sustainable urban development by the integration
of constrained cellular automata and GIS. International Journal of Geographic Information
Science 14(2): 131–152.
Lubowski, R. N., A. J. Plantinga, and R. N. Stavins. 2006. Land-use change and carbon
sinks: Econometric estimation of the carbon sequestration supply function. Journal of
Margules, C. R. and R. L. Pressey. 2000. Systematic conservation planning. Nature 405: 242–253.
Meir, E., S. Andelman, and H. P. Possingham. 2004. Does conservation planning matter in a
dynamic and uncertain world? Ecology Letters 7: 615–622.
Millennium Ecosystem Assessment. 2005. Living beyond our means: Natural assets and human
well-being. Washington, DC: Island Press.
Moilanen, A., A. M. A. Franco, R. I. Early, R. Fox, B. White, and C. D. Thomas. 2005. Prioritizing
multiple-use landscapes for conservation: Methods for large multi-species planning prob-
lems. Proceedings of the Royal Society, Series B 272: 1885–1891.
Nalle, D. J., C. A. Montgomery, J. L. Arthur, S. Polasky, and N. H. Schumaker. 2004. Modeling

joint production of wildlife and timber. Journal of Environmental Economics and Management
48(3): 997–1017.
Natural Heritage Data Center Network. 1993. Perspectives on species imperilment: A report from
the Natural Heritage Data Center Network. Arlington, VA: The Nature Conservancy.
Nelson, G. C., and D. Hellerstein. 1997. Do roads cause deforestation? Using satellite
images in econometric analysis of land use. American Journal of Agricultural Economics
79: 80–88.
Nelson, G. C., V. Harris, and S. W. Stone. 2001. Deforestation, land use, and property
rights: Empirical evidence from Darien, Panama. Land Economics 77(2): 187–205.
Nelson, E., S. Polasky, D. J. Lewis, A. J. Plantinga, E. Lonsdorf, D. White, D. Bael, and J.
J. Lawler. 2008. Efficiency of incentives to jointly increase carbon sequestration and species
conservation on a landscape. Proceedings of the National Academy of Sciences of the USA
105(28): 9471–9476.
Newburn, D. A., P. Berck, and A. M. Merenlender. 2006. Habitat and open space at risk of
land-use conversion: Targeting strategies for land conservation. American Journal of
Nicholson, E., M. I. Westphal, K. Frank, W. A. Rochester, R. L. Pressey, D. B. Lindenmayer, and
H. P. Possingham. 2006. A new method for conservation planning for the persistence of
multiple species. Ecology Letters 9: 1049–1060.
Noss, R., B. Csuti, and M.J. Groom. 2006. P. 213–251. Habitat fragmentation. In Principles
of conservation biology, 3E. eds. M.J. Groom, G. K. Meffe and C. R. Carrol. Sunderland,
MA: Sinauer Associates.
Onal, H., and R. A. Briers. 2003. Selection of a minimum boundary reserve network using inte-
ger programming. Proceedings of the Royal Society of London B, 270: 1487–1491.
Palmer, J. F. 2004. Using spatial metrics to predict scenic perception in a changing land-
scape: Dennis, Massachusetts. Landscape and Urban Planning 69(2–3): 201–218.
Plantinga, A. J. 1996. The effect of agricultural policies on land use and environmental quality.
Polasky, S., J. D. Camm, and B. Garber-Yonts. 2001. Selecting biological reserves
cost-effectively: An application to terrestrial vertebrate conservation in Oregon. Land
Economics 77(1): 68–78.
Polasky, S., E. Nelson, E. Lonsdorf, P. Fackler, and A. Starfield. 2005. Conserving species
in a working landscape: Land use with biological and economic objectives. Ecological
Polasky, S., E. Nelson, J. Camm, B. Csuti, P. Fackler, E. Lonsdorf, D. White, J. Arthur, B.
Garber-Yonts, R. Haight, J. Kagan, C. Montgomery, A. Starfield, and C. Tobalske. 2008.
Where to put things? Spatial land management to sustain biodiversity and economic pro-
duction. Biological Conservation 141(6): 1505–1524.
Provencher, B. 1995. Structural estimation of the stochastic dynamic decision problems of
resource users: An application to the timber harvest decision. Journal of Environmental
Rust, J. 1989. Optimal replacement of GMC bus engines: An empirical model of Harold Zurcher.
Econometrica 55(5): 999–1033.
Sala, O. E., F. S. Chapin, III, J. J. Armesto, E. Berlow, J. Bloomfield, R. Dirzo, E. Huber-Sanwald,
L. F. Huenneke, R. B. Jackson, A. Kinzig, R. Leemans, D. M. Lodge, H. A. Mooney,
M. Oesterheld, N. L. Poff, M. T. Sykes, B. H. Walker, M. Walker, and D. H. Wall. 2000. Global

biodiversity scenarios in the year 2100. Science 287: 1770–1774.
Stavins, R. N., and A. B. Jaffe. 1990. Unintended impacts of public investments on private deci-
sions: The depletion of forested wetlands. American Economic Review 80(3): 337–352.
Strange, N., B. J. Thorsen, and J. Bladt. 2006. Optimal reserve selection in a dynamic world.
Biological Conservation 131: 33–41.
Vane-Wright, R. I., C. J. Humphries, and P. H. Williams. 1991. What to protect? Systematics and
the agony of choice. Biological Conservation 55(3): 235–254.
Wilcove, D. S., D. Rothstein, J. Dubow, A. Phillips, and E. Losos. 2000. Leading threats to biodi-
versity: What’s imperiling U.S. species. In Precious heritage: The status of biodiversity in the
United States, eds. B. A. Stein, L. S. Kutner, and J. S. Adams, 239–254. Oxford, UK: Oxford
University Press.
Woodford, J. E., and M. W. Meyer. 2003. Impact of lakeshore development on green frog abun-
dance. Biological Conservation 110: 277–284.
Wu, F. 1998. SimLand: A prototype to simulate land conversion through the integrated GIS
and CA with AHP-derived transition rules. International Journal of Geographic Information
Science 12(1): 63–82.
Wu, F. 2002. Calibration of stochastic cellular automata: The application to rural-urban land
conversions. International Journal of Geographic Information Science 16(8): 795–818.
Wu, F., and C. J. Webster. 2000. Simulating artificial cities in a GIS environment: Urban growth
under alternative regulation regimes. International Journal of Geographic Information
Science 14(7): 625–648.
Wu, J., R. M. Adams, and A. J. Plantinga. 2004. Amenities in an urban equilibrium
model: Residential development in Portland, Oregon. Land Economics 80(1): 19–32.
C HA P T E R 16
AN ECONOMIC PERSPECTIVE
O N AG E N T- B A S E D M O D E L S
OF LAND USE AND LAND
C O V E R C HA N G E
DAW N C AS S A N DR A PA R K E R
With the advent of high-performance computing and increased availability of spatial

data, interest is increasing in the development of spatially explicit models across a wide
range of scientific disciplines. These models are being developed to address a host of
growing management challenges related to diverse problems such as urban sprawl, the
decline of former industrial cities, the challenge of ecosystem service preservation in
human-impacted landscapes, containment of invasive species, and emerging global
trends in agricultural production, such as biofuel production, yield gaps, and manage-
ment of genetically modified crops. A variety of new methods for spatial analysis and
modeling have developed in response to these opportunities and challenges. This chap-
ter reviews one family of such new models: agent-based models of land use and land
cover change (ABM/LUCC). ABM/LUCC are computational simulation models that
operate at the scale of real-world decision making and directly represent the decisions
and interactions of economic agents at that scale, over a spatially explicit and dynamic
virtual landscape. The goal of this chapter is not to replicate recent excellent reviews of
agent-based land use models, but rather to provide practical guidance and context for
land economists who wish to understand, evaluate, and construct agent-based land use
models. The chapter strives to answer the following questions, with an economist’s per-
spective in mind:
• What are agent-based land use models?

• How are they structured, how does their structure relate to standard theoretical
and statistical models in economics, and how does this structure facilitate investi-
gation of novel economic questions?
AN ECONOMIC PERSPECTIVE ON AGENT-BASED MODELS OF LULCC 403
• What novel issues need to be considered for their construction and execution,
relative to standard economic models?
• What complementarities exist between these models and other economic model-
ing and analysis methods?
• What are important future research directions for this field?
1. What Are Agent-Based Models of

Land Use Change?
Agent-based modeling (ABM) is a simulation methodology that is increasingly used

throughout the social sciences (Berry et al. 2002; Hernandez et al. 2008; Waldrop
2009). ABMs have been applied to a variety of economic questions (Tesfatsion and
Judd 2006), with some of the most visible work in the area of agent-based financial
markets (LeBaron 2006; Anufriev and Branch 2009). ABMs often represent heteroge-
neous decision-making entities and their interactions with their social and physical
environment (Parker et al. 2003; Irwin 2010). In contrast to mathematical or compu-
tational techniques traditionally used in economics, ABMs are simulation-based, not
equilibrium-based. Although models may reach equilibrium, the equilibrium results
from interactions between lower-level entities. Thus, they are suitable for modeling
domains in which the complex relationships between agent heterogeneity, interactions,
and cross-scale feedbacks render traditional equilibrium-based models analytically
intractable. They are also often applied to explore the out-of-equilibrium dynamics of
an economic system (Arthur 2006). ABMs can be used as computational laboratories
to explore future status quo trajectories for systems of interest, as well as to explore how
modified economic incentives can alter system outcomes (Tesfatsion 2006).
ABM/LUCC combine an ABM of land use change with a spatially explicit landscape,
modeling land use, conversion, management, and exchange events. Some (perhaps all)
of the agents in ABM/LUCC make decisions regarding these events in the modeled land-
scape. Although some ABM/LUCC, including Schelling’s famous model of residential
segregation (Schelling 1971) were developed before the advent of the high-performance
computing and geographic information systems (GIS), a significant number of scientific
applications began to develop during the 1990s (Kohler 2000; Gimblett 2002; Bousquet
and Le Page 2004), in parallel with major advances in high-performance computing,
object-oriented programming, and GIS. (Interestingly, the development of ABM/LUCC
has lagged, to some extent, the development of conceptually parallel individual-based
models in ecology; see Grimm and Railsback [2006]). Following the 2001 workshop on
ABM/LUCC (Parker et al. 2002), the number of scientific publications on this topic has
increased exponentially (Polhill et al. 2011). Several excellent review articles summarize
the contributions of specific models over the past decade, including coupled natural and
404 Dawn Cassandra Parker
human systems models (Parker et al. 2003; Matthews et al. 2007), agricultural models
(Schreinemachers et al. 2010), agent-based land market models (Parker and Filatova
2008), and urban land use change models (Benenson and Torrens 2004; Irwin 2010).
2. Structure, Function, and Relationship

to Standard Economic Models
Analytical equilibrium-based and econometric economic models have been used to

shed light on a variety of important economic problems. For example, the declines in
fuel consumption and the value of outlying properties predicted by traditional model-
ing approaches are robustly observed as fuel prices increase. However, the application of
such models often requires a large number of simplifying assumptions. ABMs are gen-
erally implemented in situations where the research question under study requires the
modeler to relax simplifying assumptions designed to maintain analytical tractability and
closed-form solutions. A more general, flexible model structure is possible because ABMs
use a simulation approach, rather than imposing equilibrium conditions. This simulation
approach, including its technical implications, is discussed in greater details in Section 3. In
this section, the general structure of ABM/LUCC is discussed, highlighting the ways in
which this structure can parallel, but also generalize, traditional microeconomic models.
2.1 Bringing the “Invisible Hand” to Life: An

Illustrative Example
Figure 16.1(a-b) (Nolan et al. 2009) illustrates basic conceptual differences between the
two model types. Traditional models (Panel a) analytically aggregate firm supply and
consumer demand curves (often making simplifying assumptions to ensure continuous
and monotonic functions) into market supply and demand curves. A set of equilibrium
conditions is then imposed and solved. From this solution, market prices and quantities
and corresponding welfare and income measures are derived.
Panel b illustrates a potential spatial agent-based market. As with the traditional
approach, each producer and consumer will have some rule set (which may be a bound-
edly rational form of a traditional supply or demand curve) that links current economic
conditions to selling and buying decisions. However, the representation of each decision
function remains an active aspect of the model, rather than being aggregated into supply
and demand curves. In short, ABMs strive to represent decision making at the scale at
which it occurs in the real world. Each firm and consumer will also have a fixed loca-
tion in space—and therefore a fixed spatial relationship to other economic actors, and
their decisions may be influenced by spatial factors that would be difficult to include at
a microscale in traditional models. For example, Panel b might represent a local market
for residential landscaping plants. Greenhouse firms located in the exurban environ-
ment supply plants for residential landscaping, which consumers buy to landscape their
yards. Firm 1 may decide to adopt a new integrated pest management strategy after
learning, through social interactions and observation, that firm 2 has had success with
this strategy. Firm 3, being located close to firm 1, may price certain products a bit below
Firm 1 in the hopes of gaining their potential customer base (a strategic interaction).
Alternatively, Firm 3 may offer a product not offered by Firm 1, in the hopes of captur-
ing additional dollars from potential shoppers. On the residential side, neighbors may
imitate the landscaping decisions they find attractive, potentially also transferring plant
starts between themselves. Further, their landscaping decisions may be strongly influ-
enced by social norms within the neighborhood that dictate that only certain landscap-
ing practices are acceptable.
(a)
Market price Income welfare

Market quantity measures
Analytical supply Equilibrium Analytical demand

function conditions function
Firm 1 Firm 2 Firm 3 Consumer 1 Consumer 2 Consumer 3

supply supply supply demand demand demand
(b)
Spatially and temporally

distributed realized individual Income welfare
transaction prices and measures
quantities
(Technology adoption)
(Social Network)
Simulated market
s)
tion
or)
F1 F2
avi
rac
beh
nte
C1 C2
i
ion
gic
itat
ate
F3 C3
(Str
(Im
Spatial and social Individual supply

interactions and demand
FIGURE 16.1 Comparison of the structure of traditional economic and spatial agent-based

models (Nolan et al. 2009).
Next, rather than aggregating individual supply and demand functions to obtain a
market clearing price and quantity, actual bilateral market interactions may be simu-
lated. In this simple example, residents will likely purchase plants from greenhouse
suppliers at a fixed price. In other spatial markets—for example, land or labor markets—
bargaining may occur between buyers and sellers over a final transaction price. Buyers
may also conduct an incomplete product search. For example, a buyer’s decision regard-
ing which landscaping firm to stop at may depend on a stochastic travel route through
the countryside. Although this simple example may not seem particularly significant
from an economic perspective at first glance, if landscaping plant purchases are tied, for
example, to the spread of an invasive pest, the spatial dynamics described in this exam-
ple could be critical to understanding the circumstances under which the pest could
spread and cause significant economic damage.
2.2 A Generalized Structure for ABM/LUCC

The previous section discussed essential differences between a traditional micro-
economic model and ABM/LUCC from an economic lens. To provide a more
general overview of the structure of ABM/LUCC, the next section uses the MR
POTATOHEAD (Model Representing Potential Objects That Appear in The Ontology
of Human-Environmental Actions & Decisions) framework (Parker, Brown, et al. 2008;
Parker, Entwisle, et al. 2008) to demonstrate how ABM/LUCC can generalize economic
land use models. MR POTATOHEAD is a hierarchical ontology that describes and cat-
egorizes the key components of ABM/LUCC. Such an ontology is needed for ABMs
because their model structure cannot be described completely using mathematical
equations or statistical algorithms, as is possible for most economic models. The ontol-
ogy describes model elements and their relationship to one another, but does not give
details on specific functions, causal linkages, and algorithms present in models.
MR POTATOHEAD was developed with several goals in mind. The first is to describe
the key components of an ABM/LUCC—an answer to the question, “What elements
need to be specified in order to develop a functional ABM/LUCC?” To this end, Parker,
Brown et al. (2008) identify which elements of the ontology are essential. The second
goal of MR POTATOHEAD is to provide a template that can be used to describe the
structure of a given ABM/LUCC, either for the purposes of assisting model develop-
ment or as a means of documenting existing models. In Parker, Brown et al. (2008), five
separately developed models are described using MR POTATOHEAD. Once disparate
models are described using the same ontology, MR POTATOHEAD facilitates model
comparison, as demonstrated in Parker, Entwisle et al. (2008). An extended goal of the
project is to use the MR POTATOHEAD template as part of a graphical modeling lan-
guage, so that nonprogrammers can easily develop, run, and analyze ABM/LUCC. MR
POTATOHEAD has been implemented in OWL (Web Ontology Language) using the
Protégé/OWL software (Stanford Center for Biomedical Informatics Research 2007).
For the purposes of this chapter, the main classes and select elements (the Landscape
Interfaces to other models Model operation Interaction environments
May have Must have May have
ABM/LUCC
Must have Must have Must have May have
Environment Demographics Land use decision Land exchange
FIGURE 16.2 Key top-level class elements of an agent-based models of land use change
(ABM/LUCC).
element of the Environment class, the Demographic class, the Land use Decision class,
and the Land Exchange class) are discussed in detail and are illustrated using the CMAP
software (Institute for Human and Machine Cognition 2011), which allows creation of
nested graphics. Readers are referred to previous publications for complete details of the
ontology, including examples of how different models implement the various elements.
MR POTATOHEAD contains seven main classes (Figure 16.2). The Interfaces to
Other Models, Model Operation, and Interaction Environments classes are not discussed
in detail here, although event-sequencing mechanisms are discussed later. A more
detailed discussion of the other elements follows.
The Environment class (Figure 16.3) specifies the spatial and socioeconomic elements
that influence land use decisions. It contains several elements: Landscape, Other Spatial
Elements (spatial network and neighborhood models), Non-spatial networks (social,
trade, and affiliation), Institutional/Political Rules and Constraints, Economic Structures
(local markets and economic parameters), Potential Land Uses, and Factors Affecting
Land Productivity. As discussed earlier, many of these elements, such as networks and
neighborhood relationships, are rarely included in traditional economics models. The
Landscape class (Figure 16.3) is used to describe the spatial structure over which the
model operates. Its elements essentially describe the spatial environment in GIS terms.
Although many economists now use data generated through GIS as inputs to empiri-
cal models (Bockstael 1996; Nelson and Geoghegan 2002), few economists run spatial
simulation models over a dynamic spatial landscape (Irwin 2010).
However input data are generated, implementation of an ABM/LUCC requires the
modeler to make key decisions about the spatial structure of the simulation environ-
ment. The Landscape class contains two elements: GIS (spatial data) Layers, and Spatial
Data Structure. It asks the modeler to specify whether the model is empirical or abstract,
the nature of agent-parcel relationships (one-to-one or many-to one, from both sides),
whether parcels can contain multiple management units, whether the data structure is
vector or raster, and whether parcel boundaries are fixed or vary as the simulation runs.
These details highlight the additional spatial structural detail that is possible in ABM/
LUCC relative to traditional economic models. Although some two-dimensional spatial
analytical models have been developed, to the author’s knowledge, they either operate
Institutional/Political
rules and constraints Economic structures
Non-spatial networks Potential land uses

May have
Must have
Factors affecting
Other spatial elements Environment land productivity
Must have
Landscape
Must have
GIS layers Spatial data structure
Must specify
Realism Parcel structure
Agent-parcel relationships
Spatial data types
Decision making units
FIGURE 16.3 The Environment class of agent-based models of land use change (ABM/
LUCC), with detail for Landscape and Spatial Data Structure elements.
over a limited number of cells or represent locations as dimensionless points under con-
tinuous space. (See Albers et al. [2010] and Horan and Lupi [2010] for recent examples
of novel economic spatial models.) Although the flexible spatial structures of ABM/
LUCC create many technical challenges (discussed later), they facilitate exploration of
a wide range of questions, such as urban gentrification and densification (Diappi and
Bolchi 2008; Jackson et al. 2008), effects of parcel size zoning on exurban development
(Robinson and Brown 2009), and effects of market forces and incentive policies on agri-
cultural land consolidation (Happe et al. 2008; Angel et al. 2011).
ABM/LUCC may contain a wide variety of decision-making agents. For example,
urban land use change models may represent residential buyers and sellers, businesses,
developers, and zoning boards. Agricultural models may represent many types of farm-
ing households, large commercial farming operations, input suppliers, purchasers of
agricultural outputs, extension agents, and regulatory agents. The Demographic class
(Figure 16.4) describes the characteristics of the population of agents active in the mod-
els, their demographic dynamics, and the decision-making model of each agent type.
For each agent type, the Agent class (Figure 16.4) describes the Agent Decision Model
(which should minimally include a decision about land use and/or land management),
their Internal Characteristics, and their External Resources. Each of these elements
could be designed to mirror a very traditional microeconomic model. For example, the
Agent Decision Model should specify how agents Calculate Payoffs for each land use or
Calculate payoffs Decision strategy
Must specify
Internal characteristics Agent decision model External resources
Must have
Agent
Must have
Demographics
Life cycle dynamics May have Out migration
Reproduction In migration
FIGURE 16.4 The Agent and Demographic classes of agent-based models of land use change
(ABM/LUCC).
management strategy, as well as the Decision Strategy they would use. Consistent with
a traditional economic approach, payoffs could be calculated through expected profit
or utility, and boundedly rational profit or utility maximization could be used as a deci-
sion strategy. However, alternative models, such as imitation and satisficing, can also be
implemented. Agent’s Internal Characteristics can include factors standard to econom-
ics models, such as human capital, time horizon, discount rate, and risk preference. They
can also include noneconomic factors, such as household age and composition, cultural
preferences, satisfaction thresholds, and propensity to imitate neighbors. External
Resources can also be standard economic factors, such as household labor, physical, and
financial capital, but also noneconomic factors such as status in a social network.
Moving beyond traditional models, ABM/LUCC often embed agent decision
models within a dynamic model of demographic change. In short, models are ini-
tialized with a certain population of agents, and those agents may have demographic
rules governing their growth and decline. Models may have exogenously set rates of
in-and-out-migration. They may also have endogenous reproduction, birth, and death,
and household division, governed by aging and marriage or partnership. Such dynam-
ics may be important for models of residential location and for models of agricultural
household decision making. (See, for example, Jackson et al. [2008] and Torrens [2007]
for urban land use change examples, and Parker, Entwisle et al. [2008] for examples
related to land use change in frontier regions.)
Agent decision model
Must have
Land use decision
Must have
Land use decision data

Potential land uses Neighborhood effects
Must have May have

Institutional/political Parcel accessibility
rules and constraints
Factors affecting
Economic strucures
land productivity
FIGURE 16.5 The Land Use Decision class of agent-based models of land use change (ABM/
LUCC).
In economic models of the land system, two important events are generally mod-
eled, sometimes independently and sometimes together: a land use decision and land
exchange. In MR POTATOHEAD, the Land Use Decision class (Figure 16.5) basically
serves to identify the spatial and social factors that feed into the land use/land manage-
ment decision component of the Agent Decision Model. These factors include standard
drivers of land use change, consistent with the von Thünen and Ricardian conceptual
models of land allocation, including parcel accessibility, other market influences, and
biophysical suitability. However, they can also include nonmarket factors such as neigh-
borhood effects and institutional rules and constraints.
ABM/LUCC can include representations of Land Exchange (Figure 16.6) that include,
but also go beyond, land markets. This explicit modeling of land exchange dynamics sets
ABM/LUCC aside from standard economic models. MR POTATOHEAD characterizes
land exchange as having three main elements: Suppliers of Land, Acquirers of Land, and
Exchange rules.
Suppliers of Land have a Motivation for Supply, a specification of Parcels Supplied, and
Terms Offered for land exchange. These can be purely economic. For example, farmers
may offer particular parcels for rent, at a minimum price of the shadow value of land,
as a result of constrained profit maximization (Berger 2001). Or, developers may offer
residential parcels for sale at a profit-maximizing expected price based on recent com-
parable sales (Magliocca et al. 2011). Alternatively, out-migrating bankrupt household
agents may abandon land, making it available for acquisition without cost (see Parker,
Entwisle et al. [2008] for examples).
Event sequencing Allocation mechanisms
Must have
Land exchange rules
Must have
Land exchange
Must have
Suppliers of land Acquirers of land
Motivation Motivation for

Must have Terms offered Terms offered Must have land acquisition
for supply
Parcels supplied Parcels they hope

to acquire
FIGURE 16.6 The Land Exchange class of agent-based models of land use change (ABM/
LUCC).
In parallel, Acquirers of Land have a Motivation for Land Acquisition, a set of Parcels
they hope to Acquire, and Terms Offered for parcel acquisition. Again, these can be purely
economic: a developer offering a bid for an agricultural parcel based on expectation of
profits from the subsequent sale of residences (Magliocca et al. 2011) or a resident offer-
ing a budget-constrained bid on her highest utility residential parcel (Filatova, Parker,
and van der Veen 2009; Filatova, van der Veen, and Parker 2009). However, land acqui-
sition can also be based on in-migration or the need to maintain household subsistence
(see Parker, Entwisle et al. [2008] for examples).
Exchange Rules consist of both Event Sequencing (triggers for land transfer) and
Allocation Mechanisms. In a land market model, buyers and sellers may be allocated into
the market by the modeler, either as a one-time allocation (Filatova, Parker, and van
der Veen 2009) or as a dynamic flow representing in-migration (Robinson and Brown
2009; Ettema 2010). Alternatively, they may put their house up for sale and seek a new
residence when a dissatisfaction threshold is reached (Benenson and Torrens 2004), or
they may offer a parcel up for sale when profit expectations exceed a certain threshold
(Magliocca et al. 2011). In ABM/LUCC, alternative triggers for land supply and demand
may also be implemented, such as inheritance or a bequest to a newly split household
(Parker, Entwisle, et al. 2008). In a land market model, the Allocation Mechanism is likely
to include a bilateral trade or auction mechanism. However, allocation can also occur
through an agent simply occupying a chosen parcel (in a frontier setting, for instance),
through bequest, negotiation, or even takings.
3. Novel Model Construction and

Analysis Issues
Standard methodologies for the most commonly used modeling methods in economics
are very well developed, and, in general, textbooks and courses detailing these meth-
ods are available at both the undergraduate and graduate levels. In contrast, agent-based
social science models are sufficiently new that standard methodological templates
are not available, and these models are covered only in a small number of specialized
graduate classes. Supporting texts are just now being developed (Railsback and Grimm
2012), but they are not specific to economic applications. Because ABM combines
concepts from social science, computer science, geographic information science, and
simulation modeling, a new practitioner will need to gain familiarity with many new
concepts. Due to the complex systems foundations of these models, practitioners will
also need to approach the modeling of familiar social science concepts in novel ways.
In short, many of the basic assumptions related to economic dynamics are modified in
ABM/LUCC, and these modifications have implications for model design and opera-
tion. Furthermore, practitioners will need to learn and apply new concepts from com-
puter science and simulation. In the following section, this set of novel issues is briefly
reviewed, with each concept supported through specific examples related to the eco-
nomics of land use change.
3.1 Model Design
3.1.1 Simulating Landscape Structure
As discussed earlier, ABM/LUCC can, in principle, be designed with complex and
dynamic spatial structures (vector landscapes, network and neighborhood effects,
compound agent-parcel relationships, and parcel structures that evolve dynamically).
In practice, several practical challenges arise when building ABM/LUCC. The first is
integrating GIS functionality with the ABM. Options are discussed in detail in Parker
(2005) and Castle and Crooks (2006). In general, models that incorporate GIS func-
tionality as part of the ABM have been more successful in terms of speed, performance,
and robustness than models that attempt to build an ABM within a commercial GIS.
A second challenge relates to generation of simulated model landscapes whose prop-
erties structurally resemble real-world landscapes. Many abstract ABMs operate over
fixed, raster-based landscapes, with cells of uniform size and shape. In the real world,
the size and distribution of parcel sizes is rarely uniform, and irregularities in structure
can be very important—for instance, when examining the effects of scale economies
in agriculture or the increasing gradient of parcel sizes in residential landscapes. Some
exciting new methods are evolving to generate simulated landscapes with specific
distributional properties. Morgan and O’Sullivan (2009) use quad-tree algorithms to

generate simulated urban landscapes whose parcel size distribution follows empirically
observed fractal or power-law urban land use distributions. Le Ber et al. (2009) use both
Voronoi and rectangular tessellations to simulate parcel boundaries in agricultural land-
scapes, comparing the empirical performance of each against real-world landscapes. The
most difficult outstanding spatial modeling challenge in ABM/LUCC is the modeling of
parcel division and agglomeration. Each requires a model of how a developer or zoning
board might combine, divide, and redesign a parcelized landscape. Although some prom-
ising work has been done in this area (Alexandridis and Pijanowski 2007), much more
work is needed.
3.1.2 Characterizing Agent Heterogeneity

As mentioned earlier, the ability to represent multiple sources of heterogeneity and inter-
actions in a single model is a driving motivation for the construction of ABM/LUCC.
Referring again to Figure 16.4, agents can be heterogeneous at the class level (implying a
unique set of values for any or all of the key elements: decision model, internal character-
istics, or external resources), or they can be of the same class but simply have variations
along a distribution for any of these elements. Ideally, the sources of agent heterogene-
ity that are included in an ABM/LUCC will be motivated by the research application and
corresponding research questions. However, even when these are identified, questions
remain as to how to represent and measure agent heterogeneity. Theoretically, agent het-
erogeneity can be instantiated through a set of discretely different agent types, drawing on
the concept of classes and subclasses from computer science. For example, Berger (2001)
differentiates between large commercial and smallholder household farmers. In models
of residential land markets, following Schelling (1971), agents are often endowed with dif-
ferent ethnicities (Benenson and Torrens 2004). Land developers can also be modeled as
specializing in particular residential development product types (Robinson and Brown
2009). Often these agent types are identified empirically by applying cluster or principal
components analysis or econometric methods (Valbuena et al. 2008; Schreinemachers
et al. 2009). Alternatively, agent heterogeneity can be implemented by specifying continu-
ous stochastic distributions for particular agent properties. For example, Filatova et al.
(2011) explore the effects of risk perceptions using both theoretical stochastic distribu-
tions and empirical risk perception distributions from survey data. Happe et al. (2008)
initialize their model with a population of representative farms whose characteristics are
derived from census data. Robinson et al. (2007) review additional methods for develop-
ing empirically based agent decision rules, which could also be applied to measure agent
heterogeneity.
3.1.3 Characterizing Agent Interaction

The ability to model agent-agent interaction is another prime motivation for ABM/LUCC
(Polhill et al. 2011). Agent-agent interactions can be either direct or indirect. Direct mod-
eling of land market interactions is a primary innovation for ABM/LUCC (Parker and
Filatova 2008).1 Several excellent agricultural production applications include models of

land rental markets (Balmann and Happe 2000; Berger 2001; Happe et al. 2008). Markets
for land services can also be represented. For example, Mathevet et al. (2003) model mar-
kets for duck hunting rights on agricultural landscapes. Although an obvious potential
application, to the author’s knowledge, no ABM/LUCC with strong economic founda-
tions include endogenously priced local commodity markets. Other direct interactions can
include information transfer (Berger 2001) and imitation of successful agricultural strate-
gies (Polhill et al. 2001). Indirect interactions can generally be understood as externalities.
Distance-dependent spatial externalities are a classic example (Parker and Meretsky 2004).
Others might be characterized as pecuniary externalities. For example, when market price
expectations are based on previous recent sales, a particularly high bid by a single agent can
raise prevailing market prices for all other potential buyers (Magliocca et al. 2011).
Agent-agent interactions are simple to conceive and characterize, but much more
difficult to implement empirically. For example, although transaction price data repre-
senting the final result of bargaining between potential buyers and sellers can often be
obtained, it is almost impossible to obtain data on the initial bid and ask prices of buyers
and sellers. Contingent valuation or experimental methods can be used to estimate will-
ingness to pay (WTP) and willingness to accept (WTA) functions for buyers and sellers
(Plantinga and Lewis, Chapter 15; Cho et al., Chapter 17; Messer et al., Chapter 19).
Effects of spatial externalities can also be measured using spatial hedonic methods or
direct production function methods in the case of agricultural externalities. The struc-
ture and effects of social networks and the ways in which they transmit information has
traditionally been more difficult to measure, although standard surveys can be very use-
ful in this respect. However, with the prevalence of social media, cell phones, and vol-
untary participatory information websites, new avenues are opening for data collection
(Batty et al. 2010; Onnela and Reed-Tsochase 2010).
3.1.4 Event Sequencing

With some exceptions (e.g., game theoretic or experimental models that might have
a first and second mover, individual models of optimal timing decisions, and statisti-
cal duration or hazard models), the sequence of action of agents is not explicitly repre-
sented in economic models. In contrast, in ABMs, the modeler must make deliberate
decisions as to how the sequence of agent action will or might unfold. These rules are
called event sequencing mechanisms.
Event sequencing mechanisms can be predetermined, meaning that the timing of
agent actions and interactions are specified through a set of fixed rules by the mod-
eler. Predetermined event sequencing mechanisms can be synchronous (all agents
1
Although many economists might argue that standard economic approaches already model market
interactions, as illustrated in Figure 16.1, they do not—rather they model an equilibrium based on the
assumption that trades are occurring. ABMs often model the trades themselves, producing outcomes
that differ from those that would be obtained through indirect, equilibrium approaches (Gode and
Sunder 1993; Filatova, Parker, and van der Veen 2009).
are assumed to make simultaneous decisions in each time period) or asynchronous

(only a portion of agents are allowed to be active in a given time period). In a land use
change context, Filatova, Parker et al. (2009) model synchronous bidding, in which all
active buyers simultaneously examine market conditions and place a bid on their high-
est utility parcel. In the next time period, all sellers simultaneously examine bid offers
and accept the highest bid, if it exceeds their WTA. In contrast, Parker and Meretsky
(2004) implement an asynchronous event sequencing mechanism, in which land man-
ager agents make a land allocation decision every other time period in a checkerboard
pattern. Because payoffs to land uses in this model depend on the actions of nearest
neighbors, this event sequencing mechanism avoids economically irrational oscillation
of land uses. Event sequencing mechanisms can also occur according to some stochastic
distribution, for example the “Poisson alarm clock,” in which any agent has a fixed prob-
ability of being active in a given time period. For example, in a housing market model,
a resident might evaluate the utility of her current residence in relation to alternatives
according to this random process.
For predetermined event sequencing mechanisms, there is a tradeoff between the
degree of structure/predictability of the mechanism and the degree of path dependence
introduced. Path dependence refers to sensitivity of model outcomes to initial condi-
tions and/or stochastic elements. A synchronous event sequencing mechanism, since
it has no additional stochastic elements, introduces no path dependence. However, if
agents and their decision environment are highly homogeneous, and if agents are not
modeled as forward looking, this can introduce economically irrational behavior, such
as too-frequent switching of strategies, oscillation, and cyclical behavior. (Consider for
example the classic cobweb model of agricultural supply.) Although a completely sto-
chastic event sequencing mechanism, with a reasonably small number of agents active
in each time period, can resolve this economic irrationality, it introduces a high degree
of path dependence, especially if agents and their decision environment are highly
heterogeneous. As an example, consider the case of technology adoption discussed in
Parker et al. (2003). If technology adoption follows a bandwagon model in which cer-
tain groups adopt only if they observe a given proportion of other agent adopting, the
presence of an early adopter is required to trigger a cascade of technology adoption. In
a highly stochastic model, that early adopter may appear early on in some runs, later in
others, and not at all in still others, leading to a path-dependent variety of outcomes.
In such cases, a high number of model runs may be necessary to map out the complete
potential output space of the model.
The alternative to predetermined event sequencing mechanisms is an event-driven
model. In such models, agents become active decision makers only when internal or
external conditions meet a given threshold. For example, agricultural producers may
decide to sell land when they hit a solvency constraint (Polhill et al. 2008b) or when a
family farmer dies or retires (Lynch and Lovell 2003). An urban resident may decide to
relocate when household family structure changes due to demographic transitions or
when the household becomes unsatisfied due to the demographic or income composi-
tion of the neighborhood (Benenson and Torrens 2004). Event-driven mechanisms can
obviously introduce a higher degree of path dependence into ABM/LUCC. However,

their advantage is that they can mirror real-world structure and dynamics, which are
often central to the research questions of interest. For example, event-driven models
have been used to examine agricultural land consolidation (Happe et al. 2008; Angel
et al. 2011), the emergence of spatial segregation (Benenson and Torrens 2004), and
urban gentrification (Diappi and Bolchi 2008; Jackson et al. 2008).
3.1.5 Representing Boundedly Rational Optimization

ABMs are generally applied to systems that do not have closed-form mathematical solu-
tions, implying that the modeler herself has incomplete information, before models are
run, regarding the path and final state of price and quantity outcomes (Anufriev and
Branch 2009; Nolan et al. 2009). Because of the degree of interdependencies built into
these models, agents’ decisions are contingent on the path of previous agents’ decisions,
due to stochastic initial conditions and event sequencing mechanisms. These issues
are present even for short-run equilibrium problems but are exacerbated for dynamic
problems. Even in nonstochastic environments, mapping out the state space of possible
outcomes and their best responses may be computationally intractable. (Consider, for
example, the limitations of computerized chess programs.) Thus, agents in ABMs prac-
tically must be modeled as having incomplete information—putting them into the class
of bounded rationality, as discussed by Simon (1996).2
Within this limited information context, however, agents can be modeled as opti-
mizers. In short, some model of learning or expectations formation must be formally
included in models in order to acquire an estimate of uncertain future parameters.
This mandate for boundedly rational agents is clearly illustrated through alternative
approaches to modeling residential land markets. Traditional closed-form models use
simplifying assumptions (equal utility for homogeneous agents or equal utility to an
exogenous, outside housing option) to identify housing prices, so that buyers’ bid prices
can be derived through budget-constrained utility maximization. In ABM land market
models, since both buyers and the spatial goods being traded are heterogeneous, it is
analytically impossible for a given buyer to exactly anticipate the cost of housing as a
function of its characteristics. Thus, the utility maximization problem cannot be directly
solved. One alternative (as proposed by Parker and Filatova [2008]) is to develop induc
tive models of price expectation formation and to use these expected prices to solve
for a formal demand function. Both Ettema (2010) and Magliocca et al. (2011) have
developed alternative price expectation formation models that could be used for this
purpose, modeling a role that, in the real world, is provided by real estate agents or infor-
mation sites such as Zillow. Inductive price expectation models could also be applied
to estimate future agricultural commodity input and output prices. Again, this strat-
egy would parallel the real world because actual agricultural supply decisions are based
2 Other sources of bounded rationality can and have been represented in ABMs; for example,
satisficing behavior (Gotts et al. 2003).
on incomplete estimates of future costs and prices. Although some related work has
been done for financial markets, much more research is needed to understand which
models are, in fact, consistent with how real-world agents form price expectations, how
much real-world variation there is in expectations formation mechanisms, and how
alternative models interact to influence actual market prices. Experimental economic
approaches can help to fill in this gap.
3.1.6 Modeling Learning

ABMs (and the parallel individual-based models in ecology; see Grimm [2006] for an
overview) often incorporate models of learning and adaptation—about the agent’s envi-
ronment, about the behavior of other agents, or about the success of various strategies. As
can be seen from the discussion in Section 3.1.5, price expectation formation can be a key
example of such learning in ABM/LUCC. Ettema (2010) and Magliocca et al. (2011), for
example, both model price expectation formation in residential land markets, for devel-
opers, rural sellers, and residential buyers, respectively. Agricultural agents may also
learn about the relative success of production options through both experimentation and
imitation (Gotts et al. 2003; Polhill et al. 2008b), about the cost of compliance with pest
control regulations (Carrasco et al. 2012), and about anticipated value of rental land and
the optimal bidding strategies on the land market (Kellermann and Balmann 2006).
3.1.7 Equilibrium
The review of these design issues highlights the dynamic, evolutionary nature of most
ABM/LUCC. These evolutionary dynamics may be inconsistent with the concept of
economic equilibrium (Arthur 2006; Parker and Filatova 2008). In fact, some authors
argue that economic systems, and land markets in particular, should be characterized
and studied as nonequilibrium systems. Even for equilibrium models, ABMs allow
more exploration of the path toward equilibrium than do traditional models (Nolan
et al. 2009). However, in some cases, a researcher may construct an ABM/LUCC with
the specific goal of extending an analytical equilibrium model, in which case the equilib-
rium properties of the extended model may be of interest (Caruso et al. 2007; Filatova,
Parker, and van der Veen 2009; Filatova, van der Veen, and Parker 2009, 2011; Parker
and Meretsky 2004). In such models, equilibrium can be achieved by holding the num-
ber of active agents in the model fixed in a given time period and instituting a stopping
rule when no further trades occur. This rule is consistent with the standard concept of
a short-run economic equilibrium—no additional economic activities for which gains
from trade are positive are possible for active agents. In a residential housing or agri-
cultural commodity market context, such a short-run equilibrium is consistent with a
seasonal market. (Note that agents participating in such markets may use temporally
dynamic, forward-looking models to assess future payoffs to land use.)
ABM/LUCCs generally become dynamic when the population of active agents
changes over time. For example, Robinson and Brown (2009) and Magliocca (2011)
assume growing populations of buyers in order to model expanding urban areas. For
land markets, even if net population is fixed, some exogenous entry and exit to markets
may be needed if endogenous land supply decisions are modeled. In short, since a relo-
cating buyer may need to sell his or her current property before entering the market, an
active relocation market requires that some properties be on the market at any given
time. Ettema (2010) resolves this issue by modeling exogenous entry and exit but keeps
the total population of agents fixed. In principle, if agent population characteristics were
fixed over time, such a residential land market could reach a steady state in which rates
of exit and entry were equal, and no agents currently in the market had an incentive to
relocate, given current market opportunities. Similarly, an agricultural market could be
viewed as being in a long-run equilibrium if no agents had an incentive to change their
production decisions or buy or sell land. Whether markets such as these exist in the real
world is an open question.
3.2 Experimental Design
3.2.1 Pseudo-Inductive Modeling
In most economic analysis, theoretical and empirical modeling are distinct (although
potentially logically connected) activities, and models are constructed, presented,
and analyzed separately. Theoretical models are derived deductively, with a set of
assumptions regarding model structure leading mathematically or logically to a set
of equilibrium conditions and a corresponding set of comparative static or dynamic
propositions. The derivation of a supply curve from profit maximization is a classic
example. Empirical models are generally inductive, distilling patterns or trends from
real-world data. Econometric modeling can be considered inductive because it essen-
tially is a pattern analysis technique that calibrates a set of best-fit coefficients to an
assumed set of mathematical relationships. For example, supply curves may be esti-
mated econometrically, with the expectation (derived from the theoretical model) that
the coefficient on price should be positive. Some exceptions to these generalities exist.
For example, mathematical programming (MP) models are often parameterized with
a set of empirically derived coefficients, while basing their mathematical structure on
theory. Computable general equilibrium (CGE) models also start with a mathematical
structure based on theory, but then calibrate a set of coefficients that best fit real-world
outcome data.
As with MP and CGE models and their theoretical antecedents, ABM/LUCC mod-
els can be purely theoretical or highly empirical. However, it is generally acknowledged
that ABMs cannot be classified as purely inductive or purely deductive (Axelrod 1997).
Essentially, these models begin with a set of structural assumptions, per a deductive
approach. However, rather than generating a set of equilibrium conditions, theoretical
propositions, or axioms, as would an analytical or logical model, model runs generally
generate multiple data observations—often with a structure that parallels the real-world
data that would be used to calibrate an econometric or CGE model. Such generated data
can then be analyzed inductively to search for regular patterns that can form the analog
of the theoretical propositions produced by a closed-form model.
For example, ABM land market models are built on traditional (but boundedly ratio-
nal) models of land supply and demand. These models generate a spatial and tempo-
ral landscape of successful transactions, as well as of unsuccessful land bids and sales
attempts. Each data point has associated spatial (property and accessibility) character-
istics, as well as associated characteristics and behaviors of economic agents (informa-
tion often not available in real-world data). These data can be analyzed using statistical
methods to estimate a hedonic land rent function, in parallel to real-world econometric
models (Filatova et al. 2009). Alternatively, econometric modeling that examines the
relationship between input parameters and macroscale outcomes of interest can be used
to conduct model sensitivity analysis (Happe et al. 2006). When a theoretical model is
applied in this context, the inductive model’s results can, in principle, play a similar role
to comparative statics and dynamics, producing a set of testable hypotheses that are
embedded in the estimated coefficients. For example, a negative estimated coefficient
on distance in a hedonic land rent function estimated from computational data parallels
the theoretical downward-sloping land rent gradient derived in the traditional analyti-
cal von Thünen/Alonzo model. (However, as discussed further later, particular issues
related to complex systems data—nonlinearities, thresholds, and endogeneity between
micro- and macroscale elements—can render traditional statistical approaches inap-
propriate, leading to a new set of outstanding challenges.) When an empirically param-
eterized model is applied in this context, the estimated model provides a direct target
for empirical model validation (testing to see whether the outputs of the model have a
reasonable correspondence to their real-world analogs). (See Fagiolo, Birchenhall et al.
[2007] and Fagiolo, Moneta et al. [2007] for more extensive discussion of validation in
ABMs, and Verburg et al. [2006] for an overview of validation in land use models.)
3.3 Model Construction, Execution, and Analysis

3.3.1 Software Choices and Resources
One current challenge for developers of ABM/LUCC is that no standard software exists
for ABM in the social sciences that is appropriate for large-scale scientific modeling
analysis and that does not require a high level of programming ability. These issues,
and some current popular programming environments, are discussed in more detail in
Parker et al. (2002), Nolan et al. (2009), and Castle and Crooks (2006). The models dis-
cussed in this chapter have been programmed in a variety of environments, including
Swarm, C++, RePast (Java libraries for ABM), Netlogo, and Matlab. Little or no code
sharing occurs between research groups, in spite of openness to making code available
and a consensus that code sharing might bring efficiency gains. Several factors likely
contribute to this situation. First, although funding for the development of ABM/
LUCC has increased markedly in the past decade, especially in the United States and the
European Union, it is difficult to obtain funding to develop a general code base, espe-
cially one that brings together code from separately funded projects. At the same time,
the user base is too small to support the development of commercial ABM software.
Furthermore, a wide variety of modeling approaches exist concurrently (Parker et al.
2003; Richiardi et al. 2006; Fagiolo et al. 2007), meaning that it might not be clear what
set of standard models should be included in the core of a community modeling library
or a commercial software product. Finally, models are often developed for different pur-
poses and to address different questions, thus reducing the amount of potential shared
code. Yet, other fields have successfully developed community modeling libraries that
support multiple research endeavors (Krieger 2006; Gent et al. 2011), and this author
hopes that similar standard libraries will be developed for ABM/LUCC (Parker, Brown,
et al. 2008). These would potentially substantially lower model development costs and
reduce barriers to entry to the field.
3.3.2 Communicating Model Structure and Results

As discussed extensively in other publications (Allesa et al. 2006; Richiardi et al. 2006;
Parker, Brown, et al. 2008), because ABM/LUCC cannot be expressed solely in terms
of mathematical equations or statistical algorithms, communication of model rules,
structure, and function can be a major challenge. Modelers are encouraged to publish
model code or executables along with research findings. A growing number of journals
are providing archival links to code along with electronic publication, and the Open
ABM website (openABM.org) has also been established as a model code archive for
ABMs of coupled human-natural systems. However, perusing model code is a highly
inefficient way to discern model structure, especially as the same model can be imple-
mented in many alternative software libraries. Current alternative model communi-
cation protocols include the MR POTATOHEAD framework developed specifically
for ABM/LUCC, demonstrated here, and the ODD protocol and its extensions, devel-
oped for any individual or ABM (Grimm et al. 2010; Groeneveld et al. 2012). These
two protocols have been applied to the same subset of ABM/LUCC (Parker, Brown,
et al. 2008; Polhill et al. 2008a). Many computational modelers, including this author,
however, imagine a future in which readers will not only be able to understand model
rules, but will be able to directly interact with models, including not only examining
multiple output visualizations from the published experiments, but also running alter-
native models and analyzing their output, without having to modify programming
code—and ideally without having to download models. An extensive set of program-
ming libraries that would support such modeling is described by Parker, Brown et al.
(2008), and a supporting a new format for research publication is described in detail by
Mesirov (2010).
3.3.3 Generating Data
As with any simulation model that has a stochastic element, multiple model runs
are often required for ABM/LUCC, even for a fixed set of parameters. In short, sto-
chasticity can enter models both through differing initial conditions, through event
sequencing, and through any other stochastic element, such as a demographic event,
an exogenous price, or a biophysical condition whose value might depend on a sto-

chastic draw. Ideally, Monte Carlo model runs should be conducted to completely
map outcomes for any parameter set. Often, to further complicate matters, a modeler
may want to run models for multiple parameter values, to perform sensitivity or sce-
nario analysis. Thus, a distribution of model outcomes will exist for each parameter
set, and a parameter sweep will consist of a large collection of output distributions,
resulting in a large output database whose generation is not conceptually complex
but is computationally intensive.
3.3.4 Analyzing Generated Data

A major outstanding challenge for ABM/LUCC lies in how to effectively analyze
this wealth of generated data in order to answer research questions of interest. In
theory, by following the steps given in the preceding sections (building a theoret-
ical model and generating a database of outcomes under an appropriate range of
initial conditions, random variations, and model parameters), the modeler should
produce a collection of simulated data similar to what he or she might have access
to in the real world. This should be, in principle, an opportunity and not a problem,
especially for economists. Economists love data, and most have years of rigorous
academic training in statistical data analysis methods. The problem lies in the struc-
ture of the models that generated the data, which itself dictates the structure of the
data. Complex systems are characterized by nonlinear, nonmonontic, nonstation-
ary relationships; non-Gaussian (power law) distributions; and feedbacks across
scales. This implies that the relationships between model parameters and model
outputs are not likely to be additive, linear, and monotonic, and input data are not
likely to be normally distributed (Fagiolo, Birchenhall, and Windrum 2007; Fagiolo,
Moneta, and Windrum 2007). The majority of econometric methods are developed
for data that are separable, monotonic, and stationary. Therefore, new data analysis
methods need to be developed. Again, some promising work in this area is under
way, especially related to examination of the joint influence of changes in multiple
model parameters on model outputs. Happe et al. (2006) conduct regression-based
meta-modeling using a design of experiments approach to examine the sensitivity of
agricultural structural change to agricultural household level drivers. Gimona et al.
(2011) use regression tree analysis to analyze potentially nonlinear relationships
between multiple land preservation incentives and species diversity in an ABM/
LUCC. Ligmann-Zielinska and Sun (2010) use time-dependent global sensitivity
analysis to separate independent and interaction effects of behavioral parameters
on the time-path of fragmentation in residential landscapes, including analyzing
thresholds and regime shifts. Yet again, much more basic work is needed in this area,
and, given economists’ tremendous skill and experience in statistical data analysis
for social systems, their potential contributions to the effort are substantial (Nolan
et al. 2009).
4. Conclusion
With the perspective of an economist new to ABMs of land use change in mind, this
chapter has defined and described this class of models and has offered a detailed tech-
nical discussion of potential novel issues that an economist might face in their design,
construction, and analysis. Numerous examples of existing ABM/LUCC have been used
for illustration purposes, thus providing a technically focused review of current work in
the field. To conclude the chapter, some thoughts on promising future research direc-
tions and a call for bold forward movement are offered.
4.1 Complementarities with Other Economic

Modeling Methods
As illustrated through the many examples cited in this chapter, ABMs are often used to
expand the range of research questions that can be addressed using economic model-
ing, and, in these cases, the ABM approach can serve as a substitute for more traditional
approaches. Yet the development of ABMs often takes advantage of other modeling
approaches that complement ABM. For example, simple versions of ABMs are often
developed to replicate well-established theoretical analytical models, in order to pro-
vide structural validation for the ABMs. Econometric methods can also be used to
empirically test the hypotheses generated by theoretical ABMs. Both econometric and
experimental approaches are essential in helping to develop decision models for empiri-
cally grounded ABMs. Thus, the development of economically based ABM/LUCC
provides a new opportunity for collaborative research and knowledge sharing with
better-established economic modeling methods.
4.2 Future Research Directions: Cross-Scale Modeling

The ABM/LUCC cited in this chapter operate at a single scale—generally a land par-
cel (spatial scale) or a land manager managing multiple parcels (behavioral/institu-
tional scale). These models have made important methodological advances and have
been used to answer novel research questions. Yet many of the most interesting research
questions related to the operation of spatial markets require models that operate across
scales (Irwin 2010; Chen et al. 2011).
For example, to understand the patterns of growth and decline of residential land
markets within cities, the relationship between employment centers and residential
locations must be understood. This relationship operates at a neighborhood, rather
than a parcel scale. At a regional scale, some understanding of the local economy is also
required: what jobs will be created and lost, who will migrate into and out of the urban
area as a result, and what will be their demographic and socioeconomic characteristics?
This requires models that at least interface with regional and national scale models.
A second example relates to regional and global commodity production—for exam-
ple, understanding the effects of new demand for biofuels on food commodity produc-
tion. At a local scale, agricultural production decisions are shaped by both parcel- and
household-level characteristics, such as biophysical suitability and household knowl-
edge and resources, and by the incentives provided by potential sales prices for com-
modities in external markets. These factors have been effectively modeled using ABM/
LUCC. However, at a regional, national, and global scale, local production decisions will
modify external prices. These feedbacks have not been effectively modeled using ABM/
LUCC. Furthermore, regional and global commodity supply models generally fail to
account for the impacts of local spatial and agent-level heterogeneity on commodity
production. The two modeling scales need to be brought together. In particular, work
needs to be done to explore the extent to which ABM/LUCC agricultural production
models can be integrated with higher scale computable general equilibrium models.
These efforts could potentially be informed by ongoing efforts to link integrated assess-
ment models with lower scale supply and demand models and examples of coupled
models from other domains (Energy Modeling Forum 2012; Rausch and Mowers 2012).
Finally, tradable permit models for the preservation of ecosystem services have
received much attention, both from a theoretical and applied policy perspective
(Tietenberg 2005). However, such models rarely account for important sources of spatial
heterogeneity in potential markets. Carbon markets are a potentially promising applica-
tion area. Currently, markets for carbon sequestration are a patchwork of regional and
national policies, with little coordination and uniformity of structure and regulatory
level. ABM/LUCC could potentially be used to explore how these markets might func-
tion together if global standards were implemented, but local programs were allowed to
be maintained to meet standards, thus exploring how patterns of carbon emissions and
trading would change if markets were integrated. Models could also track the develop-
ment of pollution hot spots and patterns of other ecosystem service generation (e.g.,
biodiversity preservation). However, again, higher scale market models would need to
be developed to track permit trades and market clearing at a regional scale.
4.3 A Call for Innovation

Economics tends to be a methodologically conservative discipline. This conservatism
certainly has its benefits: highly technical standards for academic training, detailed
technical scrutiny of new methods, and rigorous peer review for published work. Yet, in
a world where policy challenges are emerging at a rapid rate and unexpected global and
environmental crises are capable of destabilizing global commerce, a portfolio of tech-
nical approaches to economic analysis is needed—some standard and codified by years
of use, others novel, innovative, and even risky. It can be quite difficult for an economist
using novel methods to succeed in a purely economic academic context due to a high
degree of challenge involved in publishing analysis based on nontraditional methods in

economics journals, navigating the review and promotion process, and obtaining grant
funding from traditional sources (although new funding initiatives by the US National
Science Foundation have relaxed this last constraint).
There is wide consensus in the economics profession that economics is not a particu-
lar method but rather more generally the science of the study of scarce resources. A wide
variety of models have been used for this purpose, with new methods emerging, proving
their utility, and gaining broader acceptance. I encourage the economics community to
give reasonable consideration to any new methods that logically and rigorously study
this allocation. Given the global-scale challenges we face and the need to quickly develop
novel responses, I issue a call to the economics profession to consider, evaluate, and test
these new methods with a moderately higher degree of tolerance for risk and uncer-
tainly than is the norm in the profession. In fact, a more diverse portfolio of research
approaches could be seen as an economically rational response to new challenges and
the uncertainty they entail. The rewards for the profession—and more importantly,
for policy analysis—could be huge. As pointed out by Nolan et al. (2009), econometric
methods were once new, also, but have provided proven utility for the profession over
time (Messer et al., Chapter 19, this volume.). Experimental methods are even newer but
have also proven their utility. Much of the work cited in this chapter represents careful,
economically grounded work by classically trained urban, environmental, resource, and
agricultural economists. A subset of that work has been successfully published in eco-
nomics journals. Hopefully, these works have broken the ground for the next generation
of agent-based land use models to find a home within the economics literature.
Acknowledgments
This project has benefited from funding from and extensive discussions with the
SLUCE2 project team (funding by US NSF CNH-0813799) related to this concept of
static versus dynamic equilibrium in land markets. The manuscript content has also
benefited from discussion with many participants at Waterloo Institute for Complexity
and Innovation (WICI.CA) seminar and workshop participants.
References
Albers, H. J., A. Ando, and J. F. Shogren. 2010. Introduction to spatial natural resource and
environmental economics. Resource and Energy Economics 32(2): 93–7.
Alexandridis, K., and B. C. Pijanowski. 2007. Assessing multiagent parcelization performance
in the MABEL simulation model using Monte Carlo replication experiments. Environment
and Planning B 34(2): 223–244.
Allesa, L. N., M. Laituri, and M. Barton. 2006. An “all hands” call to the social science com-
munity: Establishing a community framework for complexity modeling using agent based
models and cyberinfrastructure. Journal of Artificial Societies and Social Simulation (4),
http://jasss.soc.surrey.ac.uk/9/4/6.html.
Angel, N., M. North, E. Tatara, C. E. Laciana, E. Weber, and F. Ruiz Toranzo. 2011. An agent
based model to simulate structural and land use changes in agricultural systems of the
argentine pampas. Ecological Modelling 222(19): 3486–3499.
Anufriev, M., and W. A. Branch. 2009. Introduction to special issue on complexity in economics
and finance. Journal of Economic Dynamics and Control 33(5): 1019–1022.
Arthur, W. B. 2006. Out-of-equilibrium economics and agent-based modeling. In Handbook
of computational economics, Vol. 2 Agent-based computational economics, eds. L. Tesfatsion
and K. L. Judd. Amsterdam: Elsevier B.V., 1551–1564.
Axelrod, R. 1997. Advancing the art of simulation in the social sciences. In Simulating social
phenomena, eds. R. Conte, R. Hegselmann, and P. Terna. Berlin: Springer, 21–40.
Balmann, A., and K. Happe. 2000. Applying parallel genetic algorithms to economic prob-
lems: The case of agricultural land markets. Paper read at “Microbehavior and Macroresults,”
IIFET 2000 Proceedings, at Corvallis, Oregon USA.
Batty, M., A. Hudson-Smith, R. Milton, and A. T. Crooks. 2010. Map mashups, Web 2.0 and the
GIS revolution. Annals of GIS 16(1): 1–13.
Benenson, I., and P. Torrens. 2004. Geosimulation: Automata-based modeling of urban phenom-
ena. London: John Wiley & Sons.
Berger, T. 2001. Agent-based spatial models applied to agriculture: A simulation tool for
technology diffusion, resource use changes, and policy analysis. Agricultural Economics
25(2–3): 245–260.
Berry, B. J. L., L. D. Kiel, and E. Elliot. 2002. Adaptive agents, intelligence, and emergent human
organization: Capturing complexity through agent-based modeling. Proceedings of the
National Academy of Sciences of the USA 99(Supplement 3): 7178–7188.
tive. American Journal of Agricultural Economics 78: 1168–1180.
Bousquet, F., and C. Le Page. 2004. Multi-agent simulations and ecosystem manage-
ment: A review. Ecological Modelling 76(3–4): 313–332.
Carrasco, L. R., D. Cook, R. Baker, A. MacLeod, J. D. Knight, and J. D. Mumford. 2012. Towards
the integration of spread and economic impacts of biological invasions in a landscape of
learning and imitating agents. Ecological Economics 76: 95–103.
Caruso, G., D. Peeters, J. Cavailhes, and M. Rounsevell. 2007. Spatial configurations in a peri-
urban city: A cellular automata-based microeconomic model. Regional Science and Urban
Economics 37(5): 542–567.
Castle, C., and A. T. Crooks. 2006. Principles and concepts of agent-based modelling for
developing geospatial simulations. In CASA working paper series. London: Center for
Advanced Spatial Analysis. Report number 110, http://www.casa.ucl.ac.uk/publications/
workingPaperDetail.asp?ID=110.
Chen, Y., E. G. Irwin, and C. Jayaprakash. 2011. Incorporating spatial complexity into eco-
nomic models of land markets and land use change. Agricultural and Resource Economics
Review 40(3): 1–10.
Cho, S-H. Hoon, S. G. Kim, and R. K. Roberts. 2014. Spatial econometric modeling of land
use change. In Oxford handbooks of land economics, eds. J. M. Duke and J. Wu, 430–451.
Diappi, L., and P. Bolchi. 2008. Smith’s rent gap theory and local real estate dynam-
ics: A multi-agent model. Computers, Environment, and Urban Systems 32(1): 6–18.
Energy Modeling Forum. 2012. Snowmass conferences: Climate change impacts and inte-
grated assessment (CCI/IA). Stanford University 2012. Available from http://emf.stanford.
edu/research/snowmass/.
Ettema, D. 2010. A multi-agent model of urban processes: Modelling relocation processes and
price setting in housing markets. Computers, Environment, and Urban Systems 35(1): 1–11.
Fagiolo, G., C. Birchenhall, and P. Windrum. 2007. Empirical validation in agent-based mod-
els: Introduction to the special issue. Computational Economics 30(3): 189–194.
Fagiolo, G., A. Moneta, and P. Windrum. 2007. A critical guide to empirical validation of
agent-based models in economics: Methodologies, procedures, and open problems.
Computational Economics 30(3): 195–226.
Filatova, T., D. C. Parker, and A. van der Veen. 2009. Agent-based urban land markets: Agent’s
pricing behavior, land prices and urban land use change. Journal of Artificial Societies and
Social Simulation 12(1): 3.
Filatova, T., A. van der Veen, and D. C. Parker. 2009. Land market interactions between het-
erogeneous agents in a heterogeneous landscape: Tracing the macro-scale effects of indi-
vidual trade-offs between environmental amenities and disamenities. Canadian Journal of
Filatova, T., A. van der Veen, and D. C. Parker. 2011. The implications of skewed risk perception
for a Dutch coastal land market: Insights from an agent-based computational economics
model. Agricultural and Resource Economics Review 40(3): 405–423.
Gent, P., G. Danabasoglu, L. Donner, M. Holland, E. Hunke, S. Jayne, D. Lawrence, R. Neale,
P. Rasch, M. Vertenstein, P. Worley, Z-L. Yang, and M. Zhang. 2011. The community climate
system model Version 4. Journal of Climate 24(19): 4973–4991.
Gimblett, H. R. (ed.). 2002. Integrating geographic information systems and agent-based
modeling techniques for simulating social and ecological processes. Oxford: Oxford
University Press.
Gimona, A., and J. G. Polhill. 2011. Exploring robustness of biodiversity policy with a coupled
metacommunity and agent-based model. Journal of Land Use Science 6(2–3): 175–93.
Gode, D., and S. Sunder. 1993. Allocative efficiency of markets with zero-intelligence trad-
ers: Market as a partial substitute for individual rationality. Journal of Political Economy
101(1): 119–137.
Gotts, N. M., J. G. Polhill, and A. N. R. Law. 2003. Aspiration levels in a land use simulation.
Cybernetics and Systems 34: 663–683.
Grimm, V., and S. F. Railsback. 2006. Chapter 1: Introduction. In Individual-based model-
ing and ecology, eds. V. Grimm and S. F. Railsback. Princeton, NJ: Princeton University
Press, 2–20.
Grimm, V., U. Berger, D. L. DeAngelis, J. G. P., J. Giske, and S. F. Railsback. 2010. The ODD
protocol: A review and first update. Ecological Modelling 221(23): 2760–2768
Groeneveld, J., B. Müller, F. Angermüller, R. Drees, G. Dreßler, C. Klassert, J. Schulze, H. Weise,
and N. Schwarz. 2012. Good modelling practice: Expanding the ODD model description
protocol for socioenvironmental agent based models. Paper read at Proceedings of the 2012
International Congress on Environmental Modelling and Software: Managing Resources of
a Limited Planet., July 2–5, Leipzig, Germany.
Happe, K., K. Kellermann, and A. Balmann. 2006. Agent-based analysis of agricultural poli-
cies: An illustration of the agricultural policy simulator AgriPoliS, its adaptation and behav-
ior. Ecology and Society 11(1): 49.
Happe, K., A. Balmann, K. Kellermann, and C. Sahrbacher. 2008. Does structure matter? The
impact of switching the agricultural policy regime on farm structures. Journal of Economic
Behavior & Organization 67(2): 431–444.
Hernandez, C., K. Troitzsch, and B. Edmonds (eds.). 2008. Social simulation technolo-
gies: Advances and new discoveries. Hershey, PA: Information Science Reference.
Horan, R. D., and F. Lupi. 2010. The economics of invasive species control and management: The
complex road ahead. Resource and Energy Economics 32(4): 477–482.
Institute for Human and Machine Cognition. 2011. Cmap Tools knowledge modeling toolkit.
Available from http://cmap.ihmc.us/.
Irwin, E. G. 2010. New directions for urban economic models of land use change: Incorporating
spatial dynamics and heterogeneity. Journal of Regional Science 50(1): 65–91.
Jackson, J., B. Forest, and R. Sengupta. 2008. Agent-based simulation of urban residential
dynamics and land rent change in a gentrifying area of Boston. Transactions in GIS 12(4):
475–491.
Kellermann, K., and A. Balmann. 2006. How smart should farms be modeled? Behavioral foun-
dation of bidding strategies in agent-based land market models. Paper read at 2006 Annual
Meeting, Queensland, Australia.
Kohler, T. A. 2000. Dynamics in human and primate societies. New York and Oxford: Oxford
University Press.
Krieger, K. 2006. Life in silico: A different kind of intelligent design. Science 312(14): 188–190.
LeBaron, B. 2006. Agent-based computational finance. In Handbook of computational econom-
ics, Vol. 2 Agent-based computational economics, eds. L. Tesfatsion and K. L. Judd, 1187–
1233. Amsterdam: Elsevier B.V.
Le Ber, F., C. Lavigne, K. Adamczyk, F. Angevin, N. Colbach, J. –F. Mari, and H. Monod. 2009.
Neutral modelling of agricultural landscapes by tessellation methods: Application for gene
flow simulation. Ecological Modelling 220(24): 3536–3545.
Ligmann-Zielinska, A., and L. Sun. 2010. Applying time-dependent variance-based global
sensitivity analysis to represent the dynamics of an agent-based model of land use change.
International Journal of Information Science 24(12): 1829–1850.
Lynch, L., and S. Lovell. 2003. Combining spatial and survey data to explain participation in
agricultural land preservation programs. Land Economics 79(2): 259–276.
Magliocca, N., E. Safirova, V. McConnell, and M. Walls. 2011. An economic agent-based model
of coupled housing and land markets (CHALMS). Computers, Environment and Urban
Systems 35(3): 183–191.
Mathevet, R., F. Bousquet, C. Le Page, and M. Antona. 2003. Agent-based simulations of inter-
actions between duck population, farming decisions and leasing of hunting rights in the
camargue (southern France). Ecological Modelling 165: 107–126.
Matthews, R. B., N. G. Gilbert, A. Roach, J. G. Polhill, and N. M. Gotts. 2007. Agent-based
land-use models: a review of applications. Landscape Ecology 22(10): 1447–1459.
Mesirov, J. P. 2010. Accessible reproducible research. Science 327: 415–416.
information and auction efficiency in ecosystem service markets. In Oxford handbook of
land economics, eds. J. Duke and J. Wu, 481–546 New York: Oxford University Press.
Morgan, F., and D. O’Sullivan. 2009. Using binary space partitioning to generate urban spa-
tial patterns. 4th International Conference on Computers in Urban Planning and Urban
Management, 1–16.
Nelson, G., and J. Geoghegan. 2002. Introduction to the special issue on spatial analysis for
agricultural economists. Agricultural Economics 27(3): 197–200.
Nolan, J., D. C. Parker, and G. Cornelis van Kooten. 2009. An overview of computational mod-
eling in agricultural and resource economics. Canadian Journal of Agricultural Economics
57(4): 417–429.
Onnela, J-P., and F. Reed-Tsochase. 2010. Spontaneous emergence of social influence in online
systems. Proceedings of the National Academy of Sciences of the USA 107(43): 18375–18380.
Parker, D. C., T. Berger, and S. M. Manson, eds. 2002. Agent-based models of land-use and
land-cover change: Report and review of an international workshop, October 4–7, 2001. Vol.
6, LUCC report series. Bloomington: LUCC Focus 1 office.
Parker, D. C., S. M. Manson, M. A. Janssen, M. Hoffmann, and P. Deadman. 2003. Multi-agent
systems for the simulation of land-use and land-cover change: A review. Annals of the
Association of American Geographers 93(2): 314–337.
Parker, D. C., and V. Meretsky. 2004. Measuring pattern outcomes in an agent-based model
of edge-effect externalities using spatial metrics. Agriculture, Ecosystems and Environment
101: 233–250.
Parker, D. C. 2005. Integration of geographic information systems and agent-based models of
land use: Challenges and prospects. In GIS, Spatial Analysis and Modeling, eds. D. J. Maguire,
M. F. Goodchild, and M. Batty. Redlands, CA: ESRI Press, 403–422.
Parker, D. C., D. Brown, J. G. Polhill, S. M. Manson, and P. Deadman. 2008. Illustrating a new
“conceptual design pattern” for agent-based models and land use via five case studies: The
MR POTATOHEAD framework. In Agent-based modelling in natural resource management,
eds. A. L. Paredes and C. H. Iglesias. Valladolid, Spain: Universidad de Valladolid, 29–62.
Parker, D. C., B. Entwisle, E. Moran, R. Rindfuss, L. Van Wey, S. Manson, L. Ahn, P. Deadman,
T. Evans, M. Linderman, S. M. M. Rizi, and G. Malanson. 2008. Case studies, cross-site com-
parisons, and the challenge of generalization: Comparing agent-based models of land-use
change in frontier regions. Journal of Land Use Science 3(1): 41–72.
Parker, D. C., and T. Filatova. 2008. A theoretical design for a bilateral agent-based land mar-
ket with heterogeneous economic agents. Computers, Environment, and Urban Systems
32(6): 454–463.
Plantinga, A. J., and D. J. Lewis. 2014. Landscape simulations with econometric-based land-use
models. In Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 380–401.
Polhill, J. G., N. M. Gotts, and A. N. R. Law. 2001. Imitative versus nonimitative strategies in a
land use simulation. Cybernetics and Systems 32(1–2): 285–307.
Polhill, J. G., D. C. Parker, D. Brown, and V. Grimm. 2008a. Using the ODD protocol for describ-
ing three agent-based social simulation models of land-use change. Journal of Artificial
Societies and Social Simulation 11(2–3). http://jasss.soc.surrey.ac.uk/11/2/3.html
Polhill, J. G., D. C. Parker, and N. M. Gotts. 2008b. Effects of land markets on competition
between innovators and imitators in land use: Results from FEARLUS-ELMM. In Social
simulation technologies: Advances and new discoveries, eds. C. Hernandez, K. Troitzsch and
B. Edmonds. Hershey, PA: Information Science Reference.
Polhill, J. G., A. Gimona, and R. J. Aspinall. 2011. Agent-based modelling of land use effects on
ecosystem processes and services. Journal of Land Use Science 6(2–3): 75–81.
Railsback, S. F., and V. Grimm. 2012. Agent-based and individual-based modeling: A practical

introduction. Princeton, NJ: Princeton University Press.
Rausch, S., and M. Mowers. 2012. Distributional and efficiency impacts of clean and renewable
energy standards for electricity. In Joint program report series. Boston: MIT Joint Program
on the science and policy of global change.
Richiardi, M., R. Leombruni, N. Saam, and M. Sonnessa. 2006. A common protocol for
agent-based social simulation. Journal of Artificial Societies and Social Simulation (1),
http://jasss.soc.surrey.ac.uk/9/1/15.html. http://globalchange.mit.edu/files/document/
MITJPSPGC_Rpt225.pdf Report number 225, 48 pp.
Robinson, D. T., and D. G. Brown. 2009. Evaluating the effects of land-use development policies
on ex-urban forest cover: An integrated agent-based GIS approach. International Journal of
Geographic Information Science 23(9): 1211–1232.
Robinson, D. T., D. G. Brown, D. C. Parker, P. Schreinemachers, M. A. Janssen, M. Huigen,
H. Wittmer, N. Gotts, P. Promburom, E. Irwin, T. B., F. Gatzweiler, and C. Barnaud. 2007.
Comparison of empirical methods for building agent-based models in land use science.
Journal of Land-Use Science 2 (1): 31–55.
Schelling, T. 1971. Dynamic models of segregation. Journal of Mathematical Sociology
1: 143–186.
Schreinemachers, P., T. Berger, A. Sirijinda, and S. Praneetvatakul. 2009. The diffusion of green-
house agriculture in northern Thailand: Combining econometrics and agent-based model-
ing. Canadian Journal of Agricultural Economics 57(4): 513–536.
Schreinemachers, P., C. Potchanasin, T. Berger, and S. Roygrong. 2010. Agent-based modeling
for ex ante assessment of tree crop innovations: Litchis in northern Thailand. Agricultural
Economics 41(6): 519–536.
Simon, H. 1996. The sciences of the artificial. Cambridge, MA: MIT Press.
Stanford Center for Biomedical Informatics Research. 2007. Protege OWL 2007. Available
from http://protege.stanford.edu.
Tesfatsion, L. 2006. Agent-based computational economics: A constructive approach to eco-
nomic theory. In Handbook of computational economics Vol. 2: Agent-based computational
economics, eds. L. Tesfatsion and K. L. Judd, 831–880. Amsterdam: Elsevier B.V.
Tesfatsion, L., and Kenneth L. Judd. 2006. Handbook of computational economics, Volume
2: Agent-based computational economics. Amsterdam: Elsevier B.V.
Tietenberg, T. 2005. Economic instruments for environmental regulation. In Economics of the
environment: Selected readings, ed. R. Stavins, 277–301. New York: W. W. Norton.
Torrens, P. M. 2007. A geographic automata model of residential mobility. Environment and
Planning B: Planning and Design 34: 200–222.
Valbuena, D., P. Verburg, and A. K. Bregt. 2008. A method to define a typology for agent-based
analysis in regional land-use research. Agriculture, Ecosystems and Environment
128(1–2): 27–36.
Verburg, P., K. Kok, R. G. Pontius, A. Veldkamp, A. Angelsen, B. Eickhout, T. Kram, A.
J. Walsh, D. C. Parker, K. Clarke, D. Brown, and K. Overmars. 2006. Modelling land-use and
land-cover change. In Land-use and land-cover change: Local processes, global impacts, eds.
E. Lambin and H. Geist, 117–131. New York: Springer Berlin Heidelberg.
Waldrop, M. 2009. A model approach. Nature 460: 667.
C HA P T E R 17
S PAT IA L E C O N O M E T R I C
MODELING OF LAND
U S E C HA N G E
SE ONG - HO ON C HO, SE U NG G Y U K I M , A N D

ROL A N D K . ROBE RT S
Land is used by humans and other living creatures and involves complex
human-environment interactions. Land uses may be broadly classified for forest,
agriculture, and urban uses, and their uses may be altered by land users’ purposes.
Understanding land use change is essential because it occurs to generate desirable and
undesirable impacts on the environment and human welfare. Theoretical and empiri-
cal modeling approaches have been developed to examine the drivers, processes, and
implications of changes in land use.
Theories of land use change conceptualize the frameworks describing changes from
one type of use to another and explain why, when, how, and where land use changes
occur under the frameworks of disciplines studying economic, environmental, and
spatial changes. Microeconomic theory-based approaches have adopted von Thunen’s
agricultural rent theory (Thünen and Heinrich 1966), Alonso’s (1972) urban land mar-
ket theory, and agent-based theories of urban and regional spatial structure (Schaffer
1999). Von Thunen’s agricultural rent theory covers location theory and the urban and
regional spatial structure of a wide range of spatial scales and provides the foundations
for Alonso’s urban land market theory. Alonso’s theory derives individual equilibria
for households based on bid-rent functions and a market clearing mechanism. The
agent-based theoretical approaches accommodate the endogeneity of the spatial distri-
bution of agents and of their associated activities.
Empirical modeling of land use change emphasizes discrete land use decisions at the
parcel or plot scale (e.g., develop or not). The discrete land use decisions conceptualize
landowners choosing to develop land if the present value of the future stream of net
returns from development is greater than the present value of the future stream of net
returns from the land remaining in its current nonurban use (Bockstael 1996). The main
Spatial Econometric Modeling of Land Use Change 431
objective of this chapter is to provide a comprehensive review and critique of the litera-
ture of empirical modeling of land use decisions, focusing particularly on the strengths
and weaknesses of different spatial econometric modeling approaches and important
future research directions. To accomplish the objective, (1) a comprehensive review of
the literature on spatial econometric modeling of land use decisions is presented, (2) a
case study to illustrate one of the approaches is developed, (3) an overall assessment of
different approaches is provided, and (4) important directions and challenges for future
research are presented.
Empirical specifications typically use binary probit or logit regression models in which
conversion from nonurban to urban use is explained by rents derived from different
land uses, such as rents derived from forest or farmland (White and Fleming 1980; Alig
1986; Alig et al. 1988; Lichtenberg 1989; Parks and Murray 1994; Hardie and Parks 1997;
Kline and Alig 1999; Plantinga et al. 1999), distances to commodity markets and ame-
nity areas (Chomitz and Gray 1996; Turner et al. 1996; Nelson and Hellerstein 1997;
Cropper et al. 1999; Cho and Newman 2005), and land use regulations (Irwin and
Bockstael 2002; Irwin et al. 2003; Libby and Sharp 2003; Miller and Vaske 2003).
A major challenge with econometric specification of discrete land use decisions is
that land conversion decisions may be co-determined through neighborhood spillover
effects. Neighbors share common characteristics, hence their decisions exhibit high
dependence among the error terms in land conversion models (Irwin and Bockstael
2001; Carrión-Flores and Irwin 2004; Cho and Newman 2005; Irwin et al. 2006). Spatial
dependence can occur due to spatially correlated land use decisions or as a consequence
of residual correlation caused by unobserved factors that are spatially dependent. Like
any other statistical problem caused by the lack of independence of the errors, the pres-
ence of spatial dependence of the errors (referred to as “spatial error autocorrelation”) in
the econometric specification of discrete land use decisions causes parameter estimates
to be inconsistent and inefficient (Carrión-Flores and Irwin 2004).
The application of spatial econometrics to discrete dependent variable models, such
as binary probit and logit models, is comparatively less developed than for models with
continuous dependent variables. Discrete spatial process models that accommodate
spatial error autocorrelation typically are based on the maximum likelihood estima-
tion (MLE) method (Case 1992), the linearized version of the generalized method of
moments (GMM) (Pinske and Slade 1998; Fleming 2004; Klier and McMillen 2008),
the spatial general linear model (GLM) method (Schabenberger and Pierce 2002), the
nonparameteric probit geographically weighted regression (GWR) model approach
(LeSage 1999; Páez 2006), or the nonparametric GMM model method (Conley 1999;
Conley and Dupor 2003).
432 Seong-Hoon Cho, Seung Gyu Kim, and Roland K. Roberts
Case (1992) assumed a block-diagonal matrix of spatial weights by taking a common

spatial component for all observations within a given boundary using the MLE for the
spatial probit model. This restrictive specification has been criticized for not account-
ing for distance-decay effects. The GMM estimator with a spatial autoregressive term
proposed by Kelejian and Prucha (1999) is considered better than MLE mainly because
(1) the number of integrals in the likelihood function equals the sample size, which is
computationally intractable when the sample size is large; and (2) MLE requires full
distributional assumptions, which can affect parameter estimation and the accuracy of
spatial predictions when the error distribution asssumption is incorrect, whereas GMM
has no such requirement.
Páez (2006) presented a nonparametric probit GWR with heteroscedastic error
terms to analyze land development by generating parameter estimates for every regres-
sion point to highlight spatial variation. Despite the benefits of the probit GWR, the
literature using GWR identifies potentially serious problems with the approach such as
(1) spatial error dependence (Leung et al. 2000; Fotheringham et al. 2002), (2) poten-
tial multicollinearity among local regression coefficients (Wheeler and Tiefelsdorf
2005), and (3) extreme coefficients, including sign reversals (Farber and Páez 2007).
Alternatively, nonparametric GMM models that allow for spatial dependence generate
consistent covariance-matrix estimators regardless of sample size (e.g., Grenander and
Rosenblatt 1957; Hall et al. 1994; Hall and Patil 1994; Conley 1999; Conley and Dupor
2003) (Grenander and Rosenblatt 1957; Hall, Fisher, and Hoffmann 1994; Hall and Patil
1994). (Grenander and Rosenblatt 1957; Hall, Fisher, and Hoffmann 1994; Hall and
Patil 1994; Conley 1999; Conley and Dupor 2003).
2. Case Study
This section showcases a case study of applying the nonparametric GMM model to dis-
crete land use decisions with 12,375 observations using a consistent covariance matrix
for the GMM estimator in the presence of spatial error autocorrelation. The case study
uses the model to evaluate maximum lot coverage as a potential policy tool for mitigat-
ing urban sprawl.1 Maximum lot coverage is the maximum percentage of impervious
surface allowed on any given lot. Lot coverage is calculated as the total amount of imper-
vious surface on the lot divided by the total lot area. For example, if the maximum lot
coverage of 20% were assigned to a residential lot, the area of the lot could not be covered
by impervious surface of more than 20%, leaving the remaining uncovered portion as
private open space. Maximum lot coverage restrictions have been implemented (1) to
maintain a consistent and compatible land use pattern for residential neighborhoods
1 The word “sprawl” first appeared in print in 1955, in the context of low-density and leapfrogging
development (Evans 1999; Rybczynski 2005). Despite divergent viewpoints on the definition of urban
and (2) to prevent excessive impervious surfaces and thus reduce the risk of drainage and
flooding problems (Pierson 2002; City of Redmond 2011). Maximum lot coverage was
chosen as a potential policy tool for mitigating urban sprawl because it explicitly utilizes
the tradeoff between lot size and public open space, ensures the provision of open space,
and can curb urban sprawl (Lichtenberg and Hardie 2007; Lichtenberg et al. 2007). To
examine the effects of maximum lot coverage on urban sprawl, landowners’ development
decisions for new residential housing at the parcel level are empirically estimated in a
land conversion model based on the conceptual framework developed in Section 2.1.
Once estimates are acquired from the land conversion model, the impacts of changes
in maximum lot coverage on development patterns are evaluated by ex ante simulations
of development patterns inside and outside the developed area that existed prior to the
emergence of urban sprawl. The ex ante simulations forecast development rates under
observed status quo and hypothetical maximum lot coverage scenarios. The ex ante
simulations suggest that an increase in maximum lot coverage encourages increased
development inside the area of non-disconnected, preexisting development that existed
prior to the emergence of urban sprawl (referred to as “preexisting development”), rel-
ative to the area outside preexisting development. With an increase in maximum lot
coverage, a greater development rate inside the area of preexisting development than
outside this area effectively serves to mitigate sprawl by encouraging development close
to preexisting development and discouraging fragmented sprawl development farther
from preexisting development.
2.1 Conceptual Framework

Household location choices are modeled by extending the work of Brueckner (1987),
Fujita (1990), Fujita and Thisse (2002), Wu (2006), and Glaeser et al. (2008). Households
are assumed to choose consumption bundle (q, s, o) to maximize utility subject to a bud-
get constraint:
max U (q, s, o)
q,s
(1)
s.t . pq + s = y

where U(⋅) is a differentiable utility function, q is the size of residential space (feet2), s
is the consumption of a composite numéraire nonhousing good, o is the size of neigh-
boring open space surrounding the residential location, p is the housing price per unit
sprawl, there is consensus that urban sprawl is well-described as the leapfrogging of development
beyond the city’s outer boundary into smaller rural settlements (Hanham and Spiker 2005). An area
of leapfrogging or fragmented development is considered an area of urban sprawl when development
occurs disjoint from existing development (Isberg 1973; Ewing 1994; Wu and Plantinga 2003; Wu 2006).
($/feet2), and y is gross household income. The indirect utility function can be implicitly
defined from the maximization problem in equation (1):
V ( p, y , o) = U (q *, y − pq *, o) = V o , (2)

where q* = arg maxq U(q, y − pq, o) is the demand function for housing. In equilib-
rium, each household has the same level of utility Vo, which is independent of its loca-
tion. Vo is also exogenous from the perspective of a single “open city,” because in- and
out-migration will equate household utility across cities (Wu 2006).
Assumptions about q are posed to analyze the effect of the policy variable, maximum
lot coverage:
q = q(l , m), (3)
where l is the lot size for a residential house and m is the maximum lot coverage imposed
by zoning. The size of residential space q is expected to be positively related with lot size
l and maximum lot coverage m because both larger lot size and maximum lot coverage
provide more residential space. Also, m is expected to be smaller in areas with more
neighboring open space; that is, m = m(o) and ∂m/∂o < 0, because maximum lot cov-
erage intends to preclude excessive structure development on each parcel and is rela-
tively stricter in low-density developed areas (e.g., agriculturally zoned districts) than
in densely designed areas (e.g., multifamily districts) (Johnston and Madison 1997;
Harrison County 2009; New York City 2009). By substituting equation (3) into (2),
equation (3) can be rewritten as:
V ( p, y , o) = U (q * (l , m(o)), y − pq * (l , m(o), o)) = V o . (4)
Equation (4) implicitly defines a market-level inverse demand function (p* = p(⋅)):
p* = p( y , o, V o ), (5)
where o = o−1(m). To see the impact of maximum lot coverage on housing demand, we
apply the implicit function theorem to equation (5):
∂V
∂p * ∂p * ∂o −1 (m)
= − ∂m = −1 ⋅ . (6)
∂m ∂V ∂o (m) ∂m
∂p *
The first ratio on the right-hand side of equation (6) is typically positive because
neighboring open space provides a positive amenity to households. The second ratio
is positive if private open space is perceived as a substitute for neighboring open space
because higher maximum lot coverage reduces private open space, and larger neighbor-
ing open space substitutes for the reduced private open space.2 Consequently, the sign
of equation (6), the effect of maximum lot coverage on housing demand reflected in the
housing price (p*), is positive. Thus, an increase in m (i.e., less stringent maximum lot
coverage) increases housing demand.
The supply side of the housing market is specified in terms of maximum lot coverage
by assuming a competitive industry with constant return to scale production technol-
ogy (Wu 2006). The developers choose the density of development to maximize profit π:
max π (d ) = p * d − r − c(d ), (7)

d
where d is the development density represented by the structure/lot size ratio, r is

site-specific land cost, and c(d) is material-labor-capital cost. The first-order condition,
p* − c′(d) = 0, implicitly defines the optimal development density d*:
d * = c ′ −1 ( p*), (8)

where c′−1(⋅) is the inverse of the marginal cost function. Developers are forced to face
two possible alternatives for the residential development decision:
developable, if d * ≤ m
undevelopable, if d * > m, (9)

which shows that developers would develop a residential house when their optimal
choice of development density is not bound by the maximum lot coverage regulation;
that is, d∗ ≤ m . A lower m sets a more stringent bar for development. Thus, the prob-
ability of development would be lower if m were lower (i.e., more stringent regula-
tion of maximum lot coverage). Because a decrease in m decreases the probability of
development and consequently decreases the supply of new housing, a decrease in m
reduces housing levels in equilibrium because both supply of new housing and demand
2
Thorsnes (2002) found that that larger residential lot sizes are viewed as a substitute for forest open
space to some extent. Cho and Roberts (2007) found different degrees of willingness to trade between
neighborhood density (representing open space availability) and lot size. Likewise, Kopits, McConnell,
and Walls (2007) found that in the urban-rural fringe, owning a private lot was preferrable to public open
space with low willingness to trade between the two. Cho et al. (2009) recently found that substitutability
between open space and lot size exists inside the city boundary.
for housing are lower if private open space is perceived as a substitute for neighboring
open space.
Because the stock of housing depends on the urban population, according to urban
spatial theory, observed land development decisions are regarded as an equilibrium
point at which supply of housing service meets consumer demand (Dipasquale 1999). To
examine the effect of maximum lot coverage on the equilibrium housing level, land devel-
opment decisions by landowners for new residential housing at the parcel level are empir-
ically estimated. Equilibrium housing levels are observed as land development decisions
under the observed status quo maximum lot coverage. A parcel-based land development
model (e.g., Bockstael 1996; Bockstael and Bell 1998; Nickerson and Lynch 2001; Bell and
Irwin 2002; Irwin and Bockstael 2002, 2004; Irwin et al. 2003; Cho and Newman 2005)
and ex ante simulations measure deviations from equilibrium housing levels following
implementation of alternative hypothetical maximum lot coverage scenarios.
2.2 Spatial Binary Model Applying a Nonparametric Estimator

Let Yi denote a binary indicator of the choice for observation i of whether or not to
develop a parcel for a single family house. Then a probit model is defined as:
P (Yi = 1 | X i ) = Φ(X ′i β), (10)

where Φ is the cumulative distribution function for the standard normal distribution;
Xi is a (k+1)×1 vector of explanatory variables including parcel information such as
parcel size, socioeconomic and environmental variables associated with the parcel’s
location, and zoning regulations (i.e., maximum lot coverage); and β is a (k+1)×1 vector
of parameters including an intercept.
The parameters in equation (10) were estimated by nonparametric methods that allow
for spatial dependence. The methods are capable of generating consistent covariance-
matrix estimators regardless of sample size. Heteroscedasticity-consistent standard
errors were estimated to remove residual spatial autocorrelation caused by codeter-
mined development decisions (e.g., clustered residential developments within subdivi-
sions). The covariance-matrix estimators were modified to allow regression disturbance
terms to be correlated across neighborhood parcels as a general function of their
Euclidean distances.
By assuming stationarity in X, which enables the joint distribution of X for any loca-
tion to be invariant to shifts in the entire set of locations, the covariance of two observa-
tions is measured by a function of Euclidian distances without directional information
(Conley 1995).
cov(X Si , X S j ) = f (|| si − s j ||), (11)

where x si and x s j are vectors of explanatory variables of observations located in si and sj

within a Euclidean space (Conley and Topa 1999).
The spatial autocovariance at distance (δ) is estimated by applying the nonpara-
metric estimator of the spatial autocovariance function proposed by Hall, Fisher, and
Hoffmann (1994):
n n
f (δ ) = ∑∑W [| δ − || s − s
i = 1 j= 1
n i j || |](X Si − X )( X S j − X ), (12)

where X is the sample mean and Wn[⋅] is a function of the sample size that concentrates
its mass at zero as the sample becomes arbitrarily large at an appropriate rate defined
by Conley and Topa (1999). The spatial covariance was estimated by an average of
cross-products between the vectors (X Si − X ) and (X S j − X ) within a given distance δ.
By employing the nonparametric estimator of the spatial autocovariance function in the
spatial GMM approach, the error term is permitted to be conditionally heteroskedastic
and spatially correlated across parcels. Once the parameters are estimated in equation
(10), for example, the marginal effect for Xk (kth explanatory variable) is calculated by
∂ Prob(Yi = 1 | Xi )/ ∂X k = Φ( X′i β )βk, where βk is the kth parameter estimated.
The predicted values for the GMM approach of the spatial probit model facilitate ex
ante comparisons between predicted probabilities generated under the observed status
quo maximum lot coverage and two hypothetical maximum lot coverage scenarios; that
is, plus and minus 10% of the status quo. In a binary model like the spatial probit model,
unequal frequencies of the two outcomes always lead to lower estimated prediction
probabilities for the less frequent outcome than for the more frequent outcome (Cramer
1999). Cramer (1999) suggested the average occurrence rate of the two outcomes as a
cutoff value, which has been used as the alternative cutoff value for the binary model
in previous empirical studies (McPherson et al. 2004; Liu et al. 2005). In this study, the
average occurrence rate of development (Yi = 1) and no development (Yi = 0) was cal-
culated as the average predicted probability of development and used as a cutoff value
for the prediction of development in the spatial probit model. The use of this occurrence
rate, instead of the conventional threshold of 0.5, accounts for unequal frequencies in
our dataset (to be presented later).
2.3 Identifying the Area of Preexisting Development

Identifying areas of preexisting development for the purpose of dividing the areas
that existed prior to and after the emergence of urban sprawl to measure the impact
of maximum lot coverage on development pattern is not straightforward. The areas of
preexisting development should reflect the areas of clustered development prior to the
emergence of urban sprawl, whereas the area outside of preexisting development should
represent the area of sprawl. A spatial break dividing these two areas is not clear because
cumulative growth of a city is done through additions to its periphery over the course of
many years.
One tentative way is simply to identify areas of parcels that were developed prior to the
duration of the model period as preexisting development. Although this makes sense if
sprawl development did not exist prior to the sample period used for model develop-
ment, measuring and comparing development rates within and outside of preexisting
development prior to and after the sample period may not be appropriate because areas
that were developed prior to the sample period may include areas with sprawl develop-
ment patterns. Under this definition, the areas of preexisting development are probably
not free from sprawl, and thus comparisons of development patterns between these two
areas may not accurately reflect the degree of sprawl.
Alternatively, a two-step approach was designed to systematically identify preexist-
ing development clusters and thus draw the spatial break. In the first step, local indica-
tors of spatial association (LISA) for the built years of parcel data were estimated. LISA
values of built years of parcel data indicate the extent of spatial autocorrelation between
the built year of a particular parcel and the built years of the parcels around it. Through
inference analysis, spatial clusters of old-built parcels (old-built parcels surrounded by
old-built parcels) were identified as clusters developed prior to the emergence of urban
sprawl. The LISA values of built years of parcel data served well for the purpose of iden-
tifying spatial breaks in the built years of parcel data.
Disjoint areas among the spatial clusters of old-built parcels identified from the first
step were removed from preexisting development in the second step. Joint spatial clus-
ters of old-built parcels were difficult to verify because continuous clusters of old-built
parcels may still have gaps (e.g., roads, industrial and commercial development, and
unusable land). Thus, buffer polygons of a specified distance around the old-built par-
cels were created using a buffer tool in ArcMap 9.3 (Environmental Systems Research
Institute 2009). The buffers were used to merge the continuous buffer areas. The buffers
of old-built parcels that were joined in continuous buffer areas were assumed to be areas
of preexisting development, and those areas that were not joined in continuous buffer
areas were assumed to be disjoint from the spatial clusters of old-built parcels.
The buffer size needed to distinguish the parcels that were inside and outside preexist-
ing development was initially unclear. Thus, the second step was repeated with buffer
polygons of different distances around the old-built parcels (i.e., 0.1- to 1-mile radii with
0.1-mile increments). After a number of trials of the procedure with different distances,
0.5 mile was chosen as the threshold. Buffer polygons of 0.5 mile were merged into a
single spatial cluster containing 95% of the clusters of old-built parcels identified using
LISA in the first step.
2.4 Study Area and Data

This study used four primary geographic information system (GIS) datasets: individual
parcel data, census-block group data, boundary data, and environmental feature data
from Nashville-Davidson County, Tennessee. The individual parcel data were obtained
from the Metro Planning Department, Nashville-Davidson County (MPD 2009) and the
Davidson County Tax Assessor’s Office. Information from 467 census-block groups was
used to reflect the socioeconomic status of neighborhoods, such as per capita income
and unemployment rate for parcels located within the boundaries of the census-block
groups. The average size of a census-block group was 721 acres, with a standard devia-
tion of 1,588 acres. Boundary data (e.g., high school districts and jurisdiction bound-
aries) were also obtained from the Metro Planning Department, Nashville-Davidson
County (MPD 2009). Environmental feature data (e.g., water bodies and golf courses)
were collected from Environmental Systems Research Institute Data and Maps 2004
(Environmental Systems Research Institute 2004). Other environmental feature data
(e.g., shape files for railroads and parks) were also acquired from MPD. Definitions and
simple descriptive statistics of the variables used in the regressions are listed in Table 17.1.
Developed parcels used for the dependent variable in the spatial probit model were
defined as single-family houses that were built in 2007. At the start of 2007, the number
of vacant parcels in Nashville-Davidson County was 20,990. Only single-family hous-
ing development in residentially zoned districts was considered in the model because
the development decision processes for other land uses (e.g., multifamily housing, com-
mercial, and industrial land uses) are influenced by different development factors and
property characteristics. Of the 12,375 parcels in residentially zoned districts, 1,603 par-
cels (or 13.0%) were developed for single-family housing in 2007. The average size of
undeveloped parcels was 1.8 acres, whereas the average size of parcels developed for
single-family housing was 0.3 acres.3 Distances between any two closest neighboring
parcels among the 12,375 observations ranged from 10 feet to 8,231 feet. The average
distance of 236 feet was used as the cutoff value for spatial correlation. Maximum lot
coverage regulated by Nashville-Davidson County ranged from 0.2 to 0.6.4
2.5 Empirical Results

The overall percentage of correct predictions was 78%, using the average probability of
development of 0.13 from the data. The rates of correct prediction were 80% for devel-
oped parcels and 65% for undeveloped parcels. The marginal effects calculated based on
the parameter estimates of the GMM spatial probit model are presented in Table 17.2.
3
We treat the lot size as a lagged exogenous variable and thus face no endogeneity problem because
all information available for decision makers, including lot size, at the time of development is collected
prior to the duration of development in 2007.
4
Developers were not allowed to cover more than 20% (0.2 maximum lot coverage), 30% (0.3
maximum lot coverage), 35% (0.35 maximum lot coverage), 40% (0.4 maximum lot coverage), 45% (0.45
maximum lot coverage), 50% (0.5 maximum lot coverage), and 60% (0.6 maximum lot coverage) for
Zoning Districts of AG, AR2a, RS80, and R80; RS40 and R40; RS20, R20, RS15, R15, and RM2; RS10,
R10, and RM4; R8 and RS7.5; R6, RS5, RM6, RM9, and RM15; and RS3.75, OR40, RM20, OR20, RM40,
RM60, and I, respectively.
Table 17.1 Variable names, definitions, and descriptive statistics

Variables Description Mean (Std. Dev.)
Develop Dummy variable for development in 2007 (1 if a 0.130 (0.336)

single family house was built in 2007, 0 otherwise)
Lot size Lot size in thousand square feet 78.485 (388.639)
Lot value per acre Assessed land value in $1,000 per acre in 2007 46.299 (85.743)
Per capita income Per capita income in $1,000 for census-block 23.027 (11.127)
group in 2000
Housing density Housing density (the number of houses per acre) 1.323 (1.284)
for census-block group in 2000
Travel time to work Average travel time (minutes) to work for 23.734 (4.358)
census-block group in 2000
Unemployment rate Unemployment rate for census-block group in 0.051 (0.045)
2000
Vacancy Vacancy rate for census-block group in 2000 0.067 (0.041)
ACT Average composite score of American College Test 17.819 (1.358)
by high school district in 2007
Water Distance in 1,000 feet to the nearest water body 6.671 (4.623)
Park Distance in 1,000 feet to the nearest park 6.283 (4.276)
Park size Size of nearest park in 1,000 square feet 5,598.627
(12,711.409)
Golf Distance in 1,000 feet to the nearest golf course 21.002 (12.782)
CBD Distance in 1,000 feet to the central business 40.093 (20.345)
district
Greenway Distance in 1,000 feet to the nearest greenway 11.003 (7.886)
Rail Distance in 1,000 feet to the nearest railroad 7.822 (6.901)
Interstate Distance in 1,000 feet to the nearest interstate 8.808 (6.565)
highway
Slope Slope in degrees where the parcel is located 4.756 (3.991)
Maximum lot coverage Maximum lot coverage assigned by zoning 0.406 (0.073)
regulation
Number of observation 12,375
The discussion here is limited to the variables that are statistically significant at the 5%
level. The marginal effects of per capita income, unemployment rate, vacancy rate, and
housing density were found to be significant. These four variables capture the socio-
economic status of neighborhoods at the census-block group level: an increase in per
capita income by $1,000 decreases the probability of development by 0.01. The negative
effect of income on development may be explained by greater supply of land that can be
developed in lower income areas. A decrease in the unemployment rate by 1 percentage
point increases the probability of development by 0.31. A decrease in the vacancy rate by
1 percentage point increases the probability of development by 0.16. The negative effects
of the unemployment and vacancy rates indicate that the economic status of the neigh-
borhood at the census-block group level is an important factor affecting the dynamics
of housing development. A decrease in housing density by 1 house per acre increases
Table 17.2 Estimated marginal effects from the land development model

Variable Marginal effect Spatial standard error
Lot size 0.002*** 0.000

Lot value per acre 0.132*** 0.005
Per capita income −0.009** 0.004
Housing density −0.010*** 0.003
Travel time to work 0.001 0.001
Unemployment rate −0.314*** 0.081
Vacancy rate −0.163** 0.068
ACT −0.001 0.002
Water −0.005 0.003
Park 0.038*** 0.005
Park size −0.001** 0.000
Golf 0.005** 0.002
CBD −0.002 0.002
Greenway 0.012*** 0.002
Rail 0.008*** 0.002
Interstate 0.004 0.002
Slope −0.002*** 0.001
Maximum lot coverage 0.203*** 0.036
** and *** indicate statistical significance at the 5% and 1% levels.
the probability of development by 0.01. The greater probability of development in lower

density housing areas reflects a pattern of urban sprawl.
Similarly, variables reflecting the properties of the parcel itself (i.e., size, value, and
proximities to local parks, golf courses, and greenway, slope) were significant. A lot
size increase of 1,000 feet2 (or 0.023 acres) increases the probability of development by
0.002. A $1,000 per acre increase in lot value increases the probability of development
by 0.13. The positive effect of lot value per acre on development reflects higher develop-
ment pressure for land with higher value (Brueckner and Kim 2003; Cho et al. 2010).
A 1,000-foot (or 0.19 miles) increase in distance to the nearest local park increases the
probability of development by 0.038. A 1,000-foot increase in distance to the nearest
golf course increases the probability of development by 0.01. The negative effects of
proximity to local parks and golf courses may be explained by the crowding of preexist-
ing residential development closer to local parks and golf courses that occurred prior to
2007. A 1,000-foot increase in distance to the nearest greenway increases the probability
of development by 0.01. The negative effect of proximity to a greenway may be explained
by greenways being built in mature residential neighborhoods; thus, new development
occurs farther away from greenways. A decrease in the slope of a parcel by 1 degree
increases the probability of development by 0.002, implying the importance of flatness
of land for residential development.
Given this context, an increase in maximum lot coverage by 1 percentage point

increases the probability of development by 0.20. This finding indicates that higher
maximum lot coverage (or less stringent maximum lot coverage) increases the prob-
ability of development. This empirical result confirms the theoretical expectations that
less stringent regulation on maximum lot coverage increases both housing supply and
demand, thus increasing the probability of development. Thus, maximum lot coverage
plays a significant role in land development decisions.
The development rates (i.e., number of parcels predicted to be developed in each area
divided by the number of parcels predicted to be developed in the entire county) and
development occurrence rates (i.e., number of parcels predicted to be developed in each
area divided by the number of vacant parcels in the residentially zoned districts in each
area) inside and outside the area of preexisting development under the observed status
quo maximum lot coverage and the two hypothetical maximum lot coverage scenarios
are reported in Table 17.3. Under the current maximum lot coverage, 59% of parcels
(3,963 of 6,725 parcels) are predicted to be developed in the area outside of preexisting
development. In contrast, only 19% of parcels (1,068 of 5,650 parcels) are predicted to
be developed inside the area of preexisting development. Thus, 21% (1,068 of 5,031)
of the parcels that were predicted to be developed occurred inside the area of preexist-
ing development, whereas 79% (3,963 of 5,031) occurred outside the area of preexisting
Table 17.3 Development rates and development occurrence rates under the

observed status quo maximum lot coverage and two hypothetical maximum lot
coverage scenarios
Development rates a Development occurrence rates b
Outside the
Maximum Inside the area area of Inside the area Outside the area
lot coverage of preexisting preexisting of preexisting of preexisting
scenarios development development development development Entire county
Current 21% 79% 19% 59% 41%

maximum lot (1068/5031) (3963/5031) (1068/5650) (3963/6725) (5031/12375)
coverage
10% lower 18% 82% 15% 57% 38%
hypothetical (845/4644) (3799/4644) (845/5650) (3799/6725) (4644/12,375)
maximum lot
coverage
10% higher 25% 75% 24% 61% 44%
hypothetical (1349/5439) (4090/5439) (1349/5650) (4090/6725) (5439/12375)
maximum lot
coverage
a
Number of parcels predicted to be developed in each area/Number of parcels predicted to be
developed in the entire county.
b Number of parcels predicted to be developed in each area/Number of vacant parcels in the
residentially zoned districts in each area.

development. The considerably higher frequency of predicted development outside the

area of preexisting development implies a pattern of urban sprawl.
The predicted development rate for the entire county falls from 41% (or 5,031 of
12,375 parcels) under the current maximum lot coverage to 38% (or 4,644 of 12,375 par-
cels) under the 10% lower hypothetical maximum lot coverage scenario (more stringent
maximum lot coverage). The decline in the development rate due to lower maximum
lot coverage occurs both inside and outside of the preexisting development area (i.e.,
19% to 15% in the preexisting development area and 59% to 57% outside of the preexist-
ing development area). Among the 4,644 parcels that are predicted to be developed in
the overall area under the 10% lower hypothetical maximum lot coverage scenario, 845
parcels (or 18%) occur inside the area of preexisting development, whereas 3,799 par-
cels (or 82%) occur outside the area of preexisting development. These results indicate a
3 percentage point decrease in the predicted development rate inside the area of preex-
isting development and a 2 percentage point increase in the development rate outside of
the area, suggesting that more stringent maximum lot coverage limits overall develop-
ment but constrains development more inside relative to outside the area of preexisting
development.
The predicted development rate for the entire county rises from 41% (or 5,031 of
12,375 parcels) under the status quo maximum lot coverage to 44% (or 5,439 of 12,375
parcels) under the scenario with 10% higher hypothetical maximum lot coverage (less
stringent maximum lot coverage). The rise in the development rate due to higher maxi-
mum lot coverage occurs both inside and outside of the area of preexisting develop-
ment area (i.e., 19% to 24% in the preexisting development area and 59% to 61% outside
of the preexisting development area). Among the 5,439 parcels that are predicted to be
developed under the scenario with 10% higher hypothetical maximum lot coverage,
1,349 parcels (or 25%) occurred inside the area of preexisting development, whereas
4,090 parcels (or 75%) occurred outside the area of preexisting development. These
results indicate a 4 percentage point increase and a 4 percentage point decrease in the
development rates inside and outside the area of preexisting development, respectively,
compared to current maximum lot coverage. The evidence suggests that less stringent
maximum lot coverage encourages overall development but provides greater encour-
agement for development inside relative to outside the area of preexisting development.
The empirical results from the spatial probit model for land development confirm
that more stringent regulation of maximum lot coverage (lower maximum lot coverage)
decreases the probability of development over the entire metropolitan county while it
results in a larger decrease in the development rate inside relative to outside the area of
preexisting development. In contrast, less stringent regulation of maximum lot coverage
(higher maximum lot coverage) increases the probability of development over the entire
county but the development rate increases less outside relative to inside the area of pre-
existing development. Thus, less stringent maximum lot coverage encourages overall
development but provides greater incentive for development inside relative to outside
the area of preexisting development.
Greater responses in the probability of development to changes in maximum lot cov-

erage inside relative to outside the area of preexisting development suggest that changes
in maximum lot coverage regulations create more profound deviations from equilib-
rium housing levels inside relative to outside the area of preexisting development. These
results imply that equivalent changes in maximum lot coverage cause greater changes
in housing demand and supply inside relative to outside the area of preexisting devel-
opment. A further implication is that the positive amenity of neighboring open space
(first ratio on the right-hand side of equation [6]‌) and the substitutability between pri-
vate open space and neighboring open space (second ratio on the right-hand side of
equation [6]) are larger inside relative to outside the area of preexisting development.
Hence, results imply that the optimal choice of development density (equation [1]) is
more likely bound by regulation of maximum lot coverage inside relative to outside the
area of preexisting development.
Modifying site-specific maximum lot coverage based on predicted development rates
inside and outside the area of preexisting development under different maximum lot
coverage scenarios provides policy makers with additional information for designing or
updating site-specific maximum lot coverage policies in their efforts to moderate urban
sprawl. For example, if local policy makers and planners wished to curb sprawl develop-
ment and revitalize the inner city, they could impose zoning regulations that increase
maximum lot coverage in the inner city and simultaneously lower maximum lot cover-
age outside the inner city. Results from this case study suggest that local policy makers
and planners should weigh differences in the spatial effects of zoning regulations when
considering maximum lot coverage as a policy tool for curbing urban sprawl.
3. Conclusion
This chapter contributes to the general understanding of spatial econometric modeling

as a tool for evaluating policies designed to influence land development patterns. The
evaluation of land use policies involves a complex process driven by spatial interactions
between changes in land use and the policies being considered (Irwin and Geoghegan
2001). Land use management practices without consideration of these interactions do
not respond to the system dynamics caused by spatial interactions. Advances in spatial
econometric modeling, including the nonparametric methods used in the case study,
allow policy makers to design land use management practices that are more effective
in stimulating the desired response from a system characterized by spatial interactions.
The econometric specification of discrete land use decisions is useful in the sense
that underlying spatial dynamic processes of land use decisions are modeled explicitly,
which allows linking the land use models with GIS characterizing the spatial pattern of
land use (e.g., land use dynamics and landscape change patterns). Plot- and parcel-level
data, such as those used in the case study, have been applied to the aforementioned
framework and much effort has been focused on high spatial dependence among the
error terms in models for land use decisions (e.g., develop or not), which makes stan-
dard probit estimation inconsistent. Methods such as MLE, GMM, GLM, probit GWR,
and nonparametric GMM are among the potential econometric techniques that accom-
modate spatial error autocorrelation in the discrete spatial process model.
The spatial probit model used in the case study to model discrete land use decisions
has typically not been estimated with MLE methods mainly due to the computational
intensity of the iterative techniques that control for both heteroscedastic and spatially
correlated errors (Carrión-Flores and Irwin 2004). For instance, Carrión-Flores and
Irwin (2004) explicitly avoided using a spatial binary model, because the iterative tech-
niques were impractical for the large sample size (9,760 observations).
Responding to this very issue, Klier and McMillen (2008) proposed a computation-
ally feasible estimator for spatial discrete-choice models. Their estimator is a linearized
version of the GMM estimator proposed by Pinske and Slade (1998), and it extends the
literature on spatial modeling by allowing a spatially weighted dependent variable to
be estimated in a discrete-choice framework. The benefits of linearization are that it
allows the model to be estimated with large sample sizes because no matrix needs to be
inverted, and estimation requires only standard probit or logit models. The approach
produces a practical estimation method with a few approximations of sample-based
moments. The shortcoming of this type of GMM estimator is that the asymptotic prop-
erties of the GMM do not hold, so that it is biased (Smirnov 2010).
Alternatively, nonparametric GMM models have been applied to discrete land use
decisions, as shown in the case study. Because nonparametric GMM models investigate
the impact of spatial error autocorrelation based on the estimators of the asymptotic
variance of the sample average, the asymptotic covariance matrix estimators derived
from large-sample approximation do not suffer from the issue of computational inten-
sity associated with MLE for the spatial probit model. In additon, nonparametric GMM
provides consistency and asymptotic normality of the GMM estimator (Conley 1995).
As shown in the case study, spatial econometric estimates of the land use change
model can be used to simulate changes in landscapes, and simulation results can reveal
different effects of land use policies on not only individual development decisions but
also on overall landscape patterns. For example, Carrión-Flores and Irwin (2004) devel-
oped a two-step approach that combines a parcel-level, discrete land use decision model
and ex ante simulations of the discrete-choice model with and without land use policies
using spatial landscape pattern metrics. This kind of a two-step approach, which was
adopted in the case study, has the advantage of being able to link between econometric
modeling and landscape patterns. Such a modeling approach also allows simulation of
the aggregate effects of landscape transition probabilities resulting from land use policy
(Newburn et al. 2006; Lewis and Plantinga 2007; Langpap et al. 2008).
Future research in spatial econometric modeling of land use decisions needs to focus
on spatiotemporal modeling. Knowing the spatial structure of land use decisions is
essential to making informed policy and planning decisions, as shown in the case study
presented in this chapter. Likewise, enhanced understanding of the temporal dynam-
ics of land use decisions is important. A better model would be a spatial-dynamic
model based on a time series of actual land cover changes with the appropriate
time-varying explanatory variables under the framework of spatial econometric mod-
eling (Geoghegan et al. 2010). For example, three levels of land use change processes
are essential for spatial-dynamic modeling: (1) the slow processes of industrial, residen-
tial, and transport construction; (2) the medium processes of economic, demographic,
and technological changes; and (3) the fast processes of mobility of labor, goods, and
information (Wegener 1994). These temporal processes are to be incorporated in the
spatial model.
An attempt at spatial-dynamic modeling could be framed as a spatial panel data
model, which is an emerging topic within the spatial econometrics literature (e.g.,
Baltagi et al. 2003; Elhorst 2003; Baltagi, Egger, and Pfaffermayr 2007; Baltagi, Kelejian,
and Prucha 2007; Kapoor et al. 2007; Anselin et al. 2008; Baltagi and Liu 2008; Baltagi
et al. 2009; Elhorst 2009; Millo and Piras 2009; Lee and Yu 2010b; 2010a, Lee and Yu
2010c; Pesaran and Tosetti 2011; Millo and Piras 2012). A rare application in modeling
discrete land use decisions is the panel data spatial logistics regression model by Frazier
and Kockelman (2005). Under their framework, both spatial autocorrelation and time
adjustment are incorporated to simulate future changes in population and land cover.
Forecasting based on simulation of land use changes is particularly useful in the frame-
work of the panel data spatial model because the spatial panel model and its correspond-
ing forecasts better fit with time series forecasting. The demand for panel data for use in
spatial-dynamic modeling is expected to increase as demand for forecasts of future land
use changes increases, particularly as interest increases in predicting the impact of cli-
mate change on land use.
Acknowledgments
The views expressed are those of the authors and cannot be attributed to University of
Tennessee.
References
Alig, R. J. 1986. Econometric analysis of the factors influencing forest acreage trends in the
southeast. Forest Science 32(1): 119–134.
Alig, R. J., F. C. White, and B. C. Murray. 1988. Economic factors influencing land use changes in
the south-central United States. U.S. Department of Agriculture Forest Service, Southeastern
Forest Experiment Station. Res. Pap. SE-272:23.
Alonso, W. 1972. A theory of the urban land market. In eds. M. Edel and J. Rothenberg, Readings
in urban economics, 104–111. New York: Macmillan.
Anselin, L., J. Le Gallo, and H. Jayet. 2008. Spatial panel econometrics. In eds. L. Matyas and P.
Sevestre, The econometrics of panel data: Fundamentals and recent developments in theory
and practice, 3rd ed., 624–660. Berlin, Heidelberg, Germany: Springer-Verlag.
Baltagi, B. H., P. Egger, and M. Pfaffermayr. 2007. Estimating models of complex FDI: Are there
third-country effects? Journal of Econometrics 140(1): 260–281.
Baltagi, B. H., P. Egger, and M. Pfaffermayr. 2009. A generalized spatial panel data model with
random effects. CPR Working paper series.
Baltagi, B., H. H. Kelejian, and I. R. Prucha. 2007. Analysis of spatially dependent data. Journal
of Econometrics 140(1): 1–4.
Baltagi, B. H., and L. Liu. 2008. Testing for random effects and spatial lag dependence in panel
data models. Statistics & Probability Letters 78(18): 3304–3306.
Baltagi, B. H., S. Heun Song, B. Cheol Jung, and W. Koh. 2007. Testing for serial correlation, spa
tial autocorrelation and random effects using panel data. Journal of Econometrics 140(1): 5–51.
Baltagi, B., S. Song, and W. Koh. 2003. Testing panel data regression models with spatial error
correlation. Journal of Econometrics 117: 123–150.
Bell, K. P., and E. G. Irwin. 2002. Spatially explicit micro-level modelling of land use change at
the rural-urban interface. Agricultural Economics 27(3): 217–232.
tive. American Journal of Agricultural Economics 78(5): 1168–1180.
Bockstael, N. E., and K. P. Bell. 1998. Land use pattern, and water quality: The effect of differ-
ential land management controls. In R. Just, and S. Netanyahu, eds. International water and
resource economics consortium Conflict and cooperation on trans-boundary water resources,
169–191. Dordrecht: Kluwer Academic.
Brueckner, J. K. 1987. The structure of urban equilibria: A unified treatment of the Muth-Mills
model. In E. S. Mills, eds. Handbook of regional and urban economics, Vol. 2, 821–845.
Amsterdam: NorthHolland.
Brueckner, J. K., and H. A. Kim. 2003. Urban sprawl and the property tax. International Tax and
Public Finance 10(1): 5–23.
Carrión-Flores, C., and E. G. Irwin. 2004. Determinants of residential land-use conversion and
sprawl at the rural-urban fringe. American Journal of Agricultural Economics 86(4): 889–904.
Case, A. 1992. Neighborhood influence and technological change. Regional Science and Urban
Economics 22(3): 491–508.
Cho, S., C. D. Clark, W. M. Park, and S. G. Kim. 2009. Spatial and temporal variation in the
housing market values of lot size and open space. Land Economics 85(1): 51–73.
Cho, S., D. M. Lambert, and R. K. Roberts. 2010. Forecasting open space with a two-rate prop-
erty tax. Land Economics 86(2): 263–280.
Cho, S., and D. H. Newman. 2005. Spatial analysis of rural land development. Forest Policy and
Economics 7(5): 732–744.
Cho, S., and R. K. Roberts. 2007. Cure for urban sprawl: Measuring the ratio of marginal implicit
prices of density-to-lot-size. Review of Agricultural Economics 29(3): 572–579.
Chomitz, K. M., and D. A. Gray. 1996. Roads, land use, and deforestation: A spatial model
applied to Belize. The World Bank Economic Review 10(3): 487–512.
City of Redmond. 2011. Site requirements for residential zones. http://www.codepublishing.
com/WA/redmond/CDG/RCDG20C/RCDG20C3025.html.
Conley, T. G. 1995. Econometric modeling of cross-sectional dependence. PhD dissertation,
University of Chicago.
Conley, T. G. 1999. GMM estimation with cross sectional dependence. Journal of Econometrics
92 (1):1–45.
Conley, T. G., and B. Dupor. 2003. A spatial analysis of sectoral complementarity. Journal of
Conley, T. G., and G. Topa. 1999. Socioeconomic distance and spatial patterns in unemployment.
C.V. Starr Center for Applied Economics, New York University.
Cramer, J. 1999. Predictive performance of the binary logit model in unbalanced samples.
Journal of the Royal Statistical Society. Series D (The Statistician) 48(1): 85–94.
Cropper, M., C. Griffiths, and M. Mani. 1999. Roads, population pressures, and deforestation in
Thailand, 1976–1989. Land Economics 75(1) 58–73.
Dipasquale, D. 1999. Why don’t we know more about housing supply? The Journal of Real Estate
Finance and Economics 18(1): 9–23.
Elhorst, J. P. 2003. Specification and estimation of spatial panel data models. International
Regional Science Review 26(3): 244–268.
Elhorst, J. P. 2009. Spatial panel data models. In M. M. Fischer, and A. Getis, eds. Handbook of
applied spatial analysis, 377–407. Berlin, Heidelberg, New York: Springer.
Environmental Systems Research Institute. 2004. Data and maps 2004. Redlands, CA: ESRI Inc.
Environmental Systems Research Institute. 2009. Arcgis 9.3. Redlands, CA: ESRI Inc.
Evans, A. 1999. The land market and government intervention. In eds. E. S. Mills, and P.
C. Cheshire, Handbook of regional and urban economics, 1637–1669. The Netherlands:
Elsevier.
Ewing, R. 1994. Characteristics, causes, and effects of sprawl: A literature review. Environmental
and Urban Issues 21(2): 1–15.
Farber, S., and A. Páez. 2007. A systematic investigation of cross-validation in GWR model esti-
mation: Empirical analysis and Monte Carlo simulations. Journal of Geographical Systems
9(4): 371–396.
Fleming, M. 2004. Techniques for estimating spatially dependent discrete choice models. In
L. Anselin, R. J. G. M. Florax, and S. J. Rey, eds. Advances in spatial econometrics, 145–168.
Heidelberg: Springer.
Fotheringham, A. S., C. Brunsdon, and M. Charlton. 2002. Geographically weighted regres-
sion: The analysis of spatially varying relationships. West Sussex, UK: John Wiley & Sons.
Frazier, C., and K. M. Kockelman. 2005. Spatial econometric models for panel data: Incorporating
spatial and temporal data. Transportation Research Record: Journal of the Transportation
Research Board 1902: 80–90.
Fujita, M. 1990. Spatial interactions and agglomeration in urban economics. In M. Chatterji,
and R. E. Kunne, eds. New frontiers in regional sciences, 184–221. London: Macmillan.
Fujita, M., and J. Thisse. 2002. Economics of agglomeration: Cities, industrial location, and
regional growth. Cambridge, UK: Cambridge University Press.
Geoghegan, J., D. Lawrence, L. C. Schneider, and K. Tully. 2010. Accounting for carbon stocks
in models of land-use change: An application to southern Yucatan. Regional Environmental
Change 10(3): 247–260.
Glaeser, E. L., M. E. Kahn, and J. Rappaport. 2008. Why do the poor live in cities the role of
public transportation. Journal of Urban Economics 63(1): 1–24.
Grenander, U., and M. Rosenblatt. 1957. Some problems in estimating the spectrum of a time
series. Berkeley: University of California Press.
Hall, P., N. Fisher, and B. Hoffmann. 1994. On the nonparametric estimation of covariance
functions. The Annals of Statistics 22(4): 2115–2134.
Hall, P., and P. Patil. 1994. Properties of nonparametric estimators of autocovariance for sta-
tionary random fields. Probability Theory and Related Fields 99(3): 399–424.
Hanham, R., and J. S. Spiker. 2005. Urban sprawl detection using satellite imagery and geo-
graphically weighted regression. In R. Jensen, J. Gatrell, and D. McLean, eds. Geo-spatial
technologies in urban environments, 137–151. Berlinthebe: Springer.
Hardie, I. W., and P. J. Parks. 1997. Land use with heterogeneous land quality: An application of
an area base model. American Journal of Agricultural Economics 79(2): 299–310.
Harrison County. 2009. What is lot coverage? Http://Co.Harrison.Ms.Us/Downloads/
Downloads%20by%20department/Zoning/Informational%20brochures/Lot%20cover-
age%20brochure%20final.Pdf).
Irwin, E. G., K. P. Bell, and J. Geoghegan. 2003. Modeling and managing urban growth at the
rural-urban fringe: A parcel-level model of residential land use change. Agricultural and
Resource Economics Review 32(1): 83–102.
Irwin, E. G., K. P. Bell, and J. Geoghegan 2006. Forecasting residential land use change. In eds.
R. J. Johnston, and S. K. Swallow, Economics and contemporary land use policy: Development
and conservation at the urban-rural fringe, 55–82. Washington, DC: Resources for the
Future.
Irwin, E. G., and N. E. Bockstael. 2001. The problem of identifying land use spillovers: Measuring
the effects of open space on residential property values. American Journal of Agricultural
Economics 83(3): 698–704.
Irwin, E. G., and N. E. Bockstael. 2002. Interacting agents, spatial externalities and the evolu-
tion of residential land use patterns. Journal of Economic Geography 2(1): 31.
Irwin, E., and N. Bockstael. 2004. Land use externalities, open space preservation, and urban
sprawl. Regional Science and Urban Economics 34(6): 705–725.
Irwin, E. G., and J. Geoghegan. 2001. Theory, data, methods: Developing spatially explicit eco-
nomic models of land use change. Agriculture, Ecosystems & Environment 85(1–3): 7–24.
Isberg, G. 1973. Controlling growth in the urban fringe. Journal of Soil and Water Conservation
28(4): 155–161.
Johnston, R. A., and M. E. Madison. 1997. From landmarks to landscapes. Journal of the
American Planning Association 63(3): 365.
Kapoor, M., H. H. Kelejian, and I. R. Prucha. 2007. Panel data models with spatially correlated
error components. Journal of Econometrics 140(1): 97–130.
Kelejian, H. H., and I. R. Prucha. 1999. A generalized moments estimator for the autoregressive
parameter in a spatial model. International Economic Review 40(2): 509–533.
Klier, T., and D. P. McMillen. 2008. Clustering of auto supplier plants in the United
States: Generalized method of moments spatial logit for large samples. Journal of Business
and Economic Statistics 26: 460–471.
Kline, J. D., and R. J. Alig. 1999. Does land use planning slow the conversion of forest and farm-
lands? Growth and Change 30(1): 3–22.
Kopits, E., V. McConnell, and M. Walls. 2007. The trade-off between private lots and pub-
lic open space in subdivisions at the urban-rural fringe. American Journal of Agricultural
Economics 89(5): 1191–1197.
Langpap, C., I. Hascic, and J. Wu. 2008. Protecting watershed ecosystems through targeted local
land use policies. American Journal of Agricultural Economics 90(3): 684–700.
Lee, L., and J. Yu. 2010a. A unified estimation approach for spatial dynamic panel data mod-
els: Stability, spatial co-integration, and explosive roots. In A. Ullah and D. E. A. Giles, eds.
Handbook of empirical economics and finance, 397. Chapman and Hall/CRC.
Lee, L., and J. Yu. 2010b. A spatial dynamic panel data model with both time and individual
fixed effects. Econometric Theory 26(2): 564–597.
Lee, L., and J. Yu. 2010c. Estimation of spatial autoregressive panel data models with fixed
effects. Journal of Econometrics 154(2):165–185.
LeSage, J. P. 1999. Spatial econometrics. Regional Research Institute, West Virginia University.
Leung, Y., C. L. Mei, and W. X. Zhang. 2000. Testing for spatial autocorrelation among the resid-
uals of the geographically weighted regression. Environment and Planning A 32(5): 871–890.
Lewis, D. J., and A. J. Plantinga. 2007. Policies for habitat fragmentation: Combining economet-
rics with GIS-based landscape simulations. Land Economics 83(2): 109–127.
Libby, L. W., and J. S. Sharp. 2003. Land-use compatibility, change, and policy at the
rural-urban fringe: Insights from social capital. American Journal of Agricultural Economics
85(5): 1194–1200.
Lichtenberg, E. 1989. Land quality, irrigation development, and cropping patterns in the
Northern High Plains. American Journal of Agricultural Economics 71(1): 187–194.
Lichtenberg, E., and I. Hardie. 2007. Open space, forest conservation, and urban sprawl
in Maryland suburban subdivisions. American Journal of Agricultural Economics
89(5): 1198–1204.
Lichtenberg, E., C. Tra, and I. Hardie. 2007. Land use regulation and the provision of open space
in suburban residential subdivisions. Journal of Environmental Economics and Management
54(2): 199–213.
Liu, C., P. Berry, T. Dawson, and R. Pearson. 2005. Selecting thresholds of occurrence in the
prediction of species distributions. Ecography 28(3): 385–393.
McPherson, J., W. Jetz, and D. Rogers. 2004. The effects of species’ range sizes on the accuracy
of distribution models: Ecological phenomenon or statistical artifact? Journal of Ecology
41(5): 811–823.
Miller, C., and J. Vaske. 2003. Individual and situational influences on declining hunter effort in
Illinois. Human Dimensions of Wildlife: An International Journal 8(4): 263–276.
Millo, G., and G. Piras. 2009. Implementation of ML estimation for spatial panels. Unpublished,
Regional Research Institute.
Millo, G., G. Piras. 2012. Splm: Spatial panel data models in R. Journal of Statistical Software
47(1): 1–38.
MPD. 2009. Metro planning department. Http://Www.Nashville.Gov/Mpc/.
Nelson, G. C., and D. Hellerstein. 1997. Do roads cause deforestation? Using satellite images in
econometric analysis of land use. American Journal of Agricultural Economics 79(1): 80–88.
New York City. 2009. New York City zoning. Http://Www.Nyc.Gov/Html/Dcp/Html/Zone/
Glossary.Shtml.
Newburn, D. A., P. Berck, and A. M. Merenlender. 2006. Habitat and open space at risk of
land-use conversion: Targeting strategies for land conservation. American Journal of
Agricultural Economics Report 88(1): 28–42.
Nickerson, C., and L. Lynch. 2001. The effect of farmland preservation programs on farmland
prices. American Journal of Agricultural Economics 83(2): 341–351.
Páez, A. 2006. Exploring contextual variations in land use and transport analysis using a probit
model with geographical weights. Journal of Transport Geography 14(3): 167–176.
Parks, P. J., and B. C. Murray. 1994. Land attributes and land allocation: Nonindustrial forest use
in the Pacific Northwest. Forest Science 40(3): 558–575.
Pesaran, H., and E. Tosetti. 2011. Large panels with common factors and spatial correlations.
Journal of Econometrics 161(2): 182–202.
Pierson, K. M. 2002. Livable communities toolkit: A best practices manual for metropolitan
regions. http://www.crcog.org/community_dev/livable_toolkit.html.
Pinske, J., and M. E. Slade. 1998. Contracting in space. Journal of Econometrics 85(1): 125–154.
Plantinga, A. J., T. Mauldin, and R. Alig 1999. Land use in Maine: Determinants of past trends
and projections of future changes, 20. Portland, OR: U.S. Department of Agriculture, Forest
Service, Pacific Northwest Research Station.
Rybczynski, W. 2005. Suburban despair. Is urban sprawl really an American menace? Slate,
November 7.
Schabenberger, O., and F. J. Pierce. 2002. Contemporary statistical models for the plant and soil
sciences, xxii, 738. Boca Raton, FL: CRC Press.
Schaffer, W. 1999. The web book of regional science. Regional Research Institute, West Virginia
University.
Smirnov, O. A. 2010. Modeling spatial discrete choice. Regional Science and Urban Economics
40(5): 292–298.
Thorsnes, P. 2002. The value of a suburban forest preserve: Estimates from sales of vacant resi-
dential building lots. Land Economics 78(3): 426–441.
Thünen, J. H., and J. Heinrich. 1966. The isolated state. Trans. Carla M. Wartenberg.
Oxford: Pergamon.
Turner, M. G., D. N. Wear, and R. O. Flamm. 1996. Land ownership and land-cover change in
the Southern Appalachian Highlands and the Olympic Peninsula. Ecological Applications
6(4): 1150–1172.
Wegener, M. 1994. Operational urban models: State of the art. Journal of the American Planning
Association 60(1): 17–29.
Wheeler, D., and M. Tiefelsdorf. 2005. Multicollinearity and correlation among local regres-
sion coefficients in geographically weighted regression. Journal of Geographical Systems
7: 161–187.
White, F. C., and F. N. Fleming. 1980. An analysis of competing agricultural land uses. Southern
Journal of Environmental Economics and Management 46(2): 288–309.
C HA P T E R 1 8
U S I N G Q UA S I - E X P E R I M E N TA L
M E T H O D S T O E VA LUAT E
LAND POLICIES
Application to Maryland’s Priority
Funding Legislation
C HA R L E S TOW E , R E BE C C A L EW I S , A N D
LOR I LY NC H
States and local governments use a variety of instruments to direct the location, inten-
sity, and timing of growth. In some cases, governmental entities desire growth to locate
in area with existing infrastructure such as roads, sewer, and schools. In others, gov-
ernmental entities cannot accommodate the number of housing structures being built
and the population growth with existing resources. In hopes of retaining open space or
a viable agricultural industry, governments may increase (decrease) or encourage (dis-
courage) the density permitted within certain areas. However, even with ongoing plan-
ning and constant review of the instruments selected, concerns about land use continue.
Establishing the effectiveness of the plethora of instruments has been plagued by the
difficulty of establishing the causality of a particular policy related to encouraging and
discouraging certain land use actions. Recently, researchers examining land use issues
have begun to explore new methods to elicit the effectiveness of land-related policies
and programs.
Empirical researchers need to establish a causal effect of programs or policies. This task is
complicated by implicit or explicit selection and potential endogeneity on the part of land
agents. In the past two decades, and especially in the field of labor economics, researchers
have been focused on the econometrics and statistical analysis of causal effects (Imbens
and Wooldridge 2009). To this end, researchers seek to evaluate how a program (or treat-
ment) affects certain entities (individuals or parcels of land) given some desired outcome
of interest. In the sciences, one randomly assigns an individual to receive the new drug or
the placebo and then compares the outcome of individuals who were “treated”—that is,
Using Quasi-Experimental Methods to Evaluate Land Policies 453
received the new drug—to those of the “control”—that is, those who received the placebo.
Complications arise, however, when researchers are unable to design a study that imple-
ments random assignment, a virtual impossibility in the field of land economics and land
use policy evaluation. Quasi-experimental methods are a class of models that utilize obser-
vational data and attempt to recreate the ideal of random assignment. Many models exist
under this framework, including instrumental variables, regression discontinuity, various
difference in difference type models, and propensity score matching (PSM).
In this chapter, we outline the land policy environment in Maryland and the issues
that arise when trying to determine causality of a policy on outcomes. We describe the
policy we use as an illustrative example. After discussing the challenges, we consider var-
ious econometric and statistical approaches, including PSM. We then demonstrate the
strengths of the PSM method in the evaluation of this particular land use policy utilizing
an extensive spatial microlevel dataset from the state. We discuss the potential areas of
concern with PSM approach. We discuss the results of the analysis and their limitations.
In the conclusion, we discuss some of the broader issues with quasi-experimental meth-
ods related to land use issues.
1. Policy Environment
Most, if not all, policy environments dealing with land use are not candidates for experimen-
tation or random assignment, and, as such, basic regression-based modeling techniques
are open to the critique of ignoring selection issues or functional form assumptions to con-
struct counterfactuals. These methods often fail to detect small biases in the data between
the so-called treated and untreated. Even among methods classified as quasi-experimental,
such as instrumental variables, producing and observing “good” instruments are rare.
Similarly with regression discontinuity approaches, clear and exogenous discontinuities
often are not available. In the context of land use policy evaluation, PSMs main advantage
is that it does not rely on exclusion restrictions (i.e., a variable that affects “treatment” status
but does not affect the outcome of interest). In fact, all variables affecting both the treatment
and the outcome should be included in the propensity score model. However, we do not
need to specify exactly how the variables impact treatment or outcome. We choose to evalu-
ate an actual policy in the state of Maryland to illustrate the use of the PSM approach most
effectively. This policy focuses financial incentives for development in targeted geographic
areas while leaving similar geographic areas without such incentives.
Generally speaking, quasi-experimental methods can be used to evaluate regulatory
or voluntary programs in which decision makers can be individual landowners or rent-
ers, counties, states, regional entities, or any other geographic aggregation.1 Treatments
1 As long as the geographic aggregation is not confounded with the outcome measurement, as might
be the case using census-defined boundaries.

454 Charles Towe, Rebecca Lewis, and Lori Lynch
could be any of the instruments or programs employed to alter the location, intensity,
or timing of land use decisions. We will employ the matching technique with land units
typically smaller than non-subdivided parcels being the implicit actors to analyze a spa-
tially defined incentive-based land use policy.
In this application, Maryland’s Smart Growth Program relies on financial incentives
to direct growth to designated areas in an attempt to curb low-density (intensity) or
sprawling development (location). Under Smart Growth, the state sought to stop “sub-
sidizing sprawl” and direct its funding for infrastructure only to locally designated and
state-approved Priority Funding Areas (PFAs). A county seeking state funds to finance
infrastructure needed to designate and have approved such geographic areas or PFAs (or
designated growth areas). The policy is similar to an urban growth boundary in design
but lacks the “teeth” of regulatory policy, such as strong zoning changes to support it.
The PFA approach utilized a combination of planning and monetary incentives to
direct housing development to these growth areas in order to make it relatively less
profitable to convert land currently in an agricultural or resource use and to constrain
urban expansion. These designated areas: (1) depict a physical line between urban and
rural areas; (2) limit expansion of services such as water and sewer infrastructures by
withholding state-level financial incentives outside PFAs; and (3) by providing infra-
structure spending inside PFAs, lower the cost of housing construction. However, one
questions whether the PFA approach has accomplished its goals. At the bottom line, one
might ask: Have PFAs caused the redirection of the housing construction, or any subset
of housing construction, within Maryland?
On the face of it, one could answer this question simply by comparing housing
starts within the PFA to those outside the PFA. However, assessing the impact of the
PFAs on housing starts is more challenging that it appears. The research question cen-
ters on comparing what would have happened to an individual land parcel in two sce-
narios; that is, one knows what has happened to housing starts for a land parcel within
the PFA, but one would also like to know what would have happened to housing starts
for the same land parcel if it had not been within a designated PFA. This land parcel
within the PFA cannot be in two states simultaneously, nor can a researcher randomly
assign which area is designated a PFA and which is not. This is the classic evalua-
tion problem, which Holland (1986) defines as the fundamental problem of causal
inference. Because we are analyzing data from a real-world policy (i.e., observational
data rather than experimental data), we face challenges in estimating causal effects
without making assumptions concerning unconfoundedness, exogeneity, ignorabil-
ity, or selection on observables (Imbens and Wooldridge 2009). For example, if PFAs
are only designated in areas where growth would otherwise have gone, the change in
financial incentives could have been inconsequential, thus not affecting the timing
or location of housing starts at all. As such, it would be an inefficient expenditure of
scarce public funds. One would find growth dominates within the PFAs, but the cau-
sality may be spurious.
This type of endogeniety is not a new issue in econometrics. In fact, these issues have
been well studied, particularly by labor econometricians. The common example in labor
economics is that individuals choosing to participate in a training program may be fun-

damentally different from those who do not. If these differences affect the labor market
outcome, such as the wage level, then the computed causal effects of the training may be
invalid. Those who chose to not participate in the identical training program may not
have the same outcome if they did. Similar concerns confront land program evaluations.
The selection of the land parcels within a designated growth area can be influenced by
factors including the political power of local communities, the open space inventory
within a county, rising land values, and other development pressures. Landowners may
only “permit” the PFA to include their land when they anticipate an increase in their
land value. And landowners outside the PFA may fight to include their land within a
PFA if they perceive their land value will decrease if placed outside the PFA due to fewer
incentives for housing development.
Analogous to the training program example, the factors that cause certain parcels or
areas to be inside or outside the PFA boundary (participate/not participate in training)
will also affect the outcome—in our case, the probability of conversion to housing. If
one estimates the impact of the PFA without considering the potential endogeneity of its
boundaries, the impact of these other factors on the probability of housing starts within
the PFA will also be included in the impact estimates of the PFA program and invalidate
the causal implications of the analysis.
Endogeniety, measurement error, and omitted variables issues can result in
included covariates being correlated with the error term in the regression. In these
cases, the estimates in an ordinary least square (OLS) regression would be biased
and inconsistent. To overcome such issues, researchers have resorted to a variety of
econometric and statistical approaches. Land program evaluation may use an instru-
mental variable approach, which relies on the presence of an instrument, or several
variables, that satisfy specific exogeneity and exclusion restrictions. As outlined in
Angrist and Krueger (2001), Wright (1928) used curve shifters such as weather and
price of substitute goods to estimate supply and demand curves. Theil (1953) intro-
duced a two-stage approach to allow the inclusion of more than one instrument to
predict the endogenous covariates. However, these necessary exogeneity and exclu-
sion restrictions have often proved quite difficult to meet. An instrument must be
correlated with the endogenous covariate but not be correlated with the error term.
Discovering such an instrument can be problematic. For example, if the instrument is
correlated with the error, estimators will remain inconsistent. Similarly, instrumental
variable that have some association with omitted variables can result in biased esti-
mators. In addition, weak instruments may result in predicted covariates with little
variation and, as such, they do not have much explanatory ability in the regression
equation (Angrist and Krueger 2001). Thus, one cannot interpret an insignificant
estimator as evidence of the lack of causality because it could be that the instrument
was too weak to elicit the causality effect.
When using standard regression methods that assume a linear approximation,
researchers must also be aware that the average treatment effects estimated can
be biased if the linear approximation in not accurate globally. Imbens and Rubin
(forthcoming, as cited in Imbens and Wooldridge 2009, 24) suggest, “as a rule of
thumb that with a normalized difference exceeding one quarter, linear regression
methods tend to be sensitive to the specification.” Imbens and Rubin (forthcom-
ing, as cited in Imbens and Wooldridge 2009) suggest comparing the covariates of
the treated and the control groups by computing the normalized differences. They
believe normalized differences, which are not sensitive to sample size, are superior to
comparing t-statistics. The problem of inference for the average treatment effect and
thus credible results is also not inherently more difficult in larger samples as the t-test
could suggest. In cases in which the normalized difference exceeding one-quarter,
PSM, which does not rely on a linear approximation, provides a more viable approach
to the analysis.
In some cases, researchers can utilize a regression discontinuity design if a common
and exogenous border of some type can be identified such that the continuity can be
exploited. Regression discontinuity can be used when one is designing an experiment
but wants to ensure the treatment is received by those in most need (i.e., the sickest,
poorest, the preferred growth area, or least educated for example) rather than randomly.
Due to the observational quality of the data, a land policy evaluation is most likely to
be used when the treated is geographically next to the control or untreated sample.
One assumed in this case that the treated parcels along the cutoff point would be quite
similar to the untreated parcel along the same point. As such, to employ the technique,
the two samples must be ordered across space, with a clear cutoff point for inclusion in
the treated or control group. One then can compare the outcomes of the treated parcel
inside the cutoff with those outside the cutoff to determine if there is a difference. In
some cases, however, one find a “fuzzy” discontinuity rather than a sharp cutoff point.
In this case, the discontinuity may be highly correlated with the treatment. This could
be due to geographic features in the land; path-dependency, such as previous land use
decisions; or political pressure. This can create many of the same problems as faced by
instrumental variables.
By design, regression discontinuity produces results along the cutoff point; that is,
local average treatment effects. These results may not be generalizable to the whole
population. Also, many regressions have fewer data points due to the necessity of being
set along the cutoff point. There could also be fewer observations outside the cutoff area
compared to more within the treated area. The lack of power may result in an insig-
nificant estimator. Regression discontinuity estimates also suffer if misspecified; for
example, functional forms that do not include existing nonlinear relationships result in
biased results.
1.1 Advantages of the Propensity Score Matching Approach

In this analysis, we use a quasi-experimental approach, PSM (Rosenbaum and Rubin
1983), which implicitly assumes endogenous PFA boundaries. Our focus is on a binary
treatment (in PFA/out of PFA), which is a common situation confronted by researchers
addressing land issues.2 In this case, we have a large and rich set of covariates that will
impact both the treatment designation (PFA status) and influence the outcome variable
of interest (number of housing starts). We employ the PSM method to evaluate the PFA
program ability to shift the location of housing starts by comparing the outcome for the
gridded landscape3 of Maryland within the PFAs with observationally equivalent grids
that are outside the PFAs (Kaza et al. 2011). The primary advantages of PSM under these
circumstances are that (1) we do not need to specify how each variable will affect selec-
tion into treatment or control, (2) we do not need to specify how each variable affects the
potential outcome, and (3) nor do we need to determine an exogenous variable to satisfy
an exclusion restriction. The PSM method has several other benefits as well. First, the
matching protocol ensures that the grids within the PFAs are matched to the grids out-
side the PFAs that are most similar to them in terms of observable characteristics. This
provides a more transparent means to limit the influence of outliers and dissimilar grids.
Second, because not all grids are equally likely to have been designated PFAs or have
housing starts, this method incorporates pretreatment covariates that may influence the
existence of a PFA designation ands housing starts into the propensity score calculation.
Third, a linear functional form is not assumed for the outcome equation, the decision
process, or the unobservable variables. As such, PSM requires fewer functional form
and homogeneity assumptions. However, we do rely on the assumption of “selection
on observables,” which implies the data are rich enough to describe the selection pro-
cess (Smith and Todd 2005). Because we have information on the outcome for treated
and control land parcels both before and after the treatment, a difference-in-difference
approach can be utilized, which should remove any selection issues derived from
time-insensitive unobservable.
2. Analytical Method
We employ the PSM method developed by Rosenbaum and Rubin (1983). PSM has not
been used to study land policy evaluation issues until recently. It is used to study the
land market effects of zoning (McMillan and McDonald 2002), the land market effects
of conservation easement restrictions (Lynch et al. 2007, 2009), land market effects of
down-zoning (Liu and Lynch 2011a), the impact of farmland preservation programs on
farmland loss (Liu and Lynch 2011b), the impact of designated preservation zones (i.e.,
Rural Legacy Areas [RLAs]) on rates of preservation (Lynch and Liu 2007), the impact
of energy policy on farm prices (Towe and Tra 2013), and the impact of development
2
Quasi-experimental methods have been extended recently to multivalued and continuous
treatments (Imbens 2000; Gill and Robins 2001; Lechner 2001; Lechner and Miquel 2005).
3 We grid the landscape at ¼-mile squares to allow for aggregation of data (e.g., housing starts) by an
area not defined by the landscape, social or geographic, of existing houses, like Census designations.
moratoria on housing starts (Bento et al. 2007). Each of these studies performed use-
ful analysis within its specific policy environments; however, the applicability of these
finding to other policy environments may be limited; this is an empirical question that
further PSM analysis will help researchers answer.
As mentioned earlier, assessing the impacts of PFAs on housing starts, like many
other land policy evaluation, is difficult because of incomplete information. Although
one can identify whether a grid is part of a PFA (is treated) or not (not treated/a control)
and the outcome (number of housing starts or the difference in housing starts pre and
post PFA) conditional on its treatment, one cannot observe the counterfactual—what
would have happened if the grid was not part of the PFA. Thus, the fundamental prob-
lem in identifying the true causal effect is constructing the unobservable counterfactu-
als for treated observations.
At this point, some simple notion serves to demonstrate the issue, let Y1 denote the
outcome in the group of grids if treatment has occurred (D = 1), and Y0 denote the out-
come for the grids of control observations (D = 0). If one could observe the treated and
the control states, the average treatment effect, τ, would equal Y1 − Y0 where Y1 equals
the mean outcome of the treatment grids and Y0 of the control grids. Unfortunately,
only Y1 or Y0 are observed for each observation. In a laboratory experiment, research-
ers solve this problem by randomly assigning subjects to be treated or not treated,
and then they construct the unobserved counterfactual. In a natural setting, however,
τ ≠ Y1 − Y0 because the treatment condition is not randomly assigned. The PSM method
demonstrates that if data justify matching on some observable vector of covariates, X,
then matching pairs on the estimated probability of selection into treatment or control
groups based on X is also justified. In our case then, within land grids with the same pro-
pensity score, the land characteristics and other variables can be treated as independent
of the treatment status. Therefore, the average treatment on the treated estimates for
grids with the same propensity scores will not be biased and can be compared.
The PSM method relies on the assumption of conditional independence, which
requires that there are no unobserved factors associated both with the treatment and the
outcomes conditional on observed covariates. The PSM method, like other economet-
ric and statistical approaches, suffers from the presence of unobservable covariates that
may not be independent to the treatment assignment or to the treatment itself.
To satisfy the conditional independence assumption (CIA) and estimate an unbi-
ased treatment effect, one must find a vector of covariates, X, such that Y0 ⊥ D | X ; or
Y0 ⊥ D | P (D = 1 | X ) where P (D = 1 | X ) ∈ (0,1) is the propensity score that an individual
self-selects into treatment groups, and ⊥ denotes independence. If CIA holds, Y0, the
outcome for the controls (D = 0), can be assigned to the corresponding treated observa-
tions (D = 1) as their unobserved counterfactuals using certain matching techniques.
This assumption may fail if the “independent” variables, the parcel characteristics, are
affected by the treatment as well. Wooldridge (2005) demonstrates that if treatment
is randomized with respect to the counterfactual outcome but not with respect to the
other variables, then CIA will be violated.
The CIA condition is quite strong. Therefore, we use the conditional mean
independence (CMI) assumption (Heckman et al. 1998) that E[Y0 | D = 1, X ] =
E[Y0 | D = 0, X ] = E[Y0 | X ], P (D = 1 | X ) ∈(0,1) to estimate the average treatment effect.
The average treatment effect on the treated sample is thus the expected difference in
outcome Y between the PFA grids and their corresponding counterfactuals (non-PFA
grids) constructed from the matched controls:
∆ TT = E(Y1 | D = 1) − E(Y0 | D = 1) = E(Y1 | D = 1) − E(Y0 | D = 0, P ( X )) (1)

For the weaker condition to hold, the set of X needs to include all of the variables that
may affect the outcome (housing starts) and the selection into the treatment state (PFA
status or not).
By using the matching algorithm, we are constructing a counterfactual pool of land
grids equivalent in the covariates for the matched treatment and control observa-
tions, and we are controlling for the effect that these factors may have on the number
of housing starts. In short, we are recreating a random experiment or, in this case, a
quasi-random experiment. As such, we must also ensure that the samples “overlap.”
The treated areas must have control areas that are observationally equivalent, such that:
0 < pr (Wi = 1Xi = x ) < 1, for all x .
Thus, the support of the conditional distribution of Xi given Wi = 0 contains the con-
ditional distribution of Xi given Wi = 1. If the covariates for the control observations
(non-PFAs) do not overlap with the covariates for the treatment observations (PFAs),
then we end up estimating an average impact only over the range where the overlap
exists. The overlap is called the common support. The propensity scores can illumi-
nate whether this assumption is satisfied. For example, if few or none of the non-PFA
grids have a high probability of being a PFA, then those grids with a high probability of
being a PFA will have few grids from which to make the counterfactual comparison(s).
Similarly, if many of the grids have propensity scores close to zero, estimating the aver-
age effect of the treatment precisely becomes more difficult (Imbens and Wooldridge
2009). Problems are also likely to arise when some grids are almost certain to receive
treatment. When the covariate distributions are different between the treatment and
control grids, the propensity scores may approach zero and 1. This might produce dif-
ferent probabilities or propensity scores for logit and probit models, yet model choice in
these cases are often ad hoc rather than well-motivated (Imbens and Wooldridge 2009).
Dehejia and Wahba (2002) suggest finding the smallest value of the estimated pro-
pensity score among the treated observations and dropping all control observations
that have an estimated propensity score less than it. By setting the relevant sample
to have common support, one eliminates those non-PFA grids that are so different
from the PFA grids that they should not be compared. The results will be sensitive to
the threshold chosen for the common support and, as such, some sensitivity analysis
may be warranted. Heckman, Ichimura, and Todd (1997) and Heckman et al. (1998)
use density functions to determine the set of treatment and control variables. Rubin
(2006) proposes using a matched sample by ordering the treated observations by their
estimated propensity score and then matching them to the nearest control grid. If one
matches without replacement, one will end up with an equal number of non-PFA and
PFA grids. One does not need to use these pairs for estimating the average treatment on
the treated—the impact results—but rather to determine an overlapping sample. This
approach can also improve the balance between covariates.
2.1 Matching Methods and Bandwidth Selection

Matching estimators construct an estimate of the expected unobserved counterfactual
for each treated observation by taking a weighted average of the outcomes of the control
observations.4 In our case, this would be an estimate of a PFA’s housing starts if it was not
a PFA based on the outcome of the non-PFAs grids. Several different matching methods
are available. For example, nearest-neighbor-only uses the control grid with the closest
propensity score to each treated grid. All matching estimators have the generic form for
estimated counterfactuals:
 

(Yio | Di = 1) = 
 ∑ w(i, j)Y jo | D j = 0 ,

 j ∈{D j = 0} 

where j is the index for control observations that are matched to the treated observa-
tion i based on estimated propensity scores (j = 1, 2, . . . J). The matrix, w(i, j), contains
the weights assigned to the jth control observation that is matched to the ith treated
observation. By using different weights in different matching estimators, one is implic-
itly making a tradeoff between efficiency and bias. All the estimators are asymptoti-
cally the same in large sample but might return different estimates in finite samples.
Nearest-neighbor matching has each PFA grid paired with the control grid whose pro-
pensity score is closest in absolute value (Dehejia and Wahba 2002). Dehejia and Wahba
(2002) and Rosenbaum (2002) both found that matching with replacement performs as
well or better than matching without replacement (in part because it increases the num-
ber of possible matches and avoids the problem that the results are potentially sensitive
to the order in which the treatment observations are matched). This may result in a con-
trol grid never being used to compute the average treatment effect if it is not the nearest
neighbor to any treated PFA grid. This reduces possible bias but is not necessarily the
most efficient.
Kernel and local linear matching techniques match each treated PFA grid with all
control grids where the estimated propensity scores fall within a specified bandwidth.
Uniform kernel, for example, gives equal weight to all control grids within the chosen
4 This subsection follows the description provided by in Liu and Lynch (2011b).
bandwidth or for a wide variety of bandwidths if implemented via multiple nearest

neighbors. The matched control grids are weighted according to the density function
of the kernel type, but more control grids are used, which permits higher efficiency but
potentially more bias. Bandwidths are centered on the estimated propensity score of the
treated observation.
In some evaluation studies, the distribution of the control and treated propensity
scores are not equivalent. For example, the estimated propensity scores for the control
grids can be asymmetrically distributed with a large tail at zero. The estimated propen-
sity scores for the treatment grids can be more evenly distributed or with a large tail
at one. Kernel matching uses the additional data, where they exist, but excludes bad
matches; thus, it can be a better choice when asymmetric distributions exist. McMillen
and McDonald (2002) suggest that the local linear estimator is less sensitive to bound-
ary effects (i.e., when many observations have a propensity score near one or zero).
The minimum mean square errors (MSE) for different matching methods and differ-
ent bandwidth combinations can be used to pick the optimal bandwidth for each kernel
type. Then, using the optimal bandwidth, one can select the optimal kernel type based
on the minimum MSE for each matching method. Finally, one may compute which
matching method to use based on the minimum MSE, given their optimal kernel type
and bandwidth. One can also employ Racine and Li’s (2004) leave-one-out validation
mechanism to choose among the matching methods. The formula for calculation of
treatment effect on treated thus is:
N   
∆ TT =
1 N 
N i=1 
Y − (
|D =1  = 1
∑ i1 io i  N ∑
Y ) 
i=1 
Yi1 −  ∑ ( ) jo j
 j∈{D j =0}
w i , j Y | D = 0 
 
 
2.2 Balancing Test
Three types of balancing test methods exist in the empirical literature: standardized
difference test, Hotelling T2 for joint equality test, and a regression-based test. The
Hotelling T2 tests the joint null of equal means of all of the variables included in the
matching between the treatment group and the matched control group. Smith and Todd
(2005) found that, in some cases, Hotelling T2 incorrectly treated matched weights as
fixed rather than random. The standardized difference test uses a t-test for equality of
the means for each covariate in the matched treated and control grids. The regression
test estimates a regression of each covariate on polynomials of the estimated propensity
scores, [P ( X )] and the interaction of the polynomials with the treatment binary vari-
l
able, D * [P ( X )]l (l, the order of the polynomial, equals 3). The treated PFA grids would
not have a different regression line than the non-PFA control grids, and the balancing
condition is satisfied if the estimated coefficients on the interaction terms are jointly
equal to zero according to an F-test.
2.3 Difference in Difference
The effect of the PFA on the likelihood of new housing starts, controlling for the over-
all trend in housing starts over the relevant time period, is accomplished using the
difference-in-difference approach or by simply replacing the outcome variable, housing
starts in 2000–2003, with the difference in housing starts from 1994–1997 to 2000–2003.
These years were selected because the legislation passed in 1997 and was implemented
in the counties by 2000.
Y Post = Y 2003 + Y 2002 + Y 2001 + Y 2000
and
Y Pre = Y1997 + Y1996 + Y1995 + Y1994
So, ∆Y = YPost-YPre is the outcome variable. This overall trend incorporates the
market-level effects of housing demands and has the added advantage of controlling for
time invariant and unobservable characteristics.
One may also conduct robustness tests by restricting to which control grids the treated
grids can be compared. This can address possible unobservable characteristics of grids
that vary spatially, by time period, or by governmental entities. For example, in this case,
intuitively, the further away the PFA grids are from the grids that are not within the PFA,
the more likely it is that their housing starts may be influenced by unobserved factors
(e.g., different land markets). For example, the Chesapeake Bay Bridge may be a psy-
chological barrier to many individuals who do not look beyond it to buy a home, irre-
spective of the time or distance of the commute. These types of unobserved factors may
bias our estimates if we use all PFA and non-PFA grids when matching whether or not
they are on the same side of the Chesapeake Bay Bridge. For robustness against possible
unobservable factors, we restrict matches on three scales: (1) any PFA and non-PFA grid
in the state (least restrictive); (2) any PFA grid and non-PFA grid within the same region
as defined in Figure 18.1; and (3) any comparable PFA grid and non-PFA grid within the
same county (most restrictive). We also limit our analysis to smaller geographic regions
of the state: Western, Central 1 and Central 2, South, Upper Eastern Shore, and Lower
Eastern Shore. See Figure 18.1 for the regional boundaries.
We first match the treatment and control observations without any restriction and
calculate the overall treatment effect. Matching over the full sample has the advantage of
providing better controls for treated grids than matching within the county where fewer
non-PFA grids would be available. We also restrict matches to grids within the same
region and county.
Maryland regions
Legend
West
Central 2
UES
LES
Central 1
South
FIGURE 18.1 Maryland regions used in the propensity score analysis to limit grid matching
to within same colored area.
In practice, we estimate the propensity score with the dichotomous dependent vari-
able of being in a PFA equal to 1 and not in PFA as zero using a logit model with the vari-
ables outlined earlier and including county fixed effects. Subsequently, we implement
nearest-neighbor matching based on the predicted value from the propensity score
regression. The average treatment on the treated (ATT) is then calculated as the differ-
ence in means between the housing starts for the treatment grids, and the control grids.
All standard errors are constructed via 1,000 replication bootstrap estimates of the ATT.
3. The Policy-Priority Funding Areas
The Smart Growth Areas Act passed in 1997 by Maryland required all counties to des-
ignate PFAs. The Maryland Department of Planning then reviewed and approved the
areas. Unless an exception or exemption is granted, the state spends growth-related
funds for new infrastructure and some revitalization and economic development pro-
grams only within the PFAs. PFAs automatically include certain areas of the state—
Baltimore City, other incorporated municipalities, areas within the Baltimore and
Washington Beltways, and designated neighborhoods, enterprise zones, and heritage
Allegany Washington Cecill

Carroll Harford
Garrett Baltimore
Frederick
Baltimore
Kent
Howard City
Montgomery Anne
Arundel Queen
Anne’s
Prince Talbot
George Caroline
Calvert
Priority Funding Area Charles
Dorchester
St. Wicomico
Mary’s
Worcester
Somerset
N
W E
0 25 50 Miles
S
FIGURE 18.2 Priority funding areas within Maryland.
areas.5 Local governments can designate additional areas as PFAs if they meet certain
criteria, based on existing and planned densities and infrastructure (Maryland Code
Annotated: State Finance & Procurement Article, §§ 5-7B-01 to -10, 2010; see appen-
dix). A map of Maryland’s PFAs is shown in Figure 18.2. Specific categories of spend-
ing for roads, housing programs, water and sewer infrastructure, state buildings, and
certain economic development incentives are defined by the statute as “growth-related”
(see appendix). Spending is constrained for certain types of projects for five agen-
cies: Transportation, Housing and Community Development, Environment, General
Services, and Business and Economic Development (Maryland Code Annotated: State
Finance & Procurement Article, §§ 5-7B-01 to -10, 2010).
As a land use regulation in its relative infancy, empirical studies of PFA impact on
housing starts have been limited. Using land use and land cover data, Shen, Liao, and
Zhang (2005) and Shen and Zhang (2007) examined the effects of PFA and PFAs Smart
Growth counterpart, RLAs, which seek to preserve resource and important ecologi-
cal lands, on land conversion in Maryland from 1992 to 1997 and from 1997 to 2002.
Using a logit model and land use land cover data, the authors found that urban develop-
ment was more likely inside PFAs and less likely in RLAs, although the effects varied by
county. These authors did not consider the endogeniety of the PFA or RLA designation.
Therefore, although PFA areas were most likely to see growth, they may have been so
even before the PFA designation. Because they did not consider the endogeneity, we
5 These areas were not included as observations for our analysis.

question whether the PFA grids were compared to proper counterfactual grids similar
across many attributes.
Hanlon, Howland, and McGuire (2009) examined the effects of PFAs on the probabil-
ity of land development in Frederick County from 2000 to 2004. They also concluded
that parcels inside PFAs were more likely to be developed than parcels outside. Lewis,
Knaap, and Sohn (2009) evaluate implementation of the statutes and development out-
comes before and after the designation of PFAs. Because state agency compliance with
reporting requirements was lax, it was difficult to assess where and how much state
funding was spent inside PFAs in accordance with the law. Using a t-test of means to
conduct before and after analysis at the county, regional, and state level, Lewis, Knaap,
and Sohn (2009) show that PFAs had little discernible impact on development patterns
after the Act went into effect.
Howland and Sohn (2007) find investments in water and sewer infrastructure were
more likely inside the PFAs than outside between 1997 and 2002. They found coun-
ties that received more state funding were more likely to invest in water and sewer
infrastructure projects inside the PFA. However, they also found that investments
in infrastructure continued outside PFAs, and some of this infrastructure received
state funds. Since 1997, Maryland provides larger tax credits and less stringent cri-
teria within PFAs for job creation than outside PFAs (Sohn and Knaap 2005). They
also found that more jobs were created inside PFAs after 1997. The differential in
job growth across the PFA, however, was small and occurred only in a few selected
industries.
4. The Data
One of the most important decisions in this type of analysis is the choice of covariates,
regardless of the analytical method used. Economic theory provides a starting point for
what broad classes of data one should employ, but, as to the exact measurement or what
specification to use, little guidance is provided. Imbens and Wooldridge (2009) suggest
more research is needed to help choose which covariates to include from a large set of
possible variables and what functional form should be employed. In this study, and in
general, the quantity and quality of data necessary to satisfy the untestable assumption
of CIA does not often include data from one source. We have collected and compiled
data theorized to impact PFA selection and housing starts in Maryland from numerous
sources. These included the Maryland Department of Planning, Maryland Department
of Assessment and Taxation, US Geological Service, US Department of Agriculture
Natural Resources Conservation Service (NRCS), Department of Transportation, and
the National Center for Smart Growth at the University of Maryland. Significant effort
has been made to attain and measure data for the relevant pre-PFA designation in 1997
Table 18.1 Number of grids within the sample, number of grids with perfect pre-
diction of priority funding area (PFA) status, and number of grids used to estimate
the model
Total number of grids in the
Total observations priority funding areas
Full sample 169,773 27, 594

Excluded as perfect predictors* 20,983 14,088
Estimation sample 148,790 13,506
* Municipalities, high-density residential, commercial, and industrial zones are automatically included
in the PFA and thus excluded from the choice set.
and in a format that is consistent across the counties in our study area.6 For the outcome
variable, the number of new homes constructed (i.e., housing starts), we chose to aggre-
gate the underlying parcel and all independent variables into a ¼-square-mile grid cell
on the landscape.
The estimation data correspond to many time invariant features of the grid, includ-
ing soils, slopes, distances to the predominant central business districts of Baltimore
or Washington, DC, distances to amenities (parks and water), and some time-variant
features including land use, land cover,7 density of housing, and number of landowners.
These variables are calculated from the neighboring grids utilizing queen contiguity as
the definition of neighbor. Tables 18.1 and 18.2 present the variables and summary sta-
tistics. These variables define the substitutable areas for development in the all areas of
the state inside and outside the PFA. Variables are primarily measured as a percentage
of the grid to account for higher home construction costs in these areas. These variables,
drainage, slopes, flooding, and soils, are constructed using the Soil Survey Geographic
Database (SURGO) classifications (Soil Survey Staff, Natural Resources Conservation
Service 2011).
Land cover measures include water, agriculture, and forest, each with its own attract-
ing and repelling effect on new housing starts. Distances to amenities, such parks, the
ocean, or lakes, are included, as are distances to interstates and state highways (urban
arterials). Other distance to amenity-based measures include an estimated travel time
to various-size Census designated places8 within an hour’s drive proxy for the accessibil-
ity of any given grid cell. The Maryland Department of Planning’s generalized zoning
category variable is also included, ranging from the typical commercial, municipal, and
6
All Maryland Counties except Queen Anne’s County are included. Queen Anne’s had incomplete
and thus usable data.
7 Measured as of 2002.
8 Concentration of population identified by the US Census Bureau.
Table 18.2 Summary statistics for 148,790 grids included in the analysis

Variable Mean Std. Dev. Max
prePFAresDev 0.087 0.368 5.088

averSalesPr_ByBG 6.879 5.739 13.815
numSales_ByBG 33.807 44.692 408.000
numOwners1997_bg 1.119 0.891 6.858
aw_slope 7.804 9.221 110.500
aw_runoff_high 0.061 0.141 1.000
aw_drain_vpd 0.063 0.188 1.000
aw_fldFreq_freq 0.056 0.153 1.000
aw_soils1_3 0.445 0.368 1.000
aw_fedland 0.027 0.150 0.998
park 0.054 0.140 0.998
aw02_lcAg 0.305 0.356 1.000
aw02_lcFor 0.396 0.376 1.000
aw02_lcWater 0.148 0.329 1.000
tt_min6 89.999 56.680 222.960
tt_num1_1hr 9.002 10.341 52.000
tt_num2_1hr 5.898 7.411 35.000
tt_num3_1hr 4.200 5.638 24.000
tt_num4_1hr 6.347 8.981 37.000
tt_num5_1hr 0.318 0.530 2.000
dist2Int 7.060 7.509 29.082
dist2Arterial 2.359 1.805 9.221
dist2ocean 10.332 13.225 50.330
dist2lake 6.132 4.055 18.336
s_prePFAresDev 0.145 0.357 4.080
s_numSales_ByBG 34.409 36.836 408.000
s_numOwners1997_bg 1.296 0.846 6.278
s_aw_slope 7.829 8.263 76.293
s_aw_runoff_high 0.061 0.108 0.928
s_aw_drain_vpd 0.062 0.137 1.000
s_aw_soils1_3 0.587 0.287 1.000
s_aw_fedland 0.027 0.129 0.998
s_park 0.054 0.111 0.990
industrial to a finer scale of least protected, moderately protected, and very-low-, low-,
medium-, and high-density residential.9 We also calculate the number of owners of the
grid (number of owners) as a proxy for density and for the number of landowners (deci-
sion makers) in the area from whom a developer may purchase land. All of these vari-
ables describing the grids attempt to explain where housing is most demanded and most
likely to occur either because of amenities, workplace commute, or cost of development.
9 See www.mdp.state.md.us/OurWork/zoningtext.htm for exact definitions.

5. Results
The estimated coefficients using OLS are reported in Table 18.3 and the logit propen-
sity score results in Table 18.4. The results of the balancing test using a t-test approach
are presented in Table 18.5. The ATTs for the propensity score approach are reported
in Tables 18.6 and 18.7. Both types of analysis find consistent evidence that PFAs have
impacted the location of housing starts statewide. In the regression, PFAs are found
to have 1.663 more housing starts than comparable non-PFA grids. Comparing the
131,401 non-PFA grids to the 12,451 PFA grids, we find when matching that PFA
grids have an average housing start level of 2.75 compared to the most observationally
equivalent non-PFAs’ 1.71. All matching algorithms are implemented in Stata 11 using
psmatch2 (Leuven and Sianesi 2003). The large sample sizes makes eliminating any sta-
tistical differences in means an insurmountable task; however, no statistical difference
appears economically important. For example, when looking at the average sales price
per square foot of house, we find PFA grids have a mean of 10.733 and those matched
grids outside the PFA a mean of 10.359. These means are statistically different, but their
absolute difference is small enough that we do not believe this will bias the average treat-
ment effect on the treated. Similarly, the number of owners was 2.774 in 1997 within the
PFA grids but only 2.53 within the non-PFA grids.
It is interesting to note the impact that the matching procedure has on the sample
counterfactual group mean housing starts. The limited evidence from other studies
compares the cumulative housing starts either in or out of the PFA and, as is obvious
from the increase in matched controls from 0.28 to 1.71 housing starts, the algorithm
does well to select more appropriate matches. The average treatment effect on the PFA
designation is 1.04 more new homes within the PFAs grids than would have occurred
otherwise. Similarly, when we use a difference-in-difference approach (pre-PFA housing
starts to post-PFA housing starts), we find comparable results, with 1.05 new housing
starts following the PFA designation. This suggests that the OLS method overestimates
the impact of the PFA designation by treating it as if it is exogenous. One might view
this as evidence of limited impact of unobservables on the estimates and as a robustness
check of the main results.
This statewide matching approach allows matches across counties so, as an addi-
tional robustness check, we also restricted matches to grids within the same counties
and found similar although slightly larger impacts. In this case, we lose both PFA and
non-PFA observations due to the unavailability of “good” matches. Some PFA grids
have no comparable matches (off the common support) and some non-PFA grids are
very far in propensity score from any PFA grids (beyond the designated bandwidth). For
the estimation, we had 130,330 non-PFA grids and 10,874 PFA grids. The average treat
ment effect was estimated to be 1.21 new housing starts within PFA grids and 1.39 for
the difference measure. This restriction captures county-specific unobservables that were
Table 18.3 Estimated coefficients using ordinary least squares regression

dependent variable: Number of housing starts
Variable Coefficients Variable Coefficients
PFA 1.663*** tt_num3_1hr 0.0363**

(0.0439) (0.0143)
prePFAresDev 0.85*** tt_num4_1hr 0.499***
(0.0314) (0.0064)
averSalesPr_ByBG 0.0333 tt_num5_1hr 0.0363**
(0.0027) (0.0425)
numSales_ByBG 0.00049 dist2Int −0.0228***
(0.0005) (0.0064)
numOwners1997_bg −1.196*** dist2Arterial −0.011
(0.0217) (0.0081)
aw_slope 0.00748*** dist2ocean −0.0044
(0.0027) (0.0051)
aw_runoff_high 0.0229 dist2lake 0.0203***
(0.1120) (0.0056)
aw_drain_vpd 0.159* s_prePFAresDev 0.00551***
(0.0874) (0.0374)
aw_fldFreq_freq −0.321*** s_numSales_ByBG −0.0878***
(0.0775) (0.0006)
aw_soils1_3 0.515*** s_numOwners1997_bg 1.369***
(0.0500) (0.0233)
aw_fedland −1.023*** s_aw_slope −0.0072**
(0.1440) (0.0035)
park −0.361*** s_aw_runoff_high −0.313*
(0.0667) (0.1600)
aw02_lcAg −3.096*** s_aw_drain_vpd 0.231*
(0.0597) (0.1340)
aw02_lcFor −2.815*** s_aw_soils1_3 −0.253***
(0.0589) (0.0682)
aw02_lcWater −2.514*** s_aw_fedland 0.17
(0.0727) (0.1700)
tt_min6 0.00077* s_park 0.272***
(0.0004) (0.0864)
tt_num1_1hr −0.0065 County Fixed Effects Yes
(0.0078)
tt_num2_1hr −0.01 Constant 3.343***
(0.0113) (−0.218)
Observations 148,790 R2 0.1000
Standard errors in parentheses

*** p < 0.01, ** p < 0.05, * p < 0.1
Table 18.4 Estimated coefficients for logit propensity score model dependent

variable: Probability of a grid being within a priority funding area (PFA)
Variable Coefficients Variable Coefficients
prePFAresDev −0.043 tt_num4_1hr 0.0368***

(0.0314) (0.0086)
averSalesPr_ByBG −0.0304*** tt_num5_1hr 0.16***
(0.0043) (0.0539)
numSales_ByBG 0.00686*** dist2Int −0.0924***
(0.0009) (0.0126)
numOwners1997_bg 0.146*** dist2Arterial −0.429***
(0.0207) (0.0153)
aw_slope −0.00202 dist2ocean −0.0328***
(0.0047) (0.0088)
aw_runoff_high 0.291** dist2lake 0.0256***
(0.1470) (0.0089)
aw_drain_vpd −0.692*** s_prePFAresDev 0.203***
(0.1770) (0.0323)
aw_fldFreq_freq 0.966*** s_numSales_ByBG 0.000521
(0.1220) (0.0006)
aw_soils1_3 −0.235*** s_numOwners1997_bg 1.457***
(0.0747) (0.0239)
aw_fedland −0.107 s_aw_slope −0.0658***
(0.1960) (0.0064)
park −0.309*** s_aw_runoff_high 1.336***
(0.0980) (0.2180)
aw02_lcAg −1.494*** s_aw_drain_vpd −1.097***
(0.0661) (0.2840)
aw02_lcFor −1.132*** s_aw_soils1_3 0.0575
(0.0613) (0.1210)
aw02_lcWater −2.735*** s_aw_fedland 1.98***
(0.1140) (0.2310)
tt_min6 0.0113*** s_park 1.433***
(0.0017) (0.1320)
tt_num1_1hr 0.153*** County Fixed Effect Yes
(0.0103)
tt_num2_1hr −0.0337** Constant −2.754***
(0.0134) (−0.441)
tt_num3_1hr −0.196*** Observations 148,790
(0.0176)
Standard errors in parentheses

*** p < 0.01, ** p < 0.05,* p < 0.1
Table 18.5 Balance test to ensure the priority funding area (PFA) grids are
similar to the control grids using t-tests
Mean
Variable Sample PFA Grids Non-PFA t-test p > |t|
prePFAresDev Unmatched 2.2871 0.15977 70.08 0

Matched 2.0271 2.434 −3.67 0
averSalesP~G Unmatched 10.767 6.9688 69.35 0
Matched 10.733 10.359 7.19 0
numSales_B~G Unmatched 80.954 32.131 108.9 0
Matched 80.497 83.477 −3 0.003
numOwners1~g Unmatched 2.8553 1.0062 232.49 0
Matched 2.7739 2.5284 12.27 0
aw_slope Unmatched 7.0611 7.0165 0.54 0.587
Matched 7.1325 7.2208 −1.2 0.229
aw_runoff_~h Unmatched 0.08145 0.05466 19.98 0
Matched 0.07736 0.08681 −4.66 0
aw_drain_vpd Unmatched 0.01137 0.06738 −31.83 0
Matched 0.01195 0.01229 −0.34 0.73
aw_fldFreq~q Unmatched 0.04637 0.0478 −1.13 0.26
Matched 0.04663 0.05492 −4.97 0
aw_soils1_3 Unmatched 0.58709 0.45339 37.22 0
Matched 0.5939 0.55033 9.98 0
aw_fedland Unmatched 0.0307 0.03112 −0.26 0.795
Matched 0.03165 0.07783 −16.57 0
Park Unmatched 0.08451 0.10951 −9.14 0
Matched 0.08701 0.09458 −2.36 0.018
aw02_lcAg Unmatched 0.10809 0.32988 −64.52 0
Matched 0.11334 0.12494 −3.71 0
aw02_lcFor Unmatched 0.26584 0.41699 −41.51 0
Matched 0.27512 0.2864 −2.71 0.007
aw02_lcWater Unmatched 0.01539 0.1263 −39.45 0
Matched 0.01608 0.02017 −3.5 0
tt_min6 Unmatched 48.806 92.612 −77.56 0
Matched 49.915 52.958 −4.95 0
tt_num1_1hr Unmatched 22.669 9.2261 129.77 0
Matched 22.255 22.044 1.27 0.204
Matched 15.433 15.432 0.01 0.995
Matched 11.309 11.313 −0.05 0.962
Matched 17.616 17.48 0.89 0.372
tt_num5_1hr Unmatched 1.0101 0.31347 132 0
Matched 0.98933 0.95179 3.8 0
dist2Int Unmatched 3.3374 7.7556 −58.47 0
Matched 3.4544 3.8915 −5.07 0
dist2Arter~l Unmatched 0.60027 2.3651 −105.93 0
Matched 0.61819 0.76512 −10.16 0
dist2ocean Unmatched 7.0073 9.6718 −21.46 0
Matched 7.0485 6.4794 6.37 0
dist2lake Unmatched 4.4154 5.7588 −35.38 0
Matched 4.4552 4.6129 −3.7 0
Table 18.6 Propensity score results for statewide comparisons, prior to 1997 and post 1997 and for within-county comparisons
Observations on support Mean housing starts by group Statewide
Controls PFA Control PFA ATT impact^
Statewide model
Off Support 3,884 1,054 Unmatched 2.65 0.28
On Support 131,401 12,451 Matched 2.75 1.71 1.04** 8,546 units
Statewide model (difference in pre- and post-PFA housing starts as outcome, a difference-in-difference PSM)
Statewide model (matches are forced to be within same county)
*** p < 0.01, ** p < 0.05, * p < 0.1

^ adjusted for cells used multiple times as counterfactual by using only the best match
PFA, Priority Funding Area; PSM, propensity score matching
Table 18.7 Propensity score results by region of the state
Observations on support Mean housing starts by group
Statewide
Controls PFA Control PFA ATT impact^
West
Off Support − 48 Unmatched 0.35 0.18
On Support 6,659 420 Matched 0.33 0.38 −0.05 (21) Units
Central 1
On Support 15,239 4,653 Matched 3.36 1.74 1.62** 4,975 Units
Central 2
Off Support 2,882 150 Unmatched 4.8 0.32
South
Off Support 792 121 Unmatched 2.41 0.54
Upper Eastern Shore
On Support 16,495 569 Matched 1.47 0.83 0.64** 121 Units
Lower Eastern Shore
On Support 24,650 1,252 Matched 1.43 1.36 0.07 29 Units
9,624 Units
*** p < 0.01, ** p < 0.05, * p < 0.1

^ adjusted for cells used multiple times as counterfactual ATT, average treatment on the treated; PFA, Priority Funding Area
not captured estimating the propensity score. However, it disregards the fact that devel-
opers face no restrictions to operate within county boundaries.
5.1 Regional Results

Although the state-level results were all statistically significant, ample evidence exists to
suggest that growth pressures are not constant across the state. One way to investigate
heterogeneity in the distribution of the outcome of interest is to estimate at a regional
(multicounty) level of aggregation. In fact, we find significant heterogeneity across the
regions of the state, with insignificant policy impacts for two (or three in the differenced
outcome) of the geographic regions we examined. In both Western Maryland and the
Lower Eastern Shore, we could not reject the null hypothesis that the PFA had no impact
on housing starts. Both of these areas have a lower number of housing starts than the
other regions and a lower demand for homes in general. In these areas, it is possible
that the PFA is not redirecting growth because growth is not sizable enough to impact
through state incentives. Perhaps, because both regions are more rural, people who
move to these regions may prefer low-density housing, and, if they were willing to move
to a denser area, would have chosen other counties or geographic regions.
However, in the Central Regions, the South, and the Upper Eastern Shore, PFAs were
found to redirect housing. In the Central region I, the average treatment effect on the
PFA grids was 1.62 housing starts. This region contains those counties under significant
development pressure and also describes counties with many decades of strong land use
planning regimes. In the Central II region, the average treatment effect was 2.73 housing
starts. This region, in contrast to the Central I region, is largely rural with significantly
less stringent land use regimes and more reliance on state funds. In terms of number of
housing starts, the Central regions experience the largest shift of housing starts com-
pared to other regions of the state. It is interesting to note a couple of points concern-
ing the magnitudes of the differenced outcome versus the common cross-sectional
post-PFA outcome measure. First, the magnitude of the average housing starts from
the matched samples suggests there are ample grids developing in both time periods.
Second, in some cases, the large differences between the cross-sectional and the differ-
enced outcome illustrate the potential influence of unobservables missed in a standard
PSM implementation premised on the selection of observables assumption.
In Southern Maryland, we also find that the PFA designation has influenced the loca-
tion of housing starts, although the magnitude is smaller. The average treatment effect
on the PFA grids was 0.96 housing starts. Similar to the Central II region, the South has
experienced significant growth pressure in the last decade and is perhaps more respon-
sive to state funding as well. The Eastern Shore had fewer housing starts overall but still
an average treatment effect of 0.64, suggesting a modest shift to housing starts inside
the PFA.
Given the current political and economic climate, it is highly unlikely that a regulatory
approach to land development policy will continue to have broad enough support for
implementation. Additionally, it is becoming clear that externalities associated with the

timing and location of residential development is not fully captured at county boundaries.10
Thus, inventive and incentive-based land use policies that cross jurisdictional boundaries,
such as Maryland’s statewide PFA policy, are likely instruments for the future of develop-
ment controls. The question is whether these types of policies can effectively steer activity.
We find evidence that the PFA policy has effectively shifted development away from
areas similar in characteristics to PFA areas but that have not been so designated. We
should be clear that this does not suggest that the PFA has encouraged infill development
or discouraged exurban expansion. Any analysis of the state housing starts will show
exurban expansion continues to occur throughout the study period. This cannot be over-
stated; we have identified the impact of the PFA where the counterfactual is land similar to
PFA-designated land that just happens to have been excluded from designation. In short,
PSM as a policy analysis tool in this context can identify whether housing starts occurred
in PFA-designated areas as opposed to comparable non-PFA areas. Econometrically, this
is the appropriate comparison group, but it might not produce the measured treatment
effect that most interests policy makers. Here, we find that, when faced with a development
decision between identical lands, developers will focus on PFA lands. Whether this due to
the availability of state funds or perhaps simply a signal that the state and the county have
agreed this is a preferred location for housing is not clear and perhaps not relevant. As an
evaluation tool, PSM is not a panacea, and one should be precise in what is identifiable
compared to what is most desirable by the land use community. We believe this overlap is
significant here.
For example, there may be lands near to existing urban or suburban communi-
ties that are not desirable for development from the societal view. Perhaps such lands
have unique habitat or other ecosystem services that are currently without a market.
These results suggest the PFA policy can effectively steer development away from those
areas to more socially acceptable locations. This type of policy has the potential to steer
near-term development at the fringe, and, used as such, it can potentially protect areas
from development. Overall, our results suggest that between 8,500 and 9,500 homes
have located inside the PFA that otherwise would have located outside the PFA during
years immediately following implementation.11
6. Conclusion
In the context of land use policy, the use of the quasi-experimental method of PSM is
still in its infancy; however, the method has gained broader appeal across other fields
10 See “Bay on the Brink” report from the Maryland Journalism School for a classic example.
11 Calculated using only the “best” match when a control is used multiple times.
of economics. The method is well suited for land use applications precisely because the
vast majority of land use data is observational and there are rarely circumstances that
produce clean instrumental variables. Land use evaluation and other applications do
not fit well within basic regression-based modeling either. Those conducting land policy
evaluations are in need of methods like PSM to address issues of selection and potential
endogeneity.
It is critical that these policies are evaluated appropriately. Unlike many other policy
evaluation environments, the land researcher is often dealing with policies that result
in permanent adjustments to the landscape, such as the location of housing, commer-
cial, or industrial activity. PSM’s ability to use observational data while limiting the
impacts of endogeneity and of functional form assumptions is a tremendous asset in
the researchers’ toolbox. On the other hand, one should note that the generalizability of
results is difficult from PSM studies, as it is in many reduced form analyses. This chap-
ter’s results explore the effectiveness of an existing policy and inform practitioners of
areas to focus on—or stay clear of—in the future. However, these results do not suggest
an optimal policy or build on a literature moving toward discovery of an optimal policy
if one exists.
Given the brief history of these methods in the land use literature, we believe many
arenas exist in which these methods can provide insights. Furthermore, the shift from
regulatory to incentive-based policies and the fiscal issues facing many state and local
entities make any information regarding existing policy impacts necessary and relevant.
Appendix
A1. Statutory Context
PFAs are perhaps the centerpiece and the most innovative of the Maryland Smart
Growth tools. Unlike urban growth boundaries in Oregon, which impose direct restric-
tions on urban development, the 1997 Smart Growth Areas Act merely restricts state
spending on statutorily defined “growth-related” programs to areas designated for
urban growth. According to the Maryland Department of Planning (MDP) website:
The 1997 Priority Funding Areas Act capitalizes on the influence of State expendi-
tures on economic growth and development. This legislation directs State spending
to Priority Funding Areas. Priority Funding Areas are existing communities and
places where local governments want State investment to support future growth.
(Maryland Department of Planning, 2009a,
http://planning.maryland.gov/ourproducts/pfamap.shtml)
A2. Geographic Scope
By statute, PFAs automatically include certain areas of the state: Baltimore City,
incorporated municipalities, areas within the Baltimore and Washington beltways,
and areas designated by the Department of Housing and Community Development
for revitalization, enterprise zones, and heritage areas. In addition to areas des-
ignated as PFAs by statute, local governments can designate additional areas as
PFAs if they meet certain criteria. (Maryland Code Annotated: State Finance &
Procurement Article, §§ 5-7B-01 to -10, 2010; Lewis, R., Knaap, G.-J., and Sohn,
J. [2009].)
Counties may designate additional areas as PFAs based on land use, developed den-
sity, zoned density, and water and sewer service criteria. Specifically, counties may
include (a) areas inside locally designated growth areas zoned for industrial use by
January 1, 1997, or served by public sewer; (b) employment areas inside locally desig-
nated growth areas served by or planned for water and sewer; (c) a community existing
prior to 1997 that is located within a locally designated growth area, served by a public/
community sewer or water system, and has an allowed, average residential density of
≥2.0 units per net acre; (d) an area outside the developed portion of an existing com-
munity, if the area has an allowed, average build-out density of ≥3.5 units per net acre;
(e) areas beyond the periphery of the developed portion of existing development that
are scheduled for public water and sewer service and have a permitted residential den-
sity of ≥3.5 units per net acre, and (f) rural villages included in the comprehensive plan
before July 1, 1998 (Lewis et al. 2009).
Counties may designate “areas other than existing communities” as PFAs based on
analyses of supply and demand. That is, counties must analyze land capacity and demand
for the present and future, and PFAs must match the amount of land needed for a clearly
defined planning horizon (Maryland Department of Planning, 1997). Although the
statutes did not specify a particular planning horizon, MDP used a 20-year horizon as a
standard benchmark.
Criteria for delineating PFAs are based on both actual and permitted densities. The
density criteria established in the 1997 bill were the subject of much debate and have
been the subject of criticism (Cohen 2002; Knaap and Frece 2007). The original ver-
sion of the bill established a permitted density threshold at 5.0 units per net acre, but
this was amended to a permitted density of 3.5 units per net acre with urging from the
Maryland Association of Counties. The Smart Growth advocacy organization 1,000
Friends of Maryland argued that the threshold was too low, given that actual densi-
ties are often lower than permitted densities (Cohen 2002; Knaap and Frece 2007).
Although the legislation contains language stating that land can be designated for
inclusion in PFAs if “the design represents a long-term development policy for pro-
moting the orderly expansion of urban growth and an efficient use of land and pub-
lic services” (Maryland Code Annotated: State Finance & Procurement Article, §§
5-7B-01 to -10, 2010), the primary criteria for designating PFAs is based on existing
or zoned densities and infrastructure capacity, rather than “orderly” plans for future
urban growth.
A3. “Growth-Related” Expenditures
As mentioned, PFAs are intended to affect growth patterns by concentrating state

spending on “growth-related” projects in PFAs. This “growth-related” spending consists
of specific programs by Maryland Department of Environment (MDE), Department
of Housing and Community Development (DHCD), Department of Business and
Economic Development (DBED), and Maryland Department of Transportation
(MDOT). By statute, a “growth-related” expenditure is “any form of assurance, guaran-
tee, grant payment, credit, tax credit, or other assistance, including a loan, loan guaran-
tee, or reduction in the principal obligation of, or rate of interest payable on, a loan or
a portion of a loan” (Maryland Code Annotated: State Finance & Procurement Article,
§§ 5-7B-01 to -10, 2010).
References
Angrist, J. D., and A. B. Krueger. 2001. Instrumental variables and the search for identifica-
tion: From supply and demand to natural experiments. Journal of Economic Perspectives
15(4): 69–85.
Bento A., C. Towe, and J. Geoghegan 2007. The effects of moratoria on residential develop-
ment: Evidence from a matching approach. American Journal of Agricultural Economics
89: 1211–1218.
Cohen, J. R. 2002. Maryland’s smart growth: Using incentives to combat sprawl. In Urban
sprawl: Causes, consequences, and policy responses, ed. G. D. Squires, 293–324. Washington,
DC: Urban Institute.
DeGrove, J. M. 2005. Planning policy and politics: Smart growth and the states. Cambridge,
MA: Lincoln Institute of Land Policy.
Dehejia, R., and S. Wahba. 2002. Propensity score matching methods for non-experimental
causal studies. Review of Economics and Statistics 84(1): 151–161.
Gill, R. D., and J. M. Robins. 2001. Causal inference for complex longitudinal data: The continu-
ous case, Annals of Statistics 29(6): 1785–1811.
Hanlon, B., M. Howland, and M. McGuire. 2009. Hotspots for growth: Land use change in a
transitional county in the U.S. Unpublished manuscript, University of Maryland.
Heckman, J., H. Ichimura, and P. Todd. 1997. Matching as an econometric evaluation esti-
mator: Evidence from evaluating a job training programme. Review of Economic Studies
64(4): 605–654.
Heckman, J., H. Ichimura, and P. Todd. 1998. Matching as an econometric evaluation estimator.
Review of Economic Studies 65(2): 261–294.
Holland, P. W. 1986. Statistics and causal inference. Journal of the American Statistical
Association 81: 945–970.
Howland, M. and J. Sohn. 2007. Will Maryland’s Priority Funding Areas Initiative contain
urban sprawl? Land Use Policy 24(1): 175–186.
Imbens, G. 2000. The role of the propensity score in estimating dose-response functions.
Biometrika 87:706–710.
Imbens, G. W., and D. B. Rubin. Forthcoming. Causal inference in statistics and the social sci-
ences. Cambridge and New York: Cambridge University Press.
Imbens, G. W., and J. M. Wooldridge. 2009. Recent developments in the econometrics of pro-
gram evaluation. Journal of Economic Literature 47(1): 5–86.
Ingram, G. K., A. Carbonell, Y. -H. Hong, and A. Flint, A., eds. 2009. Smart growth policies: An
evaluation of programs and outcomes. Cambridge, MA: Lincoln Institute for Land Policy.
Kaza, N., C. Towe, and X. Ye. 2011. A hybrid land conversion model incorporating multiple end
uses. Agricultural and Resource Economics Review 40(3): 341–359.
Knaap, G. J., and J. Frece. 2007. Smart growth in Maryland: Looking forward and looking back.
Idaho Law Review 43: 2.
Lechner, M. 2001. Identification and estimation of causal effects of multiple treatments under
the conditional independence assumption. Econometric Evaluation of Labour Market
Policies. ZEW Economic Studies 13: 43–58.
Lechner, M., and R. Miquel. 2005. Identification of the effects of dynamic treatments by sequen-
tial conditional independence assumptions. University of St. Gallen economics discussion
paper No. 2005-17.
Leuven, E., and B. Sianesi. 2003. PSMATCH2: Stata module to perform full Mahalanobis and
propensity score matching, common support graphing, and covariate imbalance testing.
Software. http://ideas.repec.org/c/boc/bocode/s432001.html.
Lewis, R., G-J. Knaap, and J. Sohn, 2009. Managing growth with priority funding areas: A good
idea whose time has yet to come. Journal of the American Planning Association 75(4): 457–478.
Liu, X., and L. Lynch. 2011a. Do zoning regulations rob rural landowners’ equity? American
Liu, X., and L. Lynch. 2011b. Do agricultural land preservation programs reduce farmland loss?
Evidence from a propensity score matching estimator. Land Economics 87(2): 183–201.
Lynch, L., W. Gray, and J. Geoghegan. 2007. Are farmland preservation programs easement
restrictions capitalized into farmland prices? What can a propensity score matching analysis
tell us? Review of Agricultural Economics 29(3): 502–509.
Lynch, L., and X. Liu. 2007. Impact of designated preservation areas on rate of preservation
and rate of conversion: Preliminary evidence. American Journal of Agricultural Economics
89(5): 1205–1210.
Lynch, L., W. Gray, and J. Geoghegan, 2009. An evaluation of working land and open space
preservation programs in Maryland: Are they paying too much? In New perspectives on
agri-environmental policies: A multidisciplinary and transatlantic approach, eds. S. Goetz
and F. Brouwer, 72–92. New York: Routledge.
Maryland Code Annotated, State Finance and Procurement Article, §§ 5-7B-01 to -10. 2010.
LexisNexis.
Maryland Department of Planning. 2009. Smart Growth Priority Funding Areas Act of 1997.
Retrieved January 5, 2009. http://www.mdp.state.md.us/fundingact.htm
McMillen, D. P., and J. F. McDonald. 2002. Land values in a newly zoned city. Review of
Racine, J. S, Q. Li. 2004. Nonparametric estimation of regression functions with both categori-
cal and continuous data. Journal of Econometrics 119(1): 99–130.
Rosenbaum, P. R. 2002. Observational Studies (2nd ed.). New York: Springer-Verlag.
Rosenbaum, P., and D. Rubin. 1983. The central role of the propensity score in observational
studies for causal effects. Biometrika 70:41–55.
Rubin, D. B. 2006. Matched sampling for causal effects. Cambridge, UK: Cambridge
University Press.
Shen, Q., Liao, J., and Zhang, F. 2005. Changing urban growth patterns in a pro-smart growth
state: The case of Maryland, 1973–2000. Working paper.
Shen, Q., and F. Zhang. 2007. Land use changes in a pro-smart growth state: Maryland, USA.
Environment and Planning A 39(6): 1457–1477.
Smith, J., and P. Todd. 2005. Does matching overcome LaLonde’s critique of nonexperimental
estimators? Journal of Econometrics 125: 305–353.
Sohn, J., and G. J. Knaap. 2005. Does the job creation tax credit program in Maryland help con-
centrate employment growth? Economic Development Quarterly 19 (4), 313–326.
Soil Survey Staff, Natural Resources Conservation Service, United States Department of
Agriculture. Soil Survey Geographic (SSURGO) Database for Maryland. http://soildat-
amart.nrcs.usda.gov
Theil, H. 1953. Repeated least squares applied to complete equation systems. The Hague: Central
Planning Bureau.
Towe C., and C. Tra. 2013. Vegetable spirits and energy policy. American Journal of Agricultural
Economics 95(1): 1–16.
Wooldridge, J. M. 2005. Violating ignorability of treatment by controlling for too many factors.
Econometric Theory 21(5): 1026–1028.
Wright, Phillip G. 1928. The tariff on animal and vegetable oils. New York: Macmillan.
C HA P T E R 19
A P P LY I N G E X P E R I M E N T S T O
LAND ECONOMICS
Public Information and Auction Efficiency
in Ecosystem Service Markets
K E N T D. M E S SE R , JO SH UA M . DU K E , A N D
LOR I LY NC H
Understanding how institutions affect resource allocation efficiency persists as a lead-

ing concern in land economics. A large body of research has been seeking to explain
individual land use decisions under various policies while accounting for complications
associated with information and heterogeneity. Empirical and theoretical approaches
offer valuable insights but some questions remain difficult to answer. Economists have
increasingly turned to experimental economics techniques in both the laboratory and
the field because of the degree of control that the researcher can provide in the setting;
their similarity to the natural-scientific process, including replicability; and the ability
to use salient financial incentives that engage research participants in a manner that
engenders greater credibility to their responses. Experiments thus are an essential tool
for economists seeking to provide the most complete advice on how institutions affect
land behavior.
This chapter has several objectives. The chapter introduces the methodological
approach of experiments to land economists, who may be unfamiliar with the technique.
Although many excellent primers exist to familiarize economists with experiments, this
chapter focuses on the design issues associated with land markets and it also highlights
the areas land economists have focused their research. Attention is given to research set-
tings where experiments might best be employed either because experiments are well
positioned to provide insight on land research questions or because experiments would
likely fill gaps in knowledge about how policy interacts with land markets. Throughout
this section, existing research findings in land economics are emphasized—especially
relating to land conservation auctions because this area has attracted a great deal of
482 Kent D. Messer, Joshua M. Duke, and Lori Lynch
recent land research—along with emerging topics. The second half of the chapter pro-
vides a more intensive understanding of economics experiments in land conservation
by conveying the results of an original experiment. The experimental research examines
how efficiently a conservation auction delivers ecosystem services under varying infor-
mation structures.
1. The Experimental Economics Method

for Land Economics
Over the past three decades, economists have increasingly turned to experimental
methods—which provide replicability, laboratory controls, and the ability to create
salient incentives—to explore important economic questions. Many initial applica-
tions were related to testing economic theories. The laboratory often provided an
ideal location for investigation because researchers could “induce” research partici-
pants with values and use salient rewards to create incentive situations that mimicked
the assumptions from theory. As challenges to traditional economic theory began to
form and the field of behavioral economics rose in prominence, experiments became
a critical research tool for testing and developing various theories of human economic
behavior.
Modeling real world behavior comes with challenges, which are difficult to sim-
plify in many settings—particularly when aspects of each individual decision are
unknown to the researcher. With experiments, the researcher is exerting control over
the environment, participants are getting paid according to the decisions they make,
and deception is not permitted by the economic experimenters’ social norms. Thus,
participants are making real decisions that affect their payoffs and not hypotheti-
cal ones like in questionnaire-based research. In an uncontrolled environment, the
researcher has difficulty explaining causality and cannot change a single condition
(treatment) to determine its marginal impact on decision-making. However, within
the laboratory setting, the investigator can design the institutional context, know the
real monetary payoffs (i.e., the payment scheme), and examine the implications of
changing one attribute (treatment) at a time. This allows for test-bedding policies
at a fine scale, thereby suggesting the mix of institutions that will lead to the greatest
social welfare in addition to identifying policy options that will likely lead to lower
welfare.
Replicability offers another element adding to experimental methods popularity. Like
other scientific disciplines, the ability to replicate one’s own or another’s results proves
the robustness of a finding. Some investigators replicate the same experiment multiple
times. This allows one to test the results with different sets of participants to determine
similarities and differences for different pools and possibly for different social groups. It
also allows one to collect enough data to estimate econometric models of behavior. By
Applying Experiments to Land Economics 483
publishing the experimental instructions and protocols along with the research paper,
other researchers can replicate the experiment to verify the results or to use the design
to aid in university teaching.1 Replicability is one advantage of laboratory experiments
as compared to pure field experiments. While researchers may attempt to duplicate field
experiments as well, their relatively more unconstrained environment may introduce
changes that are unobservable between different locations and different time periods.
While these issues of different locations and time periods can also pose problems in
laboratory studies, they tend to be less of a concern as the researcher can use the controls
of a laboratory to mitigate these factors. For instance, in a laboratory experiment the
administrator can monitor all participants simultaneously and control the rules of com-
munication. Thus, the researchers can limit the setting to having no communication
amongst participants or to permit communication under certain established rules. They
could also allow for free communication and simply record the communication for use
in a subsequent analysis. In contrast, in a field experiment, the researcher has little to
no ability to control or directly record the communication among participants and may
only be able to gather information about the nature and amount of the communication
by having participants complete self-reported, post-experiment surveys.
Friedman and Sunder (1994) outline the four types of records and documentation
that experimental researchers should keep to ensure their experiments are replicable.
One is written instructions for participants and the details of the recruitment process.
Two, researchers should keep copies of the software and the hardware used (if applica-
ble) and should make them available to other so that other researchers can replicate the
experiment. Third, researchers should maintain documentation of the lab activities (a
“log”) that includes the date and times of experiments and other relevant facts as well as
copies of the raw data. Finally, researchers should keep a record and copies of statistical
programs and code used to analyze the data.
Recognizing that existing theory sometimes provides limited policy guidance—espe-
cially in complex settings—applied economists increasingly use experiments to search
for insights to important questions. Shogren (2004, 1218) described the process as being
“like a wind tunnel to test airplane design, lab experiments provide a testbed for what
is called economic design—the process of constructing institutions and mechanisms to
examine resource allocation.” Experiments as a testbed are particularly useful in settings
where implementation of a policy change would be difficult or costly; testing alternative
policies in the laboratory first can be highly cost-effective.
Economists have also used the controls available in experiments to identify specific
behaviors of interest, such as consumer responses to foods produced with different
methods or producers’ willingness to adopt new production technologies. Additionally,
1
While the principle of replicability is a fundamental element of experimental economics, we
recognize that publishing studies that simply replicate another experiment is difficult in economics. As
authors who have also served as journal editors, we believe that this trend may be detrimental to the
creation of knowledge because researchers should have incentives to confirm the results of other studies,
ensuring proper checks and balances in our discipline’s research process.
economists have found that results from experiments can be more compelling to indus-
try leaders and policy makers as they learn about the methods and the research setting
by participating directly in educational versions of these experiments.
Data collected on naturally occurring behavior often suffer from self-selection; those
most likely to benefit from a behavior are the most likely to participate. As such, one
has little information on the behavior of those individuals who did not participate.
Experimental methods allow some control of this phenomenon. Often undergraduate
students are recruited for experiments without knowing a priori about the experiment
design or questions. Many experiments extrapolate from any real context entirely and
are marketed generically as “research on economic decision making.” In such cases, for
instance, students more interested in land use are not the most likely to participate. Nor
will students with little experience in making decisions in land markets decline to par-
ticipate. How big an obstacle this recruitment is depends on the purposes of the study. If
a goal of the study is to understand behavior of landowners with significant experience
and interest in land use, then recruiting from a general population of undergraduates
is less appropriate. However, if the research also wants to study the behavioral response
for people who are not traditionally in the land market, people who have little interest or
experience in land use decision-making such as heirs who recently inherited agricultural
land, or people who are profit maximizing under several straightforward institutions,
then perhaps a more general participant pool makes sense. As a general rule, research-
ers ought to think carefully about whether their approach to the selection of subjects
matches the purposes of the study. For instance, when experimental studies look at con-
sumers’ willingness to pay for various foods, it makes sense to recruit household shop-
pers rather than undergraduates. However, if one wants to look at trends and responses to
new music or electronics, then an undergraduate population may be more appropriate.
That said, many lab experiments do not address recruitment as systematically as tra-
ditional survey research. Like intercept surveys, experimental participant pools may be
affected by a convenience bias. In addition, those who do not participate in lab experi-
ments may systematically differ from participants, which is equivalent to nonresponse
bias in survey research.
In the past, many experiments have recruited undergraduate students as participants
in sessions conducted in rooms of computers equipped with privacy shields that serve
as experimental laboratories in universities. Many researchers are now interested in
whether college students respond similarly to other social groups. Laboratory experi-
ments have been criticized as artificial and removed from the normal decision making
situation. Many researchers now conduct experiments in the “field” with different par-
ticipant groups to evaluate the generalizability of the students’ results. As discussed in
Harrison and List (2004), there also exists a spectrum of research settings between pure
lab and pure field experiments that uses experimental techniques.
While some researchers value the seemingly higher representativeness and realism of
field experiments and may shun the relatively artificial setting of undergraduate-student
participants in a university experimental laboratory, we think that there is no simple
“one-size-fits-all” rule when it comes to applying experiments to land economics. We try
to illustrate in Figure 19.1, that each experimental design involves balancing of strengths
and weaknesses upon three dimensions: control, context, and representativeness. Figure
19.1 provides an extension of the framing of Lusk and Shogren (2007, 15), who discuss
the trade-offs between control and context in experimental designs for auctions. These
authors define these terms as follows:
“Control means the researcher has control over the environment such that no
unmeasured external force drives choices. That is, confounding of cause and effect is
eliminated.” (p. 6)
“Context implies that subjects have some contextual cues about why their decision
might matter in a bigger world.” (p. 15)
We build upon this description by adding the third dimension of representativeness,

which examines the participant pool used in the experiments and assesses how similar
the behavior of people in this pool is to the behavior of the people making the actual
economic decisions related to land economics. While several studies have shown that, in
some settings, the behavior of actual land market participants is similar to the behavior
observed by undergraduate students, we do not believe that this will always be the case.
Therefore, researchers should be aware of the inherent trade-offs among the different
design elements as they seek to define causality, achieve external (face) validity, and test
theories and policies related to land economics.
An examination of Figure 19.1 shows that on one extreme are experiments with low con-
text and representativeness but high control. These experiments usually involve student
participants (low representativeness) in a university laboratory with researcher-set induced
values (high control) and context-neutral language (represented by Point 1). This control
of participants’ values may be particularly important when the experiments are designed
to look at settings or behavior that are difficult to examine with actual data or when the
values are formed with information generally hidden from an outside observer, such as
rent seeking in auctions, adverse selection conservation auctions, or nonpoint source pol-
lution behavior. These experiments also tend to be easiest and cheapest to conduct as the
researcher can readily recruit participants from a large pool of undergraduate students
(most often economics and business majors) and generally needs to offer these students rel-
atively small financial incentives (approximately the regular hourly wage for student work-
ers on campus) to get the students engaged in competitive behavior in a laboratory setting.
The framing of the instructions is also an important design consideration. Point 1
assumes that the instructions use context-neutral language. For instance, participants
might be asked to sell “units” to a “buyer” given various market rules or to make differ-
ent “production” decisions that lead to different “payoffs” and “costs” for themselves and
others in their group. However, some researchers worry that these generic terms are dif-
ficult for participants to understand and therefore have preferred more context-specific
language (represented by Point 2). For instance, participants might be asked to sell their
“parcels of land” to the “government” or into a “conservation program” given various
market rules. However, the introduction of specific land conservation terms may also
8
High 7
4
6
3
Representative
gh
Hi
xt
nte
Co
5
2
Low
Co
ntr
ol 1
Hig
h
FIGURE 19.1 Tradeoffs in experimental control, context, and representativeness.
lead to problems if certain landowners are reluctant to sell their property to a government
agency but would consider preserving their land with a local nongovernmental organiza-
tion such as a land trust. Similarly, participants might make different “pollution abate-
ment” decisions that lead to different “profits” and “taxes” for themselves and others in
their group. While context specific language provides more reality to the experiment set-
tings—which may help participants understand the situation and thereby lessen confu-
sion and inadvertent error that leads to poor data—the researcher is likely also sacrificing
some control of the induced values as participants bring other values into the research
setting. For instance, some participants may view selling their parcels to the government
or paying taxes to be highly objectionable due to their political beliefs or personal/family
experiences, while others may view these terms positively. To help assess these types of
concerns, post-experiment surveys can test for some of these potential factors and can
control for them in subsequent data analysis. However, researchers should seek to recog-
nize potential biases in the experimental choices through well-designed survey questions.
Points 3 and 4 on Figure 19.1 show how experiments that recruit actual landowners
as participants can improve the representativeness of the study. While being harder to
recruit and—costing more on a per-participant basis, landowners can be brought into
the experiment laboratory at a university (Point 3) and participate in the same experi-
mental market with context specific instructions.2 Concerns that the landowners who
2 Landowner participants could also participate in an experiment with neutrally framed instructions;
however, it likely would be most helpful to provide these landowners directly with context specific
language as this is the setting context that the researchers are likely most interested in studying their
would participate in a research study at university setting may be sufficiently different

than other regional landowners might lead the researcher to use mobile equipment
and set-up the experimental market in a setting that could attract a more representative
sample of landowners such as a State Fair or a conference traditionally attended by land-
owners (Point 4).
Of course, any group of landowners may bring to the laboratory attitudes and social
norms that the researcher is unaware of that may unexpectedly craft their behavior in
the research setting. Therefore, the level of experimental control over the values truly
being used by the participants is likely lower. An issue of greater concern is situations
where landowners may be concerned that the results of the research will affect the poli-
cies and regulatory environment that affect them in future. In these cases, the partici-
pants may act strategically by behaving in a different manner in the experiment than
they would when facing the actual decisions. For instance, the participants in an experi-
ment, even with relatively high stakes, may behave in a more cooperative manner or
voluntarily reduce agricultural production to demonstrate that government regulation
is not necessary. In this case, the experiment’s seemingly salient rewards fail to be the
motivating factor in participants’ decisions and the researcher may unknowingly lose
control of their research setting.
Points 5, 6, and 7 on Figure 19.1 represent situations where markets are setup in the lab,
but the researcher has no control of the values. The participants’ values are endogenous
(also referred to as “homegrown” values). For land economics, it is unusual for research
budgets to allow for experiments that involve actual land transactions. Therefore, other
lower-cost items can be used to study participant behavior. For instance, participants
could be given or endowed with coffee mugs, bottles of wine, or plants where their val-
ues for the items are endogenous to the participants, and then trading could occur. In
some settings, the values for these items may be observed directly through the use of
auctions; however, in other settings knowing this true value for the item is not impor-
tant. Generally, researchers trade off control for context with endogenous values. This
type of research design may be especially helpful when looking at underlying behav-
iors. However, researchers may find that some research questions, such as studying
problems with asymmetric information, may be difficult to answer when the values are
endogenous and as such unknown. Additionally, without the control of induced values,
researchers may need to collect more data (run more experiments or collected addi-
tional information on post-experiment surveys) to overcome the noisier data and to
detect behavioral changes due to treatment effects.
While relatively expensive, research designs can be created where participants make
actual land decisions in markets with rules established by the researcher (Point 8). These
settings are clearly high in context and representativeness, though the amount of control
may be limited by both the natural policy environment and by the amount of available
behavior. For simplicity, Figure 19.1 does not show all of the scenarios with regards to context specific
and neutral instructions, but focuses on the most common.
research funding. From a policy perspective, research involving real landowners mak-
ing decisions regarding their land given different settings is likely the most convincing
form of external (face) validity. Given the likely costs associated with this research, it
makes sense that this research should build upon previous findings from research con-
ducted in simpler and less expensive settings such as those described above. A potential
cost-saving (and data-increasing) technique is to use a lottery or competitive auction to
reduce the number of accepted experimental contracts that actually are paid.
Finally, the extreme of induced-value, neutrally-framed, lab experiments with stu-
dent participants at a university (Point 1) are natural experiments that arise due to
random actions or heterogeneity in the policy environment (Point 9). While natu-
ral experiments provide a high level of context and representativeness, their level of
control can be highly variable as the researcher is constrained ex post by the changes.
Researchers are also confronted with participants self-selecting to participate based on
the policies themselves.
1.1 Technology for Data Collection

Many of the initial applications of experimental techniques involved the experiment
administrator collecting responses from participants with slips of paper and then con-
cluding the session with a brief socioeconomic survey. While computers have generally
become the tool of choice for experimental economists, the basic “pencil-and-paper”
approach can still be quite successful, especially when conducting research in the field,
such as a county fair or a developing country, where the advantages of computers are
offset by the technical logistics of a more remote setting. Mobile computer devices bring
more capabilities to remote settings and have grown more popular over time; however,
there will remain groups of participants that find new technologies to be a barrier for
participation, especially older populations and those people in developing countries
who are unfamiliar with their use.
Along with the popularity of experimental methods has come specialized and
easy-to-use computer software such as “z-Tree” (Fischbacher 2007), which is an open
source program that already has been programmed in common experimental settings
while still allowing for research manipulation of the experimental framework. Other
research has used web-based programs for the use in the lab or created macros to link
Excel spreadsheets and Access databases using Visual Basic with Applications. These
computer programs allow experimenters to collect data in real time and can provide
real-time feedback to participants. The internet has web-based survey and experimen-
tal tools to enable one to recruit large and potentially diverse participant pools for
both simple and more complex experiments allowing people in multiple settings to
participate simultaneously. While appealing in many ways, web-based experiments
may limit some aspects of the control a researcher desires. For instance, the session
administrator will be unable to observe and fully control the behavioral setting of the
participants.
1.2 Experimental Design Issues

While others have already written at length about the proper design of economic experi-
ments (see Davis and Holt 1993; Friedman and Sunder 1994; Lusk and Shogren 2007),
there are some important issues in experimental design that are worth emphasizing as
these are decisions that every research project needs to make prior to running sessions.
How much does one need to pay participants to ensure that they are motivated by the
incentives offered in the experiment? This issue, frequently referred to as the saliency
of the rewards, is critical as the administrator wants to use the controls in the labora-
tory to ensure that the marginal incentives are sufficient to overcome other factors,
such as trying to select behaviors that they think will please the experiment administra-
tor or trying to outcompete other participants but ignoring the fundamental financial
incentives established by the experimental design. As a general rule, the higher the eco-
nomic rewards, the more attention to detail you can expect from your participants. If
the researcher sets the marginal incentives such that participants can gain significantly
more money from “optimal” or “near optimal” choices compared to just “good” choices,
then one could expect that participants will dedicate more cognitive effort in determin-
ing the true optimal choice. Salient rewards will be smaller for undergraduate students
than for professionals. In our experiments, we have generally compensated undergrad-
uate students an average $15 per hour while compensation for professional participants
can be more than $50 per hour.3 Taking into account each group’s opportunity costs of
time and proximity to the laboratory or field setting before setting compensation levels
is important for encouraging broad participation in the experiments.
How does one set-up the laboratory controls to ensure that one is isolating the
behavior that one is interested in? Does one also need to employ econometric controls
to the experimental data? These issues are especially important if one is involving par-
ticipants in repeated decisions because this can provide additional statistical power,
but it also require proper statistical accounting for the treatments generating these
observations.
How important is context to the research setting? Should the instructions be written
with context-specific language or in a neutral tone? As mentioned above, an advantage
of context-specific language is that participants may have an easier time understanding
the experimental setting. However, terms such as “taxes,” “pollution,” “conservation,” or
even “government program” can often be value-laden for the participants. Therefore,
participants may not consider the economic incentives offered in the experiment to
be salient enough to overcome their attraction (or resistance) to behavior related to
these terms.
Many of the issues related to proper experimental design can also help address poten-
tial concerns that may arise from the Institutional Review Board (IRB) at the researcher’s
3 Of course, the participant’s choices and the general outcomes of the experiment either through
group behavior or random outcomes determines the actual payoff of the participant.
institution. Our experience has been that IRBs appreciate the norm in experimental
economics of not deceiving participants. Similarly, experiments can often be designed
where participants are not identified by name, but instead simply by identification
number. IRBs also tend to like that the choices in experiments are often confiden-
tial: single-blind (where the participants do not know the choices and payoffs of other
participants) or double-blind (where neither the other participants nor the administra-
tors in the session know the choices and payoffs).
The staffing of experiments also depends upon the research design. In settings involv-
ing double-blind confidentiality or multiple experiment sessions being conducted
simultaneously, the research team may need to consist of three to four administrators.
In market settings involving induced values, only one administrator may be necessary.
Nevertheless, an extra administrator can be helpful in preparing for the experiment, dis-
tributing experiment instructions and materials, addressing questions that arise from
participants, dealing with any computer problems, and assisting with the payment of
participants at the conclusion of the session.
Emerging applications of experiments to land economics include risk attitudes
of landowners, landowners’ propensity to develop their land, institutions to reduce
nonpoint source pollution in complicated geological and political situations, such as the
Chesapeake Bay in the United States, agglomeration in conservation project selection,
and the application of behavioral economics to encourage more socially beneficial land
uses. The following section reviews recent research related to auctions, including dis-
criminatory pricing versus uniform pricing, the effect of information sets on auction
efficiency, and markets for ecosystem services.
2. The Experimental Economics

Method: Applications to Land
Conservation Auctions
Experiments related to land economics often seek to mimic real-world conserva-

tion auctions, and the Conservation Reserve Program (CRP) is one of the largest in
the United States. The United States Department of Agriculture (USDA) runs the
CRP auction in each state at various times during the year. This voluntary program
for agricultural landowners is characterized by broad environmental objectives,
and the requirement that funds be allocated on a competitive basis is satisfied with
an auction. Under the CRP, a landowner submits an offer indicating the compen-
sation that she would accept to enroll land in the program for ten to fifteen years.
At the end of September 2011, the CRP had 417,386 farms with 31.2 million acres
enrolled and was paying out $1.7 billion in rental payments annually (USDA 2011).
According to USDAs Farm Service Agency (2006), CRP enrollment has led to the
abatement of 450 million tons of erosion per year, the restoration of 2 million acres
of wetlands and adjacent buffers, the reduction of 48 million metric tons of carbon
dioxide emissions, and the protection of 170,000 stream miles. The CRP has also
increased duck and quail populations, as well as other wildlife, by restoring habitat
and corridors.
Conservation auctions are also used throughout the world, and the results gen-
erally suggest that auctions are more cost effective than alternate institutions for
procuring conservation services. For instance, the CRP’s structure was adopted by
conservation agencies in Australia for the Bush Tender pilot trials (Stoneham et al.
2003) and the Auction for Landscape Recovery pilot program (Gole et al. 2005).
Stoneham et al. (2003) concluded that the amount of biodiversity benefit gained
through the first round of Bush Tender auctions in Victoria, Australia, would have
cost the government seven times as much if a fixed payment had been used instead of
an auction.
A reverse auction mechanism was used in Scotland under the Challenge Fund
scheme, and research demonstrated that the total cost would have been 33% to
36% greater under a fixed payment (CJC Consulting 2004, 63). Connor et al. (2008)
concluded that, under the same budget, a fixed payment plan would produce only
56% of the benefits achieved with auctions in the Catchment Care Australian con-
servation auction in 2004. In one of the few theoretical treatments of this issue,
Latacz-Lohmann and Van der Hamsvoort (1997) found that total emission reduction
gained by different formats of auctions ranges from 16% to 29% more than flat-rate
offer system.
2.1 Conservation Auctions in the Laboratory

Lab experiments on conservation auctions investigate how much excess rent partici-
pants secure under different treatments as a measure of auction efficiency. The setting
of many of these experiments is conservation auctions similar to those noted above.
Cason and Gangadharan (2004) examined information effects in discriminatory and
non-discriminatory pricing schemes, finding that more information available to sell-
ers about the environmental benefits of their project led to more strategic behavior
and greater rents. Schilizzi and Latacz-Lohmann (2007) found that both target- and
budget-constrained auctions performed better than fixed payments in a single period.
However, with repetition, the advantage of auctions quickly diminishes. Hellerstein
and Higgins (2010) used the CRP as the basis for land conservation auction experi-
ments. Their results showed that capping the maximum amount a landowner can
receive in environmental markets may have intuitive appeal as a way of reducing gov-
ernment expenditures but the caps may actually lead to increases in expenditures.
Hellerstein and Higgins (2010) argued that relaxing restrictions on maximum offers
from landowners could yield better results, especially when one considers the quality
of the land enrolled.
2.2 Discriminatory-Price Auctions versus

Uniform-Price Auctions
There are several types of auctions, common auctions include the “call markets” and
“double-sided auctions” used for stock markets with multiple buyers and sellers;
ascending-price “English” auctions where there is one seller and multiple buyers; and
descending-price “Dutch” auctions (see Davis and Holt [1993] and Kagel and Roth [1995]
for details on auctions). For “reverse” auctions that involve multiple sellers and one buyer,
common auctions include “discriminatory-price auctions” in which the buyer pays the
winning sellers the amount of their offers and “uniform-price auctions,” a type of Vickrey
(1961) auction, in which the buyer pays all the winning sellers a single-price based on
the either the highest-accepted offer or the lowest-rejected offer. Latacz-Lohmann and
Schilizzi (2005) concluded from their theoretical analysis that the optimal strategy for
landholders in a discriminatory-price auction is to inflate their offer above their real
opportunity cost in order to secure information rents. Further, the authors found that
the incentive to inflate offers is greatest among sellers whose costs are lowest—and there-
fore most likely to be successful—while high-cost sellers will tend to make offers that are
closer to their true costs—but they are unlikely to be selected by the auction mechanism.
Cason and Gangadharan (2004) used a laboratory experiment in which landown-
ers competed in sealed-offer auctions to obtain payments for reducing nonpoint
source pollution from land activities. One treatment was a uniform-price auction
where everyone was paid the amount of the lowest price rejected, and the other was
a discriminatory-price auction. Offers in the uniform-price treatment were within 2%
of owners’ cost, while most offers in the discriminatory-price auction were at least 8%
higher than owners’ cost. However, because the discriminatory-price auction did not
pay a single market-clearing price—rather it paid each successful seller a price equal to
his or her offer—overall, it was more efficient.
2.3 Information and Conservation Auctions

The amount of information provided to landholders before they make offers can sub-
stantially influence the efficiency of an auction as the amount of information can influ-
ence whether sellers can make strategic offers in the search for higher rents. Banerjee
et al. (2011) considered an iterative auction for selecting offers of projects adjacent to
each other on a circular grid, finding auction performance was negatively affected by
more information available to participants. Their paper shows that when participants
had more information, the conservation program paid more for the conservation units.
Cason et al. (2003) used an experiment on landowner behavior in a nonpoint source
pollution-reduction setting to test how information affected auction performance.
They conceptualized information as whether or not participants knew the buyer’s value
for the landowner/participant’s project. Their results show that participants tend to
inflate offers more for projects with a high value to the buyer. Consequently, less infor-
mation about buyer value may increase auction efficiency.4 An auction would be effi-
cient if sellers (landowners) offer their willingness to accept (reservation value) rather
than garnering rents. This could be thought of as “procurement efficiency.” Given the
assumption of homogeneous acres, the optimal auction performance would result in
the buyer obtaining the most acreage for the lowest expense. Clearly, auction efficiency
is associated with buyers (often government) optimizing their budgets. Cason and
Gangadharan (2004) reported similar results on information impacts to Cason et al.
(2003).
In some settings, the provision of information also can undermine price-induced
competitiveness traditionally assumed to occur in markets. Hong and Shum (2002)
found that in a procurement auction format where individual participants have private
and common value information, the average procurement cost can rise as the number
of participants increases instead of participants making more conservative offers. While
more information may improve seller certainty and increase competitive pressure—
which would reduce rents—participants may be more strategic and request more rent
(Rolfe et al. 2009).
2.4 Multiple Auction Rounds versus Single-Shot Auctions in

Experiments
Most experimental studies of conservation programs frame the auctions as single
independent rounds—where each round represents the beginning and end of
the world. In cases of permanent land protection from a one-time program, that
assumption is appropriate. However, single-shot auctions are not appropriate in all
instances.
Opinions differ on the use of multiple rounds of offers versus a single round.
Rolfe et al. (2009, 290) described three theoretical reasons that favor the single-shot
approach: “Incentives to reveal true opportunity costs, avoiding strategic behavior,
and minimizing administration and transaction costs.” The aforementioned stud-
ies by Cason et al. (2003) and Cason and Gangadharan (2004) used multiple rounds
with information treatments, finding that strategic bidding did occur and that multiple
rounds can generate increased administration and transaction costs.
A number of studies have favored multi-round auctions, especially because
multi-round auctions such as the CRP are the most common found in conservation
settings. McAfee and McMillan (1996) argued that interdependencies among offers
4
This result has important implications for ecosystem service markets as the buyer preferences,
such as habitat for endangered species, are often well publicized. Additionally, priorities, such as only
acquiring land that is adjacent to already protected land, may lead to higher offers from sellers to
conservation programs.
or combinatorial benefits are introduced when there are multiple rounds. Information
can be gained by sellers through multiple rounds about the suitability of their propos-
als (Rolfe et al. 2009). Klemperer (2002) reported that allowing sellers to learn about
others’ valuations through multiple rounds could make the sellers more comfortable
with their own assessments and less cautious in making offers. Other arguments for
repeat-auction designs are that participants need more than one round to understand
the auction mechanism and how to offer true valuations, as well as to learn from mar-
ket feedback (List and Shogren 1999). In contrast, Bernard (2005) argued that, in place
of repeated trials, experiments employ single-shot auctions accompanied by in-depth
instructions and practice.
Multi-round auctions may be associated with efficiency, particularly in initial rounds.
Lusk et al. (2004) suggested that in a closed, multiple-bidding, second-price auction set-
ting, offers for different quality goods will increase, and this will happen particularly
between the first and second rounds. Rolfe et al. (2009, 292) suggested that the ques-
tion of single-shot versus repeated rounds may need to be answered “on a case-by-case
basis.” Rolfe et al. (2009, 300) argues that multiple rounds tend to deliver auction effi-
ciency when landowners are “unfamiliar with the provision of conservation actions”
and “uncertain about the opportunity costs of providing actions.” In recent experi-
mental research, Fooks et al. (2012) show that participant behavior differs signifi-
cantly if situations with identical incentives are structured to be either a single-shot or
multiple-round setting.
3. Application: The Impact of Information

on Auctions for Ecosystem Services
This section describes original research, using an experiment on conservation auctions.

Governments increasingly use reverse auctions to procure ecosystem services gener-
ated from land use at the lowest cost possible. A popular mechanism is a discriminatory
reverse auction where the government purchases services from willing sellers based on
a process that selects the least expensive offers first. The price paid to each of the selected
landowners is equal to the amount of the offers they submitted. In the simplest form,
the government achieves procurement efficiency—and budgetary cost-effectiveness—
from a discriminatory reverse auction because it benefits from the competition of
the market as it selects and pays for the number of lowest-cost offers that exhaust the
budget. Procurement efficiency occurs (when land is of homogeneous quality) when
the government enrolls land from those landowners with the lowest opportunity cost;
i.e., it obtains the ecosystem services desired with the least transfers and therefore the
least potential distortion. In a similar sense, budgetary cost-effectiveness measures the
amount the government is paying for these ecosystem services. A well-functioning
auction will obtain the highest value in ecosystem services for each dollar of taxpayers’
money. Ideally, this approach encourages competition and allows the government to
“pick off ” the supply curve, driving landowners to make offers that equal their opportu-
nity costs—the minimum willingness-to-accept compensation—and thus maximizing
the conservation services per budgetary dollar expended by the government.
However, there are several reasons to be skeptical that these auctions actually work
as well as intended. A means for evaluating their performance is to measure the “rent
premium” received by the landowners: the amount of excess profit they obtain if the
government’s payment exceeds the true opportunity cost. Kirwan et al. (2005) offered
an empirical estimate of such rent premiums generated in the CRP, which uses a reverse
auction. The authors estimated that between 10% and 40% of the program expendi-
tures went to rent premiums. This estimate is consistent with the experimental results
described earlier, and thus economists continue to study auction designs in laboratory
experiments, which offer the means to test, in a controlled fashion, many design fea-
tures. Three major lines of research have developed.
One line of inquiry focuses on selection under asymmetric information with parcel
and landowner heterogeneity. Foundational theoretical work on this problem of adverse
selection includes Wu and Babcock (1996) and Smith (1995). Recent work focused on
auctions using economic experiments comes from Arnold et al. (2013). These papers
show that the existence of heterogeneity generates a systematic tendency for landowners
of lower-quality parcels to make low-priced offers, which are more likely to be selected.
Even though these successful offers cost less, they still carry substantial information rent
premiums.
A second area of economic-experimental investigation compares the performance
of discriminatory auctions with fixed-price procurement. As discussed previously, dis-
criminatory auctions have generally outperformed uniform-price auctions (Cason and
Gangadharan 2004). Yet there is concern about the efficiency of this approach in energy
markets where auctions often involving multiple rounds per day, auctions are informa-
tion rich, and sellers have some market power that comes from withholding offers for
some units from the market or for reducing supply capacity by putting down plants for
“repairs” (Rassenti et al. 2003; Vossler et al. 2009).
A third area, and the focus of our research, is the role of public and private informa-
tion in conservation auctions. Questions about how information impacts auction effi-
ciency have naturally arisen from recent research on how participants learn through
information provided (information quality) and auction repetition (experience). In
a study addressed above, Cason et al. (2003) used experiments to examine the role of
information quality and experience and found that (1) sellers’ rents increase as they gain
experience; and (2) sellers extract more rent when they know the benefit of their offer
from the buyer’s perspective (i.e., participants had more information about demand/
benefit heterogeneity). Schilizzi and Latacz-Lohmann (2007) offered experimental eco-
nomic evidence that auctions are generally more efficient than fixed-price procurement
but that the advantage dissipates over time as participants gain experience. This result
corresponds with an earlier, agent-based model by Hailu and Schilizzi (2004) that found
that seller experience led to decreased auction efficiency.
Our work extends these papers by examining the effect of experience while control-
ling for a set of policy-relevant treatments based on three levels of information. Our
conceptualization of information treatments involves market information and thus
complements the information treatments in Cason et al. (2003) involving benefit het-
erogeneity. The research question addressed here is whether market information pro-
vided to sellers affects the efficiency of a discriminatory land conservation auction and,
if so, whether the effect attenuates with experience.
3.1 Policy Setting

Government agencies that conduct conservation auctions provide various levels of
information regarding past results of auctions to the public and to future auction par-
ticipants. Our research has been influenced by an auction used yearly since the 1990s by
the Delaware Agricultural Lands Preservation Foundation (DALPF) to secure conser-
vation easements on agricultural land. The DALPF auction provides a detailed informa-
tion set that includes the following data:
1. current period budget;

2. amount of program budgets in previous years;
3. number of offers received in previous years;
4. number of offers accepted in previous years;
5. highest accepted offer in previous years;
6. lowest accepted offer in previous years; and
7. average accepted offer in previous years.
Policy makers may feel that providing such information meets general goals of fair-
ness to owners and openness and transparency in the fiscal conduct of a governmental
program. It may also promote competition among owners, which would increase auc-
tion efficiency. Previous research has suggested, however, that such detailed public
information is likely to decrease the auction’s efficiency because potential applicants
can use it to inflate their offers (Messer and Allen 2010).
In contrast to DALPF, other programs, such as the Maryland Agricultural Lands
Preservation Foundation (MALPF), provide little detailed information to the pub-
lic or to auction participants. In this case, offers to sell for the upcoming annual cycle
are made before the previous year’s results are announced. Horowitz et al. (2009), in
their analysis of 19 years of MALPF program data, found evidence of inflated offers in
this no-information setting.5 They found that on average, bids are 5 to 15% above the
5
Similarly, the Bush Tender project in Australia does not reveal information to landowner sellers
about the environmental benefits in the biodiversity preservation auction, perhaps because auctions
underlying reservation value for a landowner and show that increased competition (in
the form of lower budgets or more bidders) reduces this mark-up. They also find evi-
dence that bidders adjust for a possible “winner’s curse” by increasing their bids by 8
to 14%.6
Because actual auction processes offer various degrees of information in highly vari-
able settings that are difficult to compare, systematically generated data are needed
to understand the influence of information on auction efficiency. It could be that the
answer is program-specific—that some information interacts with a program or loca-
tional characteristics in an unobserved or unobservable manner. The most efficient auc-
tion mechanism thus would be determined on a case-by-case basis. Or, it could be that
auction efficiency depends, subtly, on exactly how much information is provided. The
results of this research suggest that information and auction efficiency may be analo-
gous to the classic children’s story of Goldilocks and the Three Bears, where the pro-
gram must identify the amount of public information that is neither “too little” nor “too
much,” but instead is “just right.” Experimental economics provides an effective plat-
form to test various types of information as treatments in a controlled setting.
3.2 Experimental Design
This research evaluates the impact of various levels of public information on (1) seller
behavior in conservation auctions; and (2) the conservation program’s overall effective-
ness. It characterizes three information regimes, extending recent experimental conser-
vation auction work by Messer et al. (2012). The experimental sessions were conducted
at the University of Delaware’s Laboratory for Experimental & Applied Economics.
Ninety participants were recruited using email addresses from students in undergradu-
ate courses in business and economics. Each experiment lasted approximately 90 min-
utes and average earnings were $25.
Each participant was randomly assigned to a group of ten participants. Each partici-
pant was assigned the role of a landholder, who owned one parcel and could sell or keep
it. Participants were seated so that private decisions were made on individual laptop
computers with privacy screens. Participants completed a consent form and then read
conducted annually in the same region could allow sellers to infer the regulator’s private information
regarding the benefits.
6
Used first to discuss auctions for mineral rights such as oil reserves purchased from the
government, a winner’s curse phenomenon is predicted, in auctions where the buyers have actual
values that are unknown to the either the other buyers or the seller (program administrator), but are
correlated (common values). Buyers also tend to make bidding decisions based on estimated values
(geological survey based) rather than market values. Theory suggests that the buyer who “wins” these
type of auctions will have bid too highly in part because they over-estimate the value of the auction
item. However, this phenomenon tends to be well-known and, as such, experienced buyers anticipate
it and adjust their bids to compensate.
the written instructions provided (see the Appendix for the actual participant instruc-
tions used in the experiment). The administrator described the experiment verbally with
the aid of presentation software to ensure consistency. No participant-to-participant
communication was allowed.
As in several experiments described previously, each round represented a
single-period game—the beginning and end of the world. The incentives in each
round were thus described as net present values, meaning that ownership returns and
offers reflected the future stream of benefits accruing from retaining or selling parcels.
Consequently, there was no incentive to wait to sell because the budget was reset and the
induced values were new in the next round. Thus, this setting did not allow for option
value of information as each round constituted an independent observation on choice
behavior in the group.7
The experiment used “induced values”8 as the opportunity cost for the land-
owner, which was conceptualized as an “ownership return” for the participant for the
100-acre parcel of land. “Experimental dollars” (hereafter denominated “$”) were
designed to match the incentives real landowners would have in actual land mar-
kets. The exchange rate between experimental dollars and real dollars was provided
to participants and used to calculate cash payments upon completion of the experi-
ment. Ownership returns were randomly selected from a uniform distribution rang-
ing from $2000 to $8000 per acre. The opportunity cost distribution was designed
to mimic agricultural land markets in the mid-Atlantic and Northeast United States,
the location of some of the largest and most active land conservation programs in the
country and a market where land values vary substantially according to development
pressure.
Participants decided whether to submit an offer to sell their parcel. If a participant
chose not to submit an offer for the parcel, then the parcel was retained and the land-
owner received payment equal to the return on ownership for the parcel. The amount
of each offer was confidential. If an offer was submitted, the participant incurred a
nominal nonrefundable transaction cost (“submission fee”) of $20. The submission
fee was designed to reflect real-world conditions associated with landowners attempt-
ing to participate in conservation markets and also to prevent the choice of submit-
ting any offer, even a very high one, from weakly dominating a strategy to submit “no
offer” under all conditions. If the offer was accepted by the conservation program, the
7
This statement does not assume that knowledge of the market was independent for each round. On
the contrary, as will be discussed later, participants learned about the market quickly and incorporated
this information into their selling decisions.
8
Induced values are the monetary values or incentives that are set by the researcher and are not
endogenously determined by the participants. Induced values have a long history of use in experimental
economics as they allow the researchers to have greater control of the research settings as the values
can be design to test behavioral hypotheses and theory (Davis and Holt 1993). This approach can
be particularly helpful in situations where in real markets the values of individuals are hidden and
participants have incentives not to truthfully reveal them.
participant received payment equal to the offer; if it was not accepted, the participant
received payment equal to the return on ownership. Accepted offers for the conserva-
tion program were determined using a discriminatory auction in which the lab admin-
istrator computer (acting as the government) selected the least expensive offers in turn
until there were no longer enough funds in the budget to buy the next least-expensive
parcel.
The conservation budget ranged from $2 million to $6 million with an average of
$4 million. A random process using a uniform distribution determined the budget
before each round.9 While the dollar amount of each budget was selected randomly,
the specific budget for each round was held constant in each experimental session. This
enabled full control of the experience effect related to budgets and facilitated data analy-
sis. Of course, the budget for each round was not known a priori to the participants in
any of the treatments.
3.3 Information Treatments

Invariant information in the experiment includes the private induced value of a parcel.
Prior to making a decision, each participant knew his or her opportunity cost of partici-
pating in the auction, i.e. the private value, mirroring the presumed private knowledge
of an actual landholder. All participants knew the distribution from which private val-
ues (theirs and others) were drawn. Although the assumption of knowing the distri-
bution is stronger than only knowing one’s own private value, this setting captured the
notion that some landowners would anticipate how readily other landowners in their
region might choose to participate in conservation auctions. The experiment also was
constructed so that all participants knew the distribution from which the program bud-
get was drawn. This assumption captured the idea that, though program budgets may
vary over time, landowners may have expectations about the range and relative likeli-
hood of high, low, or average funding levels in any given period.
Other information in the experiment varied across treatments and is presented in
Table 19.1. The information sets included:
(1) a “full” set of detailed public information that closely mimicked the seven types
of information provided by the DALPF program;
(2) a “partial” set of public information that consisted of the previous and current pro-
gram budget; and
9
The choice to have the budgets randomly determined is consistent with the situation observed with
the DALPF program and with other government programs, such as the USDA Forest Legacy Program,
which tend to have a large variation in yearly budgets (Messer and Allen, 2010; Fooks and Messer, 2012).
Of course, some conservation programs establish their budgets over longer time horizons and thus have
far less year-to-year variability. Because the setting of this research builds upon the assumption that each
round was the beginning and end of the world, using a random budget seems most appropriate.
Table 19.1 Definition of full-, partial-, and no-information treatments

Information Full Partial No
Participants’ private information

Own WTA (“ownership return” for each parcel) x x x
Previous history of sale for each own parcels x x x
Public information
Distribution of landowners ownership returns x x x
Distribution of program budgets x x x
Budgets from previous rounds x x x
Budgets for current period x x
Number of offers received from previous rounds x
Number of offers accepted from previous rounds x
Highest accepted offer from previous rounds x
Lowest accepted offer from previous rounds x
Average accepted offer from previous rounds x
Which owners sold parcels from previous rounds x
Note: Owners were identified only by number, not by name
(3) a treatment that provided “no” public information other than the budget for the
previous round.
The three information treatments were implemented in sessions of 45 rounds. Each

participant, and thus each group, participated in only one treatment so there was no
need for treatment ordering. In any given treatment, all participants were supplied
with the same information irrespective of their individual choices. Participants had
private information about their past choices, and the participants’ screens displayed
historic information on their opportunity costs, and overall market information
(if any).
3.4 Extent of the Market

The experiment consists of a policy intervention, where preservation necessarily affects the
supply of land in preserved and unpreserved states. All else equal, greater levels of preser-
vation raise the value of unpreserved land, thereby raising the opportunity cost of future
preservation. One anticipates that these price effects will be more significant factors affect-
ing behavior when budgets are larger and over time, as land increasingly migrates to a state
of protection. These arguments may form the basis for future experimental work, but they
do not affect the results of the experiment discussed in this chapter. First, this experiment
was not dynamic—each decision period represented the stream of measureable benefits
and costs accruing to the landowner and conservation buyer from the current moment
through the foreseeable future. In other words, each round represented the “beginning”
and “end” of the world from the perspective of both buyer and seller. Second, the partici-
pants were not informed whether intervention targeted a large portion of the landscape or
was so small as to constitute a marginal change in land use allocations. In other words, this
experiment could be viewed as a partial equilibrium analysis.
3.5 Experimental Data Analysis, Hypotheses, and Results

Panel data regression models can be used to assess the results of the experiment. In all treat-
ments, the unit of analysis was the round and the dependent variable was the amount of
rent (excess profits) captured by the sellers as the aggregate participant rent for the group.
The rent premium was defined as the price received by a successful seller minus the induced
value for the sold unit. A large rent meant that participants were able to secure returns that
exceeded their opportunity costs; smaller rent corresponded to auction efficiency. Panels
reflected the session in which the dependent variable was the aggregate group rent in each
round for each of the nine groups.10
The independent variables captured the key design characteristics of the experi-
ment. The primary design measures were Full Info and No Info, which reflected the
full-information and no-information treatments, while the partial-information treat-
ment was the reserve category in the regression. The variable of Budget controlled for
the impact of different budget levels. Larger budgets should lead to greater rent premi-
ums because there is less competition when the number of sellers and their opportunity
costs are fixed and the budget is bigger. The variable of Round controlled for the round
number. A positive parameter was expected because participants gained experience
in each auction treatment and were therefore likely to make fewer mistakes by over-
pricing offers and missing out on potentially profitable transactions in later rounds. In
addition, the participants could extract higher rent premiums, especially from units
with low ownership returns. Several interaction variables test for joint effects with
Round. The information-treatment interactions with Round test whether the learning
effect was magnified or attenuated in the full information and no information settings.
Also, a budget interaction with Round controlled for any synergistic impact when the
Round value and the budget were both large (providing especially large rent premium
opportunities) and when they were both small (offering relatively low rent premium
opportunities). The coefficients on the interaction variables shift the slope on learning.
Table 19.2 presents a panel data analysis of the experiment results. The total number
of observations was 396, which reflects 44 observations (rounds) over nine sessions. The
model estimation is statistically significant. The statistical results show that larger bud-
gets lead to greater participant rents. On average, one extra dollar in budget delivered
10 The first round was discarded for analysis because the participants at that point did not have any
prior round result from which to gather information.

Table 19.2 Panel data analysis of group

price premium (excess profit)
Variable Coefficient (s.e.)
Full Info 306,895.6**

(94,291.2)
No Info 306,915.6**
(94,291.2)
Budget 0.3874**
(0.0467)
Round 17,206.51**
(6264.08)
Full Info * Round -5616.68*
(2799.00)
No Info * Round -3582.35
(2799.00)
Budget * Round -0.0040*
(0.0016)
Constant -715,789.0**
(190,676.9)
Observations 396
Wald chi2 345.0
Prob > chi2 0.000
Note: ** indicates statistical significance at the

1% level; * indicates statistical significance at the 5%
level.
$0.39 in rent premium. This is a high efficiency cost and one that, by itself, calls into
question whether auctions can efficiently deliver conservation services. The magnitude
is similar to estimates from Kirwan et al. (2005) that between 10% and 40% of the CRP’s
budget were paid to landowners as information rent premiums. A second result is that
the impact of budget size on rents attenuated slightly as participants gained more expe-
rience; that is, the Budget*Round interaction was negative. This means that, over time,
greater competition drives down rents (although the effect is small).
Round had a positive impact on rents, showing that groups were able to obtain greater
rent over time as they gained experience with the auction environment and the behav-
ior of others in the group. Although it was critical to control this measure of learning
in the lab environment, it is less clear how important this result is in real-world auc-
tions. Specifically, the real-world analog would be a landholder who owns multiple units
and sells them over time or takes multiple rounds to get his or her farm accepted by the
program. Some landowners may reflect these characteristics but many would not. The
results do not indicate whether landowners who closely observed but did not participate
in early auctions would learn over time at the same rate as those who participated in the
earlier auctions.
This study suggests that the partial-information case leads to the smallest degree of
rent premiums, which implies that partial information provides the auctioneer with
the most efficient discriminatory conservation auction. The full-information treatment
and the no-information treatment inflate rents by an equivalent amount relative to the
reserve category of partial information. The magnitude of the group’s rent increase asso-
ciated with the full-information and the no-information treatments was approximately
$307,000; the average budget was $4 million, and the information effects are approxi-
mately 7.7% of the budget. Thus, those two treatments transferred public money to
landowners, instead of using this money for protecting additional ecosystem services.
Interactions between the Round variable and the information treatments led to differ-
ent results. In the full-information treatment, participants’ experience led to a decrease
in rents. Over time, conservation auctions with full information would likely become
somewhat more competitive, and the largest rent premiums would be gained in the ini-
tial rounds. However, this effect was relatively small (coefficient of –5616.68), meaning
the benefit to the sellers of having full information was not offset even at the conclusion
of the experiment. In the no-information treatment, there was no statistical decline (or
increase) in rents over time—i.e., the impact of experience captured by the Round coef-
ficient was equal in the no-information and the partial-information treatments. In the
absence of full information the entire benefit of experience is captured by the coefficient
on Round—it measures only experience.
4. Conclusion
Experimental economics techniques have been applied to a variety of land econom-

ics questions. These techniques can be useful for land researchers as they allow for
control of values for both buyers and sellers within the market, replicability among
experimental events, examination of difference between the more controlled envi-
ronment of the laboratory and the less artificial environment of the field. They permit
a researcher to test theories about land markets, to analyze particular policies and
institutional structures, to look at landowners’ willingness to participate in different
programs, to examine specific behaviors of interest, and as a method of educating the
public and the policy makers. Of particular interest has been the efficiency of various
auction mechanisms used for obtaining the ecosystem services generated by land
use and how different forms of information alter outcomes for sellers and buyers.
This chapter reviewed the relevant literature and reported the results of an original
experiment that explores these issues in the context of markets for ecosystem service
procurement.
Results of the experiment suggest that, for a discriminatory reverse auction, a

limited-information setting may lead to greater market efficiency than either the
no-information or the full-information setting tested in this research. In other words,
this study suggests that just announcing the anticipated program budget before offers
are submitted can lead to more land enrolled, i.e., more offers closer to the landown-
ers’ opportunity costs, from the perspective of the buyer. While these experiments
did not test all of the possible variations of the information set to identify the optimal
amount of public information, they do highlight how too much public information
allows participants to “game” the auction by strategically raising their offers above
their true reservation values. Furthermore, too little public information can lead par-
ticipants to inflate their offers because market competition is not fully realized and
prices are not forced to be close to the true opportunity cost. Although a limitation
of this study is that opportunity costs are treated as exogenous and market prices are
not permitted to adjust over time, the design isolates market experience and suggests
that auction efficiency may decrease as the size of the procurement budget increases
and decreases over time as sellers learn through their experience to elevate their offers
strategically to secure greater rent premiums. If one assumes there is no systematic
joint impact of experience and endogenous opportunity costs—a question for future
experiments—then the evidence suggests that these factors lead to the conservation
program paying more than it would otherwise have to and thereby reducing the pro-
vision of ecosystem services given the limits of governmental resources. These results
contribute to the ongoing research in the area of ecosystem markets and illustrate how
experiments can be applied to address important issues related to land economics.
Acknowledgments
Funding for this research was provided by the U.S. Department of Agriculture’s
Economic Research Service (ID#58-6000-7-0089). The authors appreciate Marca
Weinberg and Dan Hellerstein for their support and feedback on earlier versions of this
research. We also appreciate the research assistance from Robin Dillaway, Jacob Fooks,
Jubo Yan, and Shang Wu from the University of Delaware.
References
Arnold, M., J. M. Duke, and K. Messer. 2013. “Adverse selection in reverse auctions for environ-
mental services” Land Economics 89(3): 387-412.
Banerjee, Simanti, Shortle, James S., and Kwasnica, Anthony M., 2011. “An iterative auction for
spatially contiguous land management: An experimental analysis,” 2011 Annual Meeting,
July 24-26, 2011, Pittsburgh, Pennsylvania 103220, Agricultural and Applied Economics
Association.
Bernard, J. C. 2005. Evidence of affiliation of values in a repeated trial auction experiment.

Applied Economics Letters 12: 687–691.
Cason, T. N., and L. Gangadharan. 2004. Auction design for voluntary conservation programs.
Cason, T., Gangadharan, L., and Duke, C. 2003. A laboratory study of auctions for reducing non-
point source pollution. Journal of Environmental Economics and Management 46: 446–471.
CJC Consulting. 2004. “Economic Evaluation of the Central Scotland Forest and Grampian
Challenge Funds.” Final report for Forestry Commission Scotland. www.forestry.gov.uk/
pdf/FCchallenge.pdf/$FILE/FCchallenge.pdf
Connor, J. D., J. Ward, and B. A. Bryan. 2008. Exploring the cost effectiveness of land conser-
vation auctions and payment policies. The Australian Journal of Agricultural and Resource
Economics. 52(3): 303–319.
Davis, D. D. and C. A. Holt. 1993. Experimental economics. Princeton University Press,
Princeton, New Jersey.
Fooks, J. R. and K. D. Messer. 2012. “Maximizing conservation and in-kind cost share: Applying
goal programming to forest protection” Forest Economics. 18: 207–217.
Fooks, J., K. D. Messer and J. M. Duke. 2012. Dynamic entry, reverse auctions, and the purchase
of environmental services. Manuscript: University of Delaware.
Fischbacher U. 2007. “z-Tree: Zurich toolbox for ready-made economic experiments,”
Experimental Economics 10(2), 171–178.
Friedman, D. and S. Sunder. 1994. Experimental methods: a primer for economists,
New York: Cambridge University Press.
Gole, C., M. Burton, K. Williams, H. Clayton, D. P. Faith, B. White, A. Huggett, and C. Margules.
2005. Auction for landscape recovery: Final report. WWF-Australia, Perth.
Hailu, A. and S. Schilizzi. 2004. Are auctions more efficient than fixed price schemes when
offerers learn? Australian Journal of Management 29(2): 147–168.
Harrison, G.W. and J.A. List. 2004. “Field experiments.” Journal of Economic Literature
XLII: 1009–1055.
Hellerstein D. and N. Higgins. 2010. The effective use of limited information: Do offer maxi-
mums reduce procurement costs in asymmetric auctions? Agricultural and Resource
Economics Review 39(2): 288–304.
Hong, H. and M. Shum. 2002. Increasing competition and the winner’s curse: Evidence from
procurement. Review of Economic Studies 69: 871–898.
Horowitz J. K., L. Lynch, and Andrew Stocking, 2009. Competition-based environmental pol-
icy: An analysis of farmland preservation in Maryland. Land Economics, 85(4) 555–575.
Kagel J. H. and A. E. Roth. 1995. The handbook of experimental economics, Princeton, New
Jersey: Princeton University Press.
Kirwan, B., R. N. Lubowski, and M. J. Roberts. 2005. How cost-effective are land retire-
ment auctions? Estimating the difference between payments and willingness to accept
in the Conservation Reserve Program. American Journal of Agricultural Economics
87(5): 1239–1247.
Klemperer, P. 2002. What really matters in auction design. Journal of Economic Perspectives 16
(1): 169–189.
Latacz-Lohmann, U. and S. Schilizzi. 2005. Auctions for conservation contracts: A review of the
theoretical and emprical literature. Report to the Scottish Executive Environment and Rural
Affairs Department. Project No: UKL/001/05.
Latacz-Lohmann, U., and C. Van der Hamsvoort. 1997. Auctioning conservation con-
tracts: A theoretical analysis and an application. American Journal of Agricultural Economics
79 (2): 407–418.
List, J. A. and J. F. Shogren. 1999. Price information and bidding behavior in repeated
second-price auctions. American Journal of Agricultural Economics 81: 942–949.
Lusk, J. L., T. Feldkamp and T. C. Schroeder. 2004. Experimental auction procedure: Impact
on valuation of quality differentiated goods. American Journal of Agricultural Economics
86: 389–405.
Lusk, J. L. and J. F. Shogren. 2007. Experimental auctions: Methods and applications in economic
and marketing research. New York: Cambridge University Press.
McAfee, R. P. and J. McMillan. 1996. Analyzing the airwaves auction. Journal of Economic
Perspectives 10 (1): 159–175.
Messer, K. D. and W. L. Allen. 2010. Applying optimization and the analytic hierarchy process
to enhance agricultural preservation strategies in the State of Delaware. Agricultural and
Messer, K. D., J. M. Duke, and L. Lynch. 2012. When does public information undermine the
effectiveness of reverse auctions for the purchase of ecosystem services? Manuscript: University
of Delaware.
Rassenti, S. J., Smith, V. L., and Wilson, B. J. (2003). Discriminatory price auctions in electricity
markets: Low volatility at the expense of high price levels. Journal of Regulatory Economics,
23(2): 109–123.
Rolfe, J., Windle, J., and McCosker, J. 2009 Testing and implementing the use of multiple offering
rounds in conservation auctions: a case study application. Canadian Journal of Agricultural
Schilizzi, S. and U. Latacz-Lohmann. 2007. Assessing the performance of conservation auc-
tions: an experimental study. Land Economics 83: 497–515.
Shogren, J. F. 2004. Incentive mechanism testbeds: Discussion. American Journal of Agricultural
Economics. 86(2004): 1218–1219.
nism. American Journal of Agricultural Economics 77: (1995) 93–105.
Stoneham, G., V. Chaudri, A. Ha, and L. Strappazon. 2003. “Auctions for conservation con-
tracts: An empirical examination of Victoria’s bushtender trial.” Australian Journal of
Agricultural and Resource Economics 47 (4): 477–500.
United Stated Department of Agriculture. 2006. Conservation reserve programs: Largest
Conservation Partnership on Private Land. Farm Service Agency. Accessed on January 17,
2012. http://www.fsa.usda.gov/Internet/FSA_File/crp20trifold.pdf
United Stated Department of Agriculture. 2011. CRP Enrollment as of September 2011 and
October 2011 Rental Payments. Accessed January 17, 2012. www.fsa.usda.gov/Internet/
FSA_File/apportstate091311.pdf
Vickrey, W. 1961. “Counterspeculation, auctions and competitive sealed tenders.” Journal of
Finance 16: 8–37.
Vossler, C. A., T. D. Mount, R. Thomas, and R. Zimmerman. 2009. An experimental investiga-
tion of soft price caps in uniform price auction markets for wholesale electricity. Journal of
Regulatory Economics 36(1): 44–59.
Wu, J., and B. Babcock. Contract design for the purchase of environmental goods from agricul-
ture. American Journal of Agricultural Economics 78 (1996) 935–945.
Appendix
A1. Experiment Instructions
Welcome to this experiment in the economics of decision making. In the course of the
experiment, you will have opportunities to earn money. Any money earned is yours to
keep. Therefore, please read these instructions carefully. Please do not communicate
with other participants during the experiment.
In this experiment, you will participate in a series of market trading rounds. You and
all of the other participants in the room will assume the role of landowners and you will
be given the opportunity to sell a parcel of land. The administrator will be the buyer. In
each round, you will need to decide whether you want to keep your parcel or try to sell
it. If you decide to try to sell it, you will also need to decide the offer price for your parcel.
Below is a hypothetical example (see Figure 19.2), where three rounds have been
completed and the fourth round is just about to begin. On your computer screen, you
will note a variety of important information. Your parcel is assigned an Ownership
Return per acre, which is the amount of money that will be added to your profit if the
parcel is not sold in that round. The size of your parcel is 100 acres. In this hypotheti-
cal example, the ownership return for your parcel in the first round is $3000 per acre.
In general, your ownership returns may not be the same as those of other sellers and
will change throughout the experiment. As shown in Figure 19.3, since the possible
ownership returns were randomly determined from a uniform distribution that has a
minimum value of $2000 per acre and a maximum value of $8000 per acre, the average
ownership return can be expected to be $5000 per acre. Your ownership returns have
been determined by random prior to the start of the experiment and you will know your
ownership return at the start of each round.
Round 1 Round 2 Round 3 Round 4

Ownership Return per acre: $ 3,000 $ 6,000 $ 4,000 $ 5,000
Offer Price per acre: $ 5,000 $ 8,000 $ - $ -
Submission Cost per acre: $ 20 $ 20 $ - $ -
Number of Acres: 100 100 100
Sold: Yes No No
Submit Submit Submit Submit
Update Update Update Update

Total Profit: $ 498,000 $ 598,000 $ 400,000 $ -
FIGURE 19.2 Experimental interface for participants (screen shot).

Probability
1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000

Ownership returns
FIGURE 19.3 Distribution of ownership return.

Probability
$1,000,000 $2,000,000 $3,000,000 $4,000,000 $5,000,000 $6,000,000 $7,000,000

Buyer's budget
FIGURE 19.4 Distribution of budgets.
The buyer’s budget will be announced before the start of each round. Like with the
ownership returns, the budgets were randomly determined before the start of the
experiment. The budgets were randomly determined from a triangle distribution that
has a minimum value of $2 million, an average value of $4 million, and a maximum
value of $6 million. As shown in Figure 19.4, the most likely budget will be $4 million.
Values closer to $4 million are more likely than values further away from $4 million.
In each round, you must decide whether you want to sell your parcel and, if so, at
what price you are willing to sell it. You will pay a submission cost of $20 per acre if you
decide to submit an offer for your parcel. You can submit your offer price confidentially
by entering it into the yellow box in your spreadsheet. Then, hit “Enter” on your key-
board, and click on the “Submit” button. In this example for Round 1, the seller submit-
ted an offer price of $5000. If you elect not to submit an offer, such as in Round 3, just
leave the yellow box blank and click the “Submit” button. In this case, you do not pay the
submission cost.
After everyone has submitted their decision, the administrator will purchase as many
parcels as possible starting from the lowest offer price and moving up until the available
budget for that round is exhausted. For example, imagine that current round budget
is $300 and eight offer prices were submitted—ranked from lowest to highest: [Offer
Prices: $30, $40, $50, $60, $70, $80, $90, $100].
Parcels are purchased in order (from left to right) until the buyer does not have
enough money to purchase another parcel. In the example, the five lowest offer prices
($30 + $40 + $50 + $60 + $70) are purchased for a total of $250. None of the last three
offers are purchased, since even the lowest non-accepted offer of $80 would bring the
total cost to $330 ($250 + $80) and therefore be higher than the buyer’s budget.
A2. Determination of Profits
After all offer prices have been received, the auctioneer will determine which parcels
were purchased. You will then click on the “Update” button. There are three possible
profit scenarios:
1) Successful sellers will receive a price equal to their offer, and thus, their profits
will be their offer price for that parcel minus the submission cost.
2) Participants that submit an offer for a parcel which is too high for the available
budget will not receive their offer price, but instead their profits will be their
ownership return for that parcel minus the submission cost.
3) Profits for participants who do not submit an offer for a parcel will be their own-
ership return for that parcel.
In Round 1 of the example below (see Figure 19.5), the participant earned a total profit
of $498,000 by successfully selling the parcel for $5000 per acre and paying the submis-
sion cost of $20 per acre. In round 2, the participant earned $598,000 as they received
their ownership return of $6000, acre when they did not successfully sell their unit, but
still had to pay the submission cost of $20 per acre. Finally, in Round 3, the participant
would earn $400,000 by receiving the ownership return of $4000 per acre and not paying
the submission cost.
Your computer will calculate your profits for each parcel in each round and will
keep track of your cumulative earnings. An exchange rate of 1,200,000 to 1 will be
used to converts your earnings from experimental dollars to dollars. For exam-
ple, if you earn 24,000,000 experimental dollars will have earned $20 US to take
home today.
A3. Market Information
In addition to the information regarding your ownership returns and whether you sold
your parcels, you will receive information regarding the market (as seen in the example)
when you click on the “Update” button after each round:
Round 1 Round 2 Round 3 Round 4

Ownership Return per acre: $ 3,000 $ 6,000 $ 4,000 $ 5,000
Offer Price per acre: $ 5,000 $ 8,000 $ - $ -
Submission Cost per acre: $ 20 $ 20 $ - $ -
Number of Acres: 100 100 100
Sold: Yes No No
Submit Submit Submit Submit
Update Update Update Update
Total Profit: $ 498,000 $ 598,000 $ 400,000 $ -
Next Round's Budget: $ 2,200,000 $ 5,500,000 $ 6,000,000

This Round's Budget: $ 2,800,000 $ 2,200,000 $ 5,500,000
Part A (Profit): Number of Offers Received: 8 6 9

$ 1,496,000 Lowest Successful Offer: $ 2,500 $ 2,000 $ 2,000
Highest Successful Offer: $ 7,000 $ 5,500 $ 5,500
Exchange Rate: Average Successful Offer: $ 5,000 $ 3,500 $ 3,500
1,400,000 Seller ID Numbers: 1 2 1
2 4 2
Part A (US$): 5 6 3
$ 1.07 9 8 6
10 9
10
FIGURE 19.5 Experimental interface for participants with information (screen shot).
Before each round, the administrator will announce:
• The buyer’s budget for that round.
After each round, the administrator will announce:
• The number of offers submitted to the market,

• The price of the lowest successful offer,
• The price of the highest successful offer,
• The price of the average successful offer, and
• The computer identification numbers of participants that successfully sold their
parcels.
PA R T I V
T H E E C ON OM IC S
OF L A N D U SE L AW
A N D P OL IC Y
C HA P T E R 20
O P E N S PAC E P R E S E RVAT I O N
Direct Controls and Fiscal Incentives
E KAT E R I NA G N E DE N KO A N D DE N N I S H E F F L EY
The forces of urbanization and the quest for open space reflect a basic economic trad-
eoff. The benefits of agglomeration, communication, and exchange are well understood
and have been explored in depth by Mills (1967), Fujita, Krugman, and Venables (1999),
Glaeser (2008), and many other economists, geographers, planners, and sociologists.
Yet, although humans seem willing to endure the negative aspects of crowded living to
reap the economic and social benefits of cities, they also show a need, or at least a long-
ing, for more “elbowroom.”
Open space may include public lands such as parks, recreation areas, or national
forests, as well as privately owned parcels: farms, golf courses, or even large residential
lots. This “private open space” may limit public access yet still offer amenity benefits to
owners and neighbors (Cheshire and Sheppard 2002). In this sense, private open space
competes with public open space, both because it, too, provides externalities, but also
because it reduces land available for public open space.
Open space preservation efforts are not new, but they have recently gained momen-
tum, especially at the nonfederal level, where many states, counties, and towns actively
seek to protect or expand open space. The popularity of such initiatives has prompted
a large body of theoretical and applied research. Beyond reviewing some of the studies
and the origins, rationale, and goals of preservation efforts, we offer a simple but adapt-
able model that highlights the need to consider the long-run effects of such programs.
To illustrate the point, we focus on the use of state aid to local governments to alleviate
fiscal pressure and facilitate open space zoning. We show, however, that well-intended
policies can have unintended results when long-run adjustments occur in household
behavior and local fiscal and zoning policies—a lesson that applies to many public
efforts to shape land market outcomes.
514 Ekaterina Gnedenko and Dennis Heffley
1. Open Space Benefits and

Preservation Programs
1.1 Origins and Directions

Despite the mixed success of past preservation efforts, there has been a visible upsurge
in public desire to preserve open space. Our analysis focuses on the notion of an optimal
amount of local open space and the role of intergovernmental revenue, particularly state
aid to municipalities, as a potential way to encourage preservation. The analysis is espe-
cially relevant for the northeastern United States, where much of the land is privately
owned, subject to local land use controls, “and pressures to develop the remaining rural
land for residential and other uses are among the strongest in the country” (Parks and
Schorr 1997, 85), but the questions and issues we address apply elsewhere as well.
Interest in open space is widespread and well established. Ancient societies under-
stood the value of reserving areas for agriculture, public use, and even traditional cer-
emonies. Elizabeth I, in 1580, banned new construction within several miles of central
London (Evans 1999), and, in a similar display of royal preference for more room (or
perhaps a more secure buffer zone), picturesque landscapes and parks surrounded
17th-century Versailles and the 18th-century Pavlovsk Palace. But the demand for
open space also has strong democratic roots: the US federal government first attempted
to secure parklands for the nation’s capital in 1791, Congress established Yellowstone
National Park in 1872, and Cape Hatteras was designated the first National Seashore
in 1937. In 1988, the governors of five New England states formally recognized open
space as a key quality-of-life indicator and the foundation of a multibillion dollar tour-
ist industry. Such acknowledgments of the economic and social benefits of open space,
echoed more recently by many public officials, have signaled state and local efforts to
preserve low-density land use patterns.
The recent America’s Great Outdoors report (America’s Great Outdoors [AGO]
Initiative 2011), prepared by the US Departments of Interior and Agriculture and the
Environmental Protection Agency, calls for support of a conservation agenda and
emphasizes the joint efforts of federal, state, and local governments, but, for some
observers, this agenda is long overdue. More than two decades ago, Anas (1988,
159) expressed concern that “the American wilderness within the public lands has
shrunk to 2% of its original size, becoming a scarce, irreplaceable resource in need of
efficient marginal pricing.” Hollis and Fulton (2002, 5) similarly note that open space
protection has lagged despite “the long list of federal policies that have promoted con-
servation of open space . . . [such as] the Land and Water Conservation Fund, the
Endangered Species Act, the Clean Water and Safe Drinking Water Acts, the North
American Wetland Conservation Act, and the National Environmental Policy Act.”
They attribute this ineffectiveness to the fact that: “In many cases, . . . open space protec-
tion has been secondary to the environmental goals of these programs. Partly for this
Open Space Preservation 515
reason, most strategies to use open space to consciously shape metropolitan form have
been initiated by states or localities.”
Decentralization of land use policy and open space preservation has advantages
and drawbacks. Given the range of geographic conditions, settlement patterns, and
socioeconomic mix across states, counties, and municipalities, it is unlikely that a
national “one-size-fits-all” approach would serve any state or community very well.
Experimentation and a localized or “tailored” approach have some obvious advantages,
allowing a better fit between individual preferences and the provision of open space, as
described by Tiebout (1956) for local public goods in general. But Tiebout efficiency
requires mobility—ample opportunities for households to seek out, identify, and occupy
communities that best fit their preferences for open space, other local public goods, and
taxes. Such “frictionless” conditions are more likely to prevail in the long run, when
transaction costs hamper mobility less.
Decentralized provision of open space expands the menu of options and caters to
local tastes, but it raises other problems, and some researchers still see a vital role for
higher level or cooperative multilevel programs (Bates and Santerre 2001). One chal-
lenge in analyzing subnational preservation efforts is that “every state has its own unique
governmental structure with varying degrees of involvement in parks, recreation, and
conservation” (Betz and Cordell 1998, 9). Uncoordinated local initiatives also may face
unfavorable spillovers and free-rider problems between jurisdictions (Loomis, 2000).
Yet federal programs have their own shortcomings. In theory, they should provide a
more consistent approach to preservation, but special interest legislation (Holcomber
and Staley 2001) and budget fluctuations produce cross-sectional and intertemporal
differences in federal efforts that may need to be “smoothed” or complemented by state
and local programs.
This mixed approach may be messy, but it has strong supporters. Ostrom (2010) sees
a need for “polycentric” governance systems to address public good and common pool
resource problems, but she also stresses the importance of nongovernmental organiza-
tions and individuals in achieving successful outcomes. However, coordination is not
easily achieved when governments and constituents are able to react to policy changes.
Stavins (2011, 82) stresses the importance of formally considering responses to pro-
posed policies, noting that although economic theory “has made major contributions to
our understanding of commons problems and the development of prudent public poli-
cies, . . . government policies that have not accounted for economic responses have been
excessively costly, often ineffective, and sometimes counterproductive.”
McGonagle and Swallow (2005, 477) share this concern, noting that: “States and
municipalities have committed over $24 billion in bond issues for land conservation in
recent years, yet the structure of the land conservation industry and markets is poorly
understood.” Nelson (1998, 34) calls for systematic evaluation of farmland retention
policies since “we actually do not know what the metropolitan regional landscape would
be in the absence of externalities and market failure. Lacking this basic understanding,
there is no way in which to compare farmland preservation techniques because there
is no benchmark with which to compare them.” Poe (1999, 589) further argues that land
use policies need to shift from pure resource development to a cohesive land use/land
preservation strategy that “may also require ‘top-level’ efforts to coordinate agency
actions.” Toward this end, Duke and Lynch (2006) offer a useful taxonomy and compari-
son of 28 farmland retention techniques, falling into four major categories: regulatory,
incentive-based, governmental-participatory, and hybrid.
1.2 Hands Off or Hands On?

Support for preservation may be growing but is not universal, and even those who value
open space may not see government intervention as essential. Some researchers suggest
that market forces may suffice (Thorsnes and Simons 1999; Holcomber and Staley 2001;
McConnell et al. 2003). Many years ago, Muth (1961) showed that even in the face of
urban population and income growth, rural land conversion will be contained around
urban areas that rely on local agricultural production, but not necessarily around those
dependent on national food markets. As the urban population expands or income
grows, cities supplied with food from local sources must bid land away from farmers,
but the increase in demand for food also makes farmland more valuable, thus slowing
the conversion of rural land to housing. More recently, Holcomber and Staley (2001, 3,
9) argue that market-oriented approaches tend to maximize the value of property and
often address the loss of open space better than do zoning or other regulatory devices
that may fall prey to special interests or compromise popular goals.
On the other hand, many researchers see a proper, if limited, role for government.
Ladd (1980) cites several reasons for inefficient land use and underprovision of open
space, including its public good attributes and external benefits, irreversibility of the
development process, and “fewer benefits [from local public services] per dollar of
market valuation to owners of agricultural or open space land than to other types of
landowners.” She concludes that: “some form of government action might be desirable
to increase agricultural or open space land” (19). Even Gardner (1977), who expresses
concerns about the use of “extra-market means” to preserve farmland (1028), sees that
in “the case where market failure is most apparent—the creation of open space and envi-
ronmental amenities . . . there may be some justification for social action to remedy this
market failure” (1031). Wu and Irwin (2008, 233) note that private markets inefficiently
convert too much undeveloped land to residential (and other) uses because they “fail to
account for the environmental costs of land development, which include not only the
standard environmental damages from pollution, but also a loss of flexibility in adjust-
ing the future path of pollution due to the irreversibility of development.”
Fundamental forces like population growth and rising incomes, which increase
demands for residential and commercial land, are the major causes of open space loss and
cannot be readily contained or avoided. Market failures associated with unfettered land
development, including undervaluation of open-space amenities, unpriced traffic con-
gestion, and underpriced infrastructure costs of new development, must be remedied by
appropriate government policies (Brueckner 2000; Anas and Rhee 2006; Brueckner and
Helsley 2011). Plantinga and Ahn (2002, 128) suggest that: “One role of land-use policies
is to narrow the divergence between privately and socially optimal land allocations by
modifying the economic incentives faced by private landowners.” McGuire and Sjoquist
(2003, 8) reiterate the externality rationale for preservation by noting that “open space on
the fringe of urban areas may have value beyond its private value.”
Even opponents of publicly subsidized preservation programs endorse occasional
intervention. Bae (2007, 39) believes that although “the macro argument is unconvinc-
ing, . . . at the micro (regional) level there may be instances where a plausible case for
farmland preservation can be made.” But microlevel land use policy also has its critics.
Hollis and Fulton (2002, 46) claim that the United States’ “decentralized system tends to
encourage reactive or ad hoc open space protection at the local level and, in many cases,
large-scale acquisitions based on different strategic objectives.” They claim that local
preservation is fragmented, and neither state nor local programs are well documented,
making interjurisdictional comparisons difficult.
1.3 Preservation Strategies

Preservation strategies vary in their locus of control—national, state, local, or mixed—
and their reliance on market-oriented versus nonmarket approaches. Other than out-
right public purchase of land, zoning is perhaps the most direct form of intervention in
land markets to affect open space. In this regard, it is important to distinguish between
land use zoning (or allowable-use zoning) and density zoning restrictions. Land use zon-
ing essentially “carves up” an area into zones where certain activities, or combinations
of activities, are permitted: residential, commercial, industrial, agriculture, mixed-use,
and the like. Mixed-use zoning has enjoyed a US revival, but mixed uses often conflict
with land use zoning objectives: separating “incompatible” uses to avoid negative exter-
nalities, restricting the local supply of certain types of land, and assigning particular
activities to the “most suitable” sites.
If land use zoning is the planner’s “macro” policy instrument, density zoning is the
“micro” tool used to fine-tune land use. Land within a residential zone can be developed
in many ways, but density zoning ordinances often set lower bounds on lot size, floor
space, and set-back distances from roads and neighboring properties, as well as upper
bounds on height and lot coverage. These provisions, individually or jointly, restrict
development density and thus tend to (imprecisely) limit population, especially if con-
straints are binding.
With the two types of zoning, local authorities can preserve or even expand open
space, primarily by using land use zoning to protect or enlarge such areas (which, absent
annexation of unincorporated areas, will limit the size of the development zone) or by
relaxing density zoning to increase the “holding capacity” of the development zone and
thereby channel activity away from threatened areas. In this way, allowing smaller build-
ing lots, easing height limits, or permitting multifamily units within a residential zone
can ease the pressure on open space.
Direct agricultural zoning prevents conversion of land to other uses, but it is a blunt
instrument that may benefit the broader public at farmers’ expense. Removing the right
to develop farmland without compensating the owner may be difficult to justify. This
has fostered “voluntary” approaches, including outright land purchase, purchase of
development rights (PDR), and transferable development rights (TDR) programs.
Fee-simple public purchase of farmland or open space gives the fullest control of
land use and ensures (barring public resale) long-term protection, but it also saddles
the government with an asset that requires further maintenance or operating expendi-
tures. Land donations to communities or private land trusts avoid the acquisition costs,
but not the ongoing outlays, unless there are specific provisions made for those in the
gift. Some of these maintenance costs to governments or land trusts can be reduced, or
at least shifted, by compensating landowners to forego development, but development
rights programs also have limitations.
PDR programs use public funds to purchase separable development rights from land-
owners who, if they accept the offer, agree to forego development in perpetuity or for a
specific period. Here, since there is no well-developed market to price the development
rights, the common problem is how to determine “fair” compensation: high enough
to secure the rights, but not so high as to overpay landowners, offend taxpayers, and
jeopardize the program. Wolfram (1981) also warns that costs of implementing and
administering a PDR program may be too high to become an efficient means of pro-
viding open space. Analyzing survey opinion toward farmland preservation, Foltz and
Larson (2002, 15) conclude that “because the public supports low cost PDR programs
uniformly across geographic and socioeconomic boundaries, farmland preservation is
an issue perhaps best engaged at the state rather than a local level.” Lopez, Shah, and
Altobello (1994, 61) compare the effects of, and farmland owners’ support for, PDR and
agricultural zoning programs and conclude that PDRs “can be effective in attaining a
socially optimal allocation of land.” Liu and Lynch (2011) find that such programs are
effective in reducing the rate of farmland loss, but Nickerson and Hellerstein (2003)
note that popularity of these programs with farmers has led to oversubscription and
queuing. More public funding would ease the problem, but such programs must com-
pete with other important public services.
TDR programs seek to address this chronic “underfunding” of public open space
programs by creating a private demand for the development rights. Costonis (1973,
1975) proposed this approach to preserve historic buildings in Chicago, but the con-
cept is readily applied to open space preservation (Barrows and Prenguber 1975; Field
and Conrad 1975; Mills 1980; Thorsnes and Simons 1999). Rather than being paid by
government, landowners in the “preservation area” who agree to forego development
are issued TDRs, which can be sold for use in some “development area” to marginally
increase development density (e.g., permission to build on a somewhat smaller lot than
normal zoning would permit). Carpenter and Heffley (1982) show that getting a TDR
market to form and behave as intended may not be trivial, possibly explaining why only
about 140, mostly local, TDR programs exist throughout the United States (Walls and
McConnell 2007, 8).
1.4 Effects of Preservation Policies on Land Values

Effects of preservation policies on the price of open space depend on the nature and
permanence of the policy, as well as on the total area of the target region (Irwin 2002;
Geoghegan et al. 2003). Some programs seek to preserve farmland by enhancing
the income that can be derived from using, or even just holding, farmland. Just and
Miranowski (1993) observe a positive capitalization effect of government agricultural
support payments on farmland values for three states, whereas Weersink et al. (1999)
find the same for Ontario, Canada. Wu and Lin (2010, 2) note that “the conventional
wisdom is that because the supply of agricultural land is highly inelastic, government
payments are largely capitalized in farmland values.” Using 1997 county-level data, they
show that the major US conservation program, the Conservation Reserve Program
(CRP), “has a positive and statistically significant effect on farmland values in all
regions” (11).
But if a program does more than simply augment the income of farmers or landown-
ers, if it restricts actual or potential use of the land without sufficient compensation, the
price impact is more typically negative, although not always significant; see, for example,
Hascic and Wu (2012). Nickerson and Lynch (2001, 341) use hedonic methods to ana-
lyze Maryland’s development restrictions, imposed by permanent easement acquisition,
to show that “[a]‌lthough preserved parcels’ actual land values are lower, the effect of the
restrictions is not statistically significant.” Using a similar approach to study a morato-
rium on urban development of agricultural land near Toronto, Deaton and Vyn (2010,
954) find that “[f]armland within the Greenbelt, and in close proximity to the Greater
Toronto Area (GTA), experienced a statistically significant decline in property values.
More specifically, the negative effect occurs mainly at the urban–rural boundary: i.e.,
farmland within 5 km of the GTA.” Henneberry and Barrows (1990) find that exclu-
sive agricultural zoning impacts farm values in even more complex ways that depend
on property characteristics, whereas Hascic and Wu (2012) find evidence of land value
impacts of zoning both inside and outside the zoned areas. They also note that “a zoning
regulation may affect the value of a parcel both directly by restricting its use and indi-
rectly by affecting land use in its surrounding areas” (199).
Adverse effects of restrictions on the price of open space itself may be offset by posi-
tive capitalization of the amenity effects on other properties (Correll et al. 1978; Katz
and Rosen 1987; Do and Grudnitski 1995; Geoghegan et al. 1997; Mahan et al. 2000;
Lutzenhiser and Netusil 2001; Shultz and King 2001; Smith et al. 2002; Geoghegan
et al. 2003; Earnhart 2006; Wu and Lin 2010). This intralocal spillover may encourage
communities to preserve open space to raise property values and reap tax benefits, but
sluggish property markets may limit the approach. Geoghegan, Lynch, and Bucholtz
(2003) offer empirical evidence that certain counties in Maryland show this capac-
ity to self-finance preservation from the increase in property tax base due to the pre-
served open space. Weigher and Zerbst (1973) even cite examples where profit-seeking
landowners offer part of their land for open space to enhance the overall value of their
residential development. Bolitzer and Netusil (2000, 193) note that the “degree of open
space self-financing . . . depends on many factors including the size of the open space, the
number of homes in proximity to the open space, open-space amenities, and the local
property tax structure.” Jiang and Swallow (2006) use simulations to show that financing
open space via amenity-induced property tax increments may be more feasible if open
space is evenly distributed to benefit more homes, although presumably the amenities
also can be spread “too thin.” This tradeoff suggests the existence of an optimal spatial
pattern of open space, a challenging topic that warrants more research. Finally, adding a
healthy touch of reality to the lure of “self-financing” open space schemes, Wu, Xu, and
Alig (2012) show that when preservation costs must be covered by property taxes, both
higher tax rates and a lower level of public services may reduce property values.
2. A Model of Local Zoning for

Open Space
2.1 Motivation
Many states have initiated preservation programs, ranging from outright land purchase
and PDR programs to market-oriented TDR schemes that use private rather than public
funds to compensate landowners who forego development. However, given the paucity
of working TDR programs, state-level preservation efforts generally rely heavily on the
public purse to acquire land or stockpile development rights.
Local governments also seek to preserve open space, typically via land use controls,
but researchers have noted that fragmented local efforts can lead to suboptimal lev-
els of open space (Hollis and Fulton 2002) and inefficient leapfrog development pat-
terns (Wu and Plantinga 2003). Given the public good attributes and external benefits
of open space, the state might seem to be the right level of government to provide this
“transboundary public good” (Loomis 2000), but, in the United States, a common
public-power split often impedes or distorts open space preservation.
With numerous tax instruments at their disposal (income, sales, excise, corporate,
estate, and others), states can potentially finance preservation efforts more readily
than localities, which rely heavily on property taxes and are often precluded from
levying other taxes. In many states, though, zoning and other land use control pow-
ers have been legislatively ceded to counties, cities, or towns. Consequently, even
where states could (in good economic times) devote resources to preservation, they
may lack the legal authority or political will to actively move local landowners toward
broad public goals.
Local governments, on the other hand, may be able to facilitate open space preserva-
tion through zoning changes, but if this requires extra spending or limits local revenue
by thwarting economic development, necessary policy changes also may fail to occur
at the local level. In sum, local governments often wield the direct instruments of land
use control but lack the necessary resources, especially if landowners must be compen-
sated to forego development; state governments, by contrast, may have the tax powers
to finance preservation but often have yielded land use control powers to counties or
towns. This awkward alignment of land use authority and fiscal powers may explain why
public surveys point to a significant unmet demand, often expressed as a strong willing-
ness to pay for various types of open space (Bergstrom et al. 1985; Kline and Wichelns
1994; Duke and Ilvento 2004; Gnedenko 2009). The strength of this demand for conser-
vation, however, also depends on the policy mechanism or instrument (Johnston and
Duke 2007).
The separation or imbalance of fiscal and zoning powers has led to suggestions that
increased state aid to communities be used to encourage local governments to zone
more land for open space. Ladd (1980) and Gottlieb (2006) discuss various aspects
of such intergovernmental transfers, whereas Gnedenko (2009) offers an analyti-
cal framework for examining the optimal mix of open space and development and its
response to state aid. On the surface, the argument for more state aid seems straight-
forward: it should reduce pressure for communities to accommodate new development
as a necessary way to finance cash-starved schools and other essential public services.
Consequently, this injection of revenue ought to make preservation economically fea-
sible and politically more palatable for local governments.
It sounds simple enough, but the issue is more complex. More state aid likely affects
the local government’s fiscal decisions, allowing it to boost spending and/or lower its
property tax rate while maintaining a balanced budget. Yet, if anything, such adjust-
ments would make the community more attractive for households and firms, and this
added development pressure could make it more, not less, difficult to preserve open
space via land use (or allowable-use) zoning. Analysis of this unintended effect of state
aid on local open space requires a model that fully endogenizes local tax, spending, and
land use zoning decisions, as well as the behavior of households.
In this section we describe an extended open-city model that has these elements
and allows us to consider the long-run effects of more state aid on a community’s land
use zoning (open-space vs. residential development) and fiscal mix (property tax rate
and public spending), household consumption patterns (lot-size, structure, and other
goods), land prices, aggregate land value, and population size. The model indicates that
more state aid will not always promote additional open space, as illustrated by simula-
tions of a calibrated version of the model and further supported by empirical analysis of
town-level panel data for Connecticut.
2.2 Households
The representative household, with annual income y, selects a lot-size (x), an amount of
structure or “floorspace” (k), and a numeraire consumption good (g), but it also derives
utility (U) from local public expenditures (G) and local open space (Xo). Exogenous
community features, such as its location relative to regional job centers or proximity
to an important amenity, also affect household utility, but this vector of characteristics
(C) can be deferred to the empirical analysis. Local open space is privately owned, but
the amount within the community’s bounded land area (L) is determined by the local
government’s land use zoning mix (Xo, L−Xo). Xo and G, as well as the local property tax
rate (T), are exogenous to the household but endogenous within the extended model of
government choice.
The household sees the local policy mix (Xo,G,T) and chooses (x, k, g) to maximize
U(x, k, g; G, Xo), subject to its budget constraint: (1−a)y = (1 + T)(px + rk) + (1 + s)g,
where p and r are annual rental prices of residential land and structure, respectively; a is
the average combined (state plus federal) income tax rate; T is the effective local prop-
erty tax rate (expressed as a fraction of the annual rental value of land and structure);
and s is the state sales tax rate on nonhousing consumption. We include income and
sales tax parameters (a and s) for realism and better calibration of the model, rather than
any attempt to fully capture higher level policy making, but these parameters do offer a
link to other levels of government that could be exploited in a multilevel framework.
The household’s constrained choice problem yields general-form demands for
lot-size, structure, and other goods:
x ∗ ( y , p, r , a, s, T , G, Xo ), k ∗ ( y , p, r , a, s, T , G, Xo ), g ∗ ( y , p, r , a, s, T , G, Xo ), (1)

and the corresponding indirect utility function gives the maximum attainable utility for
a given set of parameters:
U ∗ ( y , p, r , a, s, T , G, Xo ). (2)

If the typical household with income y can achieve a level of utility Uo in the “outside
world” and is fully mobile in the long-run, and if the rental price (r) of structure is exog-
enously determined in a regional or national market, the endogenous price (p) of local
developable land must adjust to ensure the local household can do just as well, or:
U ∗ ( y , p, r , a, s, T , G, Xo ) − U o = 0. (3)

Solving (3) implicitly for p gives the local equilibrium price of land (pe):
pe ( y , r , a, s, T , G, X0 , U o ). (4)

Substituting pe back into the earlier demand for land (x*) gives the equilibrium
lot-size (xe), or:
x e ( y , r , a, s, T , G, Xo , U o ), (5)

and, if the community has zoned Xo acres of its total land area (L) for open space and the
remaining L-Xo acres for residential use, land market clearance implies that the equilib-
rium number of households (ne) is:
ne = ( L − X o ) / x e , (6)

where xe is the expression in (5).

Some discussion of the asymmetric treatment of the prices of land and structure is
needed. Many urban models merge the two inputs by assuming that households con-
sume a composite good described as “housing services,” and, in urban spatial equilib-
rium versions of such models, the price of housing services is endogenous and declines
with distance from the urban center to compensate for commuting costs. Because we
focus on the allocation of a town’s land to residential use and open space, we see some
advantage in separating land and structure. This, in turn, raises the question of how the
price of each gets determined.
Casual observation and empirical evidence support the notion that land prices are
more endogenous and subject to local variation than the price of structure. For example,
even in a small state like Connecticut, it is difficult to explain the large variation in home
prices across different towns in terms of building cost differences. Materials and labor
costs differ, but usually not enough to account for the large differences in home prices
between (or even within) towns. Land prices, on the other hand, are quite sensitive to
location-specific attributes, including fiscal disparities and neighborhood effects. We
think this approach—an endogenous price of land that serves as the open-city adjust-
ment mechanism, coupled with an exogenous price of structure—works best within this
model, but for an alternative approach, see Wu and Plantinga (2003), where housing
prices are derived as a function of land prices.
For simplicity, we ignore commercial land use in this model, but for a similar model
that examines the long-run effects of residential versus commercial zoning (but ignores
open space), see Heffley and Hewitt (1988). Strong and Walsh (2008) offer a model of
the local housing market that focuses on the private provision of open space by develop-
ers when there exist spatial spillovers, but abstracts from the local government’s role in
determining public open space and the fiscal mix.
2.3 Local Government

Households may regard local zoning and fiscal policies (Xo, G, T) as given, but, ulti-
mately, we should allow these policies to respond to exogenous events, such as a change
in state aid. In the “open-city” environment, where households may enter or exit the
community, local officials can do little to permanently affect utility. Helpman and Pines
(1977), Brueckner (1979, 1982, 1983), Yinger (1982), Ross and Yinger (1999), Scotchmer
(2002), and many others have pointed out that this fluid environment may drive local
authorities to maximize the value of local land, the fixed resource whose market price
can be influenced by local zoning, spending, and tax policies. Casual observation of offi-
cials’ concern about the impact of budget decisions and zoning changes (or variances)
on local property values—their primary tax base as well as a common source of voter
complaints—reinforces the validity of the “open-city” assumption of land value maxi-
mization as the driving force of local policy making. Fischel (1990) also points to the
potential importance of aggregate land values in assessing the efficiency of zoning.
But our focus here is open space preservation. Why would a local government seek-
ing to maximize aggregate land value ever zone any land for open space? Would not
it be optimal to just let the market operate and freely determine the amount of open
space by letting competition allocate land to its “highest and best use”? In a geographi-
cally unconstrained world, perhaps so, but when total area of the jurisdiction is fixed, it
is easy to show why land use zoning is rational and why it is normally optimal to zone
some land for open space, even if it has no amenity value for other residents. If spillovers
do exist, the case for open space is simply strengthened.
To illustrate this point, before specifying a fuller model of local government behavior,
suppose the Walrasian demands for developable land (X) and open space (Xo), respec-
tively, are p(X) and m(Xo), where pX < 0 and mXo > 0. Maximization of aggregate land
value [V(X, Xo) = p(X)X + m(Xo)Xo] in a community where land area (L) is fixed requires
equality of the marginal value functions [p(X) + XpX = m(Xo) + XomXo] and satisfaction
of the land constraint (L = X + Xo). As shown in Figure 20.1, where demands and associ-
ated marginal value functions for the two land types are shown “face-to-face” to reflect
the land constraint, V is not maximized by simply allowing the market to freely equate
prices of the two land types. Maximization of V(X, Xo) requires zoning X* acres of land
for development and Xo* = L-X* acres of land for open space, where the marginal value
functions intersect (point A). In most cases, this marginal condition implies unequal
prices, p(X*) ≠ m(Xo*), as shown by points B and C.
B
p(X*)
C
m(Xo*)
m(Xo) A
m(Xo)+XomXo p(X)+XpX p(X)
X* (developed) Xo* (open space)
L (total land area)
FIGURE 20.1 Optimal mix of open space and development.

This simple figure points to the economic value of having some open space in a com-
munity, even if it confers no external benefits to occupants of the developed area. Open
space can increase aggregate land value within the bounded community by simply lim-
iting the amount (and raising the price) of developable land. Note that even if m(Xo)
lies everywhere below p(X) in Figure 20.1, X* may still be less than L, providing an eco-
nomic rationale for some open space (i.e., Xo* > 0). Gnedenko (2009, 15) discusses this
case in greater detail.
Alternatively, one can view land use zoning as a way to differentiate land and thereby
generate higher land rents, in much the same way that product differentiation and
unequal prices allow a seller to increase revenues. Density zoning instruments (mini-
mum lot-size, minimum floor-area requirements, height limits, coverage and set-back
restrictions, etc.) further expand the local government’s capacity to differentiate land
and boost aggregate rents. White (1975), Hamilton (1975, 1978), Grieson and White
(1981), Miceli (1992a, 1992b), and others have seen zoning as the exercise of monop-
oly power by local governments. Empirical studies by Thorson (1996), Rose (1989),
and Bates (1993) support this notion. Fischel (1978, 1980, 1985) stresses a property
rights approach, whereas Wallace (1988) empirically tests the hypothesis that zoning
policies simply tend to “follow the market.” Yet, regardless of motive, zoning does not
permanently protect open space; it “fails to establish a market price for the external-
ity” (Wolfram 1981, 402) and may cause “development to be inefficiently dispersed”
(Pollakowski and Wachter 1990, 324).
The simple view of open space as a way to differentiate local land and secure monop-
oly rents from development is obviously too narrow. Open space (Xo) may further boost
local land values if it generates external benefits that enhance the market price of devel-
oped land; that is, if p(X, Xo), where pX < 0 and pXo > 0. Our extended open-city model,
which incorporates the long-run behavior of households outlined earlier, not only
allows for this externality effect, but also for the potential effects of endogenous local
spending (G) and tax rate (T) decisions on the price of developable land (pG > 0, pT < 0).
The open-city model in Section 2.2 describes the equilibrium behavior of a repre-
sentative household and yields an expression (4) for the equilibrium price of developed
land. Households were assumed to view local public policies as given, but now we focus
on how those policies (Xo, G, T) are established, conditional on household behavior.
Again invoking the open-city assumption of land value maximization, we assume that
local authorities understand the potential effects of their policy choices on the price of
developed land, as embodied in equation (4), or pe(y, r, a, s, T, G, Xo, C, Uo). For simplic-
ity, the price of open space is assumed to be constant, m(Xo) = m > 0, but this can be
relaxed.
Excluding for now the vector of community characteristics (C) and assuming utility is
Cobb-Douglas in form, Axαkβg 1−α−βGγXoδ, or equivalently:
U(x , k , g ; G, Xo ) = ln A + α ln x + β ln k + (1 − α − β ) ln g + γ ln G + δ ln X o , (7)

where A > 0, α ∈ (0, 1), β ∈ (0,1), γ > 0, and δ > 0, the resulting household demands and
indirect utility function are:
x ∗ = α(1 − a) y /(1 + T ) p (8)
k * = β(1 − a) y /(1 + T )r (9)
g ∗ = (1 − α − β)(1 − a) y / (1 + s) (10)
U ∗ = ln A + α ln[α (1 − a) y /(1 + T ) p] + β ln [(1 − a) y /(1 + T )r ] + (1 − α − β ) (11)

× ln[(1 − α − β )(1 − a) y /(1 + s)] + γ ln G + δ ln Xo

Setting (11), a specific form of (2), equal to the outside level of utility (Uo) and solving
for p gives the specific form of (3):
pe = α{[ A(1 − a) yββ (1 − α − β)(1− α −β) G γ Xoδ ]/[r β (1 + s)(1− α −β) (1 + T )(α +β) eU o ]}1/ α . (12)
Increases in income (y), local public goods (G), and open space (Xo) boost the rental
price of residential land (pe), whereas increases in the income tax rate (a), sales tax
rate (s), property tax rate (T), and the rental price of structure (r) reduce land rents.
Substituting (12) for p in (8) also gives the equilibrium lot-size (xe):
x e = {[r β (1 + s)(1−α − β ) (1 + T )β eU o ]/[ A{(1 − a) y}(1−α ) β β (1 − α − β )(1−α − β ) Gγ Xoδ ]}1/α .

(13)
In the Cobb-Douglas case, demand for structure depends on its exogenous price (r),
but is independent of the equilibrium price of land (pe), so ke = k*, just as in (9).
With Xo acres of the community’s total land area (L) zoned for open space, the equi-
librium number of households (ne) is:
e
n = (L − X o ){[ A{(1 − a) y}(1−α ) β β (1 − α − β )(1−α − β ) Gγ Xoδ ]/[r β (1 + s)(1−α − β ) (1 + T )β eU o ]1/ α }.
(14)
Note that open space appears twice in this expression and has an ambiguous long-run
effect on population: increasing Xo reduces land for residential use (L–Xo), but the ame-
nity effect of more open space allows residents to achieve the outside level of utility (Uo)
with a smaller private lot (see [13] to verify δxe/δXo < 0), so the net effect on the number
of households (ne) is ambiguous.
This information about the long-run behavior of households can be imbedded in the
choice problem facing local officials. Again, assuming that local government selects
open space (Xo), public spending (G) and a property tax rate (T), levied on both types of
land and residential structure, to maximize aggregate land value (V)—the rental value
of residential land as well as open space—subject to a balanced-budget condition and an
imbedded constraint on total land area, the constrained optimization problem can be
expressed as:

Max v( Xo , G, T , µ ) = pe (L − Xo ) + mXo + µ{R + T[ pe (L − Xo ) + mXo + ne rk e ] − G},

(15)
where pe and ne contain the policy instruments (Xo, G, T), as shown in (12) and (14),
and R is exogenous state aid to the local government. Back-substitution of the optimal
policies (Xo*, G*, T*) into (12), (13), (14), and (15) also gives final values for the rental
price of residential land, lot size, population, and aggregate rental land value in the com-
munity. The optimal value of the Lagrange variable (μ*) measures the per-dollar impact
of state aid on the community’s aggregate land value (δV*/δR).
Even with the underlying Cobb-Douglas (or log) utility function, the first-order con-
ditions for (15) are complex and do not yield reduced-form expressions for the three
instruments. In view of our earlier discussion of proposals to stimulate local provision
of open space by increasing state aid, our primary interest is in the sign of δXo*/δR. To
examine this question, Section 3 gives some simulation results for the above model. The
simulations, as well as an econometric analysis of town land use and fiscal patterns over
multiple periods, counter the notion that steering more funds to local governments will
encourage open space zoning by easing the fiscal pressures to allow more development.
3. Simulating the Effects of State Aid on

Local Open Space
3.1 Purpose
This simulation illustrates that when the long-run responses of households and local
governments are considered, simple prescriptions to encourage preservation may not
only fail to deliver, but may even be counterproductive. One such proposal that has sur-
faced in the United States, with its division of state and local powers, is to use state aid
to relieve local fiscal pressures. The underlying premise is that an exogenous inflow of
revenue will ease the pressure on local governments to generate more property tax rev-
enue from development. If so, localities may feel freer to maintain or even expand areas
zoned for open space or lower density uses such as agriculture. Unfortunately, it is not
clear that this transfer will have the desired effect on open space if we allow for endog-
enous changes in the local property tax rate, public spending, population, land prices,
and household behavior prompted by the grant.
As noted before, despite many simplifications, the model of local fiscal and land use
zoning policies, with an imbedded open-city model of household choice, does not give
reduced-form expressions that can be directly analyzed for comparative static proper-
ties. Still, a plausibly calibrated version of the model serves to illustrate the potential
effects of various policies, including the use of higher level grants to affect local land
use and fiscal decisions. The calibration is far from exact, but relatively current data for
Connecticut and its 169 townships are used to set plausible baseline parameter values
that give “ballpark” initial outcomes. Exogenous changes in the state aid parameter (R)
then induce changes in equilibrium household behavior, property values, town popula-
tion, local land use and fiscal decisions, and aggregate land value.
3.2 Simulating the Open Space Response to State Aid

Connecticut is one of America’s geographically smaller states (48th in land area), com-
posed of 169 distinct townships that vary in area, population, and socioeconomic com-
position. Each town sets local zoning policies and relies primarily on local property
taxes and state aid to finance public services, notably education. Our goal in the simu-
lation is not to replicate this patchwork quilt of communities, but to show for a typical
town how the model can be used to examine questions related to the provision of open
space. Specifically, we focus on the long-run impacts of an exogenous increase in state
aid (R), intended to relieve fiscal pressure and thereby allow towns to limit development
and zone more land for open space. Given the “localization” of open space initiatives
over the past decade or two, this seems a useful exercise, one that highlights the need to
consider the ultimate impacts, not simply the stated intent, of such policies.
The notation, definition, units of measurement, and baseline parameter values are
listed in Table 20.1, which also gives the notation, definition, and units of measurement
for endogenous variables in the open-city household submodel and the model of local
public choice.
We are interested in the response of endogenous elements of the model, particularly
the amount of land zoned for open space (Xo), to exogenous changes in state aid (R).
Using the open-city land value maximization model outlined earlier, with its nested
submodel of household choice, and the parameter values in Table 20.1, Mathematica’s
FindRoot procedure is used to compute a baseline numerical solution, shown in the
middle row of Table 20.2. State aid (R) is then allowed to vary from its baseline value of
$17 million, upward and downward, by intervals of $1 million. The effects of R on the
local government’s choice of an optimal property tax rate (T), level of local public spend-
ing (G), and the amount of land zoned for open space (Xo) are shown in the appropriate
columns of Table 20.2. Table 20.2 also shows how other endogenous elements respond
to the changes in state aid and the induced changes in local fiscal and zoning policies.
Recall that households in this model cannot choose the amount of open space zoned
by local authorities (Xo), but they do derive utility from it. Because of such amenity
effects, local officials wanting to maximize aggregate land value (V) might see an
Table 20.1 Notation and baseline parameters

Notation Definition Units Baseline value
Parameters:
L Town land area acres 19,000
m Annual rental price of open space $/acre 500
r Annual rental price of residential structure $/sq. ft. 5
(floorspace)
y Household income $/year 84,000
Uo “Outside” level of household utility constant 12
A Utility function scalar constant 1
α Post-tax expenditure share on rental cost of fraction 1/9
residential land
β Post-tax expenditure share on rental cost of fraction 2/9
housing structure
1−α−β Post-tax expenditure share on numeraire fraction 2/3
consumption good
γ Preference weight on local public spending fraction 0.10
δ Preference weight on local land zoned for open fraction 0.15
space
s State sales tax rate fraction 0.06
a Combined (state and federal) income tax rate fraction 0.20
R State aid to the local government $/year 17,000,000
Endogenous Variables:
U Household utility
x Residential land (lot-size) acres
k Housing structure (floorspace) sq. ft.
g Numeraire consumption good (price = 1) $/year
p Annual rental price of residential land $/acre
n Local population households
T Local property tax rate (fraction of annual rental fraction
payments)
G Local public spending $/year
Xo Local land zoned for open space acres
L-Xo Local land zoned for residential use acres
V Aggregate land value (rental value of residential $/year
land and open space)
μ Lagrange variable (δV*/δR)
increase in open space as a way to boost V, particularly if an external source provides

additional revenue and thereby eases the local government’s reliance on property
tax revenue from development to finance public spending. As seen in Table 20.2, the
increase in state aid (ΔR > 0) does provide this fiscal relief, allowing town officials to
reduce the tax rate (ΔT < 0) and increase public spending (ΔG > 0). But this policy
response ultimately attracts more households (Δn > 0) and increases the equilibrium
price of residential land (Δp > 0), resulting in more land being allocated to devel-
opment (Δ(L−Xo) > 0) and less to open space (ΔXo < 0). Despite the extra land for
Table 20.2 Simulating the effects of state aid
V
R p T Xo (agg.
(state U x k g (land n (tax G (open L-Xo land μ
aid)* (utility) (lot-size) (floorspace) (numeraire) price) (households) rate) (spending)* space) (developed) value)* (dV*/dR)
7.0 12.0 1.240 2,228 42,285 4,492 6,294 0.339 44.545 11,195 7,805 40.661 2.541
8.0 12.0 1.165 2,239 42,285 4,806 6,715 0.333 47.364 11,179 7,821 43.179 2.494
9.0 12.0 1.099 2,249 42,285 5,114 7,128 0.327 50.125 11,165 7,835 45.652 2.452
10.0 12.0 1.042 2,257 42,285 5,417 7,533 0.322 52.834 11,153 7,847 48.085 2.414
11.0 12.0 0.991 2,265 42,285 5,716 7,930 0.317 55.495 11,142 7,858 50.482 2.380
12.0 12.0 0.945 2,273 42,285 6,010 8,321 0.313 58.113 11,133 7,867 52.846 2.348
13.0 12.0 0.905 2,280 42,285 6,300 8,706 0.309 60.693 11,124 7,876 55.179 2.319
14.0 12.0 0.868 2,286 42,285 6,587 9,086 0.305 63.235 11,116 7,884 57.485 2.292
15.0 12.0 0.834 2,292 42,285 6,870 9,460 0.302 65.744 11,109 7,891 59.765 2.267
16.0 12.0 0.803 2,298 42,285 7,150 9,830 0.298 68.222 11,103 7,897 62.020 2.244
Baseline: 17.0 12.0 0.775 2,303 42,285 7,428 10,195 0.295 70.671 11,097 7,903 64.254 2.223
18.0 12.0 0.749 2,308 42,285 7,703 10,557 0.293 73.093 11,091 7,909 66.466 2.202
19.0 12.0 0.725 2,313 42,285 7,975 10,914 0.290 75.489 11,086 7,914 68.659 2.183
20.0 12.0 0.703 2,318 42,285 8,245 11,268 0.287 77.861 11,081 7,919 70.833 2.165
21.0 12.0 0.682 2,322 42,285 8,513 11,618 0.285 80.211 11,077 7,923 72.989 2.148
22.0 12.0 0.663 2,327 42,285 8,779 11,965 0.282 82.538 11,073 7,927 75.129 2.132
23.0 12.0 0.644 2,331 42,285 9,043 12,309 0.280 84.846 11,069 7,931 77.253 2.117
24.0 12.0 0.627 2,334 42,285 9,304 12,650 0.278 87.134 11,065 7,935 79.363 2.102
25.0 12.0 0.611 2,338 42,285 9,564 12,989 0.276 89.404 11,062 7,938 81.458 2.088
26.0 12.0 0.590 2,342 42,285 9,823 13,325 0.274 91.656 11,058 7,942 83.539 2.075
27.0 12.0 0.582 2,345 42,285 10,080 13,658 0.272 93.891 11,055 7,945 85.608 2.062
* $ millions
development, lot size shrinks (Δx < 0), but larger structures are built (Δk > 0). Given
the underlying log (or Cobb-Douglas) utility function, numeraire consumption is
unaffected (Δg = 0), but the utility gains from more structure and greater public spend-
ing compensate for the reductions in lot size and open space to provide the larger
number of households the same (Δu = 0) “outside” level of utility (Uo) as before the
increase in state aid.
The model has a variety of limitations, but it illustrates that simply channeling revenue
from state to local governments, with the expectation that communities will rationally
zone more land for open space, could actually result in less open space when long-run
adjustments occur. Is there any evidence, though, that more state aid is associated with
less open space? In the next section, we address this question by examining a unique
multiperiod dataset for Connecticut’s 169 towns.
4. Empirical Analysis
4.1 Specification
The government choice problem, expressed in equation (15) and simulated in Section
3, implies a set of general expressions for T*, G*, Xo*, and μ* that depend on the vector
of exogenous elements (y, C, L, R, Uo, A, α, β, γ, δ, s, a, r, m). Some elements of the vector
reflect the underlying model of the representative household and are either empirically
unobservable or assumed to be roughly constant across communities. Income (y), a vec-
tor (C) of other community characteristics (suppressed in the earlier analytical model),
total land area of the town (L), and state aid (R) are observable and vary considerably
across towns and over time, so these are included in our empirical specification. But
“outside” utility (Uo) is inherently unobservable and likely to be highly correlated with
income. Parameters of the utility function (A, α, β, γ, δ) also might be income-sensitive,
but again we have no systematic information about how these parameters vary spatially
or over time, so we assume that the inclusion of income sufficiently captures latent dif-
ferences in residents’ preferences. The state’s sales tax rate (s) is uniform across towns
and has not varied much over the period of our panel data analysis; the combined state
and federal average income tax rate (a) also is assumed to be time and town invari-
ant. Residential land prices (p) vary considerably across towns and over time, but are
endogenous in our model and imbedded in the local government’s choice problem. The
price of structure (r) and the price of open space (m) are more homogeneous and also
likely to be correlated with some of the town characteristics included in C (e.g., location
within the state). With these points in mind, our empirical specification of the three
reduced-form policy expressions is:
T ∗ ( y , C , L, R) (16)

G ∗ ( y , C , L , R)
(17)
Xo∗ ( y , C , L, R)
(18)
Equations (16)–(18) represent the town’s land-value maximizing choice of a local

property tax rate (T*), level of public spending (G*), and the area of town land zoned
for open space (Xo*), each expressed as a function of local per capita income (y); a vec-
tor of three town characteristics including the town’s index crime rate (C1) as a measure
of general socioeconomic conditions, minimum distance from the town to New York
or Boston (C2), both important regional focal points on opposite sides of the state, and
a dummy variable for shoreline towns (C3) to capture an important amenity for New
Englanders; total available land area of the town (L); and intergovernmental revenue
(R), the focus of our analysis. To allow for the possibility of a delayed response to state
aid, R is empirically specified as the average (Ravg) of current (R) and one-year lagged
(Rlag) values of intergovernmental revenue.
4.2 Data
For the analysis, we use a unique dataset compiled by the Center for Land Use Education
& Research (CLEAR) in the College of Agriculture and Natural Resources at the
University of Connecticut. Using data generated from satellite imagery for five separate
years (1985, 1990, 1995, 2002, 2006), CLEAR reports 12 categories of “land cover” for
each of the state’s 169 towns: (1) developed, (2) turf and grass, (3) other grass, (4) agri-
cultural field, (5) deciduous forest, (6) coniferous forest, (7) water, (8) nonforested wet-
land, (9) forested wetland, (10) tidal wetland, (11) barren, and (12) utility corridors.
The underlying model in Section 3 focuses on the allocation of local land to “develop-
ment” and potentially usable “open space,” so we exclude several categories that are not
usable or readily subject to town control—water, various wetlands, and utility corridors
(items 7–10 and 12)—and then define the remaining categories (items 1–6 and 11) as
the total land area (L) available for either open space (Xo) or development (L – Xo). The
first two categories (developed and turf grass) are generally regarded as “developed
areas,” so the other nonexcluded categories (other grass, agricultural field, deciduous
forest, coniferous forest, and barren) are treated as “open space.” Based on these defini-
tions, Table 20.3 shows the state’s overall pattern of land cover change, aggregated from
town-level data, as well as the corresponding means of the town-level data, in each of the
five available years.
Although Connecticut is the fourth most densely populated state, the table shows that
the “developed” (L – Xo) area’s share of “total available acres” (L) rose from just 24.30%
in 1985 to 28.95% in 2006, an increase of 139,255 acres or more than 217 square miles.
Similarly, the 169-town mean value of “percent developed” rose from 28.22% to 33.06%
over the same period.
Table 20.3 Connecticut land cover, 1985–2006
1985 1990 1995 2002 2006 chg 85 to 06 % chg 85 to 06
CONNECTICUT
Open Space Xo 2,205,118 2,158,576 2,138,506 2,103,539 2,079,420 −125,698 −5.7
Developed (L-Xo) 708,030 760,465 785,394 822,607 847,285 139,255 19.7
Total Available Acres L 2,913,148 2,919,041 2,923,900 2,926,146 2,926,705 13,557 0.5
Percent Open Space % Xo 75.70 73.95 73.14 71.89 71.05 −4.65
Percent Developed %(L-Xo) 24.30 26.05 26.86 28.11 28.95 4.65
169-TOWN MEANS
Open Space Xo 13048 12773 12654 12447 12304 −744 −5.7
Developed (L-Xo) 4190 4500 4647 4867 5014 824 19.7
Total Available Acres L 17238 17272 17301 17314 17318 80 0.5
Percent Open Space %Xo 71.78 69.94 69.07 67.78 66.94 −4.84
Percent Developed % (L-Xo) 28.22 30.06 30.93 32.22 33.06 4.84
DEFINITIONS LAND COVER CATEGORIES
Open Space Xo : other grass, agricultural field, deciduous forest, coniferous forest, barren
Developed (L-Xo) : developed space, turf grass
Total Available Acres L : developed space, turf grass, other grass, agricultural field, deciduous forest, coniferous forest, barren
Excluded: water, nonforested wetland, tidal wetland, utility corridors
Source: Based on satellite imagery data from the Center for Land Use Education and Research (CLEAR), College of Agriculture and Natural Resources, University of
Connecticut, Storrs, CT.
Town-level fiscal data—effective property tax rate (equalized mill rate) and local
public spending—are drawn from reports of the Connecticut Office of Policy and
Management (OPM). Income figures are from US Census data, crime rates are from the
Connecticut Department of Emergency Services and Public Protection, and the mini-
mum driving distance from each town to either New York or Boston is generated by
Google Maps.
Both the endogenous and exogenous variables in Table 20.3 show considerable varia-
tion among the 169 Connecticut towns in each of the five periods. For example, in the
most recent year (2006), land cover identified as “open space” (Xo) ranges from 528 to
32,535 acres; government spending (G) varies from $2.03 million to $460.22 million;
and the effective property tax rate (T) ranges from $4.72 to $27.89 per $1,000 of mar-
ket value. Among the explanatory variables, annual per capita income (y) ranges from
$15,739 to $99,664, and the index crime rate per hundred persons (C1) ranges from 0.19
to 8.32. Distance to the nearest major metropolitan center (C2)—New York or Boston—
ranges from 39.4 to 131.0 miles, and the shoreline dummy variable (C3) assumes a value
of zero for 145 of Connecticut’s 169 towns and one for its 24 shoreline towns. Total avail-
able land area (L), as defined earlier in our discussion of the land cover data, varies from
3,187 to 38,136 acres. The average (Ravg) of current (R) and one-year lagged (Rlag) inter-
governmental revenue (primarily state aid in most Connecticut towns) is used to cap-
ture the effect of grants on the amount of local open space. In 2006, the two-year average
figure varies from $154,175 to $233.87 million. Despite its small size, Connecticut’s
towns show considerable diversity in land use patterns, fiscal policies, and underlying
characteristics.
4.3 Results
Three reduced-form equations—for open space acreage (Xo), total town spending (G),
and the effective property tax rate (T)—were estimated in Stata using generalized least
squares regression to correct for heteroscedasticity. Table 20.5 gives the econometric
results for the pooled five-year sample of all 169 Connecticut towns, with 1985 as the base
year and dummy variables for the other four nonconsecutive years (D90, D95, D02, D06).
The adjusted R2 for the first estimated reduced-form equation indicates that more
than 92% of the town-level variation in open space (Xo), across 169 towns and over five
discrete periods (N = 845), is explained by the joint effect of factors that closely mirror
the structure of the theoretical model. The same set of explanatory variables account for
more than 78% of the variation in local public spending (G) and about 55% of the varia-
tion in effective property tax rates (T).
Controlling for other factors, more affluent towns tend to have less open space
(δXo/δy < 0). Such towns may have more “private open space” in the form of larger resi-
dential lots, but since here the defined open space consists of the combined acreage of
grassland, agricultural fields, deciduous forest, coniferous forest, and barren areas, Xo
tends to decline with higher incomes (and the accompanying larger private lots). Higher
Table 20.4 Descriptive statistics, by year (1985, 1990, 1995, 2002, 2006) and pooled data
1985 1990 1995 2002 2006 Pooled
DEPENDENT VARIABLES
Open Space (undeveloped acres) Xo* min 652 590 570 528 528 528
max 32,750 32,661 32,664 32,614 32,535 32,750
mean 13,048 12,773 12,654 12,447 12,304 12,645
Government Spending (dollars) G* min 571,330 1,334,498 1,131,349 1,600,324 2,029,022 571,330
max 203,793,895 338,885,879 398,180,852 423,442,000 460,218,044 460,218,044
mean 18,910,771 31,488,593 38,481,369 50,782,061 61,499,178 40,232,394
Property Tax Rate ($ per $1000 T* min 6.80 5.30 8.84 7.70 4.72 4.72
market value)
max 33.40 18.90 42.97 32.73 27.89 42.97
mean 16.31 11.28 17.17 17.59 14.18 15.31
REGRESSORS
Per Capita Income (dollars) y min 9,661 11,044 12,178 14,158 15,739 9,661
max 42,697 52,063 65,344 87,519 99,664 99,664
mean 16,825 21,270 25,572 32,560 36,398 26,525
Index Crime Rate (crimes per 100 C1 min 0.53 0.90 0.68 0.13 0.19 0.13
persons)
max 13.23 16.13 12.62 8.80 8.32 16.13
mean 2.69 2.85 2.56 1.74 1.63 2.29
Min. Distance to NY or BOS (miles) C2 min 39.40 39.40 39.40 39.40 39.40 39.40
max 131.00 131.00 131.00 131.00 131.00 131.00
mean 93.72 93.72 93.72 93.72 93.72 93.72
(continued)
Table 20.4 (Continued)
1985 1990 1995 2002 2006 Pooled

Shoreline Dummy (shore = 1) C3 min 0.00 0.00 0.00 0.00 0.00 0.00
max 1.00 1.00 1.00 1.00 1.00 1.00
mean 0.14 0.14 0.14 0.14 0.14 0.14
Total Available Acres (acres) L min 3,176 3,178 3,183 3,186 3,187 3,176
max 38,027 38,083 38,107 38,121 38,136 38,136
mean 17,238 17,272 17,301 17,314 17,318 17,289
Intergovernmental Revenue R min 74,935 171,386 95,438 122,674 145,548 74,935
(dollars)
max 92,107,359 150,644,618 207,770,194 242,028,000 242,112,000 242,112,000
mean 4,648,419 8,913,625 10,796,759 14,395,113 15,534,999 10,857,783
Intergovernmental Revenue 1-year Rlag min 75,183 149,684 93,821 117,745 162,802 75,183
lagged (dollars)
max 92,231,560 124,509,272 201,758,880 243,359,822 225,628,000 243,359,822
mean 4,289,922 7,797,690 10,558,514 14,139,557 14,083,370 10,152,203
Intergovernmental Revenue 2-year Ravg min 75,059 160,535 108,734 120,210 154,175 75,059
average (dollars)
max 92,169,459 137,576,945 204,764,537 242,693,911 233,870,000 242,693,911
mean 4,485,818 8,362,935 10,703,497 14,267,335 14,823,794 10,504,993
N 169 169 169 169 169 845
per capita income also is associated with more government spending (δG/δy > 0)
and a lower effective property tax rate (δT/δy < 0). The positive effect of income on G
needs little explanation and is well documented in empirical studies. However, the
negative relationship between a community’s tax rate and its residents’ average income,
even though readily observed in the raw data before controlling for other influences, is
a common source of public confusion. Because affluent communities often pay more
property taxes per head, people assume they also face a higher rate, but this typically is
not the case. Enjoying a larger property tax base per person, higher income communi-
ties can tax their residents at a lower rate and still generate enough revenue to outspend
poorer towns.
Crime rates (C1) are negatively related to open space (δXo/δC1 < 0) and positively
related to both local spending (δG/δC1 > 0) and tax rates (δT/δC1 > 0). Like most urban
areas, Connecticut cities tend to have higher crime rates, less open space, larger budgets,
and higher tax rates than do suburban or rural towns, so these estimated coefficients, all
quite significant, are consistent.
Towns further from New York or Boston (C2) typically have more open space
(δXo/δC2 > 0) and lower public spending (δG/δC2 < 0), but also higher property tax
rates (δT/δC2 > 0) than towns closer to these regional centers. The first two results are
self-explanatory. The third effect, less significant than the first two, reflects the fact that
towns farther from one of the two regional economic centers tend to have lower prop-
erty values (the effect of regional rent gradients), which must be taxed at a higher mill
rate to support a desired level of public spending.
A similar logic applies to estimated coefficients for the shoreline dummy variable, but
due to the way it is defined (C3 = 1 for shoreline towns; 0 otherwise), the correspond-
ing signs are reversed. Popular shoreline towns tend to be densely settled with limited
open space (δXo/δC3 < 0), but have relatively high levels of public spending (δG/δC3 > 0),
perhaps needed to serve nonresident visitors as well as the “locals.” Despite these higher
public outlays, coastal amenities are capitalized into property values, allowing shoreline
tax rates to be lower (δT/δC3 < 0) than in an otherwise similar inland town. All three
effects are statistically quite significant.
Controlling for other features, geographically larger towns tend to have more open
space (δXo/δL > 0); the marginal effect is large (0.856) and highly significant, with an
elasticity of open space with respect to town area of 1.170 (using the estimated coef-
ficient and pooled sample means of 12,645 acres for Xo and 17,289 acres for L). Bigger
towns also face higher outlays (δG/δL > 0), reflecting the costlier delivery of many types
of public services (utilities, school transportation costs, fire services, etc.) to more dis-
persed populations. The property tax rate is negatively related to town area (δT/δL < 0),
but the implied elasticity is small (–0.08).
The effect of a change in intergovernmental revenue (R) on the amount of open space
is of special interest. Once again, simulations in the preceding section show how an
increase in R can decrease rather than increase the amount of land optimally allocated
to open space (Xo). Using the average (Ravg) of current and one-year lagged values of R
to allow for somewhat delayed or cumulative response of the local government to extra
Table 20.5 GLS estimates of open space, public spending, and property tax rates
(five-year panel of 169 Connecticut towns)
Open Space (Xo) Spending (G) Tax Rate (T)
REGRESSORS:
Per Capita Income y −56.5464 (-6.15) 774,236 (4.93) −0.089723 (-7.48)
(dollars)
Index Crime Rate C1 −796.639 (-8.73) 10,627,551 (4.27) 0.700663 (4.46)
(crimes per 100
persons)
Min. Distance to NY C2 32.7720 (7.37) −422,810 (-5.60) 0.009548
or BOS (miles)
Shoreline Dummy C3 −1,366.11 (−5.72) 23,565,463 (5.68) −1.66773
(shore = 1)
Total Available L 0.855556 (88.04) 630.517 (4.73) −0.000068
Acres (acres)
Intergovernmental Ravg −0.000019 (-4.13) 1.33147 (4.75) 0.000000036 (2.96)
Revenue 2-year
average (dollars)
1990 Dummy D90 97.0425 (0.45) 5,472,765 (1.64) −4.79105 (−13.73)
1995 Dummy D95 8.21837 (0.04) 9,172,473 (1.86) 1.61266 (3.89)
2002 Dummy D02 −396.281 (−1.54) 19,898,008 (2.35) 3.10217 (6.17)
2006 Dummy D06 −402.057 (−1.47) 28,097,687 (3.01) 0.094199 (0.19)
Intercept −1,349.57 (−2.23) −6,469,437 (−0.72) 16.1948 (17.57)
R2 0.927 0.790 0.556

adj R2 0.926 0.787 0.551
F 1,105.94 148.73 107.31
Root MSE 2,038.76 27,684,495 3.07
Log likelihood −7,632.45 −15,673.71 −2,140.10
N 845 845 845
revenue, we find support for the negative relationship in the simulations (δXo/δR < 0).
A million-dollar increment in Ravg is associated with a 19-acre reduction in open space.
The negative coefficient is significant at the 1% level, but the implied elasticity at sample
means is quite small (–0.016). This finding suggests that increments in unrestricted state
aid are more likely to slightly reduce, rather than increase, the amount of open space.
Not surprisingly, additional state aid is strongly associated with higher local public
spending (δG/δR > 0), and the calculated elasticity is much larger in magnitude (0.348).
Contrary to simulation results, the estimated response of the property tax rate is positive
(δT/δR > 0), but again the calculated elasticity is small (0.025).
Dummy variable estimates in the first regression generally reflect the secular loss of
open space. Evidence of an upward drift in government spending, at least in nominal
terms, is seen in the estimated dummy variable coefficients for the second regression,
whereas the pattern is less distinct in the property tax rate regression. The negative
and highly significant coefficient for D90 in the latter equation likely reflects a system-
atic reduction in effective property tax rates across many Connecticut towns in the late
1980s, when market values soared and effective tax rates fell.
The contrary empirical result for δT/δR and the weaker fit of the tax rate equation
point to a potential limitation in the present analysis. In both the underlying theoreti-
cal model and the empirical specification of reduced forms, we treat intergovernmental
revenue (R) as exogenous and unrestricted. This simplifies the analysis but ignores the
fact that, in Connecticut, and in most other states, transfers to local governments are
potentially endogenous and sensitive to socioeconomic indicators such as crime rates or
per capita income. Extending the model to incorporate a “typical” aid formula and refin-
ing the empirical analysis to allow for the potential endogeneity of R would be desirable.
The question of how to treat R and the degree to which local public expenditures are
fungible or constrained by conditional grants also raises the nagging “flypaper effect”
issue, recently discussed again by Inman (2008). For these reasons, and probably others,
our empirical results should be regarded as preliminary findings.
5. Conclusion and Research Agenda
Open space preservation programs in the United States, and perhaps in most countries,
range from large-scale national initiatives to protect wilderness areas and unique natu-
ral resources, to smaller scale state and local efforts to preserve farmland, parks, and
historic sites. Both types of programs play important social and economic roles, and
it would be unwise to wholly abandon one approach for the other, but local preserva-
tion efforts are particularly important for two key reasons. First, as noted earlier, pref-
erences for open space vary, just as they do for other public goods, and allowing local
governments to “customize” their preservation policies to local conditions and prefer-
ences may enhance efficiency. Strong advocates of “regionalism” may find this notion
unappealing, opting instead for more uniformity, consolidation, and presumed scale
economies in the acquisition or management of open space, but many public finance
economists recognize the potential benefits of a more decentralized approach. Second,
local preservation programs are important simply because they are most prevalent in
areas of relatively high density. In the United States, this means the northeastern states,
coastal regions, and suburban areas surrounding larger cities—areas that account for
a modest share of total land area but a large portion of the population. Consequently,
these are precisely the areas where market pressure to convert farmland and open space
to other uses is strongest. A large portion of the population will actively benefit from,
and therefore support, open space initiatives only if they include visible state and local
programs in higher density areas.
Given the fragmented nature of US open space preservation, the analytical and
empirical results presented in this chapter offer a caution to policy makers, but
the results should be qualified. The principal simulation results show that more
intergovernmental aid, intended to facilitate open space preservation by reducing the

fiscal incentive for local authorities to zone land for development, can produce just the
opposite: more land zoned for development and less for open space. The result initially
seems counterintuitive, but a closer look at the community-wide adjustments in the
open-city model offers a plausible explanation. More state aid allows officials to simul-
taneously increase public spending and lower the property tax rate. Both adjustments
make the community more attractive to potential entrants, and, as population expands,
land rents are bid up, average lot size shrinks, and more households are accommodated.
If officials seek to maximize the total value of land within the community, this increase
in the market price of residential land, relative to that of open space, encourages them
to zone more land for development and less for open space—not the intended outcome.
There are, as usual, variants or extensions of the model that could affect these results.
First, in a less restrictive model in which the price of open space is not constant but is
instead a decreasing function of the amount of land zoned for open space, m(Xo), the
simulated decline in open space in response to more state aid would be mitigated by an
increase in its price, perhaps not reversing our finding, but likely choking it off sooner
and damping the final effect.
A second extension of the model would relax the treatment of public goods and open
space, which here are treated as pure public goods: each resident derives the full ben-
efit of public spending (G) and local open space (Xo), independent of population (n).
Borcherding and Deacon (1972), Santerre (1985), and others have relaxed this assump-
tion to allow for “crowding” or congestion effects, in which individual consumption of
the public good increases with spending (G) but decreases with the population (n) shar-
ing the public good. Bates and Santerre (2001) use this approach in their study of the
demand for locally owned open space and find it to be quite congestible. Given that G
and n, as well as Xo, are endogenous in our model, extending it to allow for congestion in
both public goods and open space would be useful.
Third, in the scenario described in our simulations, the lot size reduction that accom-
modates more households, may be limited by minimum lot size restrictions, absent in
the present model. Incorporating both land use zoning and various forms of density
zoning (minimum lot size, maximum height, etc.) would make the model even more
applicable. Again, it is useful to think of these various zoning instruments as additional
ways for a community to differentiate itself and better satisfy residents’ preferences,
thereby enhancing aggregate land value. Such restrictions, though, raise vital questions
about the distributional impacts and exclusionary effects of zoning and other land use
policies. For a contemporary analysis of the emergence of zoning and its welfare effects,
see Calabrese, Epple, and Romano (2007).
A fourth feature of the model, its assumption of land value maximization as the pub-
lic sector’s only goal, ignores more subtle objectives, such as ecological sustainability,
wildlife protection, encouragement of local food production, and the like. Open space
affects household utility in the current model, so amenity benefits are considered, but
broadening the public goal (land value maximization) to incorporate less tangible social
returns, not fully capitalized in land values, could be beneficial.
Expanding the model to incorporate other types of agents, including farmers and
nonagricultural firms, is a fifth extension that would generalize the model and allow
researchers to address more questions. How does farmland preservation affect agricul-
tural output and food prices? Are the costs of farmland preservation partly borne by
businesses in the form of higher commercial land rents? What are the cheapest and least
distortionary ways to finance open space preservation?
Finally, another possible extension of the model warrants special mention. Much
within the present model is endogenous—household behavior, local fiscal and land
use zoning decisions, land values, and population size—but linkages between the com-
munity and other localities (other than via the open-city condition), or between the
community and higher level governments, are lacking. Households in the model face
income and sales taxes, typically levied by federal or state governments, but these link-
ages take the form of exogenous parameters. For many areas, a model that details the
interaction between the land use and fiscal policies of a higher level government and a
local government, or a network of localities subject to policy spillover effects, would be
useful in addressing policy coordination questions. Lenon, Chattopadhyay, and Heffley
(1996) find empirical evidence of zoning and fiscal interdependencies between neigh-
boring towns, and Brueckner (1998) documents similar interactions for growth con-
trols, but further understanding how a network of communities also interacts with a
higher level government to determine an equilibrium pattern of land use, fiscal mix,
and community types would be valuable. Among other applications, it could be used to
study the economic and land use impacts of statewide fees or taxes designed to finance
preservation in selected subareas. A multilevel model also might be used to evaluate
Briffault’s (1996, 1115) call for “a ‘mixed strategy’ that would both reduce the signifi-
cance of existing local boundaries and create elected regionally bounded governments
to address matters of regional significance.”
As in most areas of study, the range of interesting questions about open space pres-
ervation expands rapidly as simplifying assumptions are relaxed and new elements are
added to the analysis. The present model may be a useful foundation for such extensions.
As in the workhorse monocentric model of urban economics, land use is explicit and
rents are endogenous. The monocentric model offers more spatial detail, in that users of
land select both a lot size and location (distance from the urban center), but this model
offers a simpler foundation for some of the extensions noted earlier, without abandon-
ing the spatial dimension needed to study land use issues. Some of the suggested exten-
sions are of technical interest, but others should help to inform land use policy making.
Fortunately, open space preservation is an area of land economics in which researchers’
enthusiasm is matched by public interest.
Much of the work on open space preservation has been empirical, and, as new data
emerge and new analytical and empirical techniques are developed and applied, that
work will continue to expand our understanding of how various conservation programs
really work and the public’s willingness to support them. Another development in this
area of research and in other land use studies is the increasing availability of land cover
data from satellite imagery, over sufficiently long periods of time to reveal measurable
changes. We have used this type of data for Connecticut in the empirical section, but
similar data for other states and other countries will become more commonplace.
Marrying the data with site-specific information about zoning and other land use con-
trols is a bigger task, but one that would greatly improve our understanding of the public
sector’s capacity to steer land use.
References
America’s Great Outdoors Initiative. 2011. America’s great outdoors: A promise to future genera-
tions. Washington, DC: United States Departments of Interior, United States Department of
Agriculture, The Environmental Protection Agency.
Anas, A. 1988. Optimal preservation and pricing of natural public lands in general equilibrium.
Anas, A., and H. -J. Rhee. 2006. Curbing excess sprawl with congestion tolls and urban bound-
aries. Regional Science and Urban Economics 36: 510–541.
Bae, C. -J. 2007. Containing sprawl. In Incentives, regulations and plans, eds. G.-J. Knaap,
H. K. Haccou, K. J. Clifton, and J. Frece, 36–53. Northampton, MA: Edward Elgar.
Barrows, R., and B. Prenguber. 1975. Transfer of development rights: An analysis of a new land
use policy tool. American Journal of Agricultural Economics 57: 549–557.
Bates, L. 1993. Municipal monopoly power and the supply of residential development rights.
Eastern Economic Journal 19: 173–184.
Bates, L., and R. Santerre. 2001. The public demand for open space: the case of Connecticut
communities. Journal of Urban Economics 50: 97–111.
Bergstrom, J., B. Dillman, and J. Stoll. 1985. Public environmental amenity benefits of pri-
vate land: the case of prime agricultural land. Southern Journal of Agricultural Economics
17: 139–149.
Betz, C., and H. K. Cordell. 1998. Outdoor recreation supply in the United States: A descrip-
tion of the resources, data, and other information sources. Athens, GA: USDA Forest Service,
Southern Research Station.
Bolitzer, B., and N. Netusil. 2000. The impact of open spaces on property values in Portland,
Oregon. Journal of Environmental Management 59: 185–193.
Borcherding, T., and R. Deacon. 1972. The demand for the services of non-federal govern-
ments. American Economic Review 62: 891–901.
Briffault, R. 1996. The local government boundary problem in metropolitan areas. Stanford
Law Review 48: 1115–1171.
Brueckner, J. 1979. Property values, local public expenditure and economic efficiency. Journal
of Public Economics 11: 223–245.
Brueckner, J. 1982. A test for allocative efficiency in the local public sector. Journal of Public
Brueckner, J. 1983. Property value maximization and public sector efficiency. Journal of Urban
Brueckner, J. 1998. Testing for strategic interaction among local governments: The case of
growth controls. Journal of Urban Economics 44: 438–467.
Brueckner, J. 2000. Urban sprawl: Diagnosis and remedies. International Regional Science
Review 23: 160–171.
Brueckner, J., and R. Helsley. 2011. Sprawl and blight. Journal of Urban Economics 69: 205–213.
Calabrese, S., D. Epple, and R. Romano. 2007. On the political economy of zoning. Journal of
Public Economics 91: 25–49.
Carpenter, B., and D. Heffley. 1982. Spatial-equilibrium analysis of transferable development
rights. Journal of Urban Economics 12: 238–261.
Cheshire, P., and S. Sheppard. 2002. The welfare economics of land use planning. Journal of
Correll, M., J. Lillydahl, and L. Singell. 1978. The effects of greenbelts on residential property
values: Some findings on the political economy of open space. Land Economics 54: 207–217.
Costonis, J. 1973. Development rights transfer: An exploratory essay. The Yale Law Journal
83: 75–128.
Costonis, J. 1975. The unconstitutionality of transferable development rights. The Yale Law
Journal 84: 1101–1122.
Deaton, B., and R. Vyn. 2010. The effect of strict agricultural zoning on agricultural land val-
ues: The case of Ontario’s greenbelt. American Journal of Agricultural Economics 92: 941–955.
Do, A. Q., and G. Grudnitski. 1995. Golf courses and residential house prices: An empirical
examination. Journal of Real Estate Finance and Economics 10: 261–270.
Duke, J., and T. Ilvento. 2004. A conjoint analysis of public preferences for agricultural land
preservation. Agricultural and Resource Economics Review 33: 209–219.
Duke, J., and L. Lynch. 2006. Farmland retention techniques: property rights implications and
comparative evaluation. Land Economics 82: 189–213.
Earnhart, D. 2006. Using contingent-pricing analysis to value open space and its duration at
residential locations. Land Economics 82: 17–35.
Evans, A. 1999. The land market and government intervention. In Applied urban economics,
eds. P. Chesire and E. Mills, 1637–1669. Amsterdam: Elsevier Science, North-Holland.
Field, B., and J. Conrad. 1975. Economic issues in programs of transferable development rights.
Land Economics 51: 331–340.
Fischel, W. 1978. A property rights approach to municipal zoning. Land Economics 54: 64–81.
Fischel, W. 1980. Zoning and the exercise of monopoly power: a reevaluation. Journal of Urban
Fischel, W. 1985. The economics of zoning laws: A property rights approach to American land.
Baltimore: Johns Hopkins University Press.
Fischel, W. 1990. Introduction: Four maxims for research on land-use controls. Land Economics
66: 229–236.
Foltz, J., and B. Larson. 2002. Public support for farmland preservation programs: Empirical
evidence from Connecticut. Paper presented at the Northeast Agricultural and Resource
Economics Association (NAREA) Conference, Camp Hill, PA.
Fujita, M., P. Krugman, and A. Venables. 1999. The spatial economy: Cities, regions, and interna-
tional trade. Cambridge, MA: MIT Press.
Gardner, B. D. 1977. The economics of agricultural land preservation. American Journal of
Geoghegan, J., L. Wainger, and N. Bockstael. 1997. Spatial landscape indices in a hedonic
framework: An ecological economics analysis using GIS. Ecological Economics 23: 251–264.
Geoghegan, J., L. Lynch, and S. Bucholtz. 2003. Capitalization of open spaces into housing
values and the residential property tax revenue impacts of agricultural easement programs.
Agricultural and Resource Economics Review 32: 33–45.
Glaeser, E. 2008. Cities, agglomeration and spatial equilibrium. New York: Oxford
University Press.
Gnedenko, E. 2009. Three essays on the economics of open space. Ph.D. dissertation, University
of Connecticut.
Gottlieb, P. 2006. State-aid formulas and the local incentive to chase (or shun) ratables. Urban
Studies 43: 1087–1103.
Grieson, R., and J. White. 1981. The effects of zoning on structure and land markets. Journal of
Hamilton, B. 1975. Zoning and property taxation in a system of local governments. Urban
Studies 12: 205–211.
Hamilton, B. 1978. Zoning and the exercise of monopoly power. Journal of Urban Economics
5: 116–130.
Hascic, I., and J. Wu. 2012. The cost of land use regulation versus the value of individual exemp-
tion: Oregon’s measures 37 and 49. Contemporary Economic Policy 30: 159–214.
Heffley, D., and D. Hewitt. 1988. Land-use zoning in a local economy with optimal property
taxes and public expenditures. Journal of Real Estate Finance and Economics 1: 373–391.
Helpman, E., and D. Pines. 1977. Land and zoning in an urban economy: Further results.
American Economic Review 67: 982–986.
Henneberry, D., and R. Barrows. 1990. Capitalization of exclusive agricultural zoning into
farmland prices. Land Economics 66: 249–258.
Holcomber, R., and S. Staley. 2001. Land-use planning: An overview of the issues. In Smarter
growth: Market-Based Strategies for Land-Use Planning in the 21st century, eds. R. Holcomber
and S. Staley, 1–12. Westport, CT: Greenwood Press.
Hollis, L., and W. Fulton. 2002. Open space protection: Conservation meets growth management.
Washington, DC: The Brookings Institution.
Inman, R. 2008. The flypaper effect. NBER working paper series, WP 14579. Cambridge,
MA: National Bureau of Economic Research.
Irwin, E. 2002. The effects of open space on residential property values. Land Economics
78: 465–480.
Jiang, Y., and S. Swallow. 2006. Tax increment financing for optimal open space preservation: an
economic inquiry. Paper presented at the American Agricultural Economics Association
Annual Meeting, Long Beach, CA.
Johnston, R., and J. Duke. 2007. Willingness to pay for agricultural land preservation and policy
process attributes: Does the method matter? American Journal of Agricultural Economics
89: 1098–1115.
Just, R., and J. Miranowski. 1993. Understanding farmland price changes. American Journal of
Katz, L., and K. Rosen. 1987. The interjurisdictional effects of growth controls on housing
prices. Journal of Law and Economics 30: 149–160.
Kline, J., and D. Wichelns. 1994. Using referendum data to characterize public support for pur-
chasing development rights to farmland. Land Economics 70: 223–233.
Ladd, H. 1980. Tax policy considerations underlying preferential tax treatment of open space
and agricultural land. In Property Tax Preferences for Agricultural Land, eds. N. A. Roberts
and H. J. Brown. Montclair, NJ: Allanheld, Osman & Co., Inc.
Lenon, M., S. Chattopadhyay, and D. Heffley. 1996. Zoning and fiscal interdependencies.
Journal of Real Estate Finance and Economics 12: 221–234.
Liu, X., and L. Lynch. 2011. Do agricultural land preservation programs reduce farmland loss?
Evidence from a propensity score matching estimator. Land Economics 87: 183–201.
Loomis, J. 2000. Vertically summing public good demand curves: an empirical comparison of
economic versus political jurisdictions. Land Economics 76: 312–321.
Lopez, R., F. Shah, and M. Altobello. 1994. Amenity benefits and the optimal allocation of land.
Lutzenhiser, M., and N. Netusil. 2001. The effect of open spaces on a home’s sale price.
Contemporary Economic Policy 19: 291–298.
Mahan, B., S. Polasky, and R. Adams. 2000. Valuing urban wetlands: A property price approach.
McConnell, V., M. Walls, and E. Kopits. 2003. A market approach to land preservation. Resources
150: 15–18.
McGonagle, M., and S. Swallow. 2005. Open space and public access: A contingent choice appli-
cation to coastal preservation. Land Economics 81: 477–495.
McGuire, T., and D. Sjoquist. 2003. Urban sprawl and the finances of state and local govern-
ments. In State and local finances under pressure, ed. D. Sjoquist. London: Edward Elgar.
Miceli, T. 1992a. Optimal fiscal zoning that distorts housing consumption. Journal of Real
Estate Finance and Economics 5: 323–331.
Miceli, T. 1992b. Optimal fiscal zoning when the local government is a discriminating monopo-
list. Regional Science and Urban Economics 22: 579–596.
Mills, E. 1967. An aggregative model of resource allocation in a metropolitan area. American
Mills, D. 1980. Transferable development rights markets. Journal of Urban Economics 7: 63–74.
Muth, R. 1961. Economic change and rural-urban land conversions. Econometrica 29: 1–23.
Nelson, A. 1998. Farmland preservation policies: What works, what doesn’t and what we don’t
know. In Proceedings—The performance of state programs for farmland retention, A national
research conference, 9–48. Columbus, OH.
Nickerson, C., and L. Lynch. 2001. The effect of farmland preservation programs on farmland
prices. American Journal of Agricultural Economics 83: 341–351.
Nickerson, C., and D. Hellerstein. 2003. Protecting rural amenities through farmland preserva-
tion programs. Agricultural and Resource Economics Review 32: 129–144.
Ostrom, E. 2010. Beyond markets and states: Polycentric governance of complex economic
systems. American Economic Review 100: 641–672.
Parks, P., and J. Schorr. 1997. Sustaining open space benefits in the Northeast: An evaluation of
the Conservation Reserve Program. Journal of Environmental Economics and Management
32: 85–94.
Plantinga, A., and S. Ahn. 2002. Efficient policies for environmental protection: An econo-
metric analysis of incentives for land conversion and retention. Journal of Agricultural and
Poe, G. 1999. Maximizing the environmental benefits per dollar expended: An economic inter-
pretation and review of agricultural environmental benefits and costs. Society & Natural
Resources 12: 571–598.
Pollakowski, H., and S. Wachter. 1990. The effects of land-use constraints on housing prices.
Rose, L. 1989. Urban land supply: Natural and contrived restrictions. Journal of Urban
Ross, S., and J. Yinger. 1999. Sorting and voting: a review of the literature on urban public
finance. In Handbook of regional and urban economics. Vol. 3 of Applied urban economics,
ed. P. Chesire and E. Mills, 2001–2060. Amsterdam: Elsevier Science, North-Holland.
Santerre, R. 1985. Spatial differences in the demands for local public goods. Land Economics
61: 119–228.
Scotchmer, S. 2002. Local public goods and clubs. In Handbook of public economics. Vol. 4, ed.
A. J. Auerbach and M. Feldstein, 1998–2042. Amsterdam: Elsevier Science B.V.
Shultz, S., and D. King. 2001. The use of census data for hedonic price estimates of open-space
amenities and land use. Journal of Real Estate Finance and Economics 22: 239–252.
Smith, K., C. Poulos, and H. Kim. 2002. Treating open space as an urban amenity. Resource and
Energy Economics 24: 107–129.
Stavins, R. 2011. The problem of the commons: Still unsettled after 100 years. American
Strong, A., and R. Walsh. 2008. Communities, competition, spillovers, and open space. Land
Thorsnes, P., and G. Simons. 1999. Letting the market preserve land: The case for a market-driven
transfer of development rights program. Contemporary Economic Policy 17: 256–266.
Thorson, J. 1996. An examination of the monopoly zoning hypothesis. Land Economics
72: 43–55.
Tiebout, C. 1956. A pure theory of local expenditures. Journal of Political Economy 64: 416–424.
Wallace, N. 1988. The markets effects of zoning undeveloped land: Does zoning follow the
market? Journal of Urban Economics 23: 307–326.
Walls, M., and V. McConnell. 2007. Transfer of development rights in U.S. communities: Evaluating
program design, implementation, and outcomes. Washington, DC: Resources for the Future.
Weersink, A., S. Clark, C. Turvey, and R. Sarker. 1999. The effect of agricultural policy on farm-
land values. Land Economics 75: 425–439.
Weigher, J., and R. Zerbst. 1973. The externalities of neighborhood parks: An empirical inves-
tigation. Land Economics 49: 99–105.
White, M. 1975. Fiscal zoning in fragmented metropolitan areas. In Fiscal zoning and land use
controls, eds. E. Mills and W. Oates. Lexington, MA: Heath.
Wolfram, G. 1981. The sale of development rights and zoning in the preservation of open
space: Lindahl equilibrium and a case study. Land Economics 57: 398–413.
Wu, J., and E. Irwin. 2008. Optimal land development with endogenous environmental ameni-
ties. American Journal of Agricultural Economics 90: 232–248.
Wu, J., and H. Lin. 2010. The effect of the Conservation Reserve Program on land values. Land
Wu, J., W. Xu, and R. Alig. 2012. Optimal location and size of open space: How do they affect
urban landscapes? Unpublished manuscript.
Yinger, J. 1982. Capitalization and the theory of local public finance. Journal of Political Economy
90: 917–943.
C HA P T E R 21
L A N D C O N S E RVAT I O N
I N T H E U N I T E D S TAT E S
J E F F R EY F E R R I S A N D LOR I LY NC H
1. Land Conservation: Examples from

the United States
Land conservation provides a plethora of environmental benefits: improving ecosys-

tem function, increasing habitat to endangered or threatened species, greenhouse-gas
sequestration, and more. In addition, land conservation provides concrete economic
benefits. Conserved lands improve the productivity of farmland by reducing soil ero-
sion and water contamination, as well as by improving overall soil quality. Furthermore,
natural ecosystems and the species that populate these lands are directly valued by the
general public through activities such as outdoor recreation.
Yet, in a market system, these benefits are public goods and often underprovided from
a societal viewpoint. These public goods are difficult to value and sell within land mar-
kets because they exhibit nonexcludability (it is difficult or costly to exclude people from
using the goods and services even if they have not paid for them) and nonrivalry (one
person’s use of the goods and services does not reduce another’s use). As such, these
goods and services tend to be difficult to purchase by individuals and, even when pur-
chased, the market price underestimates individuals’ willingness to pay (WTP). In mar-
kets without coordination and/or regulations, these low prices result in underprovision;
that is, too little of the goods and services will be provided.
Underprovision of these goods can result in negative externalities on a spatial basis.
For example, owners on adjacent parcels would prefer farmland and/or a conserva-
tion use on a neighboring parcel and have a positive WTP to keep it from converting.
However, the landowner of the parcel in question does not observe this uncoordinated
WTP for conservation. Rather, the landowner observes the developers offer to pur-
chase the land to convert it to a residential use. Similarly, people may value having an old
548 Jeffrey Ferris and Lori Lynch
growth redwood forest or a rainforest remain intact even if they never intend to visit or
see it, but uncoordinated markets make this difficult to fulfill through a market mecha-
nism. Beyond the difficulties of market provision, the US government has pursued many
policy interventions, such as transportation policies, educational policies, banking regu-
lations, and crime prevention or lack there of, which all affect the individual’s optimal
land use choices. By affecting development patterns, these policies may also contribute
to suboptimal levels of land conservation and the provision of ecosystem services.
Because private landowners often make decisions in their private best interest rather
than to optimize society’s welfare, governments as well as private nongovernmental
organizations (NGOs) operate to address the resulting market failures and conserve
land. Their conservation programs often seek to redirect private and public decisions
to ensure society’s welfare is considered. Several different types of policies can be used
to aid in this redirection: regulatory, incentive-based, participatory, and hybrid policies
of these three types (categorization from Duke and Lynch 2006). Regulatory techniques
adjust existing markets or define new markets such that society’s benefits from conserva-
tion are addressed. Incentive-based techniques adjust price in the existing market struc-
ture to encourage certain conservation practices or land uses through taxes (penalties)
or subsidies (Conservation Reserve Program [CRP] rental payments). Governments,
land trusts, and other nongovernmental entities use participatory techniques—buying
or selling land parcels or land rights to redirect market activities to desired conservation
goals. In some cases, these techniques are utilized together to accomplish conservation
goals. Recently, US Department of Agriculture (USDA) programs have focused on con-
servation practices on working land that are designed to improve the environmental
performance of cropped agricultural land. These are in addition to land retirement pro-
grams, which completely remove land from agricultural use.
On a governmental level, land conservation is promoted through use of participatory,
incentive-based, and regulatory initiatives with a variety of outcomes. In this chapter,
we provide an overview of contemporary programs for land conservation in the United
States. This chapter includes a discussion of the diversity of objectives in land conserva-
tion, recent research, and experiences from these programs, as well as the future of land
conservation in the United States. We first explore the types of conservation and the dif-
ferent objectives that are pursued. These include participatory conservation through the
fee-simple purchase of land, as well as through land use regulations and incentive-based
programs for both land retirement and working lands. Recent shifts, such as conserva-
tion on working lands due to economic development concerns of local communities,
are considered. Private efforts at land conservation are also explored. Some of these
employ a hybrid of techniques. For example, tax deductions are used as an incentive to
motivate participatory-type easement donations to land trusts and governmental orga-
nizations. Financing of conservation efforts is explained. We then outline two different
methods of accounting for the benefits of land conservation: ecosystem services and
societal economic values for these services. We explore the most recent literature on
land conservation evaluations to determine what research can tell us about the existing
policies and approaches. Important questions include the effectiveness of the policies in
Land Conservation in the United States 549
terms of benefits provided, prevention of development and/or other unintended conse-

quences from the conservation, impact on land values, and spatial patterns invoked. We
then draw some conclusions.
2. Types and Objectives of Land

Conservation
Land conservation is pursued by many government agencies and individuals with a

range of desired objectives. At the public level, various federal, state, and local entities
promote land conservation through a series of interconnected programs and initiatives.
These programs range in size as well as scope and seek a diversity of conservation goals.
Here, an overview of current programs related to land conservation is provided. These
programs are broken out into five main categories (applying the Duke and Lynch 2006
classification): publicly protected land conservation (participatory), regulatory require-
ments, conservation easements (participatory), working land programs (incentive
based), and private initiatives for land conservation (participatory and hybrid).
Conservation was not the foremost thought of the United States in its formative years.
Following the founding of the nation, public policy focused on the conversion of forest
and prairies into agricultural uses and, to a lesser extent, to urban uses rather than conser-
vation. Beginning with the Homestead Act of 1862 until 1934, the government transferred
10% of the land within the US to private ownership. To obtain their deed, landowners had
to demonstrate “improvement to the land” i.e., conversion to a productive use. In addi-
tion, land reclamation converted wetlands and deltas into high-quality agricultural lands.
In Colonial America, the 48 contiguous United States contained an estimated 221 million
acres of wetlands (Dahl 1990). Over a period of 200 years, the lower 48 states lost more
than half (53%) of their original wetlands (Dahl 1990). Over time, individuals and then
government entities began to realize that conservation of forest, prairies, wetlands, and
high-quality soils may be beneficial to social welfare. Efforts toward conservation moved
slowly in many cases but strengthened in the early 20th century.
3. Participatory Conservation
3.1 Publicly Protected Land

Perhaps the most visible publicly sponsored land conservation in the United States is
the network of federal, state, and local protected lands that dot the US landscape. In
1864, advocates convinced the US government to designate Yosemite Valley a publicly
protected area. Yellowstone Park soon followed, withdrawing more than 2 million acres
from private hands to preserve forest, wildlife, and minerals from human use. But it was
not until 1916 that Congress created a federal bureau, the National Park Service (NPS),
to manage lands designated under the concept of large-scale natural preservation of
areas for public enjoyment.
At the federal level, protected areas are divided into three main categories: wilder-
ness areas, national parks and national forests.1 Across the United States, there are a
total of nearly 290 million acres2 of federally protected land classified as wilderness area,
national park, or national forest. These lands exist in all 50 US states, plus the District of
Columbia, and represent approximately 17% of the total US land area. At the state level,
more than 6,600 state parks cover approximately 14 million acres of land in the United
States (Walls et al. 2009). No central database exists for information regarding local
parks, but based on an annual survey conducted by the Trust for Public Land (2011),
there are at least 22,493 individual parks covering an area of 1.5 million acres in the
United States in 100 cities. More than half of these received 1 million visitors per year.
Publicly protected lands provide the best source of permanent and dedicated lands
for ecosystems to flourish. These lands also provide a dual function as primary locations
for outdoor recreation in the United States. Although publicly protected lands may be
acquired through a variety of regulatory3 and voluntary4 methods, the vast majority of
public resources are spent in maintaining and improving existing protected land. The
authorized operating budget of the US National Parks Service alone was nearly $2.6 bil-
lion in 2012 (US Department of Interior, 2012).
3.2 Private Initiatives

Beyond government efforts at land conservation, many private NGOs pursue conserva-
tion objectives. Many of these are land trusts, which are nonprofit organizations that
operate to conserve land through conservation easements or direct fee-simple acquisi-
tion. People organize them at the local, state, and national level. In 2010, 1,723 land trusts
were active; 1,699 at the state and local level and 24 at the national level. Together, they
conserved 47 million acres, an increase of 23 million acres in the past decade (Chang
2011). Because land trusts can act more quickly than governmental agencies, they often
purchase land or the development rights that they can then convey to a governmental
agency. Although all regions of the United States have land trusts, those in the Northeast
1
Wilderness areas are afforded the highest level of protection and are designed to be completely
untrammeled by humans. National parks are managed for ecosystem function, but also allow
recreational access, and national forests are afforded the least protection and are managed primarily for
sustainability.
2
Wilderness areas encompass 109.7 million acres, national parks 79.7 million acres, and national
forests 192.9 million acres
3 Lands may be procured by the government through the right of eminent domain.
4 Some governments also receive protected lands by donation or outright purchase.
have preserved almost twice as much land as land trusts in other regions. The 2010
National Land Trust Census Report suggests that almost half of the land trusts have now
written project selection criteria to guide their selection of parcels, with important natu-
ral areas and wildlife habitat being the most important priorities (Chang 2011).
4. Regulatory Conservation Land Use

Restrictions
Society seeks to alter the pace and/or pattern of land use for various reasons and, in
some cases, has used regulatory mechanisms. Regulatory conservation methods are fre-
quently utilized when either conservation objectives require a high degree of coordina-
tion or private citizens cannot be incentivized to preserve land of their own accord.
Some of the most readily used options for promoting regulatory conservation at the
state and local level is through low-density or agricultural zoning and development
restrictions. These measures include such policies as minimum lot size or clustering
zoning restrictions, urban growth boundaries (UGBs), and adequate public facility
ordinances (APFOs) (Irwin et al. 2005). These policies can be implemented simultane-
ously. Experience with conservation zoning has been mixed; while some communities
aggressively pursue conservation zoning, others ignore it all together, which creates a
patchy network of conservation zones for wildlife to utilize. Minimum lot zoning has
also been criticized for encouraging the consumption of larger land parcels rather than
for creating conservation zones.
At the federal level, the Endangered Species Act (ESA)5 uses recovery plans that
include regulatory measures to conserve species considered endangered or threatened.
After a species receives ESA protection, the recovery plan is developed to prohibit the
harvesting and hunting of the species and to establish a “critical habitat zone” with a
habitat conservation plan. A critical habitat zone is a geographic area that is deemed
necessary for the species’ survival and may exist on either public or private lands. If a
critical habitat zone intersects private land, development of this land is significantly con-
strained by regulation and requires an extensive permitting process for any additional
construction, thus limiting human use. However, substantial political and bureaucratic
hurdles must be surmounted before a species may be guaranteed protection by the ESA,
thus limiting its application to large-scale conservation efforts.
Most recently, the Environmental Protection Agency (EPA) has been work-
ing to implement a regulatory scheme to enforce the Total Maximum Daily Loads
(TMDL) requirements in cases in which water pollution problems endure. Although
a cap-and-trade system has been proposed, the TMDL regulations require that all land
5 http://www.epa.gov/lawsregs/laws/esa.html
users must reduce their nutrient pollution, in contrast with the USDA’s incentive-based
and largely voluntary programs.
5. Incentive-Based Conservation
5.1 Voluntary Land Retirement Programs

Voluntary incentive-based initiatives are another tool used for conservation—most
often individuals are paid a subsidy to adopt a conservation practice on their land. In the
early 1990s, about one-fifth of the US land area (382 million acres) was used for crop pro-
duction and one-quarter of privately owned land for grazing land for livestock (525 mil-
lion acres) (USDA National Resources Conservation Service [NRCS] 1996). The USDA
Economic Research Service reported that 2007 cropland had increased to 408 million
acres, with grassland pasture and rangeland at 614 million acres (both public and pri-
vately owned). Urban land occupied only 61 million acres (3%) (Nickerson et al. 2011).
Given the large number of acres in agricultural landowners’ hands, conservation efforts
may be well directed to them. It was not until 1936 that the USDA established the first
programs to pay farmers to use soil conserving practices. In part, these programs sought
to support farm incomes by reducing surplus grain supplies and increasing commodity
prices. Soil conservation programs were justified by “on farm” benefits like enhancing
crop yields and preserving crop productivity. It was not until 50 years later that the 1985
Farm Bill radically altered the conservation agenda. In addition to addressing conserva-
tion in its own right, it expanded soil erosion control concerns, used “swampbuster”
provisions to reverse USDA policies on draining wetlands, and employed land retire-
ment policies to remove lands with most pressing erosion problems from production.
In addition, the cross-compliance provision specifically linked commodity payments to
conservation practices for farmers. Currently, the USDA has 20 conservation programs
that can be split into five broad categories: land retirement programs (around 48% of
the 2010 budget was spent on these); working land programs that provide cost-share
or payments for conservation assistance (with 30% of the 2010 budget); conservation
technical assistance (with 17.5% of the total); agricultural land preservation and rural
development programs (with around 3.4%); and watershed structural activities like
flood prevention work (with 1.4% of the total) (Pavelis et al. 2011).
In the United States, the CRP is by far the largest program for voluntary
incentive-based conservation contracts between the government and willing land-
owners. Under CRP, farmers are paid an annual rent to remove land from agricul-
tural production for a predetermined or occasionally indefinite period and adopt a
conservation practice. Farmers often enroll low-productivity lands, which may be
infrequently cropped to begin with, in hopes that government subsidies for conser-
vation exceed the expected profit from agricultural production. In this way, farm-
ers benefit by receiving fixed and guaranteed revenue from agricultural lands, and
society benefits6 from the ecological services provided and the improvements on these
conserved lands.
CRP is administered by the US Farm Service Agency (FSA) and is broken up into
a number of smaller programs.7 “Annual signup” is the largest constituent program,8
and it allows farmers to enroll land under contracts of 10–15 years in exchange for a
predetermined annual payment. The other subprograms work similarly but focus on
preserving high-priority conservation lands (Ferris and Siikamäki 2009). In 2010, CRP
enrolled a total of 31.3 million acres, land area approximately equivalent to the size of
the state of Mississippi, and made annual payments in excess of $1.6 billion.
In the 1700s, the 48 contiguous United States contained an estimated 221 million
acres of wetlands (but lost 53% of them over the next 200 years; Dahl 1990). To slow
the loss of wetlands on farms, the USDA also initiated the Wetland Reserve Program
(WRP)9 administered by the Natural Resource Conservation Service. In the WRP,
the government purchases long-term or perpetual easements to restore, protect, and
enhance wetlands that have been in agricultural production. The WRP enrolls only a
fraction of the land10 that is enrolled in CRP, but has also steadily grown in size in recent
history, unlike the CRP.
5.2 Working Land Programs

The USDA, in cooperation with other government agencies, generates additional con-
servation benefits through use of various working land programs. Working land pro-
grams help to promote sustainability and engage farmers hesitant to adopt long-term
land retirement. Under these programs, farmers are allowed to continue cropping agri-
cultural lands while implementing conservation practices that generate substantial eco-
logical benefits.
The USDA Natural Resources Conservation Service (NRCS) is responsible for man-
aging a majority of the working land programs in the United States. The NRCS man-
ages a network of “financial assistance programs”;11 among the largest and best studied
are the Environmental Quality Incentives Program (EQIP) and the Conservation
Stewardship Program (CSP). These programs promote a range of conservation goals
6
Farmers may also derive environmental benefit from conservation easements, in the form of
improved agricultural productivity of surrounding lands.
7
Formally, CRP is composed of five subprograms with similar objectives: general signup, continuous
signup, Conservation Reserve Enhancement Program (CREP), Farmable Wetland Program (FWP), and
the Emergency Forest Conservation Reserve Program (EFCRP).
8
More than 26.6 million acres are enrolled, and annual payments are in excess of $1.1 billion in 2010
(US Farm Service Agency 2011).
9
http://www.nrcs.usda.gov/wps/portal/nrcs/main/national/programs/easements
10 Slightly more than 200,000 acres were enrolled in 2011.
11 A full list is available at http://www.nrcs.usda.gov/wps/portal/nrcs/main/national/programs/
financial
through education, cost-sharing opportunities, and access to sustainable infrastruc-

ture development. Smaller programs include the Grassland Reserve Program (GRP),
which is a voluntary conservation program that enhances plant and animal biodiver-
sity on working grazing operations. Landowners agree to restrict future development
and cropping uses of the land but can graze the land outside of the nesting seasons of
declining bird species. The Farm and Ranch Land Protection Program (FRPP) works
with existing farmland preservation programs by providing matching funds to help
purchase development rights to retain farm and ranchland as working lands. Similarly,
the Healthy Forests Reserve Program (HFRP) helps landowners restore and protect
forests serving as habitat for endangered and threatened species, to improve diversity,
and for carbon sequestration, either permanently or via 30-year contracts. The NRCS
working land programs have the additional advantage of promoting conservation of
high-quality, highly productive agricultural land. In this way, program administrators
can help promote targeted and focused conservation goals when landowners wish to
continue farming.
6. Land Conservation Financing
6.1 Tax Revenue Finances both Participatory and

Incentive-Based Conservation
Many conservation programs on the federal and state level are financed through gen-
eral income tax revenue. Other conservation has received public support through
ballot-approved financings. Counties or states sell bonds and incur debt to finance land
conservation today. These bonds are then repaid through local or state tax income over a
20- to 30-year period. People also have approved new taxes to pay for conservation. For
example, communities may agree to a ½% increase in their sales tax or a 1% increase on
their real estate transaction tax to finance the land conservation efforts. This money can
be used for fee-simple land acquisition, conservation easement purchase, or to provide
incentives to landowners to adopt conservation practices. Figure 21.1 demonstrates the
steady increase in conservation funding initiatives until 2008, when more than $8 bil-
lion in conservation funding was approved through the ballot box. Since the economic
downturn in 2008, this funding has declined.
Some programs benefit from dedicated funds. For example, if the general public has
approved new taxes to pay for conservation, this extra sales tax or real estate transaction
tax must be used for conservation purposes. Dedicated funds are not always sufficient to
accomplish a program’s goals. And if these dedicated conservation funds are connected
to conversion of farmland, as is the case with Maryland’s agricultural transfer tax, then
conservation becomes linked to farmland exiting the industry. For example, Lynch
et al. (2007) found that in Carroll County, the conversion of $60,051 worth of farm-
land—almost 12 acres—would be needed to conserve 1 acre of farmland in the county.
$9,000,000,000
$8,000,000,000
$7,000,000,000
$6,000,000,000
$5,000,000,000
$4,000,000,000
$3,000,000,000
$2,000,000,000
$1,000,000,000
$0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
FIGURE 21.1 Conservation funds approved.
Similarly, in Baltimore County, the conversion of $76,352 or 9.6 acres would be needed
to preserve 1 acre (Lynch et al. 2007). Thus, the funding and the goals of the program can
work at cross purposes.
6.2 Tax Deductions Act as Incentive-Based Tools for

Participatory Conservation
Both governmental programs and NGOs benefit from a 1976 special exception to the rule
against deduction of partial interests in property. At the federal level, due to this exception,
individual(s) who donates a conservation easement on all or part of their land are entitled
to an income tax reduction (charitable deduction) equivalent to the fair market value of the
donation,12 subject to certain eligibility requirements,13 even though they retain ownership
of the land. The tax code allows landowners to deduct 30% of their adjusted gross income
each year over a six-year period. Depending on the existing land conservation status and
percentage of income derived from agriculture by the landowner, an individual may be eli
gible for additional income tax deductions (Internal Revenue Service [IRS] 2012). In the tax
12
The difference between the value of their land before easement and the value of the land after
easement.
13 To be eligible for donation, the easement must be in perpetuity, be held by a valid nonprofit or
government agency, and serve an approved conservation purpose. All further eligibility requirements are
listed at http://www.irs.gov/businesses/small/article/0,,id=249135,00.html
years 2006–2011, these incentive-based tools were enhanced to allow landowners to take
deductions of up to 50% of their adjusted gross income and expanded the carry-forward
period to 15 years after the year of the first deduction. These enhancements allowed more
individuals to utilize the deduction fully and appears to have encouraged additional con-
servation easement donations. The enhancement was allowed to expire at the end of 2011
but has recently been extended through 2013. Thus unless Congress extends the provisions
again, this enhanced incentive will expire December 31, 2013.
In addition to the federal deduction, as of 2012, at least 16 states14 provide some
form of income tax credits for land donation or conservation easement (Land Trust
Alliance 2012). For example, Maryland landowners receive a $5,000 credit on their state
income taxes when they donate their development rights, as well as state property tax
relief. Colorado, Georgia, New Mexico, South Carolina, and Virginia grant transferable
tax credits, which allow rural landowners to sell these credits on the open market to
high-income individuals, which may be attractive to rural landowners with relatively
modest incomes and low tax brackets.
7. Conservation Benefits: Accounting and

Economic Value
Land conservation generates benefits in two primary ways: ecosystem benefits from
land preservation (“conservation benefit accounting”) and societal benefits from pres-
ervation (the “economic value of conservation”). Although interrelated, researchers and
policy makers distinguish between these two forms of conservation measurement. In
the following section, an overview of policy and recent research is provided for both
forms of conservation benefits in the context of prominent programs for land conserva-
tion in the United States.
7.1 Conservation Benefit Accounting

Conservation benefit accounting is an area of research that has undergone a rapid evolu-
tion in recent years. Owing to the complexities of ecosystem and wildlife dynamics, the
task of determining the ecological benefits resulting from a particular land conserva-
tion program used to be insurmountable. However, as understanding of these complex
ecosystem processes have improved, and as modeling efforts gain more computational
14 These states include Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia,
Iowa, Maryland, Massachusetts, Mississippi, New Mexico, New York, North Carolina, South Carolina,
and Virginia
power, contemporary models of ecosystem benefits resulting from land conservation

provide us with more information.
In part because of the additional information, policy makers have changed the met-
rics by which they report ecosystem benefits and program successes and the design
elements of conservation policy. Previously, the two most important metrics in con-
servation benefit accounting were program budget and total enrolled acres. However,
ecosystem benefits accrue in different ways depending on the program design. Thus, an
acre of land enrolled in a working land program, such as EQIP, is not equivalent to an
acre enrolled in WRP, which is also different from an acre of national parks land. As a
matter of procedure, government conservation programs increasingly and prominently
report ecological benefits as a measure of programmatic success.
This focus on ecological benefits has an impact on conservation effectiveness as well.
Given a specific conservation goal, such as reducing soil erosion, policy makers have an
improved capacity to design programs that address this ecological issue. As a result, pol-
icy makers have begun to shift more conservation resources away from fee-simple land
purchases toward targeted easement procurement and targeted working land programs.
In a case study in Maryland, Messer (2006) found that the emphasis on fee-simple
acquisition of highly ranked (and often expensive) land conservation overlooked the
higher overall returns of purchasing lower ranked, lower cost land, as well as the benefits
of using easements. Similarly, program designs that take into account threshold impacts
of conservation function more effectively. Lynch and Liu (2007) found that the tar-
geted preservation areas set by the Maryland Rural Legacy program attract additional
preservation to these areas, thus achieving higher degrees of contiguity than would
otherwise occur.
7.2 Economic Value of Conservation

In addition to providing important ecosystem benefits, conserved land also generates
value to human society. Economic benefits from conservation may accrue in many
ways, including through direct valuation of ecosystem services on these lands and indi-
rect improvement in agricultural production, as well as through the value of outdoor
recreation. Economists estimate the direct and indirect value of conservation to society
and also seek to understand the dynamic response of large landowners, like farmers and
other members of the public, to land conservation programs.
Conservation programs generate benefits in a number of ways, and economists help
policy makers understand the value such programs add to society. Because many pro-
grams set multiple goals, researchers must consider not only the primary benefit but
also the secondary benefits of conservation in their analysis, including unintended con-
sequences. As an example, the CRP was originally established to help reduce farmland
soil erosion, but, as a secondary benefit, it also provided significant income support to
farmers with relatively marginal cropland. CRP has also realized numerous other con-
servation objectives, such as habitat restoration and additional space for recreation, as
well as carbon sequestration, among others. The conservation value of the CRP pro-
gram, then, is its direct benefit of primary conservation objectives, its secondary ben-
efits to farmers, and its tertiary benefits to other conservation objectives.
Researchers have also improved their understanding of the dynamic interaction
between human behavior and conservation outcomes. Even programs designed with the
best intent can often lead to undesired consequences. For instance, the Endangered Species
Act (ESA) prohibits the destruction of critical habitats of protected species. However, this
Act may create incentives for landowners to speed up the extraction of resources from
lands that they suspect might fall under ESA protection (Lueck and Michaels 2003). This
“shoot, shovel, and shut up” behavior, also known as the 3-S treatment, may result in spe-
cies and habitat destruction rather than species and habitat retention.
With dwindling resources for conservation, government agencies are forced to pro-
vide more societal benefits with less money. Increasingly, these government agencies are
relying on researchers to help target resources that yield the greatest amount of benefit
for the least amount of money. Experiences with program evaluations have been diverse.
Government agencies have been gradually shifting away from broad national initiatives
designed to protect the greatest amount of acres but that may not fit all locations and
toward more locally defined initiatives in which local conditions can be best addressed.
A more localized approach to conservation may increase the effectiveness of achiev-
ing specific conservation objectives, as well as engage the local population to achieve
important community goals. However, this may come at the cost of missing low-cost,
high-benefit conservation opportunities.
8. Participatory: Publicly
Protected Lands
In many ways, assessing the ecological benefits from permanently protected lands is
quite difficult. Although important for habitat maintenance and ecosystem services,
much of this land has been protected for years. Some of these lands have never been
developed or converted into nonconservation uses. As a result, researchers find it dif-
ficult to determine the counterfactual (i.e., “How would this land’s use differ if not
publicly protected?”). As a result, few assessments of the economic value of ecosystem
services provided by protected land exist. However, a substantial volume of research
seeks to evaluate the recreational value of protected areas.
Most research on the ecological impacts of protected areas has focused on the poten-
tial threat that residential development, particularly along the edges of the protected
areas, poses (Table 21.1). Gude et al. (2007) studied the potential for exurban develop-
ment in the neighborhood of Yellowstone National Park to undermine local habitats
and biodiversity. On a national scale, Radeloff et al. (2010) and Wade et al. (2010, 2011)
performed a multisystem assessment of threats to protected land from developmental
Table 21.1 Assessing threats to protected land from exurban development

Study Study program Focus Study area Findings
Gude et al. National Parks Rural development Yellowstone Measured biodiversity

(2007) encroaching on National Park responses are likely to
critical habitats in undergo between 5%
publicly protected and 40% conversion
land due to exurban
development.
Radeloff et al. Federally Housing growth near Nationwide Housing growth rates
(2010) protected land publicly protected were 7% higher on
land lands within 1 km
of protected areas
than the national
average during the
1990s. Potentially,
another 17 million
housing units will be
built within 50 km of
protected areas by
2030 (1 million within
1 km).
Wade and Federally Housing growth near Nationwide Conservation buffer
Theobald protected land publicly protected zones around publicly
(2010) land protected areas will
have decreased by 12%
from 1970 to 2030 as a
result of development.
Wade et al. Federally Assessing threats to Nationwide At least 35% of
(2011) protected land US protected areas currently protected
land is at risk from
external development.
Only 20% of currently
unprotected lands
provide opportunities
for future
conservation.
encroachment on habitat buffer zones. Radeloff et al. (2010) found that housing growth
rates were higher near protected lands. Wade and Theobald (2010) also found that buf-
fers around the parks were decreasing quickly. Similarly, Wade et al. (2011) find that
fully 35% of all protected land is threatened by external development, whereas only 20%
of unprotected land could be conserved (Table 21.1). Collectively, findings by these
researchers suggest that policy makers need to be proactive in ensuring that future
development around existing protected areas does not undermine the conservation
objectives of these preserved lands.
Table 21.2 Value of parks as an urban amenity
Crompton (2007a) Mostly urban parks Assessment of impacts of parks on Aggregations of many Park-adjacent homes generate
the tax-receipts from neighborhood studies and locations premiums of up to 20% in increased
homes property value. When evaluated in
aggregate, increased tax revenues
from proximate homes may be enough
to meet costs of developing and
maintaining parks.
Crompton (2007b) Mostly urban parks Evaluating the effect of parks on Aggregations of many After reaching a certain income
attracting companies, labor supplies, studies and areas threshold, improvements in quality
and retirees to cities of life become more important than
increases in income. Parks are key to
improving lifestyle and attracting highly
skilled employees and companies.
Ham et al. (2012) National park Ascertaining the proximate value of Pike National Forest A 1% decrease in mean distance to the
national forests to home-owners Pike National Forest increases house
prices by 6.4%.
Klaiber and Phaneuf City parks Modeling the hedonic value of parks Minneapolis-Saint Paul A 12.5% increase in regional park area
(2009, 2010) under the assumption of population Metropolitan Area, would result in a hedonic willingness to
heterogeneity and sorting Minnesota pay (WTP) per household of $30. When
sorting and heterogeneity are taken
into account, WTP is between $5.19 and
$9.36.
Poudyal et al. (2009) City parks Performed a two-stage hedonic Roanoke, Virginia A 20% increase in park size increased
assessment of the value of urban per-household consumer surplus
parks based on local home prices by more than $160. Total consumer
surplus was in excess of $6.5 million
for all 50,000 homes located within a
mile of each park.
Table 21.3 Recreational value of publicly protected lands

Baerenklau et al. National Forest Developing a Southern Annual values range

(2010) Land method to analyze California from $41 to $10,369
the recreational per hectare of forest
value of forest land, depending on
cover location and other
characteristics of
these lands.
Deisenroth, A variety of Estimating the Larimer County, Mean per-person
Loomis, and Bond protected land nonmarket Colorado consumer surplus is
(2009) classes valuation estimated to be $78
of off-road per day. Every summer,
recreation each trail is expected
to provide $219,467–
$296,876, and
off-roading provides
$796,447-$1,077,367
to the entire county.
Siikamäki (2011) State Park System Valuing recreation Nationwide State parks generate
generated at state approximately 2.2
parks around the billion hours of
nation based on nature recreation
the American Time per year. In total, the
Use Survey value of recreation
generated at all state
parks is $14 billion.
Parkland is a valuable community resource impacting the health and well-being of

many urban communities (Table 21.2). Home buyers seek proximity to parkland and
are willing to pay, with such property values being upward of 20% higher than those far-
ther away (Crompton 2007a). Using hedonic models, as well as other WTP approaches,
other researchers have consistently shown a positive impact of park proximity on
home-value, including Klaiber and Phaneuf (2009), Poudyal et al. (2009) and Ham et al.
(2012). When evaluated at the social level, parkland has the highest valued land classi-
fications per acre. Parkland has higher value as open space than as residential develop-
ment. Crompton (2007b) argues that parks are also crucial to a community’s ability to
attract companies, new labor, and retirees. All these bolster a community’s tax base and
future growth prospects, thus stimulating long-term economic growth.
Local, state, and national parks are among the premier destinations in the United
States for outdoor recreation. As such, research has been conducted to assess the value
of these assets from a recreational standpoint (Table 21.3)15 employing contingent WTP
15 Walls, Darley, and Siikamäki (2009) provide a breakdown of multisystem benefits resulting from
US city, state, and national parks.
surveys as well as revealed preference studies. In recent work, Baerenklau et al. (2010)
and Deisenroth, Loomis, and Bond (2009) studied the recreational value of feder-
ally protected lands, finding ranges from $41 to almost $300,000 per hectare of forest
land and trails within the various parks. At the state level, Siikamäki (2011) studied the
value of state parks based on aggregate time-use trends from the American Time Use
Survey taken from individuals sampled16 across the United States. This research found
that about 2.2 billion hours of nature recreation occurred at state parks each year. These
studies demonstrate the high value of protected lands, providing millions of dollars each
year in consumer surplus across the United States. From a public policy perspective, the
use-value of protected land can be an overlooked component of conservation benefits.
For many, their use of these parks for outdoor recreation explains a large part of their
WTP for such areas.
9. Conservation Reserve Program and

Wetland Reserve Program
Together, CRP and WRP have been among the most studied land conservation pro-
grams in the United States. Owing to their geographic expansiveness and the availability
of data, researchers continue to improve their understanding of the benefits result-
ing from these programs. Although CRP and WRP have slightly differing conserva-
tion focuses, with CRP focusing on agricultural land and WRP focusing on wetlands,
their similar organizational structure has led many researchers to group these two pro-
grams together. Research has evaluated the ecological benefits, estimated the economic
value of these programs, and explored how farmers behave in response to CRP/WRP
parameters.
Most researchers interested in the ecological benefits of CRP/WRP have conducted
regional assessments of program benefits and environmental performance (Table
21.4). In a series of papers, Gleason et al. (2008, 2011) estimated the quantity of eco-
system services (decreased erosion, carbon sequestration, increased bird popula-
tions) provided by CRP and WRP in the Prairie Pothole Region17 of the United States.18
Wentworth, Brittingham, and Wilson (2010) utilized data from Pennsylvania farms
to assess Conservation Reserve Enhancement Program (CREP) effect on grassland
16
Individuals were sampled using the American Time Use Survey (ATUS), which offers detailed
descriptions of the daily activities elicited from a rotating sample of individuals.
17
A loosely defined region of the US Upper Midwest that includes areas of North Dakota, Minnesota,
South Dakota, and Iowa, critically important for migratory bird nesting.
18 They find that CRP/WRP conservation in these areas significantly decreases erosion
(approximately 2.6 metric tons per acre enrolled) and carbon sequestration (1.05 metric tons per acre),
in addition to significantly increasing migratory bird population, such as ducks.
Table 21.4 Ecosystem services provided by Conservation Reserve Program

(CRP)/Wetland Reserve Program (WRP) lands
Gleason et al. CRP/WRP Estimating the The Prairie CRP/WRP

(2008, 2011) ecosystem services Pothole Region conservation in these
provided by CRP/ (PPR) areas significantly
WRP lands decreases erosion
(~2.6 metric tons
per acre enrolled)
and increases carbon
sequestration (1.05
metric tons per
acre). CRP/WRP also
significantly increases
migratory bird
population, such as
ducks.
Gallant et al. CRP/WRP Modeling historical Iowa Researchers
(2011) land use patterns determined that most
in the context of ecosystem losses
current CRP/WRP in Iowa occurred in
enrollment areas with a high
concentration of
wetlands, whereas
CRP and WRP wetland
improvements
occurred in areas
with historically few
wetlands.
Wentworth et al. CREP Evaluating the Pennsylvania These researchers
(2010) effectiveness of conclude that
CREP at maximizing Conservation Reserve
programmatic Enhancement
benefits from Program (CREP)
enrolled lands wildlife benefits could
be maximized if CREP
targeted large parcels
of land or land
adjacent to other
grassland fields.
fauna.19 Gallant et al. (2011) utilizes data regarding historical land use patterns in Iowa
as a means of evaluating CRP and WRP program effectiveness in this region.20 Research
on CRP and WRP suggests that these programs provide a substantial amount of eco-
system services. However, researchers suggest that targeting and integrating CRP/WRP
lands into the existing supply of conserved land could improve program performance.
Considerable attention has been given to evaluating the economic value of CRP
conservation (Table 21.5). Hansen and Ribaudo (2008) provided one of the most
comprehensive estimates of per-ton benefits from reduced soil erosion on CRP lands.
They provide estimates for economic benefits resulting from a variety of conservation
objectives21 and conclude that, per ton of conserved soil, societal benefits for measured
conservation objectives range between $0.83 and $26.4022 per year. Wu and Lin (2010)
analyzed the impact of CRP conservation on aggregate farmland values and found that,
nationwide, CRP increased the value of farmland between $18 and $25 per acre (about
1.3–1.8% of the land value). Vukina et al. (2008) estimated the farmer’s value of conser-
vation based on the bids each made to enroll his land into available programs. They find
that farmers with higher quality land submit higher bids. Secchi et al. (2009) evaluated
the impact that continued ethanol subsidies could have on the supply of CRP lands and
found that high corn prices and thus competition for land resources could hamper CRP
conservation benefits.
Some researchers have analyzed CRP bid mechanisms to determine cost-benefit
of the program. CRP initially focused on enrolling as much land as possible through
least-cost enrollment: farmers who offered land below a threshold value were accepted
into the program. However, starting in the 1990s, CRP has gradually shifted toward
benefit-cost targeting, which scores a prospective parcel using an Environmental
Benefit Index (EBI) and enrolls land that provides the most conservation benefit per
value of contract until the budget is exhausted (Claassen et al. 2008). Changes in bid
ding structure and parcel “ranking” have increased the efficiency of CRP since the pro-
gram’s inception in the early 1980s (Ferraro 2008). However, Vukina et al. (2008) found
that making these ranking criteria publicly available has resulted in farmers adjusting
their bids based on the quality of their own land, which undermines the cost effective-
ness of the auction mechanism.
19
These researchers conclude that CREP wildlife benefits could be maximized if CREP targeted large
parcels of land and/or land adjacent to other grassland fields.
20
In the course of their research, they determined that most ecosystem losses in Iowa occurred in
areas with a high concentration of wetlands, whereas CRP and WRP wetland improvements occurred in
areas with historically few wetlands.
21
These benefits result from reduced wind erosion, water treatment, soil productivity, reservoir
services, fisheries, and recreation, among other categories.
22 At these levels, total benefits from reduced soil erosion range between $182 million and
$5,808 million per year.
Table 21.5 Economic analysis of Conservation Reserve Program (CRP)
Study
Study program Focus Study area Findings
Hansen and Ribaudo (2008) CRP Estimating societal Nationwide Per-ton economic benefits are calculated across a variety
benefits from reduced soil of soil conservation objectives and regions. Over the entire
erosion from CRP lands United States, each ton of soil conserved results in $0.83–
around the nation $26.40 worth of benefit per year, for a total societal benefit
of $182–$5808 million per year.
Wu and Lin (2010) CRP Implementing a system Nationwide CRP increases agricultural land values between $18 and $25
of equations to ascertain per acre (~1.3–1.8% of land value). The areas with the largest
the total effect of CRP on gains include Mountain, Southern Plains, and Northern Plains
agricultural land values regions.
Vukina et al. (2008) CRP Assessing farmer value North Carolina and Farmer CRP bids are relatively competitive, with average bids
of conservation based on Georgia of ~$42.70; the actual cost of conservation made up 99% of
farmer bids to CRP total bid ($42.27) and only a $0.43 premium. Furthermore,
farmers condition bids based on the perceived quality of
the land, with bids increasing by $0.73 for each increase in
Environmental Benefit Index (EBI) score.
Secchi et al. (2009) CRP Evaluating the potential Iowa At corn prices of $196.84/ton per hectare, sediment losses
impacts to the supply increase from 0.1 tons/ha for the almost 700,000 hectare
of CRP lands based of CRP included in the EPIC analysis to almost 1.9 tons per
on increases in total hectare as almost 500,000 hectares of CRP land are put back
corn-based ethanol into production.
production
Ferraro (2008) Payment for Asymmetric information Nationwide The EBI used by CRP to rank farmer bids creates competition
Ecosystem and contract design and prioritizes funding to highly valued conservation targets.
Services for payments for However, making this information public to landowners also
environmental services may encourage rent-seeking behavior from farmers with
highly sought-after land.
Claassen et al. (2008) Payment for Analyzing the literature Nationwide Although research has shown that the EBI did increase the
Ecosystem on the effectiveness of environmental benefits of the CRP, additional improvements
Services benefit-cost targeting in in environmental cost-effectiveness of the CRP could be
CRP farmland bidding achieved by further shifting emphasis from soil productivity
maintenance to enhancing water quality and wildlife habitat.
Program administrators have incorporated the research finding into program evalu-
ations and made changes to improve the economic efficiency and achieve specific con-
servation objectives. Total enrolled land peaked in 2007 at 36.8 million acres and then
decreased to a low of 31.3 million acres in 2010, the most recent year for which CRP data
are available (US Farm Service Agency [FSA] 2011). The acreage decline results from
lower budgets and nonrenewal of expiring CRP contracts. Although land enrolled in
the general CRP signups has been decreasing, land enrolled in CREP and other sub-
programs has actually increased. CREP focuses on high-priority lands for conservation
and uses much stricter guidelines for acceptance into the program than general signup
(Ferris and Siikamäki 2009). Thus, although enrolled acres have decreased by more than
5 million acres, environmental benefits have declined at a much slower rate, and the
environmental benefits per enrolled acre have actually increased. Figure 21.2 displays
reduced nitrogen, phosphorus, and sequestered carbon dioxide per CRP acre per year23
over the time frame 2004 to 2010, as well as total CRP land during this time.
10. Conservation Through Conversion

Prevention
10.1 Regulatory Conservation Through Development

Restrictions and Redirection
Since approximately the 1950s, commercial, industrial, and housing development has
shifted from city centers to suburban and exurban communities nationwide. Brown
et al. (2005) estimates that between 1950 and 2000, urban land area24 has grown less than
1–2%, compared to 5–25% for exurban land25 during this same time frame. In response
to these rapid changes in land use patterns, state, county, and city governments imple-
ment a variety of initiatives to slow the development of resource land, shift the spatial
patterns, and protect resource lands. These initiatives may be classified as developmen-
tal restriction (regulatory) and easement procurement (participatory). Researchers seek
to assess the impact of conservation initiatives on communities, as well as understand
the economic impact of these regulations. They have conducted both ex ante and ex post
analyses of these development restrictions and easement programs.
Unlike other national-level policies, community-level conservation programs are not
applied uniformly over a large geographic area and, as such, often provide researchers
23
These figures are further scaled by the average reduction of each pollutant during this time for the
sake of comparison.
24 Less than one acre per dwelling.
25 Between 1 and 40 acres per dwelling.
with natural experiments to exploit. In this way, program effectiveness may be ascer-
tained by comparing communities with active conservation policies to neighboring
communities without such policies in place. As a result, researchers have a possible
counterfactual of how a community would have evolved if a conservation policy were
not implemented.
Many local, regional, and state governments utilize developmental restrictions to con-
serve land in the short-term. A large variety of conservation initiatives exist through-
out the United States, but the most intensely studied programs include low-density and
minimum lot size zoning, UGBs, and APFOs. Irwin et al. (2009) provide a comprehen-
sive overview of the economics of urban and rural space, including a discussion of pol-
icy options for community-led conservation initiatives. In general, research has found
that development restrictions provoke mixed outcomes. Although effective in some
instances, these policies may also have unexpected and unintended impacts.
Many researchers have found that properly implemented zoning restrictions can
reduce the probability of land conversion and, as such, provide for land conservation.
However, the end result of zoning regulations can be surprising. For instance, in an
analysis of zoning and development patterns in Calvert County, Maryland, McConnell
et al. (2006) found zoning regulations decreased the quantity of residential lots by 10%
relative to what would have otherwise occurred. However, they also find that develop-
mental patterns are affected by many factors other than zoning laws. Magliocca et al.
(2012) apply an agent-based simulation model to study the effect of minimum lot zoning
on exurban development. They find that whereas hypothetical two-acre minimum lot
Land enrolled (millions of acres) Nitrogen (millions of lbs)

Phosphorus (millions of lbs) Carbon-dioxide (millions of metric tons)
1.3 38
37
1.2
36
Percent of average reduction
35
1.1
Millions of acres
34
1 33
32
0.9
31
30
0.8
29
0.7 28
2004 2005 2006 2007 2008 2009 2010
FIGURE 21.2 2004–2010 Conservation Reserve Program ecosystem benefits per year

(FSA 2007, 2011).
Table 21.6 Low-density zoning impact on development
McConnell et al. (2006) Low-density Estimating the impact of Calvert County, Maryland Low-density zoning had a small, but
zoning, transfers oflow-density zoning and TDRs on non-negligible effect on development. If zoning
development rights development trends restrictions were relaxed only a little, more than
(TDRs) 10% more lots would have been added overall in
subdivisions facing relatively low-density limits
over the sample period.
Magliocca et al. (2012) Minimum lot zoning Simulating the development Hypothetical Urban sprawl may arise from any number of
of a hypothetical mid-Atlantic Mid-Atlantic community conditions, including variations in (i) agricultural
community with an urban core, productivity across the landscape, (ii) consumers’
imposing zoning regulations and housing preferences, and (iii) how expectations
other heterogeneous conditions of future prices are formed. Researchers found
to monitor how development evidence of “leapfrogging” of urban development
patterns adapt and preferences for large lots, implying relatively
stringent minimum lot zoning in order to prevent
subdivision.
Butsic et al. (2011) Low-density zoning, Analyzing the impact of Columbia County, After accounting for endogeneity, low-density
tax credits for low-density zoning restrictions Wisconsin zoning had no impact on development decisions,
conservation of development, accounting and Wisconsin’s tax credit system had, at best, a
for the endogeneity between weak impact on development decisions.
development and zoning
patterns
Gottlieb et al. (2009) Minimum lot zoning Studying the impact of large-lot New Jersey Large-lot zoning in excess of 4 acres actually
zoning on rural communities encouraged more development by forcing more
land into the market to accommodate the same
number of people.
Lichtenberg (2011) Minimum lot size Evaluating the impact of Baltimore-Washington A 1-acre increase in minimum lot size results
zoning, Forest minimum lot size zoning on land corridor in 0.83% increase in land area needed to
Conservation Act use patterns accommodate the current population.
zoning restrictions have little effect on land development patterns, five-acre minimum
lots significantly reduced the level of exurban development. These studies’ results, as
well as those of several others, are outlined in Table 21.6.
In many communities, zoning restrictions are criticized for failing to adequately pro-
tect land from conversion. Researchers contend that some communities implement
overly generous or poorly planned zoning restrictions, which do little to actively pro-
tect land (see Table 21.7). Butsic et al. (2011) utilized a variety of econometric models
to study the impact of low-density zoning in rural Wisconsin. They find that zoning
restrictions did not influence landowners’ decision to subdivide their property. Gottlieb
et al. (2009) studied the impact of minimum-lot zoning and open space land preserva-
tion on development patterns in New Jersey and found that, at least in the short run,
these policies may actually increase property subdivision. In Maryland, Lichtenberg
(2011) studied the impact of minimum-lot regulations and local forest conservation
laws on development patterns in the Baltimore-Washington metropolitan area. He finds
that these policies contribute to more sprawl development as more land needs to be con-
verted to accommodate the same number of households.
UGBs are also a popular policy option, whereby a community demarcates a com-
pact urban development zone but imposes stringent zoning and infrastructure expan-
sion laws outside this area. Overall, UGBs have had mixed impacts in the United States.
Although these regulations may have some impact on development patterns, their effect
is often not that anticipated by policy makers. For instance, Newburn and Berck (2006,
2011) studied residential land use change in Sonoma County, California, and found
that, although the existing UGB constrained suburban development near the UGB,
exurban development was largely unconstrained by these policies. Cunningham (2007)
found that the UGB implemented by Seattle, Washington, lowered the probability of
development for parcels outside the UGB by between 28% and 39%. Interestingly, the
author finds that low-density development restrictions may decrease the price volatility
of these lands, which in turn might speed up development, all else equal. Although the
actual effect of UGBs on community development has been mixed, some communities
have benefited from having clearly defined growth areas, while leaving more rural areas
for conservation efforts.
Maryland has been one of the most proactive states at implementing land use tools
to promote conservation. Priority Funding Areas (PFAs) are among the most promi-
nent components of its 1997 Smart Growth legislation. As an incentive-based policy,
the PFA legislation restricts state spending on growth-related programs to areas des-
ignated for urban growth (i.e., the PFAs); spending is permitted for infrastructure
improvement to water and sewer, for example. Although not identical to UGBs, PFAs
have a similar focus but use the carrot of state subsidization of infrastructure improve-
ments rather than a regulatory approach. Hanlon, Howland, and McGuire (2010)
analyze the effectiveness of Maryland’s PFAs at reducing development on agricultural
lands outside PFAs near Baltimore and Washington, DC. They find that development
pressure can be high on the agricultural land outside the PFA areas, but that Maryland’s
PFA have been effective at reducing the probability of conversion of these resource
Table 21.7 Urban growth boundaries
Newburn and Berck Urban Growth Estimating the impact of Sonoma County, California Sonoma’s UGB was effective at reducing the
(2006, 2011) Boundaries (UGB) UGB on development in probability of suburban development outside
study region the ring region, but they find evidence of
exurban development leapfrogging.
Cunningham (2007) Urban Growth Evaluating the impact of Seattle, Washington Seattle’s UGB was effective in lowering the
Boundaries Seattle’s UGB on probability probability of development of rural lands by
of development for between 28% and 39%, but it also reduced
properties outside the urban price volatility of these lands. This reduced the
center barriers to development.
Hanlon, Howland, and Priority Funding Areas Studying the impact Frederick County, Maryland They find that some of the areas with the
McGuire (2010) of Maryland’s PFA on a greatest threat of development are outside
parcel’s probability of future Maryland’s PFA and, although the program is
development not 100% effective at reducing development,
it does affect a property’s probability of
development.
Lewis et al. (2009b) Priority Funding Areas Provides an overview State of Maryland The PFA program has not entirely lived up
of 10 years’ experience to its potential, although it has had some
with Maryland’s PFA effect on reducing development in rural
program, as well as some areas; the research also provides several
recommendations for recommendation for program improvement.
improvement
lands. Lewis, Knaap, and Sohn (2009b) found that, despite some conceptual and prac-
tical limitations of the program, Maryland’s PFAs did have limited success at deter-
ring rural development. Towe, Lewis, and Lynch (Chapter 18 of this volume) find that
PFAs have 1.04 more new homes within each ¼-square-mile PFA grid than would have
occurred without the program. Similarly, using the difference-in-difference approach
(pre-PFA housing starts to post-PFA housing starts), they find comparable results of
an average of 1.05 new housing starts within the ¼-square-mile PFA grids compared to
similar non-PFA grids (Table 21.8).
Both the public and policy makers are also concerned about the impact that develop-
ment restrictions have on land values. Several teams of researchers have analyzed this
issue, again finding mixed results. Although down-zoning restricts land use options,
these restrictions may not be stringent enough to put any practical limitations on land
use (i.e., they are nonbinding). Down-zoning rural lands may create positive amenities
for neighboring lands, which causes difficulties in the analysis of the overall effect that
zoning may have on property values. Dehring and Lind (2007) find that, depending on
Table 21.8 Impact of zoning on rural home values

Dehring and Lind Low-density Estimating Dallas, Texas Depending on the

(2007) zoning the impact of property type and
down-zoning on regulation, zoning
rural home values decreases land
value by 0–21%.
Deaton and Vyn Low-density Evaluating Toronto, Canada Toronto-area
(2010) zoning the impact of properties were
down-zoning decreased in value
across a variety of by 20%, whereas
land classes and more remote
zoning restrictions properties actually
increased in value,
although this result
was statistically
insignificant.
Liu and Lynch Low-density Studying the Nine rural Resource-based,
(2011a) zoning impact of Maryland high-quality
down-zoning on Counties agricultural lands
property value, are unaffected
accounting for the by down-zoning,
interdependence whereas
between non-resource
property value based lands
and probability decrease in value
of being by between 20%
down-zoned and 50%.
the stringency of regulation, down-zoning had a potentially negative impact on vacant

lots in Dallas, Texas. On the other hand, Deaton and Vyn (2010) found that although
down-zoning decreased property value for properties near Toronto, more remote
areas actually increased in value. One major challenge to attributing the real effect of
zoning is that the probability of a parcel being down-zoned is not independent of its
property value. Thus, failure to account for this relationship (known as endogeneity)
introduces bias into resulting estimations. Liu and Lynch (2011b) analyzed the impact
of down-zoning on property values in several rural Maryland counties using both pro-
pensity score matching and instrumental variables techniques to control for endogene-
ity. They find that the overall impact is differentiated by land type, with resource-based
agricultural and forest lands unaffected by down-zoning and other rural lands having
decreases in property value. Overall, these results suggest that policy makers should be
concerned about land values when designing zoning regulation that pursues desired
conservation objectives.
Open space requirements are another popular regulatory method to encourage direct
conservation of scarce resources. Interestingly, these requirements may also have unin-
tended effects on the pace and pattern of land development and conservation. Local
governments may also mandate conservation plans and enact buffer zones around sen-
sitive resources. Sander and Polasky (2009), Bin, Landry, and Meyer (2009), and Ham
et al. (2012) have confirmed that these open space requirements increase residential
home value. Similarly, Bucholtz, Geoghegan, and Lynch (2003) find that permanently
preserved land increases adjacent parcel’s land values in most cases. However, because
they provide amenity value, these open-space lands may act as a magnet for additional
development. Lichtenberg, Tra, and Hardie (2007) and Lichtenberg (2011) studied the
impact of the Maryland’s Forest Conservation Act (FCA) open-space and tree reten-
tion requirements on development.26 They find that FCA lands crowd out other sources
of open space in Maryland and actually contribute to additional sprawl and conver-
sion of resource lands. Similarly, Towe, Nickerson, and Bockstael (2008) find that per-
manent open space adjacent to a parcel increases its rate of conversion. However, the
effect of open space on development may not be uniform across all areas. Lewis et al.
(2009a) and Zipp et al. (2011) studied how an increase in open-space requirements
(public beaches) around suburban Wisconsin area lakes actually decreased the likeli-
hood of lakefront development. These results illustrate the need to design conserva-
tion policy specific to local conditions, incorporating many policy options to achieve
desired objectives.
26
In Maryland, the Forest Conservation Act (FCA) was enacted in 1991 and mandates that any new
subdivisions or additional development of lands greater than approximately one acre must implement a
forest conservation plan to either protect existing forest cover or reestablish equal or greater forest cover
on other lands.
11. Participatory and Incentive-Based

Programs for Conversion Prevention
Communities have many tools in their arsenal to promote conservation. Unlike regu-
latory approaches, participatory and incentive-based provisions provide guaranteed
conservation (with monitoring). Because these are voluntary tools, county and local
governments find them easier to implement to preserve land resources and often use
them to complement existing developmental regulations as hybrids. As such, develop-
mental restrictions and voluntary programs are often analyzed together at the local level
(Table 21.9).
Local governments utilize agricultural and conservation easements to protect
resource lands from conversion to development. Often referred to as Purchase of
Development Rights (PDR) or Purchase of Agricultural Conservation Easements
(PACE), these programs have increased in popularity because other developmental
restrictions have not provided enough protection for resource lands from conversion
to urban sprawl. PDR/PACE programs are effective in protecting land from develop-
ment according to several recent studies. For example, using a hazard model, Towe,
Nickerson, and Bockstael (2008) find that rural landowners delay the conversion of land
by up to an additional six years, even if they do not enroll the land into the local PDR
program. Liu and Lynch (2011b) implemented a propensity score matching method to
determine the effectiveness of PDR programs across six Mid-Atlantic states. They found
that counties with PDR/PACE programs had an average rate of agricultural land conver-
sion 40–55% lower than similar counties without such a program. These results suggest
that, where appropriate, PDR programs can be effective at incentivizing local resource
landowners and lowering the rate of agricultural land conversion.
Transfers of development rights (TDR) were a popular conservation tool whereby
communities allow resource properties to sell their development rights to developers to
utilize to increase development densities elsewhere.27 In theory, the social costs of a TDR
program could be lower than a comparable PDR program by creating a competitive
market for development rights.28 Lynch and Musser (2001) found that TDR programs
were less efficient in preserving parcels with the desired characteristics than were PDR
programs. The TDR programs studied had based their allocation strategies primarily
on acreage rather than on other parcel characteristics, such as prime soils or proxim-
ity to urban areas. Few TDR programs have been successful: nine programs have been
revoked, 17 have protected no land, and only 12 programs have protected more than
1,000 acres of farmland (American Farmland Trust [AFT] 2008). However, the TDR
27
As of 2008, at least 109 municipalities had active TDR programs in the United States (Kaplowitz
et al. 2008).
28 Kaplowitz et al. (2008) administered a mail-in survey of active TDR program directors that
provided insight into the experiences of these programs.

Table 21.9 Land use easements

Towe, Nickerson, Purchase of Examining Howard County, The option to

and Bockstael Development the impact Maryland preserve farmland
(2008) Rights (PDR) of farmland delays the timeframe
preservation of development by
programs up to six years, a
on timing of reduction of 12–43%
development of median conversion
time.
Liu and Lynch Purchase of Evaluating the Six Mid-Atlantic Having a PDR program
(2011b) Development impact of PDR States decreases a county’s
Rights programs on rate of farmland
reducing farmland loss by 40–55% and
conversion decreases farmland
acres lost by 375 to
550 acres per year.
Horowitz, Lynch, Purchase of Analyzing the Carroll County, Competitive PDR
and Stocking Development efficiency of Maryland programs enrolled as
2009) Rights competition-based many as 3,000 acres
PDR programs (12%) more than a
take-it-or-leave-it
offer would have
enrolled for the same
budgetary cost. Each
additional bidder
competing for the
same resources leads
to a decreases in bid
value of 0.1–1.4%
Crompton Purchase of Examining 20 states with Among the PDR
(2009) Development state statues active PDR state statutes,
Rights to understand programs (i) almost all the
the goals of language focused on
these laws and agricultural interests;
effectiveness of (ii) did not claim any
these programs open space benefits;
at promoting and (iii) term and
conservation rescinding provisions
were authorized, as
well as in-perpetuity
purposes, even
though they offer
no enduring public
benefits.
programs inform the design of future PDR programs to increase efficiency and lower
costs. By promoting competition among landowners, PDR programs may leverage their
funding and preserve more acreage. In an analysis of the Maryland Agricultural Land
Preservation Foundation (MALPF), Horowitz, Lynch, and Stocking (2009) found that
each additional bidder decreased a farmer’s bids between 0.1% and 1.4%. However,
untargeted land use easements may not be as environmentally effective, which raises
questions regarding the most critical conservation goals. Crompton (2009) analyzed 20
states’ statutes regarding PDR programs, finding that these programs prioritized agri-
cultural interests rather than conservation goals. Although PDR initiatives have been
effective at promoting agricultural land conservation, similar programs have not been
established to focus on other conservation priorities (Table 21.10).
12. Conclusion
Conservation tools have been successful in accomplishing many of the multitude of

goals set by society and policy makers. While regulatory efforts continue, many par-
ticipatory, incentive–based, and hybrid programs have surfaced in the United States in
the past 30 years. The long-term impacts of these programs continue to be of interest.
Overall, program evaluations continue to suggest that policy makers need to consider
how to get the “biggest bang for the buck.” Evaluations tend to focus more heavily on
patterns of conservation and ecological and other benefit metrics than on the number
of acres preserved or the number of dollars expended. Similarly, because of unintended
impacts, including inducing development adjacent to conservation, further refinement
may be required to existing programs to more effectively accomplish the objectives set
by society. In addition, concerns about economic development have redirected some
land conservation from land retirement (conservation practice or restoration on to a
natural state) to working land conservation. These programs will need evaluation as well.
Similar to other resource and environmental policies, society is seeking land con-
servation programs that make appropriate tradeoffs between economic growth and
environmental benefits. The United States has made this particularly important for agri-
cultural lands from a desire to maintain a critical mass of farmland for strong local agri-
cultural economies (Lynch and Carpenter, 2003). Resources have been shifting towards
working land programs at the USDA as well. These programs have received less evalu-
ation to date. A greater willingness to target land conservation to those areas or lands
with the highest benefits has also emerged. This shift in policy has generated interest in
both metrics to use for targeting and mechanisms to ensure voluntary programs incen-
tivize targeted goals appropriately. Efforts continue on the development prevention
front, attempting to prevent and redirect housing development as well as conserve land
using programs like Maryland’s Smart Growth Program.
On the research front, land conservation research has evolved dramatically with
the advent of geographic information system (GIS) data of all kinds. These data have
Table 21.10 Open space requirements

Sander and NA Studying the Ramsey County, Starting at 1,000 m,

Polasky (2009) impact of views Minnesota the marginal implicit
and open space price for reducing
on residential the distance to the
home value nearest lake by 100
m produces a $216
increase in home sale
value.
Bin, Landry, and Riparian Buffers Evaluating Neuse River Basin in Proximity to riparian
Meyer (2009) the impact of North Carolina buffer raises the
proximity to property value by
riparian buffer 25.9%.
on rural home
value
Lichtenberg, Tra, Minimum Evaluating the Baltimore-Washington A one-acre increase
and Hardie (2007); lot size impact of FCA corridor in the required FCA
Lichtenberg (2011) zoning, Forest acreage on land results in between
Conservation urban sprawl 0.12 and 0.85 acres of
Act additional open space,
varying primarily
based on the county
and sewer access of
the subdivision.
Lewis et al. Public Determining Vilas County, Researchers found
(2009a) protection of the effect of Wisconsin evidence that public
lake-front land open-space conservation land
conservation on lake shorelines
policies on can actually reduce
residential the probability that
development privately owned
density residential parcels
subdivide and develop.
Zipp et al. (2011) Public protection Assessing the Door County, WI Open space and
land impact of private land are
open-space complements, and
conservation on open space decreases
development the likelihood of
decision development for land
classes aside from
agriculture.
allowed researchers to access more information on a spatial scale than previously

available. This allows the land use models to expand beyond the Von Thunen/Mills
variety by allowing heterogeneity to be included within the analysis. These spatially
based data have been combined with US Census data—both population and agricul-
tural—to utilize demographic and economic variables with physical factors to answer
a wide range of questions. Researchers have also utilized the USDA NRCS’s National
Resource Inventory data to answer interesting questions about recent land use change
(Karetnikov, 2012). Econometric and nonparametric models have also become more
complex, allowing for incorporation of both time and spatial dimensions of land con-
servation. A duration analysis now can incorporate both timing and spatial variabil-
ity into the approach. Similarly, genetic algorithms have allowed a seemingly infinite
number of land use options to be computed more easily. As mentioned earlier, increased
computational power has allowed further refinements to ecosystem models that permit
the framing of optimal program design.
Although significant advances have been made, many challenges to evaluation and
program design exist. Most land conservation programs have multiple goals and, often
cannot accomplish them all simultaneously without multiple mechanisms. Yet these
multiple mechanisms may interact in unexpected ways. In addition to many mecha-
nisms, land conservation is inherently spatial and policy decisions must be conducted
over multiple dimensions. Spillovers and other interdependencies will impact the
outcomes in both the time and space dimensions. Each action impacts other actions
and many other outcomes. This path dependency from both a time and space perspec-
tive can be quite difficult to resolve using many of the existing evaluation techniques.
Similarly, program design becomes fraught with difficulties. To make matters more dif-
ficult, endogeneity of program selection and program outcomes can make causality and
assignment of outcomes to a particular program mechanism difficult to determine.
However, researchers continue to make progress on these fronts. Ongoing studies
about interactions between different programs and regulations continue to progress.
Although not all of the unintended consequences of certain tools may be undesirable,
knowing how programs interact on a spatial basis is crucial to designing and perfecting
programs. Similarly, conservation costs have become an issue, especially when thresh-
old impacts exist; therefore, further information on how financing mechanisms impact
program outcomes would benefit policy makers. Heterogeneous landowners also affect
the efficiency of a program, and thus determine how much information about land-
owners is needed. How information asymmetries impact conservation outcomes and
how landowners’ behavior correlates to publicly available data, such as land use and
land cover, remain to be further considered. In the same vein, researchers often wish
to connect GIS data with survey data and need to determine mechanisms to overcome
confidentiality issues. Interdisciplinary studies that link economic and ecological ben-
efits appropriately must also be pursued. Modeling ecological systems and determining
benefits in a variety of locations is needed both to ensure targeting is well done and to
evaluate if a program is accomplishing its goals.
References
American Farmland Trust (AFT). 2008. Transfer of development rights fact sheet. Northampton,
MA: Farmland Information Center.
Baerenklau, K. A., A. Gonzalez-Caban, C. Paez, and E. Chavez. 2010. Spatial allocation of forest
recreation value. Journal of Forest Economics 16(2): 113–126.
Bin, O., C. E. Landry, and G. E. Meyer. 2009. Riparian buffers and hedonic
prices: A quasi-experimental analysis of residential property values in the Neuse River
Basin. American Journal of Agricultural Economics 91(4): 1067–1079.
Brown, D. G., Johnson, K. M., Loveland, T. R., and Theobald, D. M. 2005. Rural land use change
in the conterminous U.S., 1950–2000. Ecological Applications, 15(6): 1851–1863.
Bucholtz, S., J. Geoghegan, and L Lynch. 2003. Capitalization of open spaces into housing val-
ues and the residential property tax revenue impacts of agricultural easement programs.
Agricultural and Resource Economics Review 32(1): April 33–45.
Butsic, V., D. J. Lewis, and L. Ludwig. 2011. An econometric analysis of land development with
endogenous zoning. Land Economics 87(3): 412–432.
Chang, Katie. 2011. R. C. 2010 National Land Trust CENSUS REPORT: A look at voluntary land
conservation In America, eds. R. Aldrich and C. Soto. Land Trust Alliance. Washington, DC.
Claassen, R., R. Cattaneo, and R. Johansson. 2008. Cost-effective design of agri-environmental
payment programs: US experience in theory and practice. Ecological Economics 65: 738–753.
Crompton, J. L. 2009. How well do purchase of development rights programs contribute to park
and open space goals in the United States? World Leisure Journal 51(1): 54–71.
Crompton, J. L. 2007a. The impact of parks and open spaces on property taxes. In The economic
benefits of land conservation, ed. C. de Brun, 1–12. San Francisco: Trust for Public Land.
http://cloud.tpl.org/pubs/benefits_econbenefits_landconserve.pdf.
Crompton, J. L. 2007b. Competitiveness: Parks and open space as factors shaping a location’s
success in attracting companies, labor supplies, and retirees. In The economic benefits of land
conservation, ed. C. de Brun, 48–54. San Francisco: Trust for Public Land. http://cloud.tpl.
org/pubs/benefits_econbenefits_landconserve.pdf.
Cunningham, C. 2007. Growth controls, real options, and land development. Review of
Dahl, Thomas E. 1990. Wetlands losses in the United States—1780s to 1980s (version July
16, 1997). Washington, DC: US Department of the Interior, Fish and Wildlife Service.
Jamestown, ND: Northern Prairie Wildlife Research Center Online. http://www.npwrc.
usgs.gov/resource/wetlands/wetloss/index.htm.
Deaton, J., and R. Vyn 2010. The effect of strict agricultural zoning on agricultural land
values: The case of Ontario’s greenbelt. American Journal of Agricultural Economics
92(4): 941–955.
Dehring, C., and M. Lind. 2007. Residential land-use controls and land values: Zoning and
covenant interactions. Land Economics 83(4): 445–457.
Deisenroth, D., J. Loomis, and C. Bond. 2009. Non-market valuation of off-highway vehicle rec-
reation in Larimer County, Colorado: Implications of trail closures. Journal of Environmental
Duke, J. M., and L. Lynch. 2006. Four classes of farmland retention techniques: Comparative
evaluation and property rights implications. Land Economics 82(2):189–213.
Ferris, J., and J. V. Siikamäki. 2009. Conservation reserve program and wetland reserve pro-
gram: Primary land retirement programs for promoting farmland conservation. Washington,
DC: Resources for the Future.
Ferraro, P. J. 2008. Asymmetric information and contract design for payments for environmen-
tal services. Ecological Economics 65: 811–822.
Gallant, A. L., W. Sadinski, M. F. Roth, and C. A. Rewa. 2011. Changes in historical Iowa land
cover as context for assessing the environmental benefits of current and future conservation
efforts on agricultural lands. Journal of Soil and Water Conservation 66(3): 67–77.
Gleason, R. A., N. H. Euliss, Jr., B. A. Tangen, M. K. Laubhan, and B. A. Browne. 2011. USDA
conservation program and practice effects on wetland ecosystem services in the Prairie
Pothole Region. Ecological Applications 21(Supplement): S65–S81.
Gleason, R. A., M. K. Laubhan, and N. H. Euliss, Jr. (eds). 2008. Ecosystem services derived
from wetland conservation practices in the United States prairie pothole region with an
emphasis on the US Department of Agriculture conservation reserve and wetlands reserve
programs. US Geological Survey Professional Paper 1745, Reston, Virginia, USA.
Gorte, R. W. 2012. Wilderness laws: Statutory provisions and prohibited and permitted uses.
Washington, DC: Congressional Research Service.
Gottlieb, P. D., A. O’Donnell, T. Rudel, K. O’Neill, and M. McDermott. 2009. The impact of
large-lot zoning and open space acquisition on home building in rural communities. 2009
Annual meeting of the Agricultural and Applied Economics Association, July 26–28, 2009,
Milwaukee, WI.
Gude, P. H., A. J. Hansen, and D. A. Jones. 2007. Biodiversity consequences of alternative future
land use scenarios in Greater Yellowstone. Ecological Applications 17: 1004–1018.
Ham, C., P. A. Champ, J. B. Loomis, and R. M. Reich, 2012. Accounting for heterogeneity of
public lands in hedonic property models. Land Economics 88 (3): 444–456
Hanlon, B., M. Howland, and M. McGuire. 2010. Hotspots for growth: Land use change in a
transitional county in the US. Unpublished manuscript, University of Maryland.
Hansen, L., and M. Ribaudo. 2008. Economic measures of soil conservation benefits. Technical
Bulletin Number 1922. Washington, DC: United States Department of Agriculture.
Horowitz, J. K., L. Lynch, and A. Stocking. 2009. Competition-based environmental policy: An
analysis of farmland preservation in Maryland. Land Economics 85(4): 555–575.
Internal Revenue Service. 2012. Conservation easements techniques guide. http://www.irs.
gov/businesses/small/article/0,,id=249135,00.html.
Irwin, Elena G., Kathleen P. Bell, Nancy E. Bockstael, David A. Newburn, Mark D. Partridge,
and JunJie Wu. 2009. The economics of urban-rural space. Annual Review of Resource
Kaplowitz, M., P. Machemer, and R. Pruetz. 2008. Planners’ experiences in managing growth
using Transferable Development Rights (TDR) in the United States. Land Use Policy 25
(3): 378–387.
Karetnikov, D. A. 2012. Have the National Resources Inventories advanced conservation pol-
icy? Ph.D. Dissertation, University of Maryland.
Klaiber, H. A., and D. J. Phaneuf. 2009. Do sorting and heterogeneity matter for open space
policy analysis? An empirical comparison of hedonic and sorting models. American Journal
Klaiber, H. A., and D. J. Phaneuf. 2010. Valuing open space in a residential sorting model of the
Twin Cities. Journal of Environmental Economics and Management 60(2): 57–77.
Land Trust Alliance Conservation Resource Center. 2007. State conservation tax credits: Impact
and analysis. Boulder, CO: Land Trust Alliance. http://www.landtrustalliance.org/policy/
documents/state-tax-credits-report.pdf/view
Lewis, D. J., B. Provencher, and V. Butsic. 2009a. The dynamic effects of open-space conserva-
Management 57(3): 239–252.
Lewis R., G. J. Knaap, and J. Sohn. 2009b. Managing growth with priority funding areas: A good
idea whose time has yet to come. Journal of the American Planning Association 75:457–478.
Lichtenberg, E., C. Tra, and I. Hardie. 2007. Land use regulation and the provision of open space
in suburban residential subdivisions. Journal of Environmental Economics and Management
54: 199–213.
Lichtenberg, E. 2011. Open space and urban sprawl: The effects of zoning and forest
conservation regulations in Maryland. Agricultural and Resource Economics Review
40(3): 393–404.
Liu X. P., and L. Lynch. 2011a. Do zoning regulations rob rural landowners’ equity? American
Journal of Agricultural Economics 93: 1–25.
Liu, X., and L. Lynch. 2011b. Do agricultural land preservation programs reduce farmland loss?
Evidence from a propensity score matching estimator. Land Economics 87: 183–201.
Lueck, D. L., and J. A. Michael. 2003. Preemptive habitat destruction under the Endangered
Species Act. Journal of Law and Economics vol. XLVI (April): 27–60.
Lynch L. and J. Carpenter. 2003. Is there evidence of a critical mass in the Mid-Atlantic
Agriculture sector between 1949 and 1997? Agricultural and Resource Economics Review
32(1)(April 2003): 116–128.
Lynch, L., and X. Liu. 2007. Impact of designated preservation areas on rate of preservation
and rate of conversion: Preliminary evidence. American Journal of Agricultural Economics
89(5): 1205–1210.
Lynch, L. and W. N. Musser. 2001. A relative efficiency analysis of farmland preservation pro-
grams. Land Economics November 77 (4): 577–594.
Lynch, Lori, K. Palm, S. Lovell, and J. Harvard. 2007. Expected cost of tripling Maryland’s pre-
served acres: Using a hedonic price analysis on agricultural land values from 1997–2003. Wye
Mills, MD: Harry Hughes Center for Agroecology, University of Maryland.
Magliocca, N., V. McConnell, and M. Walls. 2012. Explaining sprawl with an agent-based model
of exurban land and housing markets. RFF discussion paper No. DP 11–33, Washington,
DC: Resources for the Future.
McConnell, V., M. Walls, and E. Kopits. 2006. Zoning, TDRs, and the density of development.
Journal of Urban Economics 59: 440–457.
Messer, K. D. 2006. The conservation benefits of cost-effective land acquisition: A case study in
Maryland. Journal of Environmental Management 79: 305–315.
Newburn, D. A., and P. Berck. 2006. Modeling suburban and rural residential development
beyond the urban fringe. Land Economics 82(4): 481–499.
Newburn, D. A., and P. Berck. 2011. Growth management policies for exurban and suburban
development: Theory and an application to Sonoma County, California. Agricultural and
Nickerson, C., R. Ebel, A. Borchers, and F. Carriazo. 2011. Major uses of land in the United States,
2007. Economic Information Bulletin No. (EIB-89). Washington, DC: US Department of
Agriculture, Economic Research Service.
Pavelis, G. A., D. Helms, and S. Stalcup. 2011. Soil and water conservation expenditures by USDA
agencies, 1935–2010. Washington, DC: Historical Insights. Natural Resources Conservation
Service, Ecological Sciences Division.
Poudyal, N. C., D. G. Hodges, and C. D. Merrett. 2009. A hedonic analysis of the demand for and
benefits of urban recreation parks. Land Use Policy 26: 975–983.
Radeloff, V. C., S. I. Stewart, T. J. Hawbaker, U. Gimmi, A. M. Pidgeon, C. H. Flather, R.
B. Hammer, and D. Helmers. 2010. Housing growth in and near United States protected
areas limits their conservation value. Proceedings of the National Academy of Sciences of the
USA 107: 940–945.
Sander, H. A., and S. Polasky. 2009. The value of views and open space: Estimates from a hedonic
pricing model for Ramsey County, Minnesota, USA. Land Use Policy 26(3): 837–845.
Secchi S., P. W. Gassman, J. R. Williams, and B. A. Babcock. 2009. Corn-based ethanol pro-
duction and environmental quality: A case of Iowa and the conservation reserve program.
Environmental Management 44: 732–744.
Siikamaki, J. 2011. Contributions of the US state park system to nature recreation. Proceedings
of the National Academy of Sciences of the USA 108: 14031–14036.
Towe, C., C. Nickerson, and N. E. Bockstael. 2008. An empirical examination of the timing of
land conversions in the presence of farmland preservation programs. American Journal of
Trust for Public Land. Center for City Park Excellence. 2011. City park facts. San Francisco: Trust
for Public Land.
USDA. 1982. A national program for soil and water conservation: 1982 final program report
and environmental impact statement: Soil and Water Resources Conservation Act, p. 18.
USDA National Resources Conservation Service (NRCS). National conservation practice
standards—NHCP | NRCS|National conservation practice standards—NHCP | NRCS.
Accessed July 16, 2012. http://www.nrcs.usda.gov/wps/portal/nrcs/detailfull/national/
technical/cp/ncps/?&cid=nrcs143_026849.
USDA National Resources Conservation Service (NRCS). 1996. America’s private land, a geog-
raphy of hope. http://www.nrcs.usda.gov/wps/portal/nrcs/detail/national/technical/nra/
rca/?cid=nrcs143_014212
USDA National Resources Conservation Service. 2001. A Resources Conservation Act
report: Interim appraisal and analysis of conservation alternatives. www.nhq.nrcs.usda.gov/
land/pubs/rca_interim.html
USDA National Resources Conservation Service. 2012. Financial assistance. http://www.nrcs.
usda.gov/wps/portal/nrcs/main/national/programs/financial
USDA National Resources Conservation Service. 2012. Wetland reserve program. http://www.
nrcs.usda.gov/wps/portal/nrcs/main/national/programs/easements
US Department of the Interior. 2012. National Park Service: Budget and appropriations. http://
www.doi.gov/budget/appropriations/2013/highlights/upload/BH071.pdf
US Environmental Protection Agency. 2012. Summary of the Endangered Species Act. http://
www.epa.gov/lawsregs/laws/esa.html
US Farm Service Agency. 2007. Conservation reserve program: 2006 Annual summary and
enrolment statistics. http://www.fsa.usda.gov/Internet/FSA_File/annual2006summary.pdf
Farm Service Agency. 2011. Conservation reserve program: 2010 Annual summary and enrol-
ment statistics. http://www.fsa.usda.gov/Internet/FSA_File/annual2010summary.pdf
Vukina, T., X. Zheng, M. Marra, and A. Levy. 2008. Do farmers value the environment?
Evidence from a conservation reserve program auction. International Journal of Industrial
Organization, 26(6): 1323–1332.
Wade, A. A., and D. M. Theobald. 2010. Residential development encroachment on US pro-
tected areas. Conservation Biology 24: 151–161.
Wade, A. A., D. M. Theobald, and M. J. Laituri. 2011. A multi-scale assessment of local and con-
textual threats to existing and potential US protected areas. Landscape and Urban Planning
101(3): 215–227.
Walls, M., S. Darley, and J. Siikamäki. 2009. The state of the great outdoors: America’s parks, pub-
lic lands, and recreational resources. Washington, DC: Resources for the Future.
Wentworth, K. L., M. C. Brittingham, and A. M. Wilson. 2010. Conservation reserve enhance-
ment program fields: Benefits for grassland and shrub-scrub species. Journal of Soil and
Water Conservation 65: 50–60.
Wu, J., and H. Lin. 2010. The effect of the conservation reserve program on land values. Land
Economics 86(1): 1–21.
Zipp, K., D. Lewis, and B. Provencher. 2011. Does open space conservation increase neighboring
development? 2011 Association of Environmental and Resource Economics Presentation.
http://www.webmeets.com/files/papers/AERE/2011/196/AERE_door_county_final_
draft2.pdf
C HA P T E R 22
EUROPEAN
AG R I - E N V I R O N M E N TA L
POLICY
The Conservation and Re-Creation of
Cultural Landscapes
IA N HOD G E
Since its introduction in the mid-1980s, agri-environmental policy has become a

major component of agricultural policy in Europe. This chapter outlines the context
of its introduction and its present position. It then reviews the research that has been
undertaken, looking particularly at scheme characteristics and evaluation. A final sec-
tion reflects on the prospects for agri-environment policy and raises some wider issues.
The primary focus of the chapter is on agri-environmental policy operated within the
European Union (EU). This is implemented as an obligatory part of the Second Pillar
of the Common Agricultural Policy under the Rural Development Regulation. Similar
policies and approaches have been implemented in other European Countries, particu-
larly in Norway and Switzerland, and the literature reflects the wider coverage of these
policies.
Government intervention in the management of rural land at a broad scale for the
provision of public good environmental values represents a novel development in pub-
lic policy, still in a relatively early stage in its implementation. It is now being pursued in
many developed countries, and the approach is being further extended in payments for
ecosystem services. The experience gained from the theoretical and empirical analysis
discussed here provides insights both for the further development of agri-environment
policy itself and for the design of payments for ecosystem services.
584 Ian Hodge
1. The Common Agricultural Policy and

Land Use Change in Europe
Agri-environment schemes were introduced into European policy in the mid-1980s.

This was a time of mounting agricultural surpluses, increasing concerns as to the bud-
getary costs of the Common Agricultural Policy (CAP), and an emerging recognition
of the damage that agricultural intensification, stimulated by the incentives established
under the CAP, was having on the environment (e.g., Shoard 1980; Bowers and Cheshire
1983). Around this same time, a coalition of interest began to emerge in place of what
had previously been a clear conflict between the interests of farming and conservation
(Lowe et al. 1986). Baldock and Lowe (1996, 12) comment: “Thus, some agricultural
policy makers have responded to environmental concerns, not necessarily through any
deep convictions, but because of a perceived coincidence between the aims of environ-
mental improvement and the need to reduce agricultural output, thereby contributing
to the alleviation of surplus and budgetary problems.” This suggests that policies to pay
farmers to reduce the level of their production intensity could have the triple benefits of
protecting the environment, reducing levels of commodity production, and thus lower-
ing the cost to government of dealing with commodity surpluses (Willis et al. 1988) and,
perhaps, of offering a new justification for government payments to farmers.
After some initial trials of schemes in the United Kingdom and the Netherlands initi-
ated under the Less Favoured Areas Directive (Countryside Commission 1984), Article
19 of the 1985 European Structures Regulation (797/85) allowed Member States to pro-
vide funding for schemes that contributed toward the introduction or continued use of
agricultural production practices while being compatible with the requirements of con-
serving the natural habitat and ensuring an adequate income for farmers. This led, for
instance, to the introduction of the Environmentally Sensitive Areas scheme in England
following the implementation of the Agriculture Act 1986, as well as to schemes in other
parts of the United Kingdom and other European Member States. It was generally not
taken up by southern Member States. The Regulation did not provide for any financial
support from the European Community budget, but this was changed in Regulation
1760/87, which provided for a maximum of 25% reimbursement from the European
Agricultural Guidance and Guarantee Fund (FEOGA).
Although the focus of these schemes has been on influencing the way in which farm
land is managed, schemes have had varying objectives across different countries and
contexts (Baldock and Lowe 1996). In some cases, essentially on the extensive margin,
the concern has been to maintain agricultural use and prevent abandonment of mar-
ginal areas. This has been a particular priority in parts of France, the Alpine countries,
Scandinavia, and parts of southern Europe. In other locations, concern has been more
to mitigate the effects of agricultural intensification, along the intensive margin, par-
ticularly of pollution associated with livestock wastes, inorganic fertilizers, and pesti-
cides. But there are concerns, too, with intensification at the extensive margin where it
European Agri-Environmental Policy 585
threatens habitats and species that are associated with more extensive land uses. In the
United Kingdom, the focus has been particularly on the protection and enhancement of
landscapes and wildlife.
2. Rural Land Management in an

Old World
Agri-environment schemes operate through environmental contracts under which gov-

ernment offers payments to farmers who agree to undertake or not to undertake certain
farming practices or forms of management. Thus, the objective is to promote a rural
environment that is farmed in a particular way, almost invariably in a way that reflects
traditional and longstanding farming practices in that locality. This represents the pre-
dominant judgment that what is valued in the rural environment is a cultural landscape
that is a product of particular forms of agricultural management. This is a feature of
an “Old World,” where, over long periods of time, agricultural practices, environmental
habitats, and community arrangements have co-evolved to generate highly valued envi-
ronments (Hodge 2000).
In this respect, the rationale for the approach is consistent with the Organisation for
Economic Co-operation and Development (OECD) definition of multifunctionality.
This is typically defined in terms of two criteria: (1) jointness between commodity and
noncommodity outputs from agriculture and (2) the public good character of the non-
commodity output (OECD 2001a). Jointness is defined by the OECD (2001a, 16) when
a firm produces two or more outputs that are interlinked so that an increase or decrease
in the supply of one output affects the levels of the others. This may arise (OECD 2001a,
30; see also Blandford and Boisvert 2005) when there are technical interdependencies in
the production process, when outputs are produced from a nonallocable input, or when
outputs compete for an allocable input that is fixed at the firm level. An alternative argu-
ment concentrates on costs, so that jointness arises from economies of scope when the
costs of production are lower when two or more outputs are produced together by the
same firm. Hagedorn (2004) refers to this as “institutional jointness.” Publicness follows
the standard assumptions of nonrivalry and nonexcludability, although, in practice,
noncommodity outputs are generally not pure public goods (Cooper et al. 2009).
The justification for paying farmers for the improvement of the environment also
makes an assumption about the allocation of property rights to the landholder. This is
that landowners have a right to undertake agricultural production activities subject to
any laws that regulate land uses relating to limits on pollution or activities that might
impose costs on third parties. This defines a reference level of property rights (Hodge
1989; OECD 2001b) and a reference environmental standard that is associated with
it. Thus, the provision of a higher environmental standard, beyond Good Farming
Practice, represents a public good for which the supplier deserves payment. This thus
586 Ian Hodge
applies the “provider gets principle” (OECD 1999), an inverse of the “polluter pays prin-
ciple” in which landholders cause environmental impacts for which they do not hold
property rights. These principles are set out by the European Commission as a basis for
the implementation of agri-environment schemes in the European Union.1
3. Agri-Environment Schemes Operated in

the European Union
Agri-environment schemes became mandatory on Member States in the European

Union under a package of measures introduced in 1992 (Council Regulation 2078/92)
accompanying the MacSharry CAP reforms. These reforms introduced partial decou-
pling through direct arable area and livestock headage payments. The schemes under this
Regulation developed from the previous agri-environment initiatives but the Regulation
made it obligatory for all Member States to implement a national agri-environment pro-
gramme, including a range of measures to generate “positive effects on the environment
and the countryside” (Regulation Article 2(1)). These included measures to reduce sub-
stantially the use of fertilizers and plant protection products, to change to more exten-
sive forms of crop production, to reduce the numbers of sheep and cattle per forage area,
to use other farming methods compatible with the requirement of the protection of the
environment and natural resources, to rear animals of local breeds in danger of extinc-
tion, to ensure the upkeep of abandoned farmland or woodlands, to set aside farmland
for at least 20 years, and to manage land for public access and leisure activities. Member
States were expected to include all of these measures unless there was a clear reason why
they should not apply. The hope was that the implementation of the Regulation would
lead to a reduction in the intensity of agriculture over a significant area of land, thus
helping to stabilize or reduce production and to ease the wider pressures on the CAP
(Baldock and Lowe 1996). The rate of take up of measures under the Regulation differed
widely between different Member States. Whitby (1996) reports data from 1995 on the
% of Utilised Agricultural Area entered into schemes, varying between Austria at 91%
and Netherlands at 3.3%. The majority of countries achieved between 12 and 25%.
Under the Agenda 2000 reforms of the CAP, provision for agri-environment schemes
was included under the Rural Development Regulation as part of Pillar II of the CAP;
Pillar I covering market support measures and direct aids, and Pillar II covering rural
development. The arrangements for agri-environment schemes were set out initially
for the period 2000–2006 in the Rural Development Regulation 1257/1999 and then,
1 http://ec.europa.eu/agriculture/envir/cap/index_en.htm The legal obligations that form the
reference level for the agri-environment measures are indicated in Article 39.3 of Regulation No
1698/2005 in terms of the relevant mandatory standards.
3500
3000
Million euros 2500
2000
1500
1000
500
0
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
FIGURE 22.1 EU budget spending on agri-environment schemes 1993–2011.
Source: European Commission, Directorate General for Agriculture and Rural Development (2005)
Agri-environment measures: Overview on General Principles, Types of Measures and Application.
Unit G-4—Evaluation of Measures applied to Agriculture, Studies.
(Initial source: EAGGF Guarantee Section, budget execution), updated from Jukka Niemi (2012), personal
communication.
subsequently for the period 2007–2013, under Regulation 1698/2005. Under these
Regulations, Member States have been required to implement Rural Development
Programmes approved by the European Commission. Figure 22.1 illustrates the growth
of spending on agri-environment schemes in Member States between 1993 and 2011.
This reflects the total spending across the numbers of Members States in the EU at any
particular time. The rapid growth in 2008 reflects the increased variety of measures
available in the current programming period.
Council Regulation 1698/2005 provided for support for rural development by the
European Agricultural Fund for Rural Development (EAFRD) in order to promote sus-
tainable rural development throughout the Community. Its approach should be com-
plementary to the market and income support policies of the CAP. It established four
Axes, representing coherent groups of measures with specific goals:
xis 1: Improving competitiveness of the agricultural and forestry sector

A
Axis 2: Improving the environment and countryside
Axis 3: The quality of life in rural areas and diversification of the rural economy
Axis 4: Leader
The European Commission set rules as to the required balance in expenditure to be

achieved between these goals. Leader represents area-based bottom-up local action
projects and is an approach to be taken in the implementation of a proportion of proj-
ects under the first three Axes. At least 5% of total expenditure across all axes must be
via Leader. Agri-environment schemes are included under Axis 2, which also includes
other measures targeting the sustainable use of agricultural and forestry land, such as
588 Ian Hodge
Agri-environment payments
Modernization of agricultural holdings
Natural handicap payments (other areas)
Natural handicap payments (mountain areas)
Adding value to rural products
Infrastructure for development and…
Implementing local development strategies
Basic services in rural areas
Village renewal and development
Setting up of young farmers
Early retirement
First afforestation of agricultural land
0 5,000 10,000 15,000 20,000 25,000
FIGURE 22.2 Planned EAFRD expenditure on main Rural Development measures in the

European Union 2007–2013 (million €).
Source: European Commission (2010). Rural Development in the European Union—Statistical and Economic
Information—Report 2010 Chapter 3. Overview of the EU Rural Development Policy 2007–2013,
http://ec.europa.eu/agriculture/agrista/rurdev2010/ruraldev.htm
250
200
150
100
50
0
Spain
Romania
France
the Netherlands
Bulgaria
Portugal
Latvia
Poland
Lithuania
Slovakia
Greece
Estonia
Belgium
Czech Republic
United Kingdom
Ireland
Cyprus
Sweden
Slovenia
Luxembourg
Finland
Malta
Austria
Denmark
Germany
Hungary
Italy
FIGURE 22.3 Total planned public expenditure on agri-environment measures 2007–2013.

Mean expenditure in € per ha per annum over total Utilised Agricultural Area in 2007.
Source: European Commission (2010) Rural Development in the European Union—Statistical and
Economic Information—Report 2010. ANNEX G—EAFRD—Overview of the financial plans. UAA from
Eurostat Agricultural Statistics
payments in areas with natural handicaps, nature conservation areas, and animal welfare
payments. At least 25% of the total contribution from EAFRD must be applied to Axis
2. The maximum contribution from EAFRD toward total expenditure on Axis 2 is 80%
of eligible public expenditure in regions covered by the Convergence Objective, the least
developed Member States and regions, and 55% in other regions. Agri-environment
payments to farmers or other land managers are granted on a voluntary basis, covering
only those commitments that go beyond the relevant mandatory standards and beyond
minimum requirements for fertilizer and plant protection product use. Payments are
granted annually, generally over periods of five to seven years, and covering additional
costs and income foregone. Where necessary, they may also cover transactions costs.
The Rural Development Regulation also states that, where appropriate, the beneficiaries
may be selected on the basis of calls for tender, applying criteria of economic and envi-
ronmental efficiency.
Planned expenditures over the 2007–13 programming period are published by the
European Commission (2010). Total Community support planned for rural develop-
ment across the 27 Member States between 2007–2013 amounted to some €96.3 billion,
of which just under a quarter was allocated to agri-environment measures. The distribu-
tion amongst the major rural development measures is shown in Figure 22.2.
The level of agri-environment expenditure planned in Rural Development
Programmes varies considerably between the Member States. Overall, just over half of
expenditure (52%) on Axis 2 is directed to agri-environment payments, but the propor-
tion ranges between, say, 86% in Belgium and 70% in the United Kingdom, as compared
with 26% in Portugal or 28% in Slovakia. The intensity of expenditure per hectare on
average across the total Utilised Agricultural Area is shown in Figure 22.3. This shows
Austria at one extreme with over €200/ha and Spain and Romania at the other with
about €15/ha.
Purvis et al. (2009) estimate that there are probably in excess of 355 EU-funded
agri-environment schemes, varying widely in terms of structure, scope and focus. The
issues covered fall predominantly across three general headings: natural resources, bio-
diversity and landscape quality. A fuller list of topics is shown in Table 22.1.
More details of the ways in which agri-environment schemes have been implemented
can be illustrated through the experiences in Austria and England.
4. ÖPUL in Austria
Austria has for a long while placed considerable emphasis on supporting land man-
agement by its farmers. Darnhofer and Schneeberger (2007) discuss the context and
operation of agri-environment measures in Austria. Farms are generally small, with
an average size of 17ha, and nearly 70% of the total agricultural area is located within
the Less Favoured Area and more than half in mountainous areas. Darnhofer and
590 Ian Hodge
Table 22.1 Frequent topics covered in EU agri-environmental (AE) schemes

AE issue Individual topics % in sample of 244 schemes*
Natural Resources Aspects of soil quality and stability 63%

Aspects of water quantity and quality
Aspects of air quality
Biodiversity Conservation of wildlife species and 73%
habitats
Protection and utilization of functional
biodiversity within farming systems
Maintenance of genetic diversity,
particularly crop varieties and livestock
breeds
Landscape quality Aesthetic appearance and cultural historic 50%
value of the countryside
Multifunctional (amenity, recreational
and educational) value and use of the
countryside
Other Food quality and safety 12%
Public health and animal welfare
Controlling natural hazards
* Individual schemes target more than one issue and so total sums to more than 100%.
Source: Purvis et al. (2009).
Schneeberger (2007, 362) comment that “despite these unfavorable farming condi-
tions, suitable management of Alpine grasslands is imperative; they contribute to the
attractiveness of the mountainous areas and to their recreational value, which are cru-
cial for the tourism industry. Their sensitive ecosystem harbors an endemic flora and
fauna, which contributes to Europe’s natural heritage, and plays an important role in
the prevention of avalanches and landslides. Alpine agriculture is thus a typical example
of ‘multifunctional agriculture,’ as its nonagricultural value in terms of environmental
benefits and maintenance of the rural infrastructure may be higher than its agricultural
production value.” Environmental considerations gained importance for agricultural
policy in the early 1970s and were a central issue in the negotiations for Austria to join
the European Community in the early 1990s.
Since 1995, agri-environmental policy has been implemented through ÖPUL, the
Austrian programme to promote agricultural production methods compatible with the
requirements of the protection of the environment, extensive production and the main-
tenance of the countryside. Participation in ÖPUL is available on a voluntary basis to
all privately owned farms with a minimum size of 2 ha. It includes measures that are
available to all farmers and others that are offered within specific regions. Currently
under the Austrian Rural Development Programme (Netzwerk Land 2011) the scheme
has continued to cover a very high proportion of the total agricultural area, at around
90%, and includes over 70% (118,000) of all farms. The programme includes 29 mea-
sures supporting extensive and environmentally friendly management of the whole
farm, cultivated landscape and nature conservation, and soil, climate and water pro-
tection. Expenditure on ÖPUL represents about 48% of total expenditure on Rural
Development and 28% of total CAP expenditure in Austria. The average payment under
ÖPUL amounts to some €220/ha/annum, of which about 50% of the funds come from
the European Union.
ÖPUL may be expected to have a major impact on the rural environment. 17% of
farmland is organic and 73% of the land is subject to reduction and prohibition of the
use of yield increasing inputs (Puchta 2011). However measurement and evaluation of
policy effects is difficult due to the problems in identifying causality between the spe-
cific policy and the environmental conditions. At the same time, ÖPUL represents an
important source of income for Austrian farmers. Darnhofer and Schneeberger (2007)
emphasize the importance of public support for proposed policies, what they term
“political efficiency”, and comment that this does not generally coincide with economic
efficiency. The political process “was guided by the choice to keep the administration
costs as low as possible, so that more funds would be available for farmers” (p. 373).
5. Environmental Stewardship in England
Agri-environment policy in the UK is operated through separate Rural Development

Programmes in the areas of England, Scotland, Wales and Northern Ireland.
Environmental Stewardship, the current scheme in England, was introduced in 2005.
There have been some clear shifts in the orientation of agri-environment policy since
its initial aim of restraining agricultural intensification to protect valued aspects of the
rural environment. The first implementation in 1986 was in designated Environmentally
Sensitive Areas (ESAs) whose objective was “to help conserve those areas of high land-
scape and/or wildlife value which are vulnerable to changes in farming practices by
offering payments to farmers willing to maintain or introduce environmentally benefi-
cial farming practices” (MAFF 1989). At that stage the focus was on avoiding change
taking place, especially on areas of land being brought into more intensive production
as a consequence of the support mechanisms offered under the CAP. ESAs were targeted
as representing particularly significant landscape values and vulnerable habitats, such as
wetlands and extensive grasslands. Agri-environment schemes extended the “voluntary
principle” that had been implemented in UK policy for management agreements within
Sites of Special Scientific Interest (SSSIs) under the Wildlife and Countryside Act 1981
and subsequently incorporated into European rules.
Over time the emphasis shifted from simply preventing change toward seeking envi-
ronmental enhancement, especially to restore environmental values that had been lost
as a consequence of agricultural intensification and technical change in the past. This
592 Ian Hodge
development in policy was reflected in the introduction of the Countryside Stewardship

Scheme (CSS), a selective scheme made available to land holders throughout the coun-
try at the start of the 1990s, and in the development of higher tier contracts to promote
environmental enhancement within the ESAs.
Environmental Stewardship replaced previous schemes in 2005. It is comprised of two
elements, Higher Level Stewardship (HLS) and Entry Level Stewardship (ELS). While
HLS essentially continues with the CSS approach, ELS is a major innovation. ELS adopts
a “broad and shallow” approach with the objective of enrolling the majority of agricul-
tural land into the scheme. This may then be seen as a third phase in agri-environment
policy that extends payments beyond the primary concentration on the extensive
margin to include payments across all agricultural land areas to alter agricultural pro-
duction intensity along the intensive margin (Hodge and Reader 2010). Under ELS,
farmers can choose from a variety of land management options, such as hedge man-
agement or the introduction of buffer strips, for which they are awarded points. It is a
whole farm scheme and, in lowland areas, farmers need to attain an equivalent of 30
points per hectare for the total area of the farm, for which they are paid £30 per hectare
per annum. There is also an Organic Entry Level Stewardship (OELS) and an Uplands
Entry Level Stewardship (UELS) for farms in Less Favoured Areas. Current take up of
agri-environment schemes is shown in Table 22.2.
As a consequence of the introduction of ELS, over two thirds of agricultural land is
now covered by agri-environment contracts. This does not mean that all of this land is
subject to active agri-environmental management. Rather it represents the total areas
of the farms that have been entered into the scheme on which some such management
is being undertaken. However, it does mean that the managers of this area have had to
Table 22.2 Take up of agri-environment schemes in England, 2012

Annual
value of public
No. of expenditure Payment
Scheme Area (ha) % of UAA agreements £M £ per ha
CSS 169,805 1.8 6,025 37.9 223.2

ESA 339,382 3.7 5,700 28.4 83.7
ELS 5,332,639 57.4 39,455 152.9 28.7
OELS 343,483 3.7 2,367 27.6 80.4
HLS (Combined) 821,331 7,898 137.5 167.4
HLS (Standalone) 106,477 1.1 1,170 23.0 216.0
Total HLS 927,808 9,068 160.5 173.0
UELS 825,058 5,503 68.3 82.8
Overall total 6,291,785 67.7 54,717 407.3 64.7
Source: Natural England (2012) Land Management Update, April 2012.

consider how they manage the agri-environment on their farms and all of the environ-
mental features on the farms have had to be recorded and must be protected.
6. The Characteristics of European

Agri-environment Schemes
Agri-environment schemes have been developed around the world in a variety of forms.
Similar concerns for the quality of the rural environment have been the focus for initia-
tives dating back to the mid 1980s. For instance the Food Security Act of 1985 in the
United States authorized the Conservation Reserve Program, with a goal of retiring
45 million acres of highly erodible land. In Australia, within a very different agricultural
policy context, the Landcare movement dates back to 1986. But a number of aspects of
the approach that has been taken in Europe have been the particular subject of research.
This section focuses on these aspects.
7. Jointness
As mentioned above, much of what is valued in the European rural environment is a

product of specific agricultural systems and as such, policy is directed toward the
encouragement of the particular forms of land management and agricultural practices
that support these environmental values. In this respect the environmental values and
agricultural commodities are treated as joint products. A detailed study by Wätzold
et al. (2008) illustrates some common characteristics and challenges of European
agri-environment measures. Their study aims to estimate an “optimal level of spe-
cies conservation” for the Scarce Large Blue butterfly (Maculinea teleius), an endan-
gered meadow-dwelling species, in Landau in Germany. The females lay their eggs on
the Large Pimpernel (Sanguisorba officialis) plant. Caterpillars fall from the plant and
are carried away by red ants to their nests where they are fed by the ants over winter.
However, plant and ants require a particular mowing regime in order to survive and
this is no longer practiced due to the introduction of modern agricultural machinery
and simultaneous and regular mowing of the meadows. The optimal level of conserva-
tion depends both on the supply and demand for the species. The supply depends on
the effect of changing the mowing regime on the performance of the butterfly popu-
lation and the costs to the farm of changing practices. The level of costs incurred on
any particular farm depends on the situations of its individual meadows and farm busi-
ness. Demand is assessed in this study in terms of the willingness of the local popula-
tion to pay for larger numbers of butterflies. The analysis is conducted at a field level
to simulate cost-effective mowing regimes to deliver a given number of butterflies. The
594 Ian Hodge
authors conclude that from amongst the range of mowing regimes studied, the regime
generating the greatest number of butterflies was optimal. In this example, biodiversity
conservation requires the maintenance of a particular agricultural system that retains
meadows that are mowed at particular times. The biodiversity is thus jointly produced
with the agricultural production. The presence of the species is specific to a particular
locality and its value may be dependent on the preferences of a local population. The
success of the species depends on the way in which decisions across different farms and
fields are coordinated. The requirement for particular spatial configurations of agricul-
tural activity is also illustrated by Bamière et al. (2011) who have developed a spatially
explicit mathematical programming model for analysis of options for the conservation
of the Little Bustard (Tetrax tetrax). This bird requires the presence of extensive tem-
porary grasslands distributed throughout the landscape. This means that a least cost
option that concentrates grassland on the least profitable farms would be unlikely to
provide sufficiently disaggregated habitat.
Brunstad et al. (2005) model multifunctional agriculture in terms of its provision
of public goods of food security and landscape. While it is recognized that in practice
it is not possible to model all the attributes that enhance the value of the agricultural
landscape, such as openness, variation, biodiversity and type of agricultural technique,
the environmental benefit is assumed to be associated with the area of land under till-
age. They conclude that agricultural production would be sub-optimal in the absence
of policy toward the environment, although they judge that in practice the level of
subsidy offered in Norway exceeds that required to optimize environmental output.
Lankoski and Ollikainen (2003) model local land use with heterogeneous land quali-
ties. They include three agri-environmental externalities: biodiversity, landscape diver-
sity and nutrient run-off. Biodiversity is enhanced through the introduction of buffer
strips around field boundaries, the aesthetic value of landscape is promoted through
a diversity of land uses, and nutrient run-off depends on both fertilizer use and buffer
strips. Thus, environmental benefits are attained through the promotion of a diversity
of cropping and by means of buffer strips that reduce the area of land under production.
Their preferred policy instruments are a fertilizer tax and a buffer strip subsidy, both of
which reduce total agricultural production. The complexity of adjustments required in
promoting environmental quality is recognized by Miettinen and Huhtala (2004) who
model the relationship between cereal production and the numbers of grey partridges.
They show that farmers should increase the area under rye, reduce the use of herbicides
and limit the partridge hunting bag in recognition of the social benefits associated with
partridge conservation, but that this reduces the private returns to farming.
Peerlings and Polman (2004) investigate the joint production of milk, wildlife and
landscape services in Dutch dairy farming using a micro-econometric profit model. The
output of wildlife and landscape services is represented by the revenue received from
government and nature organizations for participation in agri-environment schemes.
They find that wildlife and landscape services compete with milk and other outputs,
i.e. producing more milk makes the production of wildlife and landscape services less
attractive. They also conclude that economies of scope exist on a small proportion of
farms, although in practice farms do not specialize, suggesting that there may be other
factors that are not taken into account in the model. Havlik et al. (2005) consider both
complementarity and competition between agricultural production and environmen-
tal goods. They note that with regard to grassland biodiversity, agricultural produc-
tion and environmental goods can be complementary over a certain range but compete
beyond this range and that this is the case in practice for pasture stocking intensity in
the Pyrenees. They analyze the position in two Environmentally Sensitive Areas, one
where the danger is of over-intensification and another at risk from land abandon-
ment. The provision of environmental goods is modelled by introducing constraints
into a mathematical programming model that represents the requirements of par-
ticular agri-environmental contracts, assuming that keeping to the conditions of the
agri-environmental contracts will generate the specified environmental goods. The
authors conclude that there is little justification for commodity-linked instruments,
noting that both complementary and competing relationships were observed within
even a relatively small region, so that commodity price increases would generate a loss
of biodiversity in some contexts.
There has been some discussion of the interactions amongst public good outputs.
Brunstad et al. (2005) consider both landscape preservation and food security and
Lankoski and Ollikainen (2003) include both biodiversity and landscape. But there
seems to have been little analysis that has modelled the complexities of the interactions
between different environmental outputs, suggesting a need to clarify the circumstances
under which environmental production takes place. More generally with regard to
jointness, it is clear that there are complementarities between the production of agri-
cultural commodities and environmental quality in certain circumstances but competi-
tion in others. It is not a simple relationship. These relationships vary between locations
and agricultural systems and within systems at different levels of production. General
support for the prices of agricultural commodities will be very unlikely to generate
consistent environmental improvements. Rather, agri-environmental policies need
to promote detailed changes in farm practices that are specific to local environmental
objectives and farming practices.
8. Asymmetric Information
Much of the early analysis of agri-environmental schemes focussed on the issue of

asymmetric information. Agri-environment schemes are generally implemented
through voluntary environmental contracts between a government agency and a group
of heterogeneous farmers and farms. In these circumstances, the farmer has more com-
plete information about the opportunity costs of adopting the requirements of the con-
tract and has the potential to hide actions as to whether or not the contract is being
complied with. This gives rise to the problems of adverse selection and moral hazard.
These issues have been extensively analyzed in the literature. Most of this work has been
596 Ian Hodge
theoretical, often simulating plausible parameter values. Moxey et al. (1999), White
(2002) and Ozanne and White (2007) have analyzed the incentive compatibility of
alternative mechanism designs based on combinations of either input quotas or input
charges and transfer payments. This work has been extended by Ozanne and White
(2007) who demonstrate that the two approaches lead to identical outcomes in terms
of abatement levels, compensation payments, monitoring costs, probabilities of detec-
tion and social welfare. Gren (2004) compares a uniform flat-rate agri-environment
payment to all farmers with a differentiated payment under conditions of private infor-
mation available to farmers’ on their individual costs of providing and managing their
land. This information is not available to the principal. Under the differentiated pay-
ments, payment level depends on the farmer’s cost type. She concludes that the gen-
eral analytical results are indeterminate in that the relative advantages of the two policy
designs depend on second derivatives of environmental land provision cost and benefit
functions. Canton et al. (2009) focus on the impact of spatial targeting and delegation
in mechanism design on overall efficiency. Spatial targeting can improve the informa-
tion available to the principal ex ante and so simplify the trade-off between allocative
efficiency and information rents. Delegation can be seen as a means of improving the
regulator’s information because local institutions may have a better knowledge of a
farmer’s characteristics. Their approach emphasizes the redistributive effects of disag-
gregated information structures with the most efficient farmers being most likely to be
negatively affected. One study that has analyzed a specific agri-environment scheme is
Quillérou and Fraser’s (2010) assessment of Higher Level Stewardship in England. They
find that, at the regional level, the enrolment of more land from lower payment regions
for a given budget constraint has reduced the adverse selection problem through con-
tracting a greater overall area and thus providing higher overall environmental benefit.
Further, the regulator’s allocation appears to reflect differences in environmental ben-
efits thereby also reducing the adverse selection problem.
Moral hazard and compliance monitoring have also been the subject of specific analy-
sis. Ozanne et al. (2001) develop a model that demonstrates that if monitoring costs are
negligible or fixed, or farmers are highly risk averse, the moral hazard problem can be
eliminated. However, if monitoring costs depend on monitoring effort and the degree
of risk aversion is low, only a second best solution can be obtained. Fraser (2004) has
analyzed the use of targeting to reduce moral hazard. Hart and Latacz-Lohmann (2005)
report observations that the predictions of models of moral hazard problems are not
consistent with what limited observations of actual experience are available, where a
combination of low fines, low rates of checking and relatively little cheating seems to be
the norm. This might reflect farmers misjudging their subjective evaluations of small
risks of detection, or else that some farmers are basically honest and simply do not
consider cheating as an option. They adopt the latter as an assumption and develop a
model that allows for a continuum of farmer compliance costs and in which they relax
the assumption that all farmers are profit maximizers. They note that in practice, given
multiple periods, the regulator has an opportunity to learn about the characteristics of
farmers and adjust behavior accordingly. The authors also draw attention to the fact that
in practice farmers are more likely to cheat at the margin rather than to blatantly not
comply at all, i.e. they may not fulfil the conditions of the contract to the letter and the
level of penalty will be graduated in relation to the offence. In these circumstances, even
farmers who cheat marginally may contribute to the fulfilment of the environmental
target. These circumstances make real world compliance monitoring more complex
than recognized in most models and it might also be noted that in practice the costs of
monitoring compliance vary between different contract requirements. For instance, it
is simpler to monitor farmer record keeping than it is to monitor the actual adoption of
land management practices in the field.
One approach toward the problem of asymmetric information lies in the use of auc-
tions or tendering as a mechanism for creating competition amongst farmers and get-
ting them to reveal information about their costs. In principle, auctions can reduce
information rents accruing to farmers and increase the cost-effectiveness of public
goods provision, although strategic bidding behavior and high transactions costs may
reduce efficiency (Latacz-Lohmann and Van der Hamsvoort 1998). Experimental work
suggests that conservation auctions outperform fixed price schemes in a one-shot set-
ting, but that with repetition the auction loses its edge (Latacz-Lohmann and Schilizzi
2007). However, while tendering has been applied in the US and Australia, and despite
the provision for the use of tendering in the European Regulation 1698/2005, there is
little empirical experience in the context of European agri-environment policies.
9. Entry into Schemes
The theoretical models of mechanism design almost invariably assume that farmers will
enter agri-environment schemes where the financial payment exceeds the opportunity
cost. However, in practice adoption depends on a much wider variety of factors and these
have been the focus of a large number of empirical studies (e.g.Wilson 1997; Wynn et al.
2001; Wilson and Hart 2001; Vanslembrouck et al. 2002; Dupraz et al. 2003; Wossink
and van Wenum 2003; Defrancesco et al. 2008; Hynes and Garvey 2009). Analysis tends
to be based on data collected in surveys of farmers enrolling into a particular scheme,
sometimes including nonparticipants, and the analyses test relationships with a wide
variety of potential influences, often in categories such as those adopted by Wynn et al.
(2001): physical farm factors, farmer characteristics, business factors and situational
factors. From amongst these, such factors as farm size, farmer age, information avail-
able, the ease with which the requirements of the scheme can be accommodated into the
farming system, farmers’ attitudes and experience with agri-environment schemes are
commonly included. Some studies have focussed on particular aspects of the decision to
enrol. Falconer (2000) has concentrated on the transactions costs facing farmers in join-
ing an agri-environment scheme, while Polman and Slangen (2008) included variables
representing trust in government and institutional design. Frondel et al. (2012) look at
the provision of information, pointing out that while not having an unambiguous effect
598 Ian Hodge
in either encouraging or discouraging entry, it may be expected to improve the qual-

ity of decisions and help to avoid mistakes. Various theoretical approaches have been
applied, such as Morris et al. (2000) who apply innovation decision theory or Beedell
and Rehman (2000) who apply the Theory of Planned Behaviour. It is clear that farmers’
decisions are affected by a wide variety of factors and it has also been suggested that par-
ticipants who are only motivated by financial incentives may be less effective environ-
mental managers than other participants. Along similar lines, Stobbelaar et al. (2009)
note the internal motivation of organic farmers for nature conservation who were more
likely to internalize the goals of environmental policy schemes. This implies an objective
to build up cultural and social capital (Burton and Paragahawewa 2011).
These studies of actual behavior ex post are limited in terms of the scheme options
that can be considered. A small number of studies have used choice experiments in
order to explore farmers’ preferences for alternative scheme design (Ruto and Garrod
2009; Espinosa-Goded et al. 2010; Christensen et al. 2011). These studies illustrate the
ways in which farmers are willing to trade the level of payment for particular scheme
requirements and suggest levels of payment that may be required in order to attract
farmers to adopt agri-environment measures.
10. Transactions Costs
Environmental policy analysis often concentrates solely on the direct or opportunity

costs to the firm of undertaking the changes required by the policy. However, this
neglects the transaction costs faced by government in identifying and selecting policy
options, disseminating information to potential participants, negotiating and imple-
menting contracts and monitoring and enforcing compliance. Transactions costs
vary substantially depending on the type of policy being considered (Rørstad et al.
2007) and can represent a significant proportion of total policy costs. This is the case
for agri-environment measures. However, they are difficult to measure and while there
is often a lack of evidence relating to them, one study by Falconer and Whitby (1999)
reported costs varying across a wide range of agri-environment measures in Europe of
between 30% and 80% of the total policy cost.
Evidence indicates that the administrative costs when schemes are first implemented
are initially high but that they fall steadily and significantly as schemes become more
established. Statistical analysis of the administrative costs of the Environmentally
Sensitive Areas in England (Falconer et al. 2001) also suggests administrative econo-
mies of size related to scheme participation. Mettepenningen et al. (2011) have analyzed
transactions costs of agri-environment schemes in nine European countries based on
stakeholders’ perceptions. They conclude that the complexity of the schemes, the num-
ber of agri-environmental measures that need to be designed and the required precision
of the measures are the major influences on costs. They also note that a number of stake-
holders believed that high transactions costs do not correspond to their environmental
benefits. Falconer and Whitby (1999) suggest that transaction costs could be reduced
by extending the contract period, by unifying schemes, and by investigating alternative
contract mechanisms such as auctions.
McCann et al. (2005) have identified a number of categories for the transactions costs
of environmental policies: research and information, policy enactment, policy design
and implementation, support and administration, contracting, compliance monitor-
ing/detection, and prosecution/enforcement. The relatively high transactions costs in
agri-environment schemes reflect asset specificity, such as the variation in the potential
value of outputs between sites, dependence on specific inputs or the influence of the
quality of the labor input from particular farmers; the relative infrequency of transac-
tions and farmers’ lack of familiarity with them; and the uncertainty in terms of what
a farmer might be contracted to do and what the outcomes of those actions might be
(Coggan et al. 2010). A general thrust in transaction cost economics is that the purpose
of institutions is to minimize transactions costs (Williamson 1996). However, in design-
ing agri-environment schemes it is clear that alternative contractual arrangements, and
hence different levels of transactions costs, have important implications for the value
of the public goods that are generated. What is important is not the absolute level of
transaction costs, but rather the return that they bring in terms of enhanced value of
environmental outputs.
The essence of transactions costs lies in the acquisition of information. In a market
transaction, each party will assess the value of the transaction to them and negotiate a
contract accordingly. In this context, there is effectively no “exchange” and government
is acting on behalf of a wide range of potential beneficiaries who could benefit from
changes in land management. Thus government needs information on the costs and the
potential outcomes of the possible changes in management and the values attached to
those outcomes by the general public. Information is required in order to be able to pre-
dict what value can be expected to be generated from any potential environmental con-
tract, taking account of the particular nature of the land being managed, its spatial and
management context, the relationships between management actions and environmen-
tal outcomes, and the spatial and temporal context within which those outcomes will
arise. There is thus a significant information requirement for policy design and imple-
mentation, and there will still remain a high degree of uncertainty. A similar argument
applies to the completeness of contracts. More complete contracts may provide more
specific direction to land managers in particular circumstances, reducing uncertainty
and increasing the expected value of the environmental outcome. However, the devel-
opment and implementation of such contracts will, again, increase transactions costs.
There is thus a trade-off between the transactions costs of gaining better informa-
tion and the capacity to design, implement and monitor higher value environmental
contracts. In principle, the optimal level of transactions costs will be where the marginal
cost of obtaining better information is equal to the marginal value of the environmental
improvement attained, with respect to each of the categories of transaction cost noted
above. In practice, measurement of both costs and benefits is challenging, but it is an
important issue on which there has been rather little empirical research. There appears
600 Ian Hodge
to have been no analysis that has attempted to identify the optimal level of transactions
costs in agri-environment schemes.
11. Scheme Effectiveness and Evaluation
Assessment of agri-environment schemes is important in determining appropriate allo-

cations of public finance and in guiding the developments in mechanism design over
time. While noting progress in the development of agri-environment schemes, the
European Court of Auditors (2011) has recently criticized schemes for a lack of clear
objectives, insufficient differentiation of payments between farmers to reflect local
conditions, and a lack of application of procedures to select projects that represent best
environmental value for money. Evaluation faces many complex challenges, including a
lack of clear stated objectives for the policy, the limited availability of data, the challenge
of identifying causality, in defining a reliable counterfactual and in assessing addition-
ality. Some analysis that has been undertaken focuses somewhat simplistically on the
numbers of farmers participating in schemes and on the changes in farm management
practices, without evidence as to the extent to which these changes do in practice deliver
environmental benefits or the importance of these benefits. Assessment of the indirect
effects of schemes represents a further challenge for analysis.
Much of the effort that has been made to assess the impacts of agri-environment
schemes has concentrated on the effects on biodiversity and the evidence remains
controversial. A paper by Kleijn et al. (2001) suggesting that some schemes simply do
not achieve their objectives or possibly even have adverse ecological effects has been
widely cited, although it was countered by Stoate and Parish (2001) and Carey (2001)
who argued that there was other evidence that some schemes are successful. Some stud-
ies of initiatives for the conservation of particular species have demonstrated success,
such as for the cirl bunting (Peach et al. 2001) or corn bunting (Perkins et al. 2011).
Studies of schemes have demonstrated positive long-term impact in particular contexts
(e.g. Taylor and Morecroft 2009). Kleijn and Sutherland (2003) reviewed the available
evidence on European schemes and concluded that, while just over half of the studies
found an increase in species richness or abundance, research design was often inad-
equate to provide reliable results so that they could not reach a general judgment on
the effectiveness on agri-environment schemes. They did not assess potential benefits
other than biodiversity, such as reduced emissions or landscape enhancement. Kleijn
et al. (2006) reviewed agri-environmental schemes in five European countries conclud-
ing that in all countries agri-environment schemes had marginal to moderately positive
effects on biodiversity but that rare species benefited less often. Most recently Batáry
et al. (2011) have undertaken a meta-analysis of mostly European studies. They con-
clude that agri-environmental management effectively enhances species abundance in
croplands, and enhances both species richness and abundance in grasslands, regard-
less of landscape context. Whittingham (2011, 509) has commented recently that
“European agri-environment schemes have so far delivered only moderate biodiversity

gains.” However, under a range of circumstances, they can achieve substantial benefit
to both biodiversity and ecosystem services. He also observes differences in the efficacy
of schemes depending on the species/taxon concerned: plants show the strongest posi-
tive responses, followed by invertebrates, with birds and mammals showing the lowest
responses.
Burrell (2011) has recently provided a conceptual framework for the evaluation of
agri-environment schemes, differentiating between administrative, scientific and
economic approaches. Within the European Union, agri-environment schemes,
as funded under the Rural Development Regulation, are required to be evaluated
under the European Commission’s framework for evaluating rural development
policies (European Commission 2006). Höjgård and Rabinowicz (2011) have iden-
tified some weaknesses in the evaluation procedure and suggested potential improve-
ments. Primdahl et al. (2010) have reviewed the use of impact models in supporting
the design, implementation and evaluation of agri-environment schemes in the EU.
Purvis et al. (2009) have proposed a standardized approach to evaluation in terms of
an “Agri-environmental Footprint Index” (AFI). This establishes a common framework
within which the characteristics of particular schemes may be identified. Stakeholders,
experts and farmers may be involved in the identification of suitable indicator variables
which are then standardized and weighted. The AFI score for a specified sample of farms
is calculated by multiplying the indicator scores by the agreed weights. Such an approach
faces challenges in terms of data availability, the subjectivity of the weighting, and in
assessing the causality and additionality of the scheme incentives. However, it can offer
a common framework within which to assess scheme performance and to aid improved
understanding and wider debate about the implementation of agri-environment
schemes within a particular context. Westbury et al. (2011) suggest that the same meth-
odology could also be used for more routine monitoring of the environmental perfor-
mance of farming systems.
Some methods have been developed in order to generate more reliable evalua-
tions. In principle, a counterfactual might be based on the performance of a control
group of farms that do not participate in the agri-environment scheme. However, in
practice there will be selection effects in that the farms that do participate are not the
same as the farms that do not. Thus any differences represent some combination of
the causal effect of the programme and the selection effect. Pufahl and Weiss (2009)
have used a semi-parametric propensity score matching estimator combined with
a difference-in-difference approach to evaluate agri-environment programmes in
Germany. They find a positive effect on the area in agricultural use and a reduction in
chemical usage per hectare. Their analysis also shows differences amongst individual
farms but these were not addressed in detail. Chabé-Ferret and Subervie (2011) have also
used difference-in-difference analysis to estimate the causal effect of agri-environment
programmes in France, finding that the windfall effects of the programmes depend on
the specific requirements of the particular program. They then seek to integrate the
results into a cost-benefit analysis.
602 Ian Hodge
A further complication in evaluation is that the value of changes on any particular

farm may depend on its local landscape context and on the changes being made on other
farms nearby (Concepción et al. 2008). Schouten et al. (2013) have developed a spatially
explicit agent-based model for an area in the Netherlands. The agricultural landscape is
modelled as an agent-based system, taking account of both the farmers’ behavior and
the spatial configuration of the landscape. Their results indicate that when policy mak-
ers want to achieve the highest contribution to the spatial habitat network they should
consider spatially differentiated payments. They point out that their model does not
take account of transactions costs, although suggest that an auction mechanism might
provide a way of allocating payments on the basis of farm opportunity costs.
The environmental benefits arising from agri-environment schemes have been the
focus for a substantial number of economic valuation studies. Valuation has had to
rely largely on stated preference techniques, earlier studies using contingent valuation
(Garrod et al. 1994, Garrod and Willis 1995), while more recently use has been made of
choice experiments (Campbell 2007), sometimes comparing the two (Hanley et al. 1998).
Hynes et al. (2011) compare a holistic valuation of landscape using contingent valuation
with a valuation of landscape attributes using a choice experiment. They find an insig-
nificant difference between the two approaches. Madureira et al. (2007) have reviewed
valuation studies conducted in France, Germany and Portugal. They find a predomi-
nance of stated preference methods and a focus on a regional scale. However, they con-
clude that the information generated is not widely used by policy makers. Campbell et al.
(2009) have conducted a choice experiment for the value of rural landscape in Ireland
and then interpolated willingness to pay estimates (WTP) for the whole of the coun-
try. They find that WTP for rural landscape declines considerably from the rural west
of Ireland to the more urbanized and modern farm landscapes of the east. They suggest
that their results indicate that landscapes are valued primarily with regard to their active
use rather than just for their existence. Garrod et al. (2012) used a choice experiment
to assess whether individual preferences for the environmental benefits associated with
Environmental Stewardship vary across landscape types. They find that there is spatial
heterogeneity of preferences as well as a preference toward benefits delivered closest and
most accessibly to where respondents live. They conclude that spatial targeting should
also take account of the size of local populations. While considerable progress has been
made in the methodological developments for valuation, less attention has been given
to combining valuation studies with critical analyses of the issues of the counterfactuals
and additionality of agri-environment schemes (Hodge and McNally 1998).
12. Prospects for Agri-environment

Schemes
Agri-environment policy represents a new challenge to governments: to re-create or

even create new actively managed rural landscapes that generate complex mixes of
private and public goods. It takes environmental policy into a new era, beyond the regu-
lation of environmental costs to the delivery of environmental benefits. It can be seen
as the forerunner of the increasingly pervasive discussion of policies for the provision
of payments for ecosystems services. Agri-environment policy has seen considerable
development since its introduction some 25 years ago. The initial concentration on
restraining pressures for agricultural intensification has moved on toward more general
policies that aim to promote environmental enhancement. At the same time, follow-
ing decoupling of the CAP in 2005, more effort will be required from agri-environment
policy to hold land in production where the greater environmental threat is from land
abandonment (Renwick et al. 2013). The longer term position will become clearer
once the nature of the CAP to be implemented beyond 2013 is determined (European
Commission 2011).
At the same time, there are likely to be increased pressures in the future on the
management of rural land to mitigate and adapt to climate change and to do more to
promote resource conservation. It seems reasonable to expect that the world will see
generally higher and more variable commodity prices and that at the same time there
will be increased pressures to reduce the levels of public expenditure. These circum-
stances set a number of challenges for the further development of agri-environment
policy in that they extend the range of outcomes that will be sought, increase the
opportunity costs of implementing policy and reduce the public resources available to
support it. These conditions may become more apparent as the wider pressures and
constraints bind more tightly. Indeed, it might be questioned whether the use of public
funds at this scale to deliver rural public goods is to be a permanent feature of European
rural policy.
There is then a clear logic for seeking both to increase the efficiency of
agri-environment schemes while at the same time looking for alternative means by
which the required environmental standards might be delivered. The issue of efficiency
relates both to precision, in terms of the standard conditions for optimality (Vatn
2002), as well as to the optimal level of transactions costs. The evidence indicates that
agri-environment schemes can have beneficial environmental impacts, but that they
do not always do so. It is clearly important to do more to unravel which approaches
are or are not successful, from both environmental and socioeconomic perspectives.
Herzog (2005) discusses agri-environment schemes as landscape experiments, sug-
gesting the potential for research based on more formal experimental designs to test
the effects of alternative mechanisms. Whittingham (2011) argues for adaptive man-
agement, an iterative approach to decision-making that learns from the evidence that
is accumulated over time. This might suggests more local approaches to governance
through socioecological adaptive co-management (Hodge 2007). There is potential for
institutional analysis of alternative arrangements whereby local communities may be
able to deliver such management. Research has indicated potential ways in which effi-
ciency might be increased, including clarifying objectives and focussing on what soci-
ety judges to be the highest priorities, targeting schemes more directly, both in terms
of specific public goods and in terms of specific locations and co-ordination across
space, and introducing more competition in the allocation of contracts. Economic
604 Ian Hodge
theory points to potential advantages from basing payments on results rather than on
the costs of making standard changes to farming systems that are expected to gener-
ate the desired outcomes (Schwarz et al. 2008; Matzdorf and Lorenz 2010). This raises
questions as to the identification appropriate indicators (Hasund 2013) and the treat-
ment of risk. Burton and Schwarz (2013) argue that the novel risks associated with
payments for the provision of environmental goods might promote cultural/social
capital amongst farmers who develop new approaches and share knowledge about the
ecological production function. There is potential to model many of these issues as
a method of evaluating different policy approaches. In the pursuit of more effective
policies, it will be important to do more work on the scale at which agri-environment
contracts should be implemented, such as looking at the potential for closer coopera-
tion both amongst farmers and between different interest groups, especially at a local
scale (Franks 2010), and on the interactions and trade-offs among different ecosys-
tems services. Emery and Franks (2012) and Franks and Emery (2013) have examined
farmers’ willingness in principle to collaborate in agri-environment schemes and the
actual experience to date with the opportunities that are available in existing schemes.
The origins of agri-environmental policy have colored the way in which the policy
has been developed. It has generally been viewed as an offshoot of agricultural policy,
beginning with the farmer, rather than as a separate rural environmental policy, start-
ing from the objective of achieving environmental change. Whether or not it is viewed
suspiciously as a policy to disguise agricultural subsidies (Anderson 2000), support for
farm businesses remains at least as an implicit objective. More analysis is required on
the incidence and distribution of the costs of implementing agri-environment manage-
ment at the farm level. Jointness is used as a rationale for the policy in general rather
than being considered simply as an argument in favor of a particular approach to imple-
mentation in certain circumstances. An ecosystems services perspective might chal-
lenge some of the assumptions that have generally been accepted in designing policy
approaches. This raises more general questions as to the sort of agricultural systems that
can be most effective in delivering alternative combinations of ecosystem services and
commodity outputs.
There are some limits that are fundamental to the use of environmental contracts in
agri-environment schemes in their present form (Hodge 2001). These include: the prob-
lems in defining and measuring the outputs that are demanded, the near impossibility of
defining farm practices that are best suited to the delivery of environmental benefits in
individual farm circumstances, the subsequent problem of writing these as conditions
into enforceable contracts, the inevitability of some degree of asymmetric information,
and the problem that any fixed term contract will come to an end, with uncertainty as
to how land management will change in the subsequent period (Whitby 2000). There is
then an uncertainty as to the “ownership” of the environmental enhancements that have
been achieved through the expenditure of public funds. Environmental contracts com-
moditize the provision of environmental services and hence may crowd out (Frey 1997)
a culture of stewardship (Colman 1994) that might anyway have delivered at least some
of the environmental benefits at zero public cost. This might make the introduction of
agri-environmental payments effectively irreversible. These are questions for social and
institutional analysis. Environmental contracts promote a culture of service delivery
rather than entrepreneurship. Agri-environmental policy needs to be considered along-
side other policy approaches of regulation, conservation covenants, land ownership and
nonprofit and community organizations (Hodge 2001). These other approaches have
their own strengths and weaknesses, but a comprehensive agri-environmental policy
will draw from a wider range of policy mechanisms and resources than has been the case
to date. These are areas for further work.
References
Anderson, K. 2000. Agriculture’s “multifunctionality” and the WTO. Australian Journal of
Baldock, D., and P. Lowe. 1996. The development of European agri-environment policy. In
The European environment and CAP reform, ed. M. Whitby, 8–25. Wallingford, UK: CAB
International.
Bamiere, L., et al. 2011. Farming system modelling for agri-environmental policy design: The
case of a spatially non-aggregated allocation of conservation measures. Ecological Economics
70: 891–899.
Batáry, P., A. Báldi, D. Kleijn, and T. Tscharntke. 2011. Landscape-moderated biodiversity
effects of agri-environmental management: A meta-analysis. Proceedings of the Royal Society
of London B 278(1713): 1894–1902.
Beedell, J., and T. Rehman. 2000. Using social-psychology models to understand farmers’ con-
servation behaviour. Journal of Rural Studies 16: 117–127.
Blandford, D., and R. Boisvert. 2005. Non-trade concerns: Reconciling domestic policy objec-
tives with trade liberalisation. International Journal of Agricultural Resources Governance
and Ecology 4(3/4): 277–291.
Bowers, J., and P. Cheshire. 1983. Agriculture, the countryside and land use: An economic cri-
tique. London: Methuen.
Brunstad, R., I. Gaasland, and E. Vardal. 2005. Multifunctionality of agriculture: An inquiry
into the complementarity between landscape preservation and food security. European
Review of Agricultural Economics 32(4): 469–488.
Burrell, A. 2011. Evaluating policies for delivering agri-environmental public goods. Paper pre-
sented at OECD Workshop on the Evaluation of Agri-environmental Policies, Braunschweig,
June 20–22.
Burton, R., and U. Paragahawewa. 2011. Creating culturally sustainable agri-environment
schemes. Journal of Rural Studies 27: 95–104.
Burton, R., and G. Schwarz. 2013. Result-oriented agri-environmental schemes in Europe and
their potential for promoting behavioural change. Land Use Policy 30: 628–641.
Campbell, D. 2007. Willingness to pay for rural landscape improvements: Combining mixed
logit and random-effects models. Journal of Agricultural Economics 58(3): 467–483.
Campbell, D., W. Hutchinson, and R. Scarpa. 2009. Using choice experiments to explore the
spatial distribution of willingness to pay for rural landscape improvements. Environment
and Planning A 41: 97–111.
606 Ian Hodge
Canton, J., S. De Cara, and P-A. Jayet. 2009. Agri-environment schemes: Adverse selection,
information structure and delegation. Ecological Economics 68: 2114–2121.
Carey, P. 2001. Schemes are monitored and effective in the UK. Nature 414: 687.
Chabé-Ferret, S., and J. Subervie. 2011. Estimating the causal effects of the French
agri-environmental schemes on farmers’ practices by difference in difference matching.
Paper presented at OECD Workshop on the Evaluation of Agri-environmental Policies,
Braunschweig, June 20–22.
Christensen, T., A. B. Pedersen, H. O. Nielsen, M. R. Mørkbak, B. Hasler, and S. Denver, S. 2011.
Determinants of farmers’ willingness to participate in subsidy schemes for pesticide-free
buffer zones—A choice experiment study. Ecological Economics 70: 1558–1564.
Coggan, A., S. Whitten, and J. Bennett. 2010. Influences of transactions costs in environmental
policy. Ecological Economics 69: 1777–1784.
Colman, D. 1994. Ethics and externalities: Agricultural stewardship and other behaviour.
Concepción, E., M. Diaz, and R. Baquero. 2008. Effects of landscape complexity on the ecologi-
cal effectiveness of agri-environment schemes. Landscape Ecology 23: 135–148.
Cooper, T., K. Hart, and D. Baldock. 2009. The provision of public goods through agriculture in
the European Union. Report prepared for DG Agriculture and Rural Development. Contract
30-CE-0233091/00-28, Institute for European Environmental Policy, London.
Countryside Commission. 1984. Application of the Less Favoured Areas (LFA) Directive in The
Netherlands, CCP 167, Countryside Commission, Cheltenham.
Darnhofer, I., and W. Schneeberger. 2007. Impacts of voluntary agri-environmental measures
on Austria’s agriculture. International Journal of Agricultural Resources, Governance and
Ecology 6(3): 360–377.
Defrancesco, E., P. Gatto, F. Runge, and S. Trestini. 2008. Factors affecting farmers’ participa-
tion in agri-environmental measures: A northern Italian perspective. Journal of Agricultural
Economics 59(1): 114–131.
Dupraz, P., D. Vermersch, B. Henry de Franham, and L. Delvaux. 2003. The environmental sup-
ply of farm households: A flexible willingness to accept model. Environmental and Resource
Emery, S., and J. Franks. 2012. The potential for collaborative agri-environment schemes in
England: Can a well-designed collaborative approach address farmers’ concerns with cur-
rent schemes? Journal of Rural Studies 28: 218–231.
Espinosa-Goded, M., J. Barreiro-Hurlé, and E. Ruto. 2010. What do farmers want from
agri-environment scheme design? A choice experiment approach. Journal of Agricultural
Economics 61(2): 259–273.
European Commission. 2006. Handbook on common monitoring and evaluation framework.
Brussels: Directorate General for Agriculture and Rural Development.
European Commission. 2010. Rural development in the European Union—Statistical and eco-
nomic information—Report 2010. http://ec.europa.eu/agriculture/agrista/rurdev2010/
ruraldev.htm
European Commission. 2011. CAP Reform—an explanation of the main elements.
MEMO/11/685.
European Court of Auditors. 2011. Is agri-environment support well designed and managed?
Special Report No. 7, European Court of Auditors, Luxembourg.
Falconer, K., and M. Whitby. 1999. The invisible costs of scheme implementation and adminis-
tration. In Countryside Stewardship: Farmers, policies and markets, eds. G. Van Huylenbroeck
and M. Whitby, 67–88. Amsterdam: Pergamon.
Falconer, K. 2000. Farm-level constraints on agri-environmental scheme participation: A trans-
actional perspective. Journal of Rural Studies 16: 379–394.
Falconer, K., P. Dupraz, and M. Whitby. 2001. An investigation of policy administrative costs
using panel data for the English Environmentally Sensitive Areas. Journal of Agricultural
Economics 52(1): 83–101.
Franks, J. 2010. Boundary organizations for sustainable land management: The example of
Dutch environmental co-operatives. Ecological Economics 70: 283–295.
Franks, J., and Emery, S. 2013. Incentivising collaborative conservation: Lessons from existing
Environmental Stewardship Scheme options. Land Use Policy 30: 847–862.
Fraser, R. 2004. On the use of targeting to reduce moral hazard in agri-environmental schemes.
Frey, B. 1997. Not just for the money: An economic theory of personal motivation. Cheltenham,
UK: Edward Elgar.
Frondel, M., P. Lehmann, and F. Wätzold. 2012. The impact of information on landowners’
participation in voluntary conservation program—Theoretical considerations and empiri-
cal evidence from an agri-environment program in Saxony, Germany. Land Use Policy
29: 388–394.
Garrod, G., K. Willis, and C. Saunders. 1994. The benefits and costs of the Somerset Levels and
Moors ESA. Journal of Rural Studies 10(2): 131–145.
Garrod, G., and K. Willis. 1995. Valuing the benefits of the South Downs Environmentally
Sensitive Area. Journal of Agricultural Economics 46(2): 160–173.
Garrod, G., E. Ruto, K. Willis, and N. Powe. 2012. Heterogeneity for the benefits of Environmental
Stewardship: A latent-class approach. Ecological Economics 76: 104–111.
Gren, I-M. 2004. Uniform or discriminating payments for environmental production on
arable land under asymmetric information. European Review of Agricultural Economics
31(1): 61–76.
Hagedorn, K. 2004. Multifunctional agriculture: An institutional interpretation. In Proceedings
of the 90th EAAE seminar, Multifunctional agriculture, policies and markets: Understanding
the critical linkages, Abstract, 17–20, Part 2, INRA, Rennes.
Hanley, N., D. MacMillan, R. Wright, C. Bullock, I. Simpson, D. Parsisson, and B. Crabtree. 1998.
Contingent valuation versus choice experiments: Estimating the benefits of Environmentally
Sensitive Areas in Scotland. Journal of Agricultural Economics 49: 1–15.
Hart, R., and U. Latcz-Lohmann. 2007. Combating moral hazard in agri-environmental
schemes: A multi-agent approach. European Review of Agricultural Economics 32(1): 75–91.
Hasund, K. P. 2013. Indicator-based agri-environmental payments: A payment-by-result
model for public goods with a Swedish application. Land Use Policy 30: 223–233.
Havlik, P., P. Veysset, J. M. Boisson, M. Lherm, and F. Jacquet. 2005. Joint production under
uncertainty and multifunctionality of agriculture: Policy considerations and applied analy-
sis. European Review of Agricultural Economics 32(4): 489–515.
Herzog, F. 2005. Agri-environment schemes as landscape experiments. Agriculture, Ecosystems
and Environment 108: 175–177.
Hodge, I. 1989. Compensation for nature conservation. Environment and Planning A
21(8): 1027–1036.
608 Ian Hodge
Hodge, I., and S. McNally. 1998. Evaluating the Environmentally Sensitive Areas: The value of
rural environments and policy relevance. Journal of Rural Studies 14(3): 357–367.
Hodge, I. 2000. Agri-environmental relationships and the choice of policy mechanism. The
World Economy 23(2): 257–273.
Hodge, I. 2001. Beyond agri-environmental policy: Towards an alternative model of rural envi-
ronmental governance. Land Use Policy 18: 99–111.
Hodge, I. 2007. The governance of rural land in a liberalised world. Journal of Agricultural
Economics 58(3): 409–432.
Hodge, I., and M. Reader. 2010. The introduction of entry level stewardship in England: Extension
or dilution in agri-environment policy? Land Use Policy 27(2): 270–282.
HÖjgård, S., and E. Rabinowicz. 2011. Evidence based agri-environmental policies—Can
institutionalized evaluation procedures provide useful input? The Swedish experience.
Paper presented at OECD Workshop on the Evaluation of Agri-environmental Policies,
Braunschweig, June 20–22.
Hynes, S., D. Campbell, and P. Howley. 2011. A holistic vs. an attribute-based approach to
agri-environmental policy evaluation: Do welfare estimates differ? Journal of Agricultural
Economics 62(2): 305–329.
Hynes, S., and E. Garvey. 2009. Modelling farmers’ participation in an agri-environmental
scheme using panel data: An application to the Rural Environmental Protection Scheme in
Ireland. Journal of Agricultural Economics 60: 546–562.
Kleijn, D., F. Berendse, R. Smit, and N. Gilissen. 2001. Agri-environment schemes do not effec-
tively protect biodiversity in Dutch agricultural landscape. Nature 413, 723–725.
Kleijn, D., and W. Sutherland. 2003. How effective are European agri-environment schemes on
conserving and promoting biodiversity? Journal of Applied Ecology 40: 947–969.
Kleijn, D., et al. 2006. Mixed biodiversity benefits of agri-environment schemes in five European
countries. Ecology Letters 9: 243–254.
Lankoski, J., and M. Ollikainen. 2003. Agri-environmental externalities: A framework for
designing targeted policies. European Review of Agricultural Economics 30(1): 51–75.
Latacz-Lohmann, U., and C. Van der Hamsvoort. 1998. Auctions as a means of creating a mar-
ket for public goods from agriculture. Journal of Agricultural Economics 49(3): 334–345.
Latacz-Lohmann, U., and S. Schillizzi. 2007. Quantifying the benefits of conservation auctions.
EuroChoices 6(3): 32–39.
Lowe, P., G. Cox, M. MacEwen, T. O’Riordan, and M. Winter. 1986. Countryside conflicts: The
politics of farming, forestry and conservation. Aldershot, UK: Gower.
Madureira, L., T., Rambonilaza, and I. Karpinski. 2007. Review of methods and evidence for
economic valuation of non-commodity outputs and suggestions to facilitate its application
to broader decisional contexts. Agriculture, Ecosystems and Environment 120: 5–20.
McCann, L., B. Colby, K. Easter, A. Kasterine, and K. Kuperan. 2005. Transaction cost measure-
ment for evaluation environmental policies. Ecological Economics 52: 527–542.
MAFF. 1989. Environmentally sensitive areas. London: HMSO.
Matzdorf, B., and J. Lorenz. 2010. How cost-effective are results-oriented agri-environment
measures? An empirical analysis in Germany. Land Use Policy 27: 535–544.
Mettepenningen, E., V. Beckmann, and J. Eggers. 2011. Public transaction costs of
agri-environmental schemes and their determinants—Analysing stakeholders’ involve-
ment and perceptions. Ecological Economics 70: 641–650.
Miettinen, A., and A. Huhtala. 2004. On joint production of cereals and grey partridges in
Finland. Paper presented at 90th European Association of Agricultural Economists Seminar,
Rennes.
Morris, J., J. Mills, and I. Crawford. 2000. Promoting farmer uptake of agri-environment
schemes: The Countryside Stewardship arable options scheme. Land Use Policy 17: 241–254.
Moxey, A., B. White, and A. Ozanne. 1999. Efficient contract design for agri-environmental
policy. Journal of Agricultural Economics 50(2), 187–202.
Netzwerk Land. 2011. What farmers are doing for the environment: The Austrian
Agri-environmental Programme ÖPUL. Vienna: Netzwerk Land.
OECD. 1999. Cultivating rural amenities: An economic development perspective.
Paris: Organisation for Economic Co-operation and Development.
OECD. 2001a. Multifunctionality: Towards an analytical framework. Paris: Organisation for
Economic Co-operation and Development.
OECD. 2001b. Improving the environmental performance of agriculture: Policy options and mar-
ket approaches. Paris: Organisation for Economic Co-operation and Development.
Ozanne, A., T. Hogan, and D. Colman. 2001. Moral hazard, risk aversion and compliance
monitoring in agri-environmental policy. European Review of Agricultural Economics
28(3): 329–347.
Ozanne, A., and B. White. 2007. Equivalence of input quotas and input charges under asym-
metric information in agri-environment schemes. Journal of Agricultural Economics
58(2): 260–268.
Peach, W., L. Lovett, S. Wotton, and C. Jeffs. 2001. Countryside Stewardship delivers cirl bun-
tings (Emberiza circlus) in Devon, UK. Biological Conservation 101: 361–374.
Peerlings, J., and N. Polman. 2004. Wildlife and landscape services production in Dutch
dairy farming: Jointness and transaction costs. European Review of Agricultural Economics
31(4): 427–449.
Perkins, A., H. Maggs, A. Watson, and J. Wilson. 2011. Adaptive management and targeting
of agri-environment schemes does benefit biodiversity: A case study of the corn bunting
Emberiza calandra. Journal of Applied Ecology 48: 514–522.
Polman, N., and L. Slangen. 2008. Institutional design of agri-environmental contracts in
the European Union: the role of trust and social capital. NJAS: Wageningen Journal of Life
Sciences 55(4): 413–420.
Primdahl, J. et al. 2010. Current use of impact models for agri-environment schemes and
potential for improvements of policy design and assessment. Journal of Environmental
Puchta, A. 2011. The evaluation of the Austrian Agri-environmental programme. Paper pre-
sented at OECD Workshop on the Evaluation of Agri-environmental Policies, Braunschweig,
June 20–22.
Pufahl, A., and C. Weiss. 2009. Evaluating the effects of farm programmes: Results from pro-
pensity score matching. European Review of Agricultural Economics 36(1): 79–101.
Purvis, G., G. Louwagie, G. Northey, S. Mortimer, J. Park, A. Mauchline, J. Finn, J. Primdahl,
H. Vejre, J. P. Vesterager, K. Knickel, N. Kasperczyk, K. Balazs, G. Vlahos, S. Christopoulos,
and J. Peltola. 2009. Conceptual development of a harmonised method for tracking change
and evaluating policy in the agri-environment: The agri-environmental footprint index.
Environmental Science and Policy 12: 321–337.
610 Ian Hodge
Quillérou, E., and R. Fraser. 2010. Adverse selection in the Environmental Stewardship
Scheme: Does the Higher Level Stewardship Scheme design reduce adverse selection?
Renwick, A., T. Jansson, P. Verburg, C. Revoredo-Giha, W. Britz, A. Gocht, and D. McCracken.
2013. Policy reform and agricultural land abandonment. Land Use Policy 30: 446–457.
Rørstad, P. K., A. Vatn, and V. Kvakkestad. 2007. Why do transactions costs of agricultural
policy vary? Agricultural Economics 36: 1–11.
Ruto, E., and G. Garrod. 2009. Investigating farmers’ preferences for the design of
agri-environment schemes: A choice experiment approach. Journal of Environmental
Planning and Management 52: 631–647.
Schouten, M., P. Opdam, N. Polman, and E. Westerhof. 2013. Resilience-based governance
in rural landscapes: Experiments with agri-environment schemes using a spatially explicit
agent-based model. Land Use Policy 30: 934–943.
Schwarz, G., A. Moxey, D. McCracken, S. Huband, and R. Cummins. 2008. An analysis of the
potential effectiveness of a Payment-by-Results approach to the delivery of environmental pub-
lic goods and services supplied by Agri-Environment Schemes. Report to the Land Use Policy
Group, UK. Macaulay Institute, Pareto Consulting and Scottish Agricultural College.
Shoard, M. 1980. The theft of the countryside. London: Maurice Temple Smith.
Stoate, C., and D. Parish. 2001. Monitoring is underway and results so far are promising. Nature
414: 687.
Stobbelaar, D. J., J. C. J. Groot, C. Bishop, J. Hall, and J. Pretty. 2009. Internalization of
agri-environmental policies and the role of institutions. Journal of Environmental
Management 90: S175–S184.
Taylor, M., and M. Morecroft. 2009. Effects of agri-environment schemes in a long-term eco-
logical time series. Agriculture, Ecosystems and Environment 130: 9–15.
Vanslembrouck, I., G. van Huylenbroech, and W. Verbeke. 2002. Determinants of the willing-
ness of Belgian farmers to participate in agri-environment measures. Journal of Agricultural
Vatn, A. 2002. Multifunctional agriculture: Some consequences for international trade regimes.
European Review of Agricultural Economics 29(3): 309–327.
Wätzold, F., N. Lienhoop, M. Dreschler, and J. Settele. 2008. Estimating optimal conservation in
the context of agri-environmental schemes. Ecological Economics 68: 295–305.
Westbury, D., J. Park, A. Mauchline, R. Crane, and S. Mortimer. 2011. Assessing the environ-
mental performance of English arable and livestock holdings using data from the Farm
Accountancy Data Network (FADN). Journal of Environmental Management 92: 902–909.
Whitby, M. 1996 The prospect for agri-environmental policies within a reformed CAP. In The
European Environment and CAP Reform, ed. M. Whitby, 227–240. Wallingford, UK: CAB
International.
Whitby, M. 2000. Challenges and options for the UK agri-environment. Journal of Agricultural
Economics 51(3): 317–332.
White, B. 2002. Designing voluntary agri-environmental policy with hidden information and
hidden action: A note. Journal of Agricultural Economics 53(2): 353–360.
Whittingham, M. 2011. The future of agri-environment schemes: Biodiversity gains and eco-
system delivery? Journal of Applied Ecology 48: 509–513.
Williamson, O. 1996. The mechanisms of governance. Oxford: Oxford University Press.
Willis, K., J. Benson, and C. Saunders. 1988. The impact of agricultural policy on the costs of
nature conservation. Land Economics 64(2): 147–157.
Wilson, G. 1997. Factors influencing farmer participation in the Environmentally Sensitive

Areas scheme. Journal of Environmental Management 50: 67–93.
Wilson, G., and K. Hart. 2001. Farmer participation in agri-environment schemes: Towards
conservation-oriented thinking? Sociologia Ruralis 41: 254–274.
Wossink, G., and H. van Wenum. 2003. Biodiversity conservation by farmers: Analysis of actual
and contingent participation. European Review of Agricultural Economics 30: 461–485.
Wynn, G., B. Crabtree, and J. Potts. 2001. Modelling farmer entry into the environmentally
sensitive areas in Scotland. Journal of Agricultural Economics 52: 65–82.
C HA P T E R 2 3
AG R I - E N V I R O N M E N TA L
POLICIES
A Comparison of US and EU Experiences
RO G E R C L A AS SE N , JO SE PH C O OPE R ,
C R I ST I NA S A LV ION I , A N D
M A RC E L L A V E RON E SI
Agriculture is more than just the production and sale of commodities; it also pro-
duces many intended and unintended positive and negative byproducts. Negative
byproducts, or disamenities, include nutrient and pesticide runoff, soil erosion, air pol-
lution, and the loss of biodiversity (ERS 2006). The positive byproducts, or amenities,
provided by agriculture can be relatively tangible goods such as open space and scenic
vistas, whereas others, such as the spiritual or symbolic value of preserving our farming
heritage, are more abstract and nonpecuniary (Cooper et al. 2005). Many environmen-
tal amenities or disamenities of agricultural production affect society as a whole and
have a social benefit or cost much greater than the private benefit or cost affecting those
involved in agriculture. In such cases, there is an economic rationale for society to sub-
sidize the environmental amenity (or tax an environmental disamenity) to produce the
desired level of environmental protection.
The United States and the European Union have a long history of agri-environmental
programs. In the 1980s, agri-environmental programs began to play a larger role in fed-
eral farm policies, in part due to greater concern about environmental damage from agri-
cultural production. Agri-environmental programs are likely to play a vital role in future
EU and US farm policy debates. In this chapter, we outline and compare EU and US
agri-environmental programs. We then overview what is known about the environmental
and land use impacts of these programs. We follow with a discussion of and EU data sources
that are key to the analysis of agri-environmental programs and their land use impacts.
Agri-Environmental Policies 613
1. Different Instruments of

Environmental Protection
Both the United States and the European Union rely primarily on a mixture of three
types of programs to address agri-environmental issues: voluntary incentive-based
programs, regulatory programs, and cross-compliance programs. 1 Other policy instru-
ments in use include in-kind technical assistance and facilitative measures such as
organic certification and labeling standards.
Agri-environmental incentives are payments to the farmer to adopt environmen-
tally sound practices or to retire environmentally sensitive land from production.
The advantage of incentives is that they lower resistance from farmers to adopting
the desired practices or retiring land. Incentive payments can also facilitate target-
ing of conservation program effort to farms where relatively large benefits—relative
to costs—can be achieved. The disadvantage of incentives is the cost to taxpayers.
Incentives can also have the effect of expanding production, so that even if the disame-
nities produced by each farm (or on each field) decrease, more farms (or fields) may
now produce disamenities.2
Regulatory requirements, or standards, represent an involuntary (or mandatory)
approach to improving agri-environmental performance. Unlike policy choices in
which farmer participation is uncertain, regulations simply require that all farmers par-
ticipate. This feature can be particularly important if the consequences of not changing
practices are drastic or irreversible. The ban on the production and application of the
chemical DDT is one such example. However, regulatory requirements are a blunt tool
and can be the least flexible of all policy instruments, requiring that producers reach
a specific environmental goal or adopt specific practices without regard for cost or
environmental effectiveness, which may vary significantly across farms but are seldom
known by regulators. Consequently, regulation can be less flexible and less efficient than
economic incentives. Regulatory requirements are used sparingly in both the European
Union and the United States.
Cross-compliance requires that farmers use practices that meet a basic environmen-
tal standard as a condition of eligibility for other government programs that farmers
may find economically desirable, such as those that provide income support payments.
Technically, cross-compliance is a voluntary, indirect, incentive-based instrument, but
because it represents a standard for receiving a subsidy, in practice, it may not strictly be
1
Only a brief overview is provided here; for a more detailed overview of the economic instruments
pertaining to US agri-environmental policy, see Claassen et al. (2001) and other papers at http://www.
ers.usda.gov/Briefing/ConservationAndEnvironment/.
2 A firm that would be unprofitable under a tax may be made profitable by an incentive or subsidy
(Baumol and Oates, 1988). Although a tax may drive a firm out of a competitive industry, an incentive
may increase entry and induce expansion in competitive outputs.
614 Roger Claassen et al.
perceived as voluntary, particularly when the existing subsidy represents an important

share of total farm income. It may be difficult for a farmer to forego cross-compliance
when the value of the existing subsidies exceeds the farmer’s costs of adopting the
mandated practices.3 An advantage of cross-compliance programs is that less gov-
ernment outlay is required than with subsidies to address environmental problems.
A key disadvantage is that not all farmers receive program payments that are subject
to cross-compliance. In the United States, for example, federally subsidized crop insur-
ance, which is an increasingly important component of US government support for
agriculture, is not currently subject to cross-compliance. Moreover, high market prices
for commodity crops have reduced some US commodity program payments. If pay-
ments become low enough, farmers might forgo participation in these programs rather
than use the practices entailed by cross-compliance (ERS 2012). Moreover, compliance
requirements, if expensive to apply, may also undercut other program objectives, such
as income support.
Other measures facilitate conservation without providing financial assistance directly
to producers. For example, the United States provides in-kind conservation assistance
to farmers through planning and technical assistance in the development and imple-
mentation of conservation practices to address specific agri-environmental problems
(known as conservation technical assistance or CTA). Another example is eco-labeling
(e.g., organic certification) that allows producers to differentiate their products in
the marketplace and, possibly, command a higher price. The key advantage of these
approaches is low cost (at least in relation to payment programs); the key disadvantage
is that off-site environmental damages are not explicitly addressed, although producer
benefits (e.g., from reduced soil erosion) may also lead to off-site benefits.
2. US Agri-Environmental Policy
Regulation is used sparingly in US federal agri-environmental policy. The Federal

Insecticide, Fungicide, and Rodenticide Act (FIFRA) regulates the availability and use of
agricultural pesticides. Regulatory authority under the Clean Water Act (CWA) has also
been used in some instances to regulate loss of sediment, and influx of nutrients, patho-
gens, and pesticides into water. For example, CWA authority has been used to regulate
3 Government payments can account for a large share of net farm income but vary widely across
regions; for example, in 2010, in the United States, they account for less than 20% of net farm income
in the Pacific region and 40% or more in the Southern Plains (Figure 1.11, ERS 2011). Farm payments
have been in place long enough in both the European Union and the United States that they are largely
capitalized into the value of land (Duffy et al. 1994; Barnard et al. 1997; Roberts 2004; Kirwan 2009). In
the United States, crop insurance premium subsidies are probably also capitalized into land values to
some extent. For many producers, the ability to purchase land, pay cash rent, or receive favorable interest
rates on loans can depend significantly on receiving farm commodity program payments and purchasing
(subsidized) crop insurance.
wetland drainage, occasionally preventing farmers from draining wetlands or requir-

ing their mitigation. In 2008, the US Environmental Protection Agency (EPA) regulated
effluent discharge from Confined Animal Feeding Operations (CAFOs), although only
those CAFOs that discharge or plan to discharge must seek a permit—other CAFOs can
self-certify that they do not and will not discharge (US EPA, undated). At the state level,
a patchwork of regulation applies mostly to large CAFOs. For example, a number of
states in the Chesapeake Bay watershed (Delaware, Maryland, Virginia, Pennsylvania,
and New York) require livestock farms to devise and apply nutrient management plans.
Arguably, regulatory policies affecting agricultural land use in the United States are
more significant at the state and local levels. In particular, zoning regulations are largely
set by local jurisdictions, such as counties. These may place restrictions on nonagricul-
tural uses in certain areas. Many states have right-to-farm rules that can protect the right
of farmers to use standard farm management practices, some of which nonagricultural
neighbors on the rural-urban fringe could otherwise seek to restrict (Hellerstein et al.
2002). For the sake of brevity, however, this chapter focuses on US federal policies affect-
ing agricultural land use.
Economic incentive programs are the backbone of US federal agri-environmental
policy, accounting for more than $5.5 billion in federal spending during fiscal year 2010
(OMB, 2012). We place these programs in one of three broad categories: land retirement
(LR), working land (WL) conservation, and agricultural land preservation.
2.1 Land Retirement

The Conservation Reserve Program (CRP) offers annual payments and cost sharing
to establish long-term, resource-conserving cover on environmentally sensitive land.
Contracts are for 10–15 years. Economic use of the land is limited during the contract
period, but landowners retain the right to return land to crop production at the end of
the contract.
The CRP includes several components. Through the general signup, producers can
enroll whole fields or whole farms but must compete for enrollment because acre-
age offered often exceeds the number of acres that can be added to CRP. Continuous
signup offers enrollment without competition for certain high-priority practices (e.g.,
field-edge filter strips or wetland restoration). Through the Conservation Reserve
Enhancement Program (CREP; a part of continuous signup), the US Department of
Agriculture (USDA) can enter into partnership with states to support practices designed
to address a specific environmental problem, such as water quality in a specific river
or lake.
The Wetlands Reserve Program (WRP) provides cost sharing and long-term or per-
manent easements for restoration of wetlands on agricultural land. Through WRP, the
USDA can offer a wide range of easement and contract options, ranging from perma-
nent easements to relatively short-term contracts (e.g., 10-year agreements). Through
the Wetlands Reserve Enhancement Program (WREP), the USDA can partner with
states and nongovernmental organizations (NGOs) to carry out high-priority wet-

land protection, restoration, and enhancement activities. During the easement period,
landowners retain land ownership and rights to recreational uses, such as hunting and
fishing.
LR dominated federal agricultural conservation spending between 1985 and 2002. At
the end of 2011, roughly 10% of US cropland—32.2 million acres—was enrolled in CRP
(29.7 million acres) and WRP (2.6 million acres). Total CRP acreage has declined from
a peak of 36.4 million acres and is currently limited to a maximum of 32 million acres.
Program acreage and expenditures have also been shifted gradually from retirement of
whole fields or whole farms to partial field practices through continuous signup. At the
end of 2011, continuous signup (including CREP) accounted for 56% of CRP contracts,
32% of annual payments, and 18% of CRP acreage.
WRP enrollment is currently capped at 3.041 million acres. Unlike CRP, the WRP
acreage cap was increased in the 2008 Farm Act from 2.275 million acres. Program
enrollment is dominated by permanent easements, which account for roughly 80% of
WRP acreage.
2.2 Working Land Conservation

The Environmental Quality Incentives Program (EQIP) provides technical assistance,
cost sharing, and incentive payments to assist livestock and crop producers with adop-
tion of a wide range of more environmentally friendly production practices or best
management practices (BMPs). The Wildlife Habitat Incentives Program (WHIP) pro-
vides cost sharing to landowners and producers to develop and improve wildlife habi-
tat. The Conservation Stewardship Program (CSP) provides payments to producers for
maintaining or adopting a wide range of structural and land management practices that
address a variety of local and/or national resource concerns. Unlike other USDA con-
servation programs, CSP payments are based on environmental benefits, as estimated
using a series of indices to measure the potential value of installing or adopting a spe-
cific practice in a specific location. Producers can also receive payments based on the
ongoing application of previously adopted practices. These payments are also based on
an estimate of environmental benefits but are lower than payments for newly adopted
practices. Also, unlike other programs, payments can be based on going beyond basic
treatment of environmental problems to a higher or “enhanced” level of environmental
performance.
Before 2002, funding for WL conservation was modest compared to LR. For EQIP,
the largest US WL program, the 2002 Farm Act authorized a five-fold increase in fund-
ing over previous levels, funding levels that have been maintained in the 2008 Farm Act
at least through fiscal year 2011, when it was $1.238 billion. The 2008 Farm Act directed
the USDA to enroll 12.77 million acres per year in the CSP at an average annual cost (to
the government) of $18 per acre. Program spending increased from $9 million in fiscal
year (FY) FY2009 to $390 million in FY2010 and $601 million in FY2011. This program
replaces the Conservation Security Program (authorized in the 2002 Farm Act), which
will end when existing contracts have expired.
2.3 Agricultural Land Preservation

The Farm and Ranch Land Protection Program (FRPP) funds the purchase of devel-
opment rights (i.e., purchases of easements) on agricultural land in urban fringe areas,
thus preserving it for agricultural production. Under the 2008 Farm Act, FRPP has been
funded at $121 million in FY2009, $150 million in FY2010, and $175 million in FY2011.
The 2002 Farm Act extended FRPP eligibility to land with “historically important land
areas and structures,” providing one instance where US policy clearly attempts to pre-
serve “positive” environmental amenities (e.g., open space, scenic vistas, or small-scale
farms). In general, these types of “environmental” goals are left to other US federal or
state programs.4 The European Union, conversely, supports such amenities of agricul-
ture as part of EU-wide agri-environmental policy, although the European Commission
has limited control in the design and operation of specific programs.
The Grassland Reserve Program (GRP) assists owners, through long-term rental
agreements or easements (i.e., voluntary sale or donation of specific use rights to land),
in restoring grassland and conserving virgin grassland while maintaining areas for live-
stock grazing and hay production. GRP supports enhancement of plant and animal bio-
diversity and protection of grasslands under threat of conversion to cropping, urban
development, and other activities. Funding has been $48 million, $100 million, and
$79 million for fiscal years 2009, 2010, and 2011, respectively.
2.4 Compliance Mechanisms

Compliance mechanisms require farmers to conserve soil on highly erodible land and con-
serve wetlands to be eligible for federal agricultural payments, including commodity sup-
port payments (see, e.g., Cooper 2010, for an overview of commodity support payments;
see Claassen et al. 2004, for an overview of programs potentially subject to compliance
sanctions). Producers can become ineligible for commodity payments if they:
• convert wetlands to make agricultural production possible (a provision widely known

as “swampbuster”); or
• produce crops on highly erodible land (HEL) without applying an approved conser-
vation system.
4 See American Farmland Trust http://www.farmland.org/programs/protection/

farmland-information-center.asp for more information on US state farmland protection programs.
The conservation measures required under this program come closest in the United States
to representing a basic level of “good farming practice” or environmental compliance such
as exists in the European Union.
Since producers must pay the costs of compliance, it is difficult to quantify expenditures
in comparison with direct incentive programs such as EQIP. Costs include applying an
approved conservation system or the opportunity cost of not using HEL or wetlands for
crop production. Some practices, such as conservation tillage, have probably lowered pro-
duction costs for some (but not all) producers who have adopted them as part of a conser-
vation system. The ERS (2006, Section 5.3) presents more detail about the benefits and costs
of conservation compliance.
3. EU Agri-Environmental Policy
EU agri-environmental policy uses a combination of voluntary, regulatory, and

cross-compliance programs to achieve environmental goals, similar in general principles to
US agri-environmental policy. These programs can be summarized into two main catego-
ries: basic legal standards and the Common Agricultural Policy (CAP).
“Basic legal standards” are regulatory rules that apply to all EU Member States and their
farmers. Farmers must comply with these environmental regulations without receiving
any compensation for doing so. The EU Nitrate Directive5 is an example of a basic legal
standard that applies specifically to agriculture. Farmers have to meet the requirements of
the Framework Directive on the Sustainable Use of Pesticides and of the Water Framework
Directive,6 as well as of those to ensure food traceability (General Food Law, EC/178/2002)
and a whole range of far-reaching animal welfare rules.
The CAP consists into two pillars: Pillar I encourages farmers and other land man-
agers to protect the environment and fight climate change through direct payments
that are decoupled from production but linked to environmental requirements via
cross-compliance; Pillar II deals with Rural Development Programs, which explic-
itly include agri-environmental measures, such as payments to farmers in return for
adoption of desired farm management practices. One of the objectives of the ongo-
ing CAP revision is to offer the opportunity to reinforce measures for “green growth”
and to make the CAP more effective in providing existing environmental benefits. The
EU Commission is now proposing a further layer of green measures to add to existing
measures, as we will discuss in Section 3.6 (“Support for Nonproductive Investments”).
later.
Cross-compliance is a mechanism that links direct payments to compliance by farm-
ers with basic standards concerning the environment, food safety, animal and plant
5 Legal basis: Directive 1991/676/EEC.

6 Legal basis: Directive 2000/60/EC.
health, and animal welfare, as well as with the requirement of maintaining land in good
agricultural and environmental condition. Cross-compliance includes:
1. Five sets of environmental concerns—for a total of 18 legislative standards—in

the field of the environment, food safety, animal and plant health, and animal
welfare, which are covered by Statutory Management Requirements (SMR).
These requirements have a long history and apply to all farmers (even those not
receiving the Pillar I support subject to cross-compliance).
2. Good Agricultural and Environmental Conditions (GAEC) are the obligations
for farmers receiving CAP payments to keeping land in good agricultural and
environmental condition, as defined by EU legislation. GAECs are compul-
sory for Member States—in many cases, they are simply the translation of EU
environmental legislation or other EU requirements into specific national obli-
gations.7 GAEC standards are designed to prevent soil erosion, maintain soil
organic matter and soil structure, ensure a minimum level of maintenance, avoid
the deterioration of habitats, protect and manage water, and require that the ratio
of permanent pastures at a national level is maintained within certain limits.
All farmers receiving direct payments, agri-environmental payments forming part of

the Rural Development Policy (RDP), and certain wine payments are subject to com-
pulsory cross-compliance.8 Cross-compliance represents the “baseline” or “reference
level” for agri-environment measures.
Farmers are expected to observe GAEC without receiving direct compensation for
doing so. However, unlike basic legal standards, the European Commission does not
mandate good farming practices but allows each Member State to decide what a good
farming practice is. Member States can make good farming practices mandatory or
cross-compliant by tying the adoption of such practices to CAP payments in a process
known as “modulation.” “Modulation” refers to the process whereby a proportion of
CAP direct aids is rechanneled into agri-environmental programs.
3.1 Agri-Environmental Measures

In 2005, the RDP, the so-called Pillar II of the CAP, was restructured into three thematic
“axes” for the period 2007–2013.9 Axis 1 aims to improve the competitiveness of the
7
As a matter of fact, many GAECs are already contemplated in the directives mentioned. Directives
lay down certain end results, but Member States are free to decide how to meet these goals. The inclusion
of GAEC in a regulation, that is, in a binding legislative act that must be applied in its entirety across the
EU, requires all farmers to meet the same obligations.
8 Legal basis: Council Regulation 73/2009 and Commission Regulation 1122/2009.
9 Council Regulation (EC) No. 1698/2005 of September 20, 2005, on support for rural development
by the European Agricultural Fund for Rural Development (EAFRD) lays down the general rules
agricultural and forestry sector; Axis 2, the environment and the countryside; and Axis
3, the quality of life in rural areas and diversification of the rural economy. Given the
topic of this chapter, we focus our discussion on Axis 2. The three priority areas of inter-
vention in Axis 2 of the RDP are “biodiversity and the preservation and development of
high nature value farming and forestry systems and traditional agricultural landscapes;
water; and climate change” (European Commission 2006).
Axis 2 includes measures aimed at contributing to the implementation of Natura
2000’s “Network10” of protected areas (definitions of EU-specific terms such as
“Natura 2000” follow); the commitment made during a European Council meeting
in Gothenburg in June 2001 to reverse the decline of the EU’s biodiversity by the year
2010; the objectives laid down in the Water Framework Directive; the Kyoto Protocol
targets for climate change mitigation; and measures targeting the sustainable use of
forestry land.
In contrast to the United States, where agri-environmental policy mainly targets
the reduction of negative externalities produced by agriculture, from its origins, EU
agri-environmental policy has rewarded farmers both for reducing negative external-
ities and for using farming practices that can help provide public goods (Baylis et al.
2008). The rationale for supporting land management practices that provide public
goods is that intensification and concentration of land used in the most competitive
areas and the marginalization or abandonment of land use in less competitive areas
result in continued declines in many species and habitats, increased water scarcity, and
significant problems with soil erosion and loss of soil organic matter. The RDP also sup-
ports social public goods, such as food security, farm animal welfare and health, and,
especially, rural vitality (Cooper et al. 2009).
Central to this discussion of EU rural development is the assumption that eco-
nomically and socially vibrant rural areas can help to promote the continuation of
agriculture and forestry which, in turn, are important in providing the environmen-
tal public goods on which many sectors—such as rural tourism and recreation—
depend (Cooper et al. 2009; ENRD 2010; Hart et al. 2011). The concept of “high
nature value farming” has recently come to play a central role in the discourse of
researchers and policy makers regarding EU rural development. This concept was
introduced in the early 1990s and stems from a growing recognition that the con-
servation of biodiversity and of heritage landscapes in the EU depends on the con-
tinuation of low-intensity farming systems (Baldock et al. 1993). However, these
governing RDP for the period 2007–2013, as well as the policy measures available to Member States and
regions. The RDPs that the Member States and regions prepared for the period 2007–2013 are currently
under implementation.
10
The Natura 2000’s EU-wide network of nature protection areas was established under the 1992
Habitats Directive. The aim of the network is to assure the long-term survival of Europe’s most valuable
and threatened species and habitats. It is comprised of Special Areas of Conservation (SAC) designated
by Member States under the Habitats Directive and also incorporates Special Protection Areas (SPAs),
which they designated under the 1979 Birds Directive.
traditional management systems are disappearing due to agricultural intensification

and rural abandonment. Therefore, the strategic approach to maintaining biodiver-
sity across the EU includes both the protection of particular habitats or species and
the maintenance of low-intensity land uses that are more favorable to wildlife and to
maintaining biological and landscape diversity.
Axis 2 measures are largely voluntary, contractual, and co-financed, and are delivered
within a strategic framework that links policy action to European, national, regional,
and local needs. Agri-environment measures are, at present, the only compulsory mea-
sure under Pillar II. Member States have a wide degree of discretion in implementing
these measures, permitting them to be tailored to different agronomic and environ-
mental circumstances. Payments per hectare or farmer can vary substantially from one
Member State to another. In the Annex to the Rural Development Regulation, maxi-
mum ceilings are set for each measure. In some cases, these ceilings can result in pay-
ments below actual costs.
The incentive payment levels are based on cost incurred and income foregone by the
farmer for participating in the agri-environmental measure. Their calculation normally
takes into account only variable costs or income forgone resulting from the participa-
tion in agri-environmental programs. A general condition for payments under Axis
2 is that farmers respect the relevant EU and national mandatory requirements (i.e.,
cross-compliance).
The Axis 2 measures targeting the sustainable use of agricultural land include
Agri-Environmental Payments (AEPs), Natural Handicap Payments (NHPs) to
farmers in mountainous areas or with other natural impediments to adoption of
modern agricultural practices (less favored areas [LFAs]), Natura 2000 payments,
payments linked to the Water Framework Directive, and other new measures related
to animal welfare. We now outline the EU agri-environmental policy measures in
more detail.
3.2 Agri-Environmental Payments

AEP schemes are designed to encourage farmers to protect and enhance the environ-
mental attributes of their land. These provide payments to farmers in return for services
such as carrying out agri-environmental commitments that go beyond the application
of standard good farming practices. Each scheme has at least one of two broad objec-
tives: reducing environmental risks associated with modern farming and preserving
natural and cultivated landscapes. Specifically, these schemes include one category of
measures related to productive land management and one related to nonproductive
land management. The first category includes measures aimed at (1) input reduction;
(2) organic farming; (3) extensification of livestock; (4) conversion of arable land to
grassland and rotation measures; (5) undersowing and cover crops, strips, and prevent-
ing erosion and fire; (6) actions in areas of special biodiversity/nature interest; (7) rear-
ing of rare local breeds and the preservation of plant genetic resources; (8) maintenance
of existing sustainable and extensive systems; (9) farmed landscape; and (10) water use
reduction. The group of measures related to nonproductive land management includes
land set-aside managed for environmental purposes, upkeep of abandoned farmland
and woodland, maintenance of countryside and landscape features, and public access.
3.3 Natural Handicap Measures

Natural Handicap Measures aim to provide an aid to farmers operating in LFAs, that
is, in mountainous areas and in other areas with specific natural handicaps. LFA is a
longstanding measure of CAP. The logic of intervention in the LFA scheme has under-
gone a significant evolution since its inception in 1975. Originally, the scheme explicitly
addressed rural depopulation as a socioeconomic objective. In 2003, the implementa-
tion of the LFA scheme was subject to criticisms in a report of the European Court of
Auditors (Court of Auditors 2003), particularly in regard to the designation of inter-
mediate LFAs and the lack of aid targeting. Starting in 2005, within the new strate-
gic approach adopted for the RDP for 2007–2013, LFA payments, renamed Natural
Handicap Payments, became part of Axis 2 of RDP. The NHP scheme has a strong focus
on land management and aims to maintain the countryside and promote sustainable
agriculture, which delivers public goods such as valuable landscapes, biodiversity, soil
conservation, and fire prevention in areas where farming is difficult.
3.4 Natura 2000 Payments

Natura 2000 payments aim to compensate farmers for the loss of income or the cost of
extra management obligations necessary to deliver the objectives of the Natura 2000
Network. Similar considerations apply for the payments linked to Directive 2000/60/
EC.
3.5 Animal Welfare Payments

Animal welfare payments are granted to farmers who commit to applying animal
welfare standards that go beyond basic legal requirements. These commitments are
described in Article 27 (7) of Commission Regulation (EC) No. 1974/2006. Any animal
welfare commitment must provide upgraded standards in at least one of the following
areas: water and feed closer to the animals’ natural needs; housing conditions, such as
space allowances, bedding, natural light; outdoor access; absence of systematic mutila-
tions, isolation, or permanent tethering; and prevention of pathologies mainly deter-
mined by farming practices or/and keeping conditions.
3.6 Support for Nonproductive Investments

Support for nonproductive investments provides financial aid to farmers and other land
managers who make on-farm agri-environmental, nonproductive investments; that
is, investments that enhance the public amenity value of high nature value areas (e.g.,
NATURA 2000 areas). Examples of nonproductive investments include planting hedge-
rows; planting trees for wind-breaks/shelters; establishing grass margins; establishing
grasslands for land use changes and nature conservation; and establishing green cover.
Regarding the rural development budget, 44% of the European Agricultural Fund for
Rural Development (EAFRD) funding for the 2007–2013 period (about €43 billion) has
been allocated by Member States to Axis 2 measures (“improving the environment and
the countryside”) (EC 2011). The CAP Health Check in 2008 assigned additional fund-
ing to five new “challenges,” including biodiversity. The planned spending for the agri-
environment payments over the 2007–2013 programming period amounts to €22.5
billion, representing half of the budget devoted to the environmental axis of rural devel-
opment policy. LFA payments—in and outside mountainous areas—total €13.4 billion.
These three measures account for 84% of all funds under Axis 2, and it is estimated that
they will result in nearly 7 million agri-environment agreements over the 2007–2013
period, bringing approximately 42 million hectares (24% of total utilized agricultural
area) under some form of environmental management. In addition, €472 million will be
spent on Natura 2000 measures on farmland and €111 million on Natura 2000 measures
on forestry land.
Regarding Axis 2 measures, AEPs represent the EAFRD policy instruments with the
highest financial allocation in most Member States. At the EU-27 level, these represent
52.5% of the EAFRD contribution allocated to this axis, and its share is higher than 70%
in Belgium (at 82.6%), the United Kingdom (at 74.4%), and the Netherlands (at 72.1%).
In the recently added Member States, the share within Axis 2 is higher than 55% in
Bulgaria (56%), Estonia (63.1%), and Hungary (67%).
In addition to cross-compliance and the agri-environmental measures under the
RDP, Article 68 of Regulation 73/2009 allows Member States to retain up to 10% of their
previously coupled payment ceilings under Pillar I for specific supports to farming and
quality production. These retained payments can be used to:
• protect the environment, improve the quality and marketing of products, or for
animal welfare support;
• help farmers producing milk, beef, sheep, goats, and rice in economically vul-
nerable or environmentally sensitive areas, as well as for economically vulnerable
types of farming;
• top-up existing entitlements in areas where land abandonment is a threat;
• support risk assurance in the form of contributions to crop insurance premiums; and
• contribute to mutual funds to combat animal and plant diseases.
The CAP is due to be reformed by 2013. After a public debate on October 12, 2011, the
Commission presented a set of legal proposals in the CAP Towards 2020. Following a
debate in the European Parliament and the Council, approval of the different regula-
tions and implementing acts is expected by the end of 2013, with the goal of having the
CAP reform in place by January 2014.
One of the objectives of the new CAP is to improve efficiencies of support granted to
farmers for practices aimed at environmental and climate change considerations. As for
Pillar I, 30% of direct support is proposed to be made conditional on “greening” (i.e., envi-
ronmentally supportive practices). According to the proposal, 30% of the total amount of
resources devoted to direct payments in each Member State is constrained by the fulfilment
of three mandatory measures: to maintain on-farm permanent grassland, to diversify crops
in order to improve biodiversity, and to devote 7% of the Utilized Agricultural Area (UAA)
to “Ecological Focus Areas” (EFA) (including terraces, buffer strips, hedges, and set-aside
areas). The only actors who would not be submitted to these constraints are organic pro-
ducers and farmers who accept the simplified scheme (“small farmers” scheme).11
The impact of greening measures on the income of European farms is estimated to
be, on average, a €43 increase per hectare of potentially eligible area, although it may
vary widely according to region and farming systems (European Commission 2011).
Concerns about the negative impacts of greening on the competitiveness of the EU agri-
cultural sector have been expressed by the major farmer associations (Copa-Cogeca
2012), as well as by the majority of the national ministries of agriculture of the Member
States. In contrast, environmental NGOs such as BirdLife and the World Wildlife
Fund (WWF) believe that the greening measures do not go far enough, and they pro-
pose replacing the crop diversification measure with a real crop rotation requirement,
increasing the EFA to 10% of the agricultural area at farm level, and including a more
stringent definition of permanent grassland (BirdLife International 2012). Academics
and researchers have also contributed to this debate, and there is wide agreement
that the strategy of green payments proposed by the European Commission could be
improved (Groupe de Bruges 2012; Mahé 2012; Westhoek et al. 2012).
RDPs remain the key element of the new CAP for delivering public goods. In line
with Europe 2020 and the overall CAP objectives, the sustainable management of natu-
ral resources and climate action have been restated as one of the three long-term strate-
gic objectives of RDP, along with improved competitiveness of agriculture and balanced
territorial development (Loriz-Hoffmann 2012) of rural areas.
Instead of three axes linked to economic, environmental, and social issues with mini-
mum spending requirements for each axis, the new programming period will have six
priorities: fostering knowledge transfer and innovation; enhancing competitiveness;
promoting food chain organization and risk management; restoring, preserving, and
enhancing ecosystems; promoting resource efficiency and a transition to a low-carbon
11 http://ec.europa.eu/agriculture/analysis/perspec/cap-2020/impact-assessment/annex2en.pdf
economy; and promoting social inclusion, poverty reduction, and economic develop-
ment in rural areas.
Under the proposals, Member States are still required to devote at least 25% of
their rural development budget to land management and climate change mitigation.12
However, the proposed rural development budget will suffer a decline in real terms for
the period 2014–2020. This could be countered in those countries that take advantage
of the option to move 10% of Pillar I funds to Pillar II. Conversely, some Member States
will be allowed to transfer funding from Pillar II to Pillar I. Agri-environmental and cli-
mate payment schemes will have greater flexibility in contract design and will be linked
to adequate training/information. In addition, new measures have been proposed to
promote organic farming.
4. Impacts of US Agri-Environmental

Programs on Land Use: A Review of the
Empirical Literature
The focus of this review is on an empirical analysis of actual agri-environmental poli-

cies—and in particular, US federal policies—rather than on generic or stylized policies.
However, some exceptions are made in cases in which the results may be particularly
illuminating for US federal policy. Much more work has been done on CRP than on
the federal WL conservation programs that we addressed earlier in this chapter. In this
review, we do not attempt to include every academic paper ever written on US federal
agri-environmental programs but focus rather on including examples covering the vari-
ous general themes that turned up during our literature review.
Ideally, the government would target land for enrollment into agri-environmental
programs in a manner that maximizes the social benefits of multiple environmental
outcomes, subject to budget constraint. However, as noted by Babcock et al. (1996),
the solution to this complex problem requires quantification of the physical tradeoffs
between the various environmental benefits and a social value function that can cap-
ture the marginal rate of substitution of the various benefits, something that is gener-
ally not feasible in practice. Reichelderfer and Boggess (1988), Babcock et al. (1996),
Wu and Boggess (1999), Feather, Hellerstein, and Hansen (1999), Wu, Zilberman, and
Babcock (2001), and Feng et al. (2006) all provide examples of approaches to targeting,
and although these cannot fully quantify the benefits, they at least permit a capturing of
some of the physical and economic tradeoffs.
12
http://europa.eu/rapid/pressReleasesAction.do?reference=MEMO/11/685&format=HTML&aged
=0&language=EN&guiLanguage=en
Benefit-cost targeting has been a key component of US agri-environmental incentive

programs since 1991, when the Environmental Benefits Index (EBI) was added to the
CRP program. The index focuses mostly on maintaining soil productivity, water quality,
and wildlife habitat. Feather, Hellerstein, and Hansen (1999) estimated that use of the
EBI increased CRP benefits by 80% over a nontargeted CRP enrollment. The research
also indicated that benefits could be increased by emphasizing water quality and wild-
life habitat over the preservation of soil productivity. Similar indices are included in the
EQIP, CSP, and other programs.
Competitive bidding has also been a part of CRP since the early 1990s. Producers
who offer to take less than the (field-specific) bid limit or who waive cost-sharing of
cover establishment can improve their EBI score (which includes a cost factor) and
their chances of being selected for enrollment. Kirwan et al. (2005) argue that as farm-
ers and landowners have become increasingly familiar with the EBI—which is largely
unchanged since 1997—they have also been less likely to offer bids below their bid limit.
Applicants offering land with high environmental benefits (EBI environmental scores
are known before bids are finalized) and those offering low-value land have been par-
ticularly reluctant to offer bid discounts.
In the context of LR programs, slippage is the extent to which enrollment of land
induces production change on land not enrolled. For example, LR may reduce crop pro-
duction, thus increasing crop prices and encouraging farmers to convert other land (say
pasture or range) to crop production. To the extent that this slippage occurs, the total
environmental benefits associated with LR decline. Wu (2000), Roberts and Bucholtz
(2005), and Wu (2005) econometrically examine CRP using National Resources
Inventories (NRI) and Census of Agriculture data to test for the presence of slippage.
Although there is some disagreement among the results of these studies, they appear to
generally support the notion that the CRP does cause some slippage to occur.
WL conservation programs can also cause a slippage effect. Lichtenberg and
Smith-Ramírez (2011) econometrically model the effect of a state-level BMP
cost-sharing program that Maryland designed to address water quality problems in the
Chesapeake Bay region. The allocation of land between pasture/wildlife habitat, contour
farming/strip cropping, and cover crops was examined using farm-level data for both
recipients and nonrecipients and found that farmers who received cost sharing allocated
larger shares of their cropland to contour farming/strip cropping and to cover crops and
less to the more environmentally benign pasture/wildlife habitat. Hence, the environ-
mental quality improvements from this WL conservation program are likely offset to
some degree. In another study of slippage potentially being associated with WL pro-
grams, Wallander and Hand (2011) examine how participation in EQIP by irrigators
affects water application rates and decisions to expand or reduce a farm’s irrigated acre-
age. Using farm-level panel data from a national survey of irrigators taken in 1998, 2003,
and 2008, they estimate changes in water application rates and irrigated acreage that
result when a farm receives EQIP payments. Their results suggest that, for the average
farm participating in EQIP between 2004 and 2008, EQIP payments may have induced
reduced water application rates but also may have increased total water use and led to
an expansion in irrigated acreage. Khanna, Isik, and Zilberman (2002) and Lichtenberg
(2004) also found that conservation subsidies can, in some instances, worsen environ-
mental quality by giving farmers an incentive to expand production onto more erodible
or otherwise environmentally fragile or more highly polluting land.
A political economy concern with LR is that it could have a negative impact on rural
economies due to lowering local expenditures on farming-related businesses. Sullivan
et al. (2004) estimated the impact that high levels of enrollment in the CRP have had on
economic trends in rural counties since the program’s inception in 1985 and through
2004. The results of the report’s growth model and quasi-experimental control group
analysis indicated no discernible impact by the CRP on aggregate county population
trends. The report found that aggregate employment growth may have slowed in some
high-CRP counties, but only temporarily. High levels of CRP enrollment appear to have
affected farm-related businesses over the long run, but growth in the number of other
nonfarm businesses moderated CRP’s impact on total employment. If CRP contracts had
ended in 2001, their simulation models suggest that roughly 51% of CRP land would have
returned to crop production and that spending on outdoor recreation would decrease
by as much as $300 million per year in rural areas. However, the resulting impacts on
employment and income varied widely among regions having similar CRP enrollments,
depending on local economic conditions. Bangsund et al. (2004) found that for CRP land
in North Dakota, increased hunting revenues partially offset agricultural revenue, but the
study did not estimate the regional economic impacts of either.
Some analysis has addressed the relationship between WL and LR programs. LR and
WL programs can potentially compete with each other, given that they are mutually
exclusive in the use of land, thus suggesting that these programs should be operated in a
coordinated fashion. Feng et al. (2005) find that interactions between CRP and WL pro-
grams may be significant. For instance, based on results of an econometric analysis of
1997 NRI data, their simulation analysis finds that acres enrolled in CRP at a given rental
rate would be about half in the presence of subsidies for conservation tillage. Using the
Environmental Policy Integrated Climate (EPIC) simulation model in conjunction with
their econometric model, they find that the presence of both large WL and LR programs
can result in more environmental benefits (and income transfers) than an LR-only pro-
gram can achieve.
Furthermore, WL and LR programs can target the same environmental benefits—
a concept known as “stacking.”13 An interesting question, albeit hard to analyze, is to
determine the cost differences of each—using either one exclusively or in combina-
tion—in achieving a given target. Feng et al. (2006) develop a formal theoretical frame-
work combining an economic model and the EPIC model to simulate the marginal costs
13
See Gillenwater (2012) and Cooley and Olander (2012) for a detailed discussion of “stacking” in
the context of ecosystem services. Stacking occurs when multiple programs provide payments for the
same ecosystem. Stacking can provide multiple revenue streams for landowners and encourage them
to manage their lands for multiple ecosystem services (Cooley and Olander 2012). However, if the
programs are not well coordinated, it may also lead to a net loss of services (ibid.) or cost inefficiencies.
of targeting an environmental objective (carbon sequestration) by LR alone, LR and WL

in combination, and WL alone. Using this model with NRI data for Iowa, they find that,
for achieving small carbon sequestration benefits, WL alone and an optimal budget split
between WL and LR have the lowest marginal costs, but LR alone has the lowest mar-
ginal costs for achieving higher levels of sequestration. The results were similar when
they modeled multiple environmental targets.
One might expect that the environmental benefits of LR increase the longer the land
stays in retirement, assuming the alternative is likely to be the land returning back to
crop production if the CRP contract expires and is not renewed. Roberts and Lubowski
(2007) use parcel-level data from the USDA’s NRI to examine actual land use choices fol-
lowing expiration of CRP contracts between 1995 and 1997 in the contiguous 48 states.
These data reflect choices made by landowners who opted out of the CRP early, chose
not to extend or renew their contracts, or submitted new contract bids that scored too
low to be re-enrolled in the program. Based on the econometric analysis of actual land
use decisions, they predicted that about 58% of all CRP acres would have converted to
crop production in 1997 had they all, in fact, exited. Given that alternatives to crop pro-
duction for lands with expiring contracts (e.g., pasture, range, or forest) are likely less
environmentally damaging than crop production, CRP exits would likely increase envi-
ronmental damages. Their results also suggested that the opportunity cost of enrolling
land in CRP is higher for newly enrolling land as compared to re-enrolling land. If so,
this result suggests that targeted signing bonuses for first-time enrollees would increase
the longer term impacts of CRP and perhaps other incentive-based land use programs.
Agricultural commodity support and agri-environmental programs can be
expected to have interaction effects, with the former potentially having negative envi-
ronmental impacts that may offset the environmental benefits associated with the
agri-environmental programs. Alternatively, agri-environmental programs can be used
to offset the potentially negative environmental impacts of commodity support pro-
grams. Using NRI data and crop production patterns and crop yields taken from unpub-
lished USDA/National Agricultural Statistics Service (NASS) files, as well as other
data gathered from 1982 to 1992, Goodwin and Smith (2003) conduct an econometric
analysis of the relative impacts of CRP and agricultural commodity support on erosion
level. They find that CRP enrollment decreases erosion (as one would expect), Federal
Crop Insurance causes a relatively small increase in erosion, but other government com-
modity support programs cause a relatively large increase in soil erosion. Conceivably,
some of the negative impacts of government commodity support could be offset with
increased support to WL conservation programs but, of course, at additional cost.
In considering potential interaction effects between agri-environmental programs
and commodity support, these may be strongest between LR and commodity support
given, in that they are mutually exclusive uses of land. Lubowski et al. (2008) use NRI
data in an econometric analysis of land use and land characteristics on nonfederal lands
conducted at 5-year intervals from 1982 to 1997 over the entire United States, exclud-
ing Alaska. In assessing the factors that affect land use decisions, they found strong evi-
dence that cropland declines over that period were due to falling crop net returns and
the existence of CRP. However, they also document the opposing influences of federal
agricultural commodity payments (see, e.g., Cooper 2010 and ERS 2009b for a discus-
sion of commodity payments). In particular, by raising farm income and/or lowering
revenue risk through commodity support payments, the government increased acre-
age in crops and directly competed with itself in providing incentives for landowners to
retire environmentally sensitive cropland under the CRP.
Not only may federal commodity support and CRP compete for land, CRP may com-
pete with—or “crowd-out”—nonfederal conservation programs for crop land, such as
land trusts. For example, Parker and Thurman (2011) examine the effects of US fed-
eral land programs (CRP and WRP) on private conservation using county-level panel
regressions between 1990 and 2000. They use econometric analysis based on a Bayesian
approach modeling private conservation acres held from several conservation NGOs as
a function of CRP and WRP enrollment, acres in federal lands, demographic data, and
other variables. They find some evidence of a small but measurable crowding-out effect
from CRP land holding on the land trusts examined.
Of course, no discussion of agriculture is complete without a mention of biofuels. In
particular, growing demand for biofuels production—mostly ethanol from corn in the
United States—raises crop prices by increasing the demand for corn. Given the highly
inelastic supply of land for crops, increasing corn acreage will increase the prices of
other crops that are displaced by the increase in corn acreage. Increasing crop prices due
to biofuels demand likely increases the total amount of land in crop production, poten-
tially increasing negative environmental impacts as land is taken out of pasture and
range and put into crops (e.g., Searchinger et al. 2008)—a change that has implications
for WL conservation programs. The increasing crop prices have the potential for lower-
ing re-enrollment rates in the CRP as farmers seek to expand planted acres, with conse-
quent negative environmental impacts. For example, for central US grassland enrolled
in the CRP for 15 years, Fargione et al. (2008) found that converting it to corn ethanol
production creates a biofuels carbon debt that would take approximately 50 years to
repay if subsequently replanted to perennial systems. Secchi et al. (2009) construct CRP
land supply curves for various corn prices and then, using the EPIC model, estimate the
environmental impacts of cropping land exiting CRP land. EPIC provides edge-of-field
estimates of soil erosion, nutrient loss, and carbon sequestration. They find that incre-
mental environmental impacts increase dramatically as higher corn prices bring into
production more and more environmentally fragile land. Hence, maintaining current
levels of environmental quality (as defined by EPIC) will require substantially higher
spending levels on LR and WL programs.
Conversely, CRP land could, in principle, be a supplier of feedstocks for
cellulosic-based production of biofuels, at least providing that harvesting of these feed-
stocks is consistent with conservation efforts. The renewable fuels standard (RFS) in
the Energy Independence and Security Act of 2007 mandates an increasing amount of
cellulosic biofuels production yearly through 2022, even though, to date, the EPA has
granted waivers to this mandate. Additionally, cellulosic biofuels can be used to sub-
stitute for corn-based ethanol under the RFS, although it is not currently economically
feasible to do so. If cellulosic-based production were to become feasible to the extent

that it could be used to actually meet the cellulosic biofuels mandate and even substitute
for corn ethanol, and assuming that much of the feedstock harvesting could be done on
CRP land, then the pressure on landowners to not re-enroll CRP land could decrease.
Mapemba et al. (2007) use a multiregion, multiperiod, mixed integer mathematical pro-
gramming model to determine cost to deliver a flow of feedstock to a biorefinery. They
found that restrictions on harvest days based on environmental impact concerns—for
example, impacts on bird habitat—can significantly increase feedstock delivery costs.
Careful management of the harvesting season will be necessary to balance environmen-
tal costs versus feedstock production costs.
The papers discussed here tend to be based on analysis of actual (revealed preference)
data. Federal WL programs tend to offer incentive payments that do not vary intertempo-
rally or cross-sectionally for the adoption of BMPs. As such, revealed preference data pro-
vide little guidance on how enrollment rates would change with different payment rates.
Such knowledge would help in adjusting payment rates to achieve desired adoption levels
for BMPs. Even if payments have variation, as in the case of the CRP, the stated preference
approach using hypothetical incentive offers could be useful if there is concern that the
actual offers are systematically higher than farmers’ minimum willingness to accept (WTA).
One approach to assessing how farmers would respond to higher or lower payments
than actually offered is to use a stated preference approach based on survey question-
naires. Examples of such approaches as applied to US federal WL programs include
Cooper (2003) and Cooper and Keim (1996), although they have been applied to CRP
as well (e.g., Cooper and Osborn 1998). As expected, the farmer’s WTA to enroll in the
WL programs, and to enroll more acres, increases with increasing incentive payments.
Cooper (1997) provides an example of how to add revealed preference data—from those
who have adopted the practices without program enrollment—in conjunction with the
stated preference data.14
There is relatively little empirical research on transaction costs associated with
agri-environmental programs. The research that does exist, however, indicates
that transaction costs may by quite large. Because US programs require produc-
ers to propose the application of specific practices in specific fields to address specific
resource concerns, a great deal of conservation planning must be done to complete an
agri-environmental program application. McCann and Easter (2000) used data from
the mid-1990s to estimate that NRCS transaction costs (for conservation planning,
14 Although, in general, survey questions eliciting WTA may not be as incentive compatible as
willingness to pay (WTP) questions, the latter are generally not an appropriate format for eliciting
enrollment behavior from farmers because the expectation is that agri-environmental programs
will pay them for conservation activities and that farmer will not have to pays fines for not adopting
conservation practices. However, in the stated preference work cited here, the survey instruments made
clear that the analysis was being conducted on behalf of the government. Strategic response bias in the
upward direction—i.e., saying “no” to enrollment rates that exceed true minimum WTA—could be
a welfare-lowering strategy because biasing WTA estimates in the upward direction could lower the
government’s interest in providing the program.
construction oversight, related overhead) were, on average, 38% of total conservation

costs (including transaction costs, financial assistance, and private costs for practice
installation or adoption). The estimate excludes producer costs related to developing
the conservation program application and a broader set of costs (e.g., research, policy
development) that could be considered a part of overall transaction costs (McCann et al.
2005). Farmer transaction costs associated with US agri-environmental program appli-
cation have not been estimated (but see McCann 2009).
Finally, a recent study addresses the issue of additionality in US conservation pay-
ment incentive programs (Mezzatesta et al. 2011). Additionality is the degree to which
conservation payments leverage the adoption of practices that would not have been
adopted in the absence of these payments. Using propensity score matching with data
from a survey of Ohio farmers, they estimate that additionality is high for structural and
vegetative practices (e.g., filter strips and cover crops) but not for conservation tillage.
Roughly 92% of filter strips, 87% of cover crop acres, but only 18% of conservation till-
age acreage were found to be additional.
5. Impact of EU Agri-Environmental

Programs: A Review of the
Empirical Literature
There is an increasing debate on whether EU agri-environmental programs actually

deliver the expected outcomes (e.g., Kleijn and Sutherland 2003; EC-DGAGRI 2005;
Kleijn 2006; Boatman et al. 2008; Hodge and Reader 2010; ECA 2011). However, the
impact assessment of EU agri-environmental programs is mainly based on “administra-
tive” evaluations produced by public officials rather than on scientific research (Burrel
2011). The increasing budget stringency and the growing competition in the use of
diminishing public funds have led the European Commission to adopt the Common
Monitoring and Evaluation Framework (CMEF) (EC 2006). CMEF is a guide for the
administrative evaluation of rural development policies in the current programming
period 2007–2013. The “administrative” evaluation proposed by the CMEF relies on
indicator-based evidence and analyses three types of outcomes: (1) “outputs” of pro-
gram activities, such as number of participants in a scheme or hectares of land con-
verted; (2) “results,” that is, direct consequences for program participants, such as
changes in management practices resulting from scheme participation (e.g., change to
organic farming, training participation); and (3) “impacts,” that is, longer lasting effects
of the intervention with direct or indirect relevance to the program’s overall objec-
tives and their attainment (e.g., improved water quality). This framework has gained
importance over time, and it is now an integral part of the policy process. The empirical
evidence produced by the application of the CMEF is used for the administrative evalu-
ation of RDPs to feed the policy monitoring cycle and to correct possible inefficiencies.
The scarce scientific evidence in the impact assessment of EU agri-environmental

programs is supported by Primdahl et al. (2010), who examine a sample of 60
agri-environmental schemes implemented in EU Member States (Denmark, Finland,
Germany, Greece, Hungary, Ireland, and the United Kingdom) within the 2000–2006
RDP period. These schemes mainly deal with natural resources (primarily water
resources, 31%), whereas only a limited proportion of schemes deals with biodiversity
(7%) and landscape (8%). This study shows that only a minority of schemes (15%) are
based on scientific research that uses quantitative causal models of relationship between
agricultural practices and environmental outcomes. In the majority of cases, impacts
had been assessed on the basis of so-called commonsense impact models, reflecting
policy design that was based more on general beliefs about causal relationships than on
scientific evidence. Boatman et al. (2008) present a meta-analysis of the environmen-
tal benefits provided by UK agri-environmental schemes and find that, in general, they
were successful in attaining the desired environmental benefits. In particular, addition-
ality is shown to be present for landscape, biodiversity (particularly for plants and birds,
less for mammals and invertebrates), and habitat objectives. However, they identify
some cases in which progressive deterioration was evident, as in the cases of grassland
and heather moorland.
The lack of scientific studies on the impact assessment of EU agri-environmental
programs is partly explained by the fact that the objectives of EU agri-environment
schemes are often unclear, imprecise, or too many and thus difficult to identify (Hodge
2001; Primdahl et al. 2003; Bartolini et al. 2005; Finn et al. 2007; ECA 2011). Even when
the quantification of the outcomes is possible, baseline environmental data for before
agri-environment payments were implemented are often missing. Primdahl et al. (2010)
find that, in absence of any comprehensive environmental baseline study prior to the
implementation of the scheme, participation has been used as an indicator of the envi-
ronmental effectiveness of agri-environmental schemes. In addition, in most cases,
the time frame for the achievement of the policy objectives is missing or is too short to
evaluate effects within time horizons of 5–10 years on issues such as biodiversity and
groundwater (Boatman et al. 2008; Primdahl et al. 2010).
Several academic studies analyze the drivers of agri-environmental measures’ adop-
tion (Crabtree et al. 1998; Delvaux et al. 1999; Wynn et al. 2001; Dupraz et al. 2002;
Vanslembrouck et al. 2002; Wossink and van Wenum 2003; Defrancesco et al. 2008;
Bertoni et al. 2008; Borsotto et al. 2008; Defrancesco et al. 2008; Barreiro-Hurlé et al.
2010; Giovanopoulou et al. 2011). The level of farmer participation is usually inter-
preted as a measure of the success of agri-environmental measures. Among the factors
that negatively affect measure uptake, private transaction costs that reduce the net pay-
ment received by farmers have been identified as one of the most important (Falconer
2000). Empirical research has shown significant variations in private transaction costs
(Hackl et al. 2007; Ducos et al. 2009), both across different agri-environmental schemes
(Rørstad et al. 2007) and within single schemes (Falconer and Saunders 2002; Rørstad
et al. 2007; Mettepenningen et al. 2009). In addition, a meta-analysis of environmental
impacts of EU agri-environmental schemes finds a positive effect on the participants’
behavior with respect to three indicators: nitrogen (N) fertilizer, livestock density, and

area of grassland (Oltmer et al. 2000).
Although the academic literature dealing with impact assessment of actual policies
is still scarce, many optimization and simulation economic models have been proposed
to show, for example, how crop farms may respond to changes in incentives for the
agri-environmental measure (Hansen and de Frahan 2011) or changes in other aspects
of the designs of agri-environmental contracts (Bamière et al. 2011). Hynes et al. (2008)
use simulated farm population microdata merged with habitat land cover data within a
geographic information system (GIS) framework to examine what type of habitats are
actually being protected under the Irish Rural Environmental Protection Scheme in
2005. They find that habitats such as forests or shallow waters are more likely to be pro-
tected than, for example, dry grassland or cut fen, although dry grassland is a dominant
land cover type in Ireland. They also show that such a program could lead to poor target-
ing of benefits; that is, the program may be targeting farmers with marginal quality land
for which there is no need of environmental improvement.
Only recently has the scientific community started to investigate, using quantitative
methods such as quasi-experimental approaches (Pufahl and Weiss 2009; Chabé-Ferret
and Subervie 2011), the additionality of program impacts relative to the counterfactual
on what would have happened in the absence of the policy. Pufahl and Weiss (2009)
assess the impact of agri-environment programs on input use and farm output in
Germany, and they find a significant positive effect on the area farmed or grazed, par-
ticularly with respect to the latter. The need to reduce the density of cattle livestock to
become eligible for agri-environmental payments and the fact that program payments
are given on a per-area base can explain the increase in farm size.
Compared with the reference group, participants show a reduction in expendi-
tures for farm chemicals (fertilizers, pesticides) per hectare. They do not find any sig-
nificant effects on farm productivity (sales per hectare), capital endowment, off-farm
labor, or cattle livestock units per farm. The insignificant effect on productivity might
be explained by the extensification process undertaken. Chabé-Ferret and Subervie
(2011) provide a disaggregated estimation of the effects of five agri-environmental pro-
grams on environmentally relevant practices for a nationally representative sample of
French farmers. They find that the impact of payment schemes for organic farming is
significantly large, whereas the impact of payment schemes for crop diversity is mixed.
These results can be explained by the fact that schemes for crop diversification enable
the entry of farmers who, in the absence of the agri-environmental scheme, would
have highly diverse rotations. Future studies should include a cost-benefit analysis of
agri-environmental programs by translating the causal effects of the agri-environmental
programs in monetary terms.
Table 23.1 provides a summary comparison of US and EU agricultural policy, focus-
ing on the US federal level and EU level, respectively. The first column lists the policy
categories, as defined broadly, and the adjacent columns categorize the regulations,
cross-compliance, and agri-environmental payment approaches for the European
Union and the United States that are associated with each policy objective. The table
Table 23.1 Comparison of EU and US federal agri-environmental policies and programsa
EU policies and programs US policies and programs
Regulation (basic Agri-environmental Agri-environmental
Policy objective legal standards) Cross-compliance payments Regulation Cross-compliance payments
Preserve and restore Natura 2000 Statutory Clean Water Act Swampbuster Conservation Reserve
natural areas Management (Wetland drainage) Program
including wildlife Requirements; Payments linked to Conservation Reserve
habitat Good Agricultural Directive 2000/60/EC Enhancement Program
and Environmental (CREP)
Conditions; Natura 2000 Wetland Reserve
Members set precise Payments Program
requirements which Support for Wildlife Habitat
serve as a "reference" Nonproductive Incentives Program
level or minimum Investments;
standard for receipt AEP supporting
of direct payments biodiversity/nature
under Pillar I of the
CAP.
AEP supporting Grassland Reserve
biodiversity/nature Program
and the conversion
of arable land to
grassland
Reduce environmental EU Nitrate Directive AEP linked to Federal Inceticide, Sodbuster Conservation Reserve
risk/damage Directive 2000/60/EC Fungicide, and Program
from agricultural Framework Support for Rodenticide Act Environmental Quality
production Directive on the Nonproductive (pesticide regulation) Incentives Program
Sustainable Use of Investments
Pesticides
Water Framework Natura 2000 Confined Animal Conservation
Directive Payments; AEP Feeding Operations Stewardship Program
supporting (CAFOs) regulations
extensification of
livestock
Promote animal Animal Welfare Animal Welfare **b ** **
welfare Basic Legal Payments
requirements
Preserve agricultural ** Natural Handicap ** ** Farm and Ranchland
landscape and Payments; AEP Protection Program
prevent land supporting farmed
abandonment landscapes and the
maintenance of the
countryside and
landscape features
Food traceability General Food Law ** ** ** **
** =No policy or program

Note: AEP is Agri-environmental Payments
a This table does not address local policies that may exist outside the rubric of EU-level and U.S. Federal policy.
b The U.S. does have Federal Law protecting animal welfare (e.g., Animal Welfare Act, Horse Protection Act, Twenty-Eight Hour Law, and Humane Methods of
Slaughter), but there have few explicit links land use and agri-environmental issues.
shows that the European Union and the United States both have policies in each of the
three policy categories to achieve the same general objective. The two policy objec-
tives for which the United States potentially does not have land use polices at the fed-
eral level are related to animal welfare and protection of farming in mountainous areas.
Although there is US federal law protecting animal welfare (e.g., Animal Welfare Act,
Horse Protection Act, Twenty-Eight Hour Law, and Humane Methods of Slaughter),
these have few explicit links to land use and agri-environmental issues. In another dif-
ference, the US FRPP funds the purchase of development rights on agricultural land in
urban fringe areas, thus preserving it for agricultural production, whereas the European
Union does not have EU-level mechanisms for explicitly protecting agricultural land in
urban fringe areas. However, this difference may be semantic and depends on the defini-
tion of “urban fringe”; certainly, the EU’s NHPs have addressed agricultural landscape
preservation on lands relatively close to urban areas. In addition, land use regulations
operating outside the rubric of US federal level and EU-level agricultural policy can pre-
serve agricultural landscapes. With regards to specific policy tools, the European Union
does not appear to make use of easements as a mechanism, unlike in the United States
(e.g., via the WRP).
6. Data Availability in the United States

and European Union
Major USDA datasets that are often used to assess land use and other implications of
federal agri-environmental programs include the NASS Agricultural Census and a vari-
ety of NASS surveys (e.g., agricultural chemical use, agricultural prices, crop yield, crop
acreage) whose data are disseminated through the Quick Stats data tool, the Agricultural
Resource Management Survey (ARMS), and the Conservation Effects Assessment
Project (NRI-CEAP) Cropland Survey. Agricultural Census and Quick Stats provide
aggregate data (county and above) on farm activities such as acreage and production,
with the former being more detailed but updated at longer intervals than the latter.
ARMS is the USDA’s primary source of information on the financial condition, pro-
duction practices, and resource use of America’s farm businesses and the economic
well-being of America’s farm households. CEAP is a multiagency effort to quantify the
environmental effects of conservation practices and programs and develop the science
base for managing the agricultural landscape for environmental quality.
Increasingly, administrative and geospatial data also play a role in assessing the land
use impacts of policy changes. Contract data for individual programs (e.g., CRP) may
offer information on variations in the opportunity cost of practice adoption or insight
on bidding behavior (Kirwan, Lubowski, and Roberts, 2005). Geospatial data on land
use and soil properties have been critical in studies of land use change and the effect of
LR programs (Lubowski, Plantinga, and Stavins 2008). The NRI includes information
on land use and land quality for more than 800,000 points of nonfederal, rural land in
the contiguous US. The USDA Cropland Data Layer provides land use interpretation of
satellite imagery at 30-meter resolution.
One emerging issue is data integration—the process of combining data from differ-
ent sources to increase opportunities for policy research. Combining farm survey and
conservation program contact data, for example, could provide insight into the role of
agri-environmental program incentives in land use and conservation practice adop-
tion decisions. Although the promise of data integration is substantial, there are a num-
ber of barriers to its realization. Bohman and Claassen (2011) identify four potential
barriers: (1) a lack of identifiers suitable for linking, particularly the lack of accurate
geo-referencing on field-level data; (2) concerns about increasing the risk of disclosure
for confidential survey data; (3) “informed consent” requirements that could mean
agencies would need to revise notices about uses of collected data so program applicants
and survey respondents are informed about possible plans to link administrative and
survey data; and (4) the fact that survey and administrative data are often collected at
different spatial scales.
Major EU datasets often used to assess land use and other implications of their
agri-environmental programs include the European Network for Rural Development
(ENRD), the Coordination of Information on the Environment (CORINE), the Farm
Structure Survey (FSS), the Survey on Agricultural Production Methods (SAPM), and
the Farm Accountancy Data Network (FADN).
The ENRD, set up by the European Commission in 2008 to help Member States
implement their RDPs in an efficient manner, produces annual information on the
RDPs’ progress at the EU scale based on RDP monitoring data made available by the
Commission. The snapshots show the current state of play of the EU rural development
policies, highlight connections between resources and outcomes, and provide users
with informed insights. Information is provided for all EU-27 Member States, covering
88 national and regional programs.
The CORINE is a European program initiated in 1985 by the European Commission.
It is aimed at gathering information relating to the environment on certain priority top-
ics—air, water, soil, land cover, coastal erosion, biotopes, and more—for the European
Union. The Corine Land Cover (CLC) is a map of the European environmental land-
scape based on interpretation of satellite images. It provides comparable digital maps of
land cover for each country for much of Europe. The CLC is useful for environmental
analysis and for policy makers.
The FSS helps assess the agricultural situation across the EU by monitoring trends
and transitions in the structure of agricultural holdings, while also modeling the
impact of external developments or policy proposals. Two kinds of FSS are carried out
by Member States: a basic survey (full-scope Agricultural Census) every 10 years and
several sample-based intermediate surveys carried out every 2 or 3 years between the
censuses to provide harmonized information about land use on all holdings of 1 ha or
more. Topics covered include area farmed, area under various types of crops, numbers
of livestock, the farm workforce, rural development, and the extent of involvement in
nonagricultural activities (such as tourism, forestry, etc.).
EC Regulation No. 1166/2008 also outlined a one-off satellite Survey on Agricultural
Production Methods (SAPM). The SAPM complements the FSS and collects infor-
mation on soil tillage methods, landscape features, animal grazing, animal hous-
ing, manure application, manure storage and treatment, and irrigation. Both FSS and
SAPM are statistically representative at the level of NUTS 2 (“Nomenclature for Units of
Territorial Statistics”).
The FADN is the primary source of economic data at the farm level in the EU. It is a
European system of sample surveys that takes place each year and collects structural
and accountancy data relating to farms. FADN includes only commercial farms, that is,
farms which are large enough to provide the major income-generating activity for the
farmer and a level of income sufficient to support his or her family. In practical terms, to
be classified as commercial, a farm must exceed a minimum economic size, the thresh-
old depending on the country and the year. Commercial farms cover the most relevant
part of agricultural activity in each EU Member State, accounting for approximately
40–50% of FSS farms. FADN is the only source of microeconomic data that is harmo-
nized, meaning that accounting principles are the same in all EU Member States; thus,
FADN can be used to make comparison between Member States. It also provides infor-
mation about payments received by the farmers participating in agri-environmental
schemes (LFA and AEMs) (EC 2009). However, the FADN does not provide data on
quantities of inputs used. Instead, only expenditures on nutrients and other chemicals
are collected.
The data collected by FSS, SAPM, and FADN provide much useful information for
the evaluation of agri-environmental schemes. Together, FSS and SAPM cover a large
fraction of the data required for the AEIs and is fully representative of the farming com-
munity. However, they do not cover information on nutrients and other external farm
inputs (apart from irrigation water) flows. The FADN gathers information about nutri-
ents and other chemicals use, but only in monetary terms and for commercial farms.
Linking satellite, survey, and administrative data at the individual level is receiving
increasing attention (Selenius et al. 2011). Although the promise of data integration
is substantial, the four potential barriers to this integration identified by Bohman and
Claassen (2011) apply.
7. Conclusion
The United States and the European Union have many similar types of
agri-environmental programs and goals, especially when it comes to preventing nega-
tive environmental byproducts such as soil erosion, overuse of chemical pesticides and
fertilizers, and abuse of environmentally sensitive areas such as wetlands and wildlife
habitats. Moreover, both the European Union and the United States offer flexibility in
meeting the specific environmental needs of individual communities. In the United

States, flexibility is given to the producer, whereas, in the EU, it is more likely given to
the Member State. Either way, economically optimal management of these programs is
based on notions of economic efficiency, in which program parameters are chosen with
the aim of equating marginal benefits and their marginal costs. In practice, however,
program designs may also be motivated by various political economy goals (e.g., income
transfer).
However, there are also important differences between EU and US programs. The
emphasis of EU agri-environmental programs on maintaining landscape features and
the explicit focus on preventing land abandonment have little counterpart in US fed-
eral agri-environmental policy. US policy is largely focused on reducing the negative
externalities of agriculture, rather than on maintaining or enhancing the positive exter-
nalities. However, it would be misleading to provide a takeaway message that Europeans
have a higher preference for focusing on the amenities of agriculture relative to its disa-
menities than do Americans. To make such an assessment be informative would require
considering agri-environmental program choices within the context of all the available
substitutes and complements for agricultural and other activities.
In the past decade, both the European Union and the United States have moved for-
ward with plans to expand their agri-environmental programs. Before 2002, funding
for WL conservation was modest compared to LR. For the EQIP, the largest WL pro-
gram, the 2002 Farm Act authorized a five-fold increase in funding over previous lev-
els, funding levels that have been maintained in the 2008 Farm Act, at least through
fiscal year 2011. Under the 2008 Farm Act, spending on the new CSP (which replaced
the Conservation Security Program that authorized in the 2002 Farm Act) increased
from FY2009 to FY2011. The funding situation for many US agri-environmental pro-
grams for 2013 and on is unknown at this writing. In the European Union, the 2003
CAP reform increased the CAP’s focus on the interactions between agriculture and the
environment by shifting some funds from support regimes to environmental programs
through making “modulation” compulsory (EC 2003). The new reform of the CAP due
by 2013 is expected to improve the environmental and climate change performance of
the CAP (i.e., to further green the CAP payments).
Although our literature review demonstrates that a substantial body of work exists
on the impacts of agri-environmental programs, a number of important issues have
not been addressed or addressed in only a handful of studies. To date, we have iden-
tified only three papers that have addressed the issue of additionality in a US context
and only one (Mezzatesta, Newburn, and Woodward, 2011) that addresses these issues
in major USDA conservation programs. Liu and Lynch (2011) and Lynch, Gray, and
Geoghegan (2007) address additionality in the context of federal-state farmland preser-
vation programs. The literature review for the European Union also shows that research
on additionality is in its early stages there (Pufahl and Weiss 2009; Chabé-Ferret and
Subervie 2011).
As already noted, transaction cost analysis in conservation programs has also
received very little attention in the literature (McCann and Easter 2000). To date, there
are no studies of producer transaction costs for the United States, although existing
work does show that agency transaction costs are large. Given that program applica-
tion can be a lengthy process, producer transaction costs may also be large and could
be a barrier to conservation program participation. The issue of transaction costs has
been more deeply explored in the European Union, and the findings there show that
significant variations in private transaction costs (Hackl et al. 2007; Ducos et al. 2009)
exist both across different agri-environmental schemes (Falconer and Saunders 2002;
Rørstad et al. 2007) and within single schemes (Rørstad et al. 2007; Mettepenningen
et al. 2009).
Finally, there is a large literature on conservation practice adoption but very few arti-
cles actually address the role of federal cost-sharing and incentive payments. Cooper
and Keim (1996) and Lohr and Park (1995), for example, base their studies on stated
responses to hypothetical payments for conservation practices. The empirical analysis
in Lichtenberg (2004) is based on cost sharing for structural soil conservation practices
provided by the state of Maryland. Many other studies (e.g., Wu et al. 2004) use simula-
tion models to estimate the effect of payments that lower the costs for production sys-
tems, which include conservation practices.
These studies consider the role of hypothetical payments in leveraging practice adop-
tion rather than payments actually offered by the US government. Although these
studies are valuable, they do not necessarily yield information on the role of existing
programs in the adoption of conservation practices. In the European Union, optimi-
zation and simulation economic models have been proposed to explain, for example,
how farmers respond to changes in incentives for adopting agri-environmental mea-
sure (e.g., Hansen and de Frahan 2011) to changes in other aspects of the designs of
agri-environmental contracts (e.g., Bamière et al. 2011).
Acknowledgments
The views expressed herein are those of the authors and do not necessarily represent
those of the Economic Research Service or the US Department of Agriculture.
References
American Farmland Trust. 1998. The farmland protection toolbox: Fact sheet. http://www.
farmlandinfo.org/fic/tas/tafs-fptool.html
Babcock, B., P. Lakshminarayan, J. Wu, and D. Zilberman. 1996. The economics of a public fund
for environmental amenities: A study of CRP contracts. American Journal of Agricultural
Economics 78(November 1996): 961–971.
Baldock, D., G. Beaufoy, G. Bennett, and J. Clark. 1993. Nature conservation and new directions
in the EC common agricultural policy: The potential role of EC policies in maintaining farm-
ing and management systems of high nature value in the community. London: Institute for
European Environmental Policy.
Baumol, W. and W. Oates. 1988. The theory of environmental policy, 2nd ed. Cambridge
University Press, Cambridge.
Bamière, L., P. Havlik, F. Jacquet, M. Lherm, G. Millet, and V. Bretagnolle. 2011. Farming sys-
tem modeling for agri-environmental policy design: The case of a spatially non-aggregated
allocation of conservation measures. Ecological Economics 70: 891–899.
Bangsund, D. A., N. M. Hodur, and L. Leistritz. 2004. Agricultural and recreational impacts of
the conservation reserve program in rural North Dakota, USA. Journal of Environmental
Management 71(4): 293–303.
Barnard, C. H., G. Whittaker, D. Westenbarger, and M. Ahearn. 1997. Evidence of capitalization
of direct government payments into US cropland values. American Journal of Agricultural
Economics 79(5): 1642–1650.
Barreiro-Hurle, J., M. Espinosa-Goded, and P. Dupraz. 2010. Does intensity of change matter?
Factors affecting adoption of agri-environmental schemes in Spain. Journal of Environmental
Planning and Management 53(7): 891–905.
Bartolini, F., V. Gallerani, M. Raggi, and D. Viaggi. 2005. Contact design and targeting for the
production of public goods in agriculture: The impact of the 2003 CAP reform. Proceeding
of 11th EAAE congress: The future of rural Europe in the global agri-food system,
Copenhagen. August 24–27, 2005.
Baumol, W., and W. Oates. 1988. The theory of environmental policy, 2nd ed. Cambridge
Baylis, K., S. Peplow, G. Rausser, and L. Simon. 2008. Agri-environmental policies in the EU
and United States: A comparison. Ecological Economics 65 :753-764.
Bertoni, D., D. Cavicchioli, R., Pretolani, and A. Olper. 2008. Agri-environmental measures
adoption: new evidence from Lombardy Region. Paper prepared for the 109th EAAE
Seminar “The CAP after the Fischler reform: National implementations, impact assessment
and the agenda for future reforms,” Viterbo, Italy, November 20–21st, 2008.
Birdlife International. 2012. Reform proposals for the common agricultural policy. Birdlife
Europe Briefing.
Boatman, N., C. Ramwell, H. Parry, N. Jones, J. Bishop, P. Gaskell, C. Short, J. Mills, and J.
Dwyer. 2008. A review of environmental benefits supplied by agri-environment schemes.
FST20/79/041. Report to the Land Use Policy Group, UK.
Bohman, M., and R. Claassen. 2011. Drowning in data, coming up dry: Making connections for
meaningful water policy analysis. Presented at the World Statistics Congress, August 21–26.
Burrel, A. 2012. Evaluating policies for delivering agri-environmental public goods. Paper pre-
sented at the OECD workshop on the Evaluation of Agri-environmental Policies, Paris, June
20–22, 2011. http://www.oecd.org/dataoecd/2/25/48185525.pdf
Borsotto P., R. Henke M. C. Macrì, and C. Salvioni. 2008. Participation in rural landscape con-
servation schemes in Italy. Landscape Research 33(3): 347–363.
Chabé-Ferret, S., and J. Subervie. 2011. Estimating the causal effects of the French
agro-environmental schemes on farmers practices by difference in difference matching.
Paper presented at the OECD workshop Evaluation of Agro-Environmental Policies, Paris,
June 20, 2011.
Claassen, R. 2012. The Future of Environmental Compliance Incentives in U.S. Agriculture, EIB-
94. Washington DC: Economic Research Service, United States Department of Agriculture.
Claassen, R., L. Hansen, M. Peters, V. Breneman, M. Weinberg, A. Cattaneo, P. Feather, D.
Gadsby, D. Hellerstein, J. Hopkins, P. Johnston, M. Morehart, and M. Smith. 2001.
Agri-environmental policy at the crossroads: Guideposts on a changing landscape. AER-794.

Washington, DC: Economic Research Service.
Claassen R., V. Breneman, S. Bucholtz, A. Cattaneo, R. Johansson, M. Morehart. 2004.
Environmental compliance in agricultural policy: Past performance and future potential,
agricultural economic report, AER-832. Washington, DC: Economic Research Service, US
Department of Agriculture.
Cooley, D., and L. Olander. 2012. Stacking ecosystem services payments: Risks and solutions.
Environmental Law Reporter, 42 ELR 10150.
Cooper, J., and R. Keim. 1996. Incentive payments to encourage farmer adoption of water qual-
ity protection practices. American Journal of Agricultural Economics 78(February): 54–64.
Cooper, J. 1997. Combining actual and contingent behavior data to model farmer adop-
tion of water quality protection practices. Journal of Agricultural and Resource Economics
22(July): 30–43.
Cooper, J., and T. Osborn. 1998. The effect of rental rates on the extension of conservation reserve
program contracts. American Journal of Agricultural Economics 80(February): 184–194.
Cooper, J. 2003. A joint framework for analysis of agri-environmental payment programs.
American Journal of Agricultural Economics 85(November, 2003): 976–987.
Cooper, J., J. Berstein, V. Vasavada, and J. C. Bureau. 2005. The environmental by-products of
agriculture: International policy responses. In Global agricultural policy reform and trade—
Environmental gains and losses, ed. Joseph Cooper, 11–38. Cheltenham, UK: Edward Elgar.
Cooper, T., K. Hart, and D. Baldock. 2009. The provision of public goods through agricul-
ture in the European Union. Report prepared for DG Agriculture and Rural Development,
Contract No 30-CE-1233091/00-28. London: Institute for European Environmental Policy.
Cooper, J. 2010. Average crop revenue election: A revenue-based alternative to price-based com-
modity payment programs. American Journal of Agricultural Economics 92(4): 1214–1228.
Copa-Cogeca. 2012. The common agricultural policy after 2013; The reaction of EU farmers
and agri-cooperatives to the commission’s legislative proposals. Report prepared by Copa
and Cogeca (European Farmers and European Agri-cooperatives), Brussels.
Court of Auditors, European Union. 2003. Special report No. 4/2003 concerning rural devel-
opment: Support for less-favoured areas, together with the Commission’s replies. EUR-Lex
Official Journal of the European Union C 151 Vol. 46. http://eurlex.europa.eu/JOHtml.do?ur
i=OJ:C:2003:151:SOM:EN:HTML
Crabtree, B., N. Chalmers, and N.-J. Barron. 1998. Information for policy design: Modelling
participation in a farm woodland incentive scheme. Journal of Agricultural Economics
49(3): 306–320.
Defrancesco, E., P. Gatto, S. Trestini, and F. Runge. 2008. Factors affecting farmers’ participa-
tion in agri-environmental measures: Evidence from a case study. Journal of Agricultural
Economics 59(1): 114–131.
Delvaux, L., B. Henry de Frahan, P. Dupraz, and D. Vermersch. 1999. Adoption d ́une MAE et
consentement á recevoir des agriculteurs en région wallone. Economie Rurale 249: 71–81.
Ducos, G., P. Dupraz, and F. Bonnieux. 2009. Agri-environment contract adoption under
fixed and variable compliance costs. Journal of Environmental Planning and Management
52(5): 669–687.
Duffy, P. A., C. R. Taylor, D. Cain, and G. J. Young. 1994. The economic value of farm program
base. Land Economics 70: 318–29.
Dupraz P., I. Vanslembrouck, F. Bonnieux, and G. Van Huylenbroeck. 2002. Farmers’ participa-
tion in European agri-environmental policies. Paper prepared for presentation at the Xth
EAAE Congress Exploring Diversity in the European Agri -Food System, Zaragoza, Spain,
August 28–31, 2002.
EC-DGAGRI. 2005. Agri-environment measures overview on general principles, types of mea-
sures, and application. Unit G-4—Evaluation of Measures applied to Agriculture, Studies,
Directorate General for Agriculture and Rural Development, European Commission.
ERS. 2002. Farm policy: Title II—Conservation. Washington, DC: Economic Research Service,
United States Department of Agriculture. http://www.ers.usda.gov/Features/FarmBill/
Titles/TitleIIConservation.htm
ERS. 2006. Agricultural resources and environmental indicators. Washington, DC: Economic
Research Service, United States Department of Agriculture. http://www.ers.usda.gov/
Publications/AREI/EIB16/
ERS. 2009a. WTO domestic support notifications. Washington, DC: Economic Research
Service, United States Department of Agriculture. http://www.ers.usda.gov/db/Wto/
AMS_database/
ERS. 2009b. Briefing room farm and commodity policy: Program provisions. Washington,
DC: Economic Research Service, United States Department of Agriculture. http://www.ers.
usda.gov/Briefing/FarmPolicy/ProgramProvisions.htm
ERS. 2011. Agricultural income and finance outlook, AIS-91. Washington, DC: Economic
Research Service, United States Department of Agriculture.
European Commission. “Modulation and other financial transfers from EAGF to EAFRD”,
Directorate-General for Agriculture and Rural Development, European Commission, 2003.
European Commission. 2006. Rural development 2007–2013. Guidance document Handbook
on Common Monitoring and Evaluation Framework. Brussels: DG for Agriculture and
Rural Development, European Commission. http://ec.europa.eu/agriculture/rurdev/eval/
guidance/document_en.pdf
European Commission. 2009. Rural development (2000–2006) in EU farms. Unit L3 D
agri.l.3(2009)212727. Brussels: DG for Agriculture and Rural Development, European
Commission.
European Commission. 2011. Common agricultural policy towards 2020—Impact assessment.
SEC(2011) 1153 final. Brussels: Author.
European Court of Auditors. Is agri-environment support well designed and managed? Special
Report No. 7/2011. http://eca.europa.eu/portal/pls/portal/docs/1/8760788.PDF
European Network for Rural Development (ENRD). 2010. Public goods and public interven-
tion. Final report of Thematic Working Group 3. Brussels: Author.
Falconer, K. 2000. Farm-level constraints on agri-environmental scheme participation: A trans-
actional perspective. Journal of Rural Studies 16(3): 379–394.
Falconer, K., and C. Saunders. 2002. Transaction costs for SSSIs and policy design. Land Use
Policy 19(2): 157–166.
Fargione, J., J. Hill, D. Tilman, S. Polasky, and P. Hawthorne. 2008. Land clearing and the biofuel
carbon debt. Science 319(5867): 1235–1238.
Feather, P., D. Hellerstein, and L. Hansen. 1999. Economic valuation of environmental benefits
and the targeting of conservation programs: The case of the CRP. Agricultural Economics
Report No. 778. Washington, DC: USDA, ERS.
Feng, H., C. L. Kling, K. Lyubov, A. Kurkalova, S. Secchi, and P. W. Gassman. 2005. The con-
servation reserve program in the presence of a working land alternative: Implications for
environmental quality, program participation, and income transfer. American Journal of
Feng, H., L. A. Kurkalova, C. Kling, and P. Gassman. 2006. Environmental conservation in agri-
culture: Land retirement vs. changing practices on working land. Journal of Environmental
Economics and Management 52(2): 600–614.
Finn, J. A., D. Bourke, I. Kurz, and L. Dunne. 2007. Estimating the environmental performance
of agri-environmental schemes via use of expert consultations. Final report of the ITAES
project, Institute National de le Recherche Agronomique, Rennes, France. http://merlin.
lusignan.inra.fr/ITAES/website/Publicdeliverables
Gillenwater, M. 2012. What is additionality? Part 3: Implications for stacking and unbundling.
Discussion Paper No. 003. Silver Spring, MD: Greenhouse Gas Management Institute.
Giovanopoulou, E., S. A. Nastis, and E. Papanagiotou. 2011. Modeling farmer participation
in agri-environmental nitrate pollution reducing schemes. Ecological Economics, Elsevier
70(11): 2175–2180.
Goodwin, B. K., and V. H. Smith. 2003. An ex post evaluation of the conservation reserve,
federal crop insurance, and other government programs: Program participation and soil
erosion. Journal of Agricultural and Resource Economics 28(2003): 201–216.
Groupe de Bruges. 2012. A cap for the future!? Why we need a better CAP that can face the
challenges of today and tomorrow. http://www.groupedebruges.eu/html/publicationscap-
ref.html
Hackl, F., M. Halla, and G. J. Pruckner. 2007. Local compensation payments for
agri-environmental externalities: A panel data analysis of bargaining outcomes. European
Review of Agricultural Economics 34(4): 295–320.
Hansen, K., and B. Henry de Frahan. Evaluation of agro-environmental policy through a cali-
brated simulation farm model. Paper presented at the EAAE 2011, August 30 to September
2, 2011, Zurich, Switzerland.
Hart, K., D. Baldock, P. Weingarten, B. Osterburg, A. Povellato, F. Vanni, C. Pirzio-Biroli, and
A. Boyes. 2011. What tools for the European agricultural policy to encourage the provision
of public goods. Study prepared for the European Parliament, IP/B/AGRI/IC/2010_094,
Brussels.
Hellerstein, Daniel, Cynthia Nickerson, Joseph Cooper, Peter Feather, Dwight Gadsby, Daniel
Mullarkey, Abebayu Tegene, and Charles Barnard. 2002. Farmland protection: The role of
public preferences for rural amenities. Agricultural Economic Report No. 815. Washington,
DC: Economic Research Service, US Department of Agriculture.
Hodge, I. 2001. Beyond agri-environmental policy: Towards an alternative model of rural envi-
ronmental governance. Land Use Policy 18: 99–111.
Hodge, Ian, and Mark Reader. 2010. The introduction of Entry Level Stewardship in
England: Extension or dilution in agri-environment policy? Land Use Policy 27 (2): 270–282.
Hynes, S., N. Farrelly, E. Murphy, and C. O’Donoghue. 2008. Modelling habitat conservation
and participation in agri-environmental schemes: A spatial microsimulation approach.
Ecological Economics 66(2–3): 258–269.
Khanna, M., M. Isik, and D. Zilberman. 2002. Cost-effectiveness of alternative green payment
policies for conservation technology adoption with heterogeneous land quality. Agricultural
Economics 27(2): 157–174.
Kirwin, B., R. Lubowski, and M. Roberts. 2005. How cost-effective are land retirement auc-
tions? Estimating the difference between payments and willingness to accept in the conser-
vation reserve program. American Journal of Agricultural Economics 87: 1239–1247.
Kirwan. B. 2009. The incidence of US agricultural subsidies on farmland rental rates. Journal of
Political Economy 117(1), 2009: 138–164.
Kleijn, D., and W. J. Sutherland. 2003. How effective are European agri-environment schemes in
conserving and promoting biodiversity? Journal of Applied Ecology 40: 947–969.
Kleijn, D. 2006. Guidelines for the evaluation of agri-environment schemes. Manuscript,
Wageningen University. http://www.ncp.wur.nl/NR/rdonlyres/EA88E46A-1B26-4A04-A
F51-4FBB5AC50619/52783/Guidelines_evaluation_ae_schemes2.pdf
Lichtenberg, E. 2004. Cost-responsiveness of conservation practice adoption: A revealed pref-
erence approach. Journal of Agricultural and Resource Economics 29(3): 420–435.
Lichtenberg, E., and R. Smith-Ramírez. 2011. Slippage in conservation cost sharing. American
Liu, X., and L. Lynch. 2011. Do agricultural land preservation programs reduce farmland loss?
Evidence from a propensity score matching estimator. Land Economics 87(2): 183–201.
Lohr, L., and T. Park. 1995. Utility-consistent discrete-continuous choices in soil conservation.
Land Economics 71(4):474–490.
Loriz-Hoffmann, J. 2012. The CAP and balanced territorial development. DG AGRI unit
G1: Consistency of rural development policy. The CAP towards 2020—taking stock with
civil society. European Commission, Brussels.
Lynch, L., W. Gray, and J. Geoghegan. 2007. Are farmland preservation program easement
restrictions capitalized into farmland prices? What can a propensity score matching analysis
tell us? Applied Economic Perspectives and Policy 29(3): 502–509.
Lubowski, R. N., A. J. Plantinga, and R. Stavins. 2008. What drives land-use change in the
United States? A national analysis of landowner decisions. Land Economics 84(4): 529–550.
Mahé, L. P. 2012. Do the proposals for the CAP after 2013 herald a ‘major’ reform? Policy paper
No. 53. Paris: Notre Europe.
Mapemba, L. D., F. M. Epplin, C. M. Taliaferro, and R. L. Huhnke. 2007. Biorefinery feedstock
production on conservation reserve program land. Applied Economic Perspectives and Policy
29(2): 227.
McCann, L., and K. W. Easter. 2000. Estimates of public sector transaction costs in NRCS pro-
grams. Journal of Agricultural and Applied Economics 32(3): 55–563
McCann, L. 2009. Transaction costs of environmental policies and returns to scale: The
case of comprehensive nutrient management plans. Review of Agricultural Economics
31(3): 561–573.
Mettepenningen, E., A. Verspecht, and G. Van Huylenbroeck. 2009. Measuring private transac-
tion costs of European agri-environmental schemes. Journal of Environmental Planning and
Management 52(5): 649–667.
Mettepenningen, E., V. Beckmann, and J. Eggers. 2011. Public transaction costs of
agri-environmental schemes and their determinants—Analysing stakeholders’ involve-
ment and perceptions. Ecological Economics 70(4): 641–650.
Mezzatesta, M., D. Newburn, and R. Woodward. 2011. Additionality and the adoption of farm
conservation practices. Selected paper prepared for presentation at the Agricultural &
Applied Economics Association’s 2011 AAEA and NAREA joint annual meeting, Pittsburgh,
PA, July 24–26, 2011.
Oltmer K., P. Nijkamp, R. Florax, and F. Brouwer. 2000. A meta analysis of environmen-
tal impacts of agri-environmental policies in the European Union. Discussion paper TI
2000-083/3, Tinbergen Institute.
OMB. 2012. The appendix, budget of the United States government, fiscal year 2013. Washington,
DC: Department of Agriculture, Office of Management and Budget, Executive Office of the
President. http://www.whitehouse.gov/omb/budget/Appendix
Parker, D. P., and W. N. Thurman. 2011. Crowding out open space: The effects of federal land
programs on private land trust conservation. Land Economics 87: 202–22.
Primdahl, J., B. Peco, J. Schramek, E. Andersen, and J. J. Onate. 2003. Environmental effects
of agri-environmental schemes in Western Europe. Journal of Environmental Management
67: 129–138.
Primdahl, J., J. P. Vesterager, J. A. Finn, G. Vlahos, L. Kristensen, and H. Vejre. 2010. Current use
of impact models for agri-environment schemes and potential for improvements of policy
design and assessment. Journal of Environmental Management 91: 1245–1254.
Pufahl, A., and C. R. Weiss. 2009. Evaluating the effects of farm programs: Results from propen-
sity score matching. European Review of Agricultural Economics 36(1): 79–101.
Reichelderfer, K., and W. G. Boggess. 1988. Government decision making and program per-
formance: The case of the conservation reserve program. American Journal of Agricultural
Economics 70(1):1–11.
Roberts, M. 2004. Effects of government payments on land rents, distribution of payment. In
Decoupled payments in a changing policy setting, eds. Mary E. Burfisher and Jeffrey Hopkins.
Agricultural economic report No. (AER838). Washington, DC: USDA. http://www.ers.
usda.gov/publications/aer838/
Roberts, M. J., and S. Bucholtz. 2005. Slippage in the Conservation Reserve Program or spuri-
ous correlation? A comment. American Journal of Agricultural Economics 87(1): 244–250.
Roberts, M. J., and R. N. Lubowski. 2007. Enduring impacts of land retirement policies: Evidence
from the Conservation Reserve Program. Land Economics 83(4): 516.
Rørstad, P. K., A. Vatn, and V. Kvakkestad. 2007. Why do transaction costs of agricultural poli-
cies vary? Agricultural Economics 36(1): 1–11.
Searchinger, T., R. Heimlich, R. Houghton, F. Dong, A. Elobeid, J. Fabiosa, S. Tokgoz, D. Hayes,
and T. Yu. 2008. Use of US croplands for biofuels increases greenhouse gases through emis-
sions from land-use change. Science 319(5867):1238.
Secchi, S., P. Gassman, J. Williams, and B. Babcock. 2009. Corn-based ethanol produc-
tion and environmental quality: A case of Iowa and the Conservation Reserve Program.
Environmental Management 44(4): 732–744.
Selenius J., L. Baudouin, and A. Kremer (Eds.). 2011. Farm data needed for agri-environmental
reporting. Methodologies and working papers. Luxembourg: EUROSTAT.
Sullivan, P., D. Hellerstein, L. Hansen, R. Johansson, S. Koenig, R. Lubowski, W. McBride,
D. McGranahan, M. Roberts, S. Vogel, and S. Bucholtz. 2004. The conservation reserve
program: Economic implications for rural America. Agricultural economic report No.
AER-834. Washington, DC: USDA.
US EPA. Undated. “Concentrated Animal Feeding Operations (CAFO) - Final Rule”, US
Environmental Protection Agency. http://cfpub.epa.gov/npdes/afo/cafofinalrule.cfm
Vanslembrouck, I., G. Van Huylenbroeck, and W. Verbeke. 2002. Determinants of the will-
ingness to participate in agri-environmental measures. Journal of Agricultural Economics
50(3): 489–511.
Vasavada, U., S. Warmerdam, and W. Nimon. 2001. Green box policies and the environment.
Briefing paper, URAA Issues Series. Washington, DC: ERS, US Department of Agriculture,
WTO Briefing Room. http://webarchives.cdlib.org/sw1wp9v27r/http://ers.usda.gov/
Briefing/wto/environm.htm
Wallander, S., and M. S. Hand. 2011. Measuring the Impact of the Environmental Quality
Incentives Program (EQIP) on irrigation efficiency and water conservation. Conference
paper, 2011 meetings of the Agricultural and Applied Economics Association, Pittsburgh,
Pennsylvania.
Westhoek H., H. van Zeijts, M. Witmer, M. van den Berg, K. Overmars, S. van der Esch, W. van
der Bilt. 2012. Greening the CAP: An analysis of the effects of the European Commission’s
proposal for the Common Agricultural Policy. PBL note. The Hague: Netherlands
Environmental Assessment Agency.
Wossink, G. A., and J. H. Van Wenum. 2003. European biodiversity conservation by farm-
ers: Analysis of actual and contingent participation. Review of Agriculture Economics
30(4): 461–485.
Wu, J. J., and W. G. Boggess. 1999. The optimal allocation of conservation funds. Journal of
Wu, J. 2000. Slippage effects of the conservation reserve program. American Journal of
Agricultural Economics 82(4):979–992.
Wu, J. J., D. Zilberman, and B. Babcock. 2001. Environmental and distributional impacts of
conservation targeting strategies. Journal of Environmental Economics and Management
41(3): 333–350.
Wu, J., R. Adams, C. Kling, and K. Tanaka. 2004. From micro-level decisions to land-
scape changes: An assessment of agricultural conservation policies. American Journal of
Agricultural Economics 86(February):26–41.
Wu, J. J. 2005. Slippage effects of the Conservation Reserve Program: Reply. American Journal
Wynn, G., B. Crabtree, and J. Potts. 2001. Modeling farmer entry into the environmentally sen-
sitive area schemes in Scotland. Journal of Agricultural Economics 52(1): 65–82.
C HA P T E R 24
S T I G M AT I Z E D S I T E S A N D
U R BA N BROW N F I E L D
REDEVELOPMENT
JOE L B. E I SE N
This chapter addresses the “stigmatized sites” located in urban areas in the United
States and Europe and the “brownfields” redevelopment programs aimed at removing
the stigma and promoting remediation and reuse of these sites. Although the European
Union has put regulatory frameworks in place (Pahlen 2004), the United States has led
the global effort to address brownfields redevelopment (Eisen 1996; Sarni 2009; Davis
2011), and the discussion in this chapter will focus on American models for brownfields
remediation and reuse.
Typically, the term “brownfields” has come to refer primarily to abandoned or unde-
rused urban sites (Eisen 1996; Paull 2008; Wernstedt et al. 2010; US Environmental
Protection Agency 2011b), often located in declining cities with industries that have
ceased operations (for example, the “Rust Belt” cities in the Northeast and Midwest
of the United States) (Robertson 1999; US Environmental Protection Agency 2011c).
Brownfields can be found throughout the nation, in rural and suburban areas, as well
as in cities, but urban sites have attracted the most attention. These sites have often had
a number of owners and a long history of industrial or commercial uses (Eisen 2007).
Frequently, the former owners are not in possession of the sites (and, often, no lon-
ger in existence), and the sites are owned by cities or other public entities (Eisen 1996;
Hollander 2009).
A brownfield site may be a small parcel, but many brownfield sites are the larger prop-
erties that once were the former “crown jewels” of the cities in which they are located (US
Environmental Protection Agency 2005). In many cities in the United States, Europe,
and elsewhere, brownfields are among the most visible urban properties, such as rot-
ting hulks of abandoned steel mills or other manufacturing facilities, formerly grand
railroad stations no longer carrying passengers and sitting idle, and other neglected
properties (Wernstedt et al. 2004). These can be large, prominent sites located in the
Stigmatized Sites and Urban Brownfield Redevelopment 649
urban core near railroads, highways, other forms of transportation, and the bulk of the
city’s population (Eisen 1996). They frequently attract attention and interest in rede-
velopment from a wide range of public and private sector entities that may play roles in
their redevelopment, including real estate developers, investors, business enterprises,
nonprofit organizations, government representatives, and elected officials (Wernstedt
et al. 2004).
What are the optimal use and societal benefits of redevelopment at a brownfield
site? Brownfields redevelopment has many potential benefits. Reinvesting in an urban
core can be the linchpin of a strategy to thwart sprawl (unchecked growth in subur-
ban and exurban areas) and preserve open space (Paull 2008). In recent years, the idea
of sustainability has gained traction as a means for pursuing a more holistic approach
to urban redevelopment that may include brownfields remediation and reuse, among
other strategies (Eisen 1999). Another challenge that brownfields redevelopment strate-
gies may help address is the urgent need to reduce greenhouse gas (GHG) emissions
to address climate change. In the United States, the second largest share of GHG emis-
sions comes from transportation, and a large part of that comes from urban commuters.
Redevelopment of brownfield sites, if done properly, could spur a decrease in emissions
by reducing the amount of vehicle miles traveled (Wernstedt et al. 2004).
The challenges to redeveloping brownfield sites are as numerous as those present at
any urban site. However, brownfield sites are not properly priced for current develop-
ment, in large part because they carry a stigma reflecting the possible presence of envi-
ronmental contamination (Davis 2011). The primary attribute and added challenge to
development of a brownfield site, as compared to other urban sites, is that it is com-
monly believed that one or more entities contaminated brownfield sites in the past,
making decisions that did not require them to reflect the full social costs of pollution,
but that the extent of the contamination and added costs are unknown.
In the mid- to late-1980s, the idea began to take shape that the stigma associated
with brownfield sites was not a result of larger societal forces, such as changes in con-
sumer preferences or residential patterns, but was instead a byproduct of governmen-
tal laws and programs designed to force the remediation of contaminated sites (such
as CERCLA, the “Superfund law,” in the United States) (Eisen 1996). There are few
reliable estimates of the number of brownfield sites, due to many factors, including
the imprecision of data collection and the uncertainty whether any specific site car-
ries the stigma of potential environmental contamination. Unofficial estimates of total
brownfield sites in the United States are based on incomplete lists dating to the 1980s,
including state inventories and the EPA’s CERCLIS database that identified poten-
tially contaminated sites. Based on these figures, it is often stated that there may be
from 400,000 to more than a million in the United States alone (National Association
of Local Government Professionals and Northeast-Midwest Institute 2004; Wernstedt
et al. 2010; Davis 2011). Recent figures are more precise. For example, a 2010 report
from the US Conference of Mayors, based on a survey of 150 major cities in 41
American states, identified a total of more than 22,000 sites in these cities alone (US
Conference of Mayors 2010).
650 Joel B. Eisen
At brownfield sites, there is a daunting information asymmetry for would-be devel-

opers. Many brownfields sites sit abandoned for a decade or more without any envi-
ronmental investigation, so it is often difficult to discern the extent of contamination
or whether they would be subject to the requirements prevailing under environmen-
tal cleanup laws (Eisen 1996). Once the potential and uncertain costs of environmental
monitoring and other policy costs (e.g., dealing with local land use authorities in the
redevelopment process) are factored in, developers’ reluctance to become involved with
these sites is understandable.
1. Brownfields and the Broader

Context of Urban Redevelopment
Redevelopment of brownfield sites cannot be considered in a vacuum, but must instead

be examined against the broader context of urban redevelopment activities (Robertson
1999). The idea that a city that has fallen into decline and decay can stop or reverse that
slide through revitalization efforts is not new to the twenty-first century (Kunstler 1993).
Nor is it a new idea that some cities that face deplorable conditions eventually regain
their prominence or that others fail to do so and are consigned to the dustbin of history.
The causes of urban decay in the modern era are well chronicled (Bradbury et al. 1982;
Duany et al. 2001; Hollander 2009). A city may experience deindustrialization when its
dominant manufacturing industry declines due to adverse business conditions, leading
to vacancies in commercial and industrial areas, a declining tax base, high unemploy-
ment, and other indicia of decline (Hollander 2009). A city’s geographic advantage may
fade if the advantage conferred no longer works in the city’s favor due to technologi-
cal obsolescence or other factors (as in the case of Buffalo when the railroads carried
freight traffic more expeditiously than the Erie Canal) or by construction of a transpor-
tation artery that bypasses it. After World War II, public policy at all levels of govern-
ment encouraged building of housing in the suburbs, and urban residents migrated out
as a result, further contributing to declines in economic activity in central core cities
(Bradbury et al. 1982; Duany et al. 2001; Hollander 2009).
Continuation of a city’s decay may appear inevitable. A center city area may decline
as the outer areas grow, no matter what redevelopment activities are undertaken.
This, of course, would suggest that it is futile to engage in redevelopment activities.
However, the arc of a city’s slide is often debatable. There have been substantial efforts
made to revitalize inner cities in the United States, and demographic trends suggest
that, in some cities, these efforts have had some success because some Americans have
moved back into the cities and made them desirable again (Kromer 2010). Although
some speak of decline and rebirth as evidence that a city “lifecycle” exists, this theory
is neither universally accepted nor reliable as a marker for brownfields redevelopment
(Hollander 2009).
When policy makers contemplate potential policies for addressing urban decline, two
intriguing questions present themselves. The first is “what are the measures of decline?”
Is decline measured in statistics about growth rates, population, employment, and so
forth? Is it visual, or is it measured in more comprehensive ways (such as a perception
among residents that standards of living have declined)? The answers to this are myriad
but can help define the goals of redevelopment activities. The metrics a city chooses to
measure success of redevelopment are presumably those it should pursue in its focus at
individual brownfield sites. A second question is more properly oriented to the urban
institutional architecture: “from whose perspective is decline measured?” Who is enti-
tled to control the destiny of urban redevelopment activities? The city’s government?
The entity that proposes to undertake redevelopment activities? The community in and
surrounding the area(s) planned for redevelopment? All of the above, in partnership?
The answer to this question is important because different actors may have diverging
ideas about the ideal plan for transforming an urban brownfield site (Eisen 1999).
There are no easy answers to these questions. Indeed, it may be the case that a par-
ticular site has been the locus of attention on more than one occasion and has been a
component of more than one type of development strategy. It may even be the case that
previous redevelopment strategies have been responsible for hastening the decline of a
site (Kunstler 1993). Over time, American and European cities have engaged in experi-
mentation, embracing numerous ideas about how to redevelop their cities. Not all of
these programs were successful. The urban renewal programs of the 1950s in the United
States rehabilitated some neighborhoods but exacerbated problems in others by tar-
geting “slums,” displacing residents, and creating public housing projects that became
symbols of urban failure (Kunstler 1993; Kromer 2010). Rather than improve residents’
standards of living, urban renewal and other policies (notably federal highway assis-
tance and mortgage insurance programs) often contributed to out-migration from cities
and further decay (Bradbury et al. 1982; Duany et al. 2001). The construction of high-
ways through urban neighborhoods often split them, hastening their decline. Rust Belt
cities in the United States declined, with migration taking place to the suburbs and new
Sun Belt cities (Kunstler 1993).
Starting in the 1980s and continuing since then, cities have often pursued redevel-
opment through mega-projects: stadiums, convention centers, shopping districts and
cultural hubs, and other attractions they view as essential to attract other activities
for redevelopment (Duany et al. 2001; Kromer 2010). The evidence on these projects
is mixed, with some being successes in attracting development to co-locate with them
and others being expensive failures (Kromer 2010). Gentrification of urban neighbor-
hoods with upscale housing projects, parks, and other amenities is controversial and
often results in the same types of urban displacement as urban renewal strategies, shift-
ing the locus of economic decline to suburbs (some of which now have the same types
of urban concerns as the core cities) or more depressed areas of cities (Kunstler 1993;
Duany et al. 2001). One important question then becomes whether governments at
all levels should continue to subsidize these transformative activities. This continuing
debate has the effect of making policy support inconsistent. To take just one of many
652 Joel B. Eisen
examples, rehabilitation tax credits helped promote redevelopment activities between

the mid-1970s and 1980s, but when they were rolled back in the 1986 tax code revision,
they lost their effectiveness (Davis 2011).
This historical approach suggests that a city should be careful to take a long-term
approach and view its abandoned or vacant brownfield sites as community and eco-
nomic opportunities, connecting redevelopment initiatives to broader community
visions and revitalization priorities (Eisen 1999; US Conference of Mayors 2010). When
a city focuses on a specific urban site and considers whether to pursue a new land use at
that site, the optimal strategy would recognize the role of that site in an overall plan for
urban redevelopment. As noted earlier, a wide variety of socioeconomic forces may be
responsible for the decay of a city and of specific sites in that city. It may be difficult or
impossible to separate the problems that led to decline at individual sites from citywide
or economy-wide trends, and therefore these broader trends should be addressed rather
than pursuing parcel-by-parcel redevelopment (Eisen 1999). A city should also ideally
recognize that it would need to revisit its plans and strategies even after they have been
implemented because, as already noted, even well intentioned urban redevelopment
strategies sometimes come to be viewed in hindsight as mistakes.
2. Brownfields and Their Developers
There may be important reasons to consider development at specific abandoned and

underused brownfield sites. These sites may have advantages over other locations.
For example, some existing geographic advantages that led predecessor businesses to
locate there may be helpful to new enterprises, such as proximity to transportation links
(Bartsch 1996; Eisen 1996; Robertson 1999). Yet, in many American and European cit-
ies, those who would become involved with transforming these sites to new, productive
purposes (who will be termed “developers” for the remainder of this chapter, although
they are not always real estate developers) have the choice of foregoing redevelopment
altogether and selecting sites at suburban and exurban locations, which can be more
attractive options than urban redevelopment for a variety of reasons (Eisen 1996; Davis
2011). So-called “greenfield” locations can often be developed without the need to deal
with whatever infrastructure may exist at the urban site.
For many developers, the cost of redeveloping a brownfield site can be higher than
that of a greenfield site. They must incur upfront investigation costs that can cause delays
they do not encounter with undeveloped greenfields. Much of the infrastructure at a
brownfield site is not likely to be suitable for the developer’s intended purpose, which
will be different from the preceding use, so new infrastructure must be built at a higher
cost. However, it is not always the case that a greenfield is optimally priced for devel-
opment in comparison to a brownfield site. The greenfield conversion itself can create
negative externalities that may not be reflected in the site’s purchase price. Moreover,
if the greenfield site in question has been the site of prior development activities, the
costs of environmental remediation and monitoring might even negate the apparent
cost advantage of developing there (Boyd et al. 1996). And, as is often the case in trans-
formative activities such as residential remodeling, it is also possible that some existing
infrastructure can be preserved at a brownfield site, thus lessening the cost of redevelop-
ment (Paull 2008).
From the outset of the development of brownfields law and policy, much was therefore
made of the need to provide incentives to attract developers to pursue redevelopment
efforts at urban sites. In this perspective, the developers’ interests are elevated over those
of other actors, including, for example, the residents of the surrounding community
(Eisen 1996). Developers often find it necessary to invest their own capital, but they also
typically can obtain resources for a site’s evaluation, remediation, and reuse from a wide
variety of federal and state agencies in the form of site assessment grants, loans, training
and education programs, and tax and other financing incentives. As discussed later, the
lesser cleanup standards at brownfield sites also operate as a sort of financial incentive
to developers, cutting the costs of remediation. For their part, cities have been willing
to provide prospective developers with tax breaks, create special districts, and establish
other incentives for them to take on the task of transforming sites (Rosenberg 2000).
Studies have consistently found that one of the major drivers of success in brown-
fields redevelopment is the extent to which these sources of public funds are available
to leverage the developer’s investment (US Environmental Protection Agency 2005;
US Conference of Mayors 2010). This can be especially important when the develop-
er’s investment is front-loaded, as, for example, when site assessment and remediation
costs threaten to exceed the current values of the brownfield sites in question (National
Association of Local Government Professionals and Northeast-Midwest Institute
2004). Conversely, some have argued that developers should not be extended financial
incentives for actions they would take without the incentives, especially if a brownfields
program makes less than full remediation of environmental contamination a possibility
(Eisen 1996). The frequent response to this argument is that, at many sites, any reme-
diation and reuse activities are preferable to allowing the sites to remain abandoned or
underused (Davis 2011).
The prevailing justification for providing incentives to developers and tailoring the
brownfields remediation system to their needs has been that developers savvy enough to
understand the risks they might be taking on might balk at contacting state and federal
environmental authorities, fearing the worst-case scenario of encountering toxic sub-
stances at the sites and being required to undertake multimillion dollar cleanups (Boyd
et al. 1996). This argument is relatively straightforward. One cannot know in advance
what might be present at a site that has a history of industrial or commercial uses and is
therefore contaminated to some extent (Davis 2011). If investigating the conditions at a
brownfield site, let alone making decisions about how to remediate them, might expose
a developer to liability, then the environmental costs associated with a particular site
cannot be quantified ahead of time (Davis 2011). Contemplating a worst-case scenario
can lead to project cost estimates that threaten to jeopardize project profitability, if spi-
raling remediation costs are factored in.
654 Joel B. Eisen
The prevailing brownfields narrative focuses on assumptions about the present con-
dition of the site and the idea that environmental laws are the problem, not the solution
(Eisen 1996). It does not usually take into account whether there might be any connec-
tion between the conditions at a brownfield site and urban redevelopment activities of
the near past. Nor does it account for any risk minimization tools that might be avail-
able to developers, such as environmental insurance. Finally, it does not account for
an interesting paradox with respect to any specific site. Although the condition of the
site is assumed to be an unquantifiable unknown, this is often not the case. At the very
least, it is typically known that a brownfield site is not one that has been targeted for
enforcement action. In the United States, the sites on the National Priorities List (NPL)
are those eligible for funding from the government’s Superfund and are the highest pri-
orities for cleanup. Most brownfields programs targeted sites that were not on the NPL
(Bartsch 1996; Eisen 1996; Geltman 2000; Davis 2011).
In this evolving policy framework, the need to attract developers takes precedence
over other means for steering redevelopment of the site. Legal and policy evolution
could well have addressed developers’ cost concerns directly by simply providing signif-
icant funding for site investigations that would delineate the extent of the risk. The fed-
eral and state governments did make funds available for this purpose, but brownfields
advocates sought more than that. One outgrowth of the focus on developers was that
if the chief concern was the overreach of the environmental laws governing cleanup of
hazardous waste sites, the natural tendency was to go beyond providing funds to evalu-
ate sites and instead to suggest that these laws be relaxed to permit more streamlined
redevelopment activities (Eisen 1996).
3. The Relationship Between

Brownfields and State and Federal
Environmental Laws
There is nothing new about attributing urban decline to governmental actions that were
intended to revitalize cities. Today, as noted earlier, the urban renewal of the 1950s in the
United States is widely regarded as a series of projects that, although well intentioned,
led to failure that exacerbated the decline of many cities. With brownfields, the prevail-
ing narrative that drove the creation of laws and policies for remediation and reuse was
altogether different: that the stigma originated indirectly, from supposedly unintended
consequences of actions that governments have taken for a beneficial purpose under
the environmental laws (Eisen 1996; Robertson 1999; Davis 2011). The problem was
not only that the extent of contamination at a brownfield site was unquantifiable (which
could have been addressed with widespread use of site assessment grants), but also that
attempting to evaluate brownfield sites might subject developers to strict liability for full
cleanups at the sites regardless of their lack of prior involvement with the sites (Geltman
2000; Hollander et al. 2010; Davis 2011).
Laws such as the US Comprehensive Environmental Response, Compensation, and
Liability Act (CERCLA, more popularly known as the “Superfund law” for the fund cre-
ated under it to remediate hazardous waste sites) require full remediation of danger-
ous conditions at hazardous waste sites (Robertson 1999; Rosenberg 2000; Davis 2011).
The CERCLA cleanup process can take years and millions of dollars to complete (Eisen
1996; Davis 2011). Because CERCLA has been interpreted to fasten joint and several
liability on those responsible for the toxic conditions at the sites, including owners and
“operators” of the sites, an entity that becomes involved with a brownfield site might face
the full price tag of remediation even if it did nothing to cause the contamination there
(Eisen 1996; Davis 2011). Knowing that price tag ahead of time is next to impossible
because the CERCLA cleanup standard is determined through a lengthy process (which
includes a number of requirements that apply under other federal environmental laws)
that cannot be completed before site investigation and assessment activities are com-
pleted (Eisen 1996; Robertson 1999; Davis 2011).
Throughout the 1980s, courts in the United States had strengthened CERCLA, con-
sistently interpreting it to give the Environmental Protection Agency (EPA) more power
and more authority to investigate and remediate sites and confer on the EPA and pri-
vate parties more tools to fasten liability on those responsible for the sites (known in
Superfund parlance as “potentially responsible parties,” or PRPs). By the end of the
decade, the regulated community viewed CERCLA as a regulatory scheme with unprec-
edented and unfortunate power and breadth (Davis 2011). In addition to joint and sev-
eral liability (which, as noted earlier, can fasten the entire price tag of a cleanup on a
single PRP), CERCLA features strict liability without regard to fault, so the EPA need
not prove that a PRP intended to dump waste at a site or was reckless about doing so
(Eisen 1996; Rosenberg 2000). It is only necessary to connect the PRP with the site, for
example, by showing that a company made wastes of the sort that were dumped at the
site (Davis 2011).
The EPA has sweeping powers to ensure that a site is remediated. For example, it
can use “notice letters” to force PRPs to discover and identify other PRPs, and it can
issue unilateral administrative orders with onerous penalties for their violation (Davis
2011). The liability provisions were so broad that their net captured a wide group of enti-
ties, whether or not they had any involvement in the actual dumping of waste at the
sites, and thousands of companies became PRPs (Davis 2011). Current property own-
ers, for example, make up one category of PRPs, whether or not they owned the sites
at the time of disposal, which potentially subjected brownfields developers to liability
(Eisen 1996; Rosenberg 2000). In the 1980s and 1990s, CERCLA had extremely lim-
ited defenses, although a brownfields defense was finally added in 2002 (Small Business
Liability Relief and Brownfields Revitalization Act [SBLRBRA] 2002; Eisen 2007). That
defense protects otherwise innocent prospective purchasers who did not know and had
no reason to know that hazardous substances were disposed of at the site and who take
due care to protect against foreseeable contamination. This requires the purchaser to
656 Joel B. Eisen
have made “all appropriate inquiry” to discern the extent of contamination. This term is
defined in the EPA’s regulations and typically requires some form of investigation at the
site (US Environmental Protection Agency 2011b).
By the end of the 1980s, it was apparent that the Superfund law had spurred identi-
fication of many contaminated sites around the country, and some were in the process
of being remediated. However, this was not an efficient process because it took years to
move a site through the labyrinthine CERCLA cleanup framework. For this reason and
others, CERCLA had created more litigation and work for lawyers than any other state
or federal environmental statute.
In the 1980s, CERCLA successes were still largely to take place in the future, and the
voices of backlash decried the slow progress of cleanups at hazardous waste sites and
criticized the statute itself as harsh and even arbitrary at times (Eisen 1996). This was
different from the public clamor a decade earlier, spurred by high-profile incidents such
as the discovery of toxic contamination at Love Canal in western New York, which had
led to the creation of the CERCLA regime. At the inception of the CERCLA program,
the overriding purpose of the law was ensuring that sites such as Love Canal were reme-
diated (Davis 2011). Throughout the 1980s and 1990s, proponents of the law success-
fully argued that any systematic exception to the joint and several liability framework
would weaken it and frustrate the Congressional intent to ensure that dangerous sites be
remediated (Eisen 1996).
Although amendments to CERCLA and EPA policies and interpretations of the
statute attempted to alleviate some of the burden faced by PRPs, they were limited in
scope (Davis 2011). The 1986 amendments added an “innocent landowner” defense to
CERCLA that might have protected those, like brownfields developers, who purchased
sites after waste dumping took place and were therefore not responsible for contaminat-
ing them. However, this defense was not widely used because meeting its requirements
was difficult (Rosenberg 2000). Among other requirements, it called for those seeking to
prove their innocence to demonstrate that they had engaged in “all appropriate inquiry”
before purchasing the site. Most courts interpreted this to mean that if prospective pur-
chasers had not discovered the contamination before purchasing the sites, they prob-
ably had not conducted sufficient inquiries (Schnapf 2007). This was exactly the type
of activity that brownfields developers were loath to undertake before purchasing sites.
Thus, rather than protect brownfields developers, the shortcomings of the innocent
landowner defense appeared to be a primary reason why more reforms to the CERCLA
structure were needed (Eisen 1996).
As a result, it is important to note that the advocacy for brownfields remediation and
reuse became more intense at the same time that a backlash was taking place against
CERCLA and burgeoning hazardous waste cleanup schemes in Europe. In general, it is
difficult to separate the growing clamor at the time for relief for brownfields developers
from the general calls for softening the tough liability-based approach of the Superfund
law. Brownfields advocates sought partnerships with environmental agencies rather
than adversarial enforcement-based relationships, shorter cleanup processes with more
finality (including releases or other forms indicating that the brownfields purchaser
would not face liability), and lesser cleanup standards that, in some cases, would allow
less costly means of addressing contamination at the site (e.g., so-called institutional
controls, such as fences and warning signs). That these were the same sorts of changes
sought generally for the harsh and unyielding Superfund scheme did not go unnoticed
(Eisen 1996).
Economic conditions in the broader US economy added fuel to the reform fire. The
stock market crash of 1987, savings and loan collapse, and ensuing recession that lasted
through the early 1990s prompted calls for development activities to spur job creation,
and policy makers in the United States and Europe increasingly turned their attention
to revitalizing urban cores (Bartsch 1996). In this struggling macroeconomic environ-
ment, the emerging cracks in the societal consensus that remediation of hazardous
waste sites was always an unalloyed good grew even wider. In the United States, brown-
field sites tended to be concentrated in states that were hit hardest by the recession of the
1990s, and, for many policy makers, it was important to put these sites to productive use
(Bartsch and Collaton 1997). A frequently voiced concern at the time was that CERCLA
and its state analogues required a level of cleanup that went too far at sites such as these
(Eisen 1996).
The term “brownfield” itself was an invention of these times, being first coined at a
1992 Congressional field hearing in Ohio (Bartsch 1996). Although “brownfield” was
meant as a counterpoint to “greenfield,” a term then in vogue to describe untouched
sites in the suburbs and exurbs, it had an unmistakably pejorative cast to it. It suggested
that these sites did, in fact, have a stigma associated with them—no one eagerly associ-
ates with “brown” sites—and advocates for change maintained that this stigma could
only be removed with reforms to environmental laws, although other legal reforms
would be necessary as well (Bartsch 1996). One natural response to the clamor for
brownfields reforms might have been that the calls for less regulation, and that it be
more transparent and easier to comply with, were the typical unjustified response of
a regulated community to environmental regulation. In this respect, developers and
their advocates often did themselves no favors by calling for “streamlined” regulations
to empower redevelopment (Eisen 1996). Therefore, at least some who were involved
in the brownfields law and policy development process recognized that any relief
would have to be carefully moderated and tailored rather precisely to the brownfields
situation.
An ironic twist on the situation was that while brownfields developers were said to
fear the federal Superfund scheme, far more brownfields sites were subject to the reach
of state environmental laws (Bartsch and Collaton 1997). There was widespread recog-
nition from the early days of the development of the regulatory frameworks for brown-
fields remediation and reuse that most sites would be addressed by the states (Bartsch
and Collaton 1997). In the United States, virtually every state had created a state law
analogue to the Superfund law, with comparable features (Wernstedt et al. 2010; Davis
2011). Some states, particularly in the Northeast, had property transfer laws as well,
requiring evaluation (and remediation if necessary) of potentially contaminated sites
prior to their transfer (Eisen 1996; Bartsch and Collaton 1997; Davis 2011). Therefore,
658 Joel B. Eisen
brownfields policy reform began in the states, where individual states could tailor their
programs to their own specific needs (Bartsch 1996).
4. The Rise of “Voluntary Cleanup

Programs”
These conditions all made for an environment ripe for change, and it is not surprising
that, between the late 1980s and mid-1990s, many states overhauled their environ-
mental cleanup and property transfer laws (Eisen 1996; Eisen 2007), and the European
Union later developed standards for brownfields remediation (Pahlen 2004).
Because the states were the drivers of brownfields legal changes, most activity in state
brownfields programs before 2002 took place without significant reforms to the federal
Superfund law, although some federal programs, such as the EPA’s use of “prospective
purchaser agreements,” did make some progress toward protecting brownfields devel-
opers (Eisen 1996; Rosenberg 2000; Davis 2011). Thus, prospective brownfields devel-
opers during this time were in an interesting situation, to say the least. They could obtain
comfort from state environmental agencies that no enforcement actions would be taken
against them if they proceeded to investigate conditions at sites and remediate them if
necessary, but there was always the potential that they might face federal liability (par-
ticularly if the cleanup in a state brownfields program was to a level less complete than
required under the federal standard) (Eisen 1996). Eventually, however, many thou-
sands of sites were processed successfully through state brownfields programs despite
what some saw as the specter of federal liability. States negotiated with the EPA to cre-
ate memoranda of understanding that secured a level of comfort for sites addressed in
their programs (Eisen 1996; Davis 2011). It was also evident that changes enhancing the
performance of state programs were almost guaranteed to eventually lead to changes
to the Superfund law itself because the state laws being changed were analogous to the
federal law.
The state brownfields programs are generally known as “voluntary cleanup programs”
(VCPs) because their central feature is that developers voluntarily come to the states and
initiate dialogues intended to lead to productive remediation and reuse of brownfields
sites (Eisen 1996; Davis 2011). This differentiates the operational paradigm of these pro-
grams sharply from normal enforcement-driven CERCLA models, in which a PRP’s
first contact with the government is typically an adversarial notice that it faces liability
(Eisen 1996). The intent of this model was to make the entire remediation and reuse
process more flexible and less confrontational from the developer’s perspective and to
create a working relationship between state regulators and developers (Dana 2005).
The features of VCPs differ widely, but most include three central attributes (Eisen
1996; Gerrard 1998 and Supp. 2006; Robertson 1999; Rosenberg 2000; Wernstedt et al.
2010; Davis 2011).
4.1 Streamlined Administrative Processes

In VCPs, the steps between identification of a site as one to which the program will apply
and final remediation and reuse are far less in number and shorter in duration than in
enforcement-driven models. Often, developers were put in control of many steps of the
process (in some states, by hiring a licensed environmental professional to administer
the entire cleanup) (Eisen 1996). One major distinction of the process from the normal
CERCLA model deserves special mention: the role of the community surrounding a
brownfield site. In CERCLA cleanups, public participation in cleanup activities is man-
datory and proceeds on a legally defined parallel track to the remediation itself; it can be
an involved, complex process, depending on the nature of the sites. In VCPs, by contrast,
many states empowered the developers themselves to determine on their own whether
and how to involve the public in decision making at the sites (Eisen 1996).
4.2 Risk-Based Cleanup Standards

In VCPs, the end use to which sites will be put (such as commercial, industrial, or
residential) is factored into the risk assessment of the sites, leading to standards that
typically require less than complete remediation. Often, developers could cut costs by
adopting remedies that were less comprehensive, such as entombing soils at brownfield
sites rather than removing and treating them. In the states that empowered developers
to employ licensed environmental professionals, these professionals were authorized to
decide the level of cleanup standards within defined parameters, which took these criti-
cal decisions out of the hands of state environmental authorities (Eisen 1996; New Jersey
Institute of Technology 2011).
4.3 Liability Protection

States offered a number of means for developers to secure protection against future
enforcement actions by their environmental agencies, ranging from “no further action”
letters (statements of intent that developers would not face liability in the future) to full
releases from liability. These forms of protection do not offer any shield against federal
liability but were typically viewed as sufficient for developers to continue with redevel-
opment activities (Eisen 1996).
The federal government has stepped in to assist brownfield developers as well. The
2002 amendments to CERCLA provided liability protection from CERCLA for devel-
opers of brownfields sites through a prospective purchaser exemption (SBLRBRA
2002). The EPA subsequently issued regulations to clarify how the protection would
be obtained; for example, by requiring the prospective purchaser to allow access to
and cooperate with regulators and to exercise care in dealing with the prior releases
660 Joel B. Eisen
of contaminants at the site so as not to exacerbate the problem (US Environmental

Protection Agency 2005; Eisen 2007; Schnapf 2007). The new law also supported the
continued use of VCPs by restricting federal actions under CERCLA against developers
who remediate sites under VCPs (SBLRBRA 2002).
The EPA implemented the 2002 law by providing additional federal grants and other
tools to address the contamination of brownfields sites (US Environmental Protection
Agency 2011a). These resources include grant programs for characterization of
brownfields sites and site analytical tools that can help make the process of environ-
mental investigation and remediation less onerous and costly (Hollander et al. 2010;
US Environmental Protection Agency 2011a). Beyond the EPA’s activities, federal gov-
ernmental agencies at all levels have adapted existing programs in such areas as infra-
structure, housing, and community development to promote site redevelopment and
reuse (Robertson 1999; Sarni 2009; Hollander et al. 2010). Interagency partnerships
have also become productive means of addressing important issues (US Environmental
Protection Agency 2011a).
5. Public Input in the Brownfields

Remediation and Reuse Process
The use to which a brownfields site would be put after its remediation has not been
central to the decision-making architecture of most VCPs, except insofar as it has fac-
tored into the choice of the cleanup standard (Eisen 1996). It has been assumed that the
developer controlled the choice of land use, in that it would not have approached the
state in the first instance without a plan for site redevelopment. A VCP could, and in
many cases, did, therefore sanction site plans that related only to the individual brown-
fields sites and bore little relationship (if any at all) to any comprehensive plan for urban
redevelopment (Eisen 1999). There were exceptions to this, as some American cities
eschewed parcel-by-parcel redevelopment in favor of more comprehensive approaches
(US Conference of Mayors 2010). Also, some states created area-wide brownfields ini-
tiatives, in which environmental regulators and state development agencies worked
together to address multiple brownfields in the same community (van Hook et al. 2003),
but these states were in the minority. There is nothing about most VCPs that requires
consistency with a holistic vision for the future of the particular city (Eisen 1999).
From the start, VCPs focused on user friendliness: alleviating the regulatory burden
for developers, not creating relationships with communities (Davis 2011). As a result,
public input in deciding the future of brownfields sites has often been limited (Eisen
1996). If residents in the community surrounding a brownfield site wanted a park
instead of a mega-project, they had little power to influence the decision, and it was, in
fact, rare for a prospective developer to change the plans for the site. In many cities, there
were no obvious stakeholders to voice the concerns of the residents near brownfields
sites, let alone engage in discussions with developers that had already prepared plans
for the sites and approached state environmental authorities (Davis 2011). It is rare for
residents living near brownfields sites to demonstrate the level of political mobilization
that can derail projects, and the surrounding communities much more often consist of
those who have historically been marginalized in urban planning efforts (Speiss 2008).
Moreover, financial concerns are usually more important to a brownfields developer
than satisfying the needs of the community (Wernstedt et al. 2010).
As VCPs have matured, savvy developers have turned to citizen “steering com-
mittees” or other community-based groups, particularly relying on existing groups
that have knowledge and expertise of the affected communities (National Center For
Neighborhood and Brownfields Redevelopment 2008; Speiss 2008). Studying the expe-
riences over the years with VCPs, the EPA and others have come to view this identi-
fication and engagement with community groups, and the resulting degree of trust
and consensus, as important to the success of brownfields reuse projects (National
Association of Local Government Professionals and Northeast-Midwest Institute 2004;
Speiss 2008; US Conference of Mayors 2010; US Environmental Protection Agency
2011b).
For this reason, and because the remediation and reuse of brownfields sites often has
dramatic social and economic impacts on the surrounding area, community planners
have viewed public outreach as an increasingly important feature of the process (Speiss
2008). A community group can bring a wider focus on community redevelopment than
is possible at a site where remediation and reuse is governed solely by the developer
(Sarni 2009). However, it should be emphasized that VCPs do not typically require this
sort of public participation in the brownfields remediation and reuse process (Eisen
1996). Thus, the danger when public meetings or other means of involvement are con-
ducted is that the level of participation can fall far short of meaningful input (Speiss
2008; Sarni 2009) or that the participation process can fail to identify a group that speaks
meaningfully for the affected community (Davis 2011).
6. The Link Between Brownfields

Redevelopment and Climate Change,
“Smart Growth,” and Sustainability
Brownfields redevelopment activities have gone beyond mere remediation to pursue

a broader agenda much more closely related to urban sustainability (De Sousa 2008;
Lewis 2008; Sarni 2009; Hollander et al. 2010). Given the link between brownfields
redevelopment and reuse to preservation of greenfields spaces, there have been active
links and synergies between the “smart growth” movement and brownfields redevelop-
ment, particularly in addressing urban sprawl (Hollander et al. 2010; US Environmental
Protection Agency 2011b). “Smart growth” is an umbrella-like term referring to strategies
662 Joel B. Eisen
to pursue development that address negative externalities of unchecked suburban and

exurban sprawl by preserving unspoiled land and protecting natural resources (Pollard
2000; Salkin 2002). It is one thing to articulate that brownfields redevelopment is con-
sistent with smart growth principles, but quite another to achieve it in practice. The fact
that brownfields projects use existing urban land and, at times, some of the existing
infrastructure, does not necessarily make that growth “smart” (Eisen 2007; Paull 2008).
An evolving long-term trend in brownfields redevelopment is connecting brown-
fields with actions to address climate change (Paull 2008). There are many opportuni-
ties for reuse of brownfield sites that would not only conserve greenfields acreage but
also deploy green technologies to reduce carbon emissions and produce other benefits
(Lewis 2008; Sarni 2009). Employing green design and construction techniques in con-
junction with overhauling existing buildings may conserve energy and feature sustain-
able building materials and creative waste reduction strategies. Some brownfields sites
have been transformed into urban greenways or “brightfields” (sites for solar arrays
to produce electricity) (Lewis 2008; Sarni 2009). To move forward on climate change
action, cities and localities are increasingly forming sustainability or environmental
quality departments that integrate energy reduction, deployment of clean energy, and
other steps to reduce carbon footprints with existing land use planning. These depart-
ments, and their climate action plans and initiatives, have become important to integrat-
ing brownfields redevelopment planning into community-wide development agendas
(Portland Bureau of Planning and Sustainability 2009).
Of course, there may be political, legal, or other obstacles to pursuing this broader
approach to urban redevelopment. In some cases, the balkanization of urban pol-
icy making is a large part of what led to the city’s decline in the first place, and simply
forming a sustainability department may not be enough to override institutional iner-
tia (Sarni 2009). Often, multiple institutions in a city have effective responsibility for
urban planning, so it should be no surprise when these institutions cannot act nimbly
to decide what to do with the brownfield sites they might choose to redevelop or when a
developer cannot ascertain which entity it must deal with (Davis 2011).
7. Conclusion: Successes and Challenges

of Brownfields Redevelopment Programs
Decades after their inception, brownfields redevelopment programs are mature envi-
ronmental programs with many successes (US Environmental Protection Agency
2011b). In numerous American cities, sites that were abandoned for many years have
been reclaimed for productive reuses. From their inception, VCPs grew quickly, and a
number of success stories have been touted throughout the nation (Davis 2011). One
well-known positive impact on the environment of brownfields remediation and reuse
is the conservation of land at greenfield locations saved from development. Building on
urban sites, with their existing footprints and infrastructure, can alleviate the damaging
impacts of sprawl by conserving acres of greenfields (Deason et al. 2001; Paull 2008).
Other benefits of brownfields redevelopment include reductions in vehicle miles trav-
eled, storm water runoff, and air pollution (National Association of Local Government
Professionals and Northeast-Midwest Institute 2004; Wernstedt et al. 2004; Paull 2008).
Studies have also found that brownfields redevelopment can have a positive impact on
property values in the areas surrounding brownfields sites at sites where the change in
land use has yielded increased values (Paull 2008). Researchers using hedonic analysis
have found that the stigma of a contaminated site depresses nearby property values and
that removal of the stigma through expedited remediation would increase sites’ present
values (Kaufman and Cloutier 2006; Messer et al. 2006).
An ongoing concern is that, although VCPs have been successful, there are still
numerous cities where brownfields continue to remain abandoned or underused.
Many claim that far more brownfield sites remain to be addressed than have proceeded
through the state VCP processes to completion (Paull 2008; Davis 2011). In many com-
munities, the incentives for brownfields remediation and reuse have not succeeded
in attracting investments because fear of environmental liability is not the only bar-
rier to successful redevelopment (Robertson 1999; Auld et al. 2011). Some cities have
brownfields that are too small to be suitable for the sort of projects that interest many
developers (Wernstedt et al. 2010). These cities often also lack institutional capabili-
ties for redevelopment activities (Auld et al. 2011). For example, they have less expe-
rience in matching prospective developers with opportunities to remediate and reuse
brownfields sites because they have less skill at marketing their communities and iden-
tifying local officials who are sufficiently skilled to navigate VCP processes. Some pro-
spective brownfields developers (such as nonprofit groups) lack experience in dealing
with environmental agencies and can be overwhelmed by the VCP process. Making
assistance available to these developers by, for example, assisting them with site evalu-
ation methods, can be important to the success of their redevelopment activities (US
Environmental Protection Agency 2005; New Jersey Institute of Technology 2011).
Another issue that has arisen at times, and is likely to recur, is whether remediation and
reuse activities at individual brownfields sites have achieved final control of environmen-
tal contamination. The efficacy of some cleanups has come under fire years afterward as
contamination has been discovered (Eisen 2007). Without aggressive provisions for revis-
iting sites that prove to be problematic in the future, there is no guarantee that the finality
craved by brownfields developers will not come at a high societal cost. Cleanups in VCPs,
with their risk-driven calculation of environmental harms, do not achieve complete erad-
ication of all risks at the site. The typical means for these state programs to protect against
backsliding is inclusion of reopeners in the laws creating them. These provisions allow
state environmental departments to require developers or their successors to pursue addi-
tional cleanup activities at a later date if changed conditions warrant it (Eisen 2007).
Reopeners in VCPs have several drawbacks as protection against future discoveries
of environmental contamination. First, and perhaps most obvious, the changed con-
ditions, that is, the release of contaminants posing a threat to human health and the
664 Joel B. Eisen
environment, would already have taken place by the time a state’s environmental agency
became involved in attempting to prevent them. Second, the reopeners are typically
limited in scope because VCPs are designed to alleviate the regulatory burden faced by
developers, not to ensure full and complete remediation of brownfields sites. A broad
reopener provision allowing a state environmental agency to intervene at a brownfields
site without a demonstration of imminent harm is typically perceived as a deterrent to
initial finality of processing sites through a VCP.
Perhaps most importantly, any reopener provision depends both on resources avail-
able in the future to state environmental agencies and on the willingness of those agen-
cies to tackle problems at sites they believed were successfully addressed in the past. The
problem of resources to devote to enforcement is especially problematic in tight budget-
ary climates because cutbacks in state budgets can lead to a slower pace of cleanups and
less vigilant oversight of brownfields sites (National Association of Local Government
Professionals and Northeast-Midwest Institute 2004; Eisen 2007; US Environmental
Protection Agency 2011b). Few state environmental agencies put oversight and moni-
toring of completed brownfields sites ahead of normal enforcement actions in their
priorities. However, they would do well to devote more resources to oversight because
some sites have been processed through state VCPs without any systematic examination
of long-term impacts (Eisen 2007).
This highlights another shortcoming of VCPs: they generally lack structures for eval-
uation of sites over time to assess lifecycle impacts of brownfields remediation and reuse
(Wernstedt et al. 2004; Auld et al. 2011). The VCP process focuses on the present-day
problem of transforming an abandoned or underused site into a locus for commerce,
and, as a result, methodologies to evaluate long-term impacts are only beginning to be
developed, years after VCPs have been operating (Wernstedt et al. 2004; Auld et al. 2011).
Deciding on modalities for assessing whether redevelopment activities at brownfields
sites are beneficial can take the form of determining consistency with third-party veri-
fication systems such as the LEED system, the popular green building certification sys-
tem that promotes sustainable building and development practices (Paull 2008). LEED’s
methodology takes positive note of buildings that are constructed on brownfields sites
and also values features that are often positive advantages of existing brownfields sites,
such as proximity to existing transportation systems (Paull 2008; Sarni 2009). However,
a different approach will be necessary to determine whether brownfields policies meet
the criteria outlined in climate action plans or, indeed, have been beneficial as a whole as
a strategy for urban redevelopment (Auld et al. 2011).
Therefore, although VCPs have undeniably been responsible for successes and ben-
efits, the final verdict on brownfields programs has yet to be rendered and will only be
made in hindsight, years after the initial development decisions have been made. This is
consistent with the societal perspective on other urban redevelopment programs, which
can change as programs have impacts not foreseen by their initial drafters. Given the
relative lack of in-depth research from economists on the impacts of brownfields pro-
grams, there is considerable room for targeted work that evaluates the value that these
programs add at individual sites and the merits of specific redevelopment strategies.
References
Auld, Ronell, et al. 2011. Assessing brownfield sustainability: Life cycle analysis and carbon
footprinting. http://www.cmu.edu/steinbrenner/brownfields/Current%20Projects/sus-
tainability.html
Bartsch, C. 1996. Coming clean for economic development: A resource book on environmental
cleanup and economic development opportunities. http://www.nemw.org/cmclean.htm
Bartsch, C., and E. Collaton. 1997. Brownfields: Cleaning and reusing contaminated properties.
Santa Barbara, CA: Praeger.
Boyd, J., W. Harrington, and M. K. Macauley. 1996. The effects of environmental liability on
industrial real estate development. Journal of Real Estate Finance and Economics 12(1): 37–58.
Bradbury, K. L., A. Downs, and K. A. Small. 1982. Urban decline and the future of American cit-
ies. Washington, DC: Brookings Institution Press.
Dana, D. A. 2005. State brownfields programs as laboratories of democracy? New York University
Environmental Law Journal 14: 86–107.
Davis, T. S. (ed.). 2011. Brownfields: A comprehensive guide to redeveloping contaminated prop-
erty, 3rd ed. Chicago: American Bar Association.
Deason, J. P., G. W. Sherk, and G. A. Carroll. 2001. Public policies and private decisions affecting
the redevelopment of brownfields: An analysis of critical factors, relative weights and areal
differentials. http://www.gwu.edu/~eem/Brownfields/
De Sousa, C. 2008. Brownfields redevelopment and the quest for sustainability. Bingley,
UK: Emerald Group Publishing.
Duany, A., E. Plater-Zyberk, and J. Speck. 2001. Suburban nation: The rise of sprawl and the
decline of the American dream. New York: North Point Press (Macmillan).
Eisen, J. B. 1996. “Brownfields of dreams?” Challenges and limits of voluntary cleanup pro-
grams and incentives. University of Illinois Law Review 1996: 883–1032.
Eisen, J. B. 1999. Brownfields policies for sustainable cities. Duke Environmental Law & Policy
Forum 9: 187–229.
Eisen, J. B. 2007. Brownfields at 20: A critical reevaluation. Fordham Urban Law Journal
34: 721–756.
Geltman, E. Glass. 2000. Recycling land: Understanding the legal landscape of brownfield devel-
opment. Ann Arbor: University of Michigan Press.
Gerrard, M. B. (ed.). 1998 & Suppl. 2006. Brownfields law & practice: The cleanup & redevelop-
ment of contaminated land. New York: Matthew Bender.
Hollander, J. B. 2009. Polluted & dangerous: America’s worst abandoned properties and what can
be done about them. Burlington: University of Vermont Press.
Hollander, J., N. Kirkwood, and J. Gold. 2010. Principles of brownfield regeneration: Cleanup,
design, and reuse of derelict land. Washington, DC: Island Press.
Kaufman, D. A., and N. Cloutier. 2006. The impact of small brownfields and greenspaces on
residential property values. Journal of Real Estate Finance and Economics 33(1): 19–30.
Kromer, J. 2010. Fixing broken cities: The implementation of urban development strategies.
New York: Routledge Press.
Kunstler, J. H. 1993. The geography of nowhere: The rise and decline of America’s man-made land-
scape. New York: Touchstone Press.
Lewis, G. 2008. Brown to green: Sustainable redevelopment of America’s brownfield sites.
http://www.nemw.org/index.php/policy-areas/brownfields/brownfields-sustainable-ur
ban-redevelopment-and-energyreports-papers
666 Joel B. Eisen
Messer, K. D., W. D. Schulze, K. F. Hackett, T. A. Cameron, and G. H. McClelland. 2006. Can

stigma explain large property value losses? The psychology and economics of Superfund.
Environmental and Resource Economics 33(3): 299–324.
National Association of Local Government Professionals & Northeast-Midwest Institute.
2004. Unlocking brownfields: Keys to community revitalization. http://www.nalgep.org/
publications/
National Center For Neighborhood & Brownfields Redevelopment. 2008. Building capac-
ity: Brownfields redevelopment for community-based organizations. http://policy.rutgers.
edu/brownfields/projects/Manual_Building_Capacity.pdf
New Jersey Institute of Technology. 2011. Brownfield site contamination investigation. http://
www.njit.edu/tab/managing/pre-development/contamination-investigation.php
Pahlen, G. 2004. RESCUE: Regeneration of European sites in cities and urban environments.
http://www.grc.engineering.cf.ac.uk/events/rescue2/pdfs/R2-MS2-GP.pdf
Paull, E. 2008. The environmental and economic impacts of brownfields redevelopment. http://
www.nemw.org/images/stories/documents/EnvironEconImpactsBFRedev.pdf
Pollard, O. A., III. 2000. Smart growth: The promise, politics, and potential pitfalls of emerging
growth management strategies. Virginia Environmental Law Journal 19: 247–285.
Portland Bureau of Planning and Sustainability. 2009. Climate action plan 2009. http://www.
portlandonline.com/bps/index.cfm?c=49989
Robertson, H. G. 1999. One piece of the puzzle: Why state brownfields programs can’t lure
businesses to the urban cores without finding the missing pieces. Rutgers Law Review
51(5): 1075–1132.
Rosenberg, R. 2000. Community resource guide for brownfields redevelopment. Williamsburg,
VA: Center for Public Policy Research, College of William and Mary.
Salkin, P. E. 2002. Smart growth and sustainable development: Threads of a national land use
policy. Valparaiso University Law Review 36: 381–412.
Sarni, W. 2009. Greening brownfields: Remediation through sustainable development.
New York: McGraw-Hill Professional.
Schnapf, L. 2007. The new “all appropriate inquiries” rule. The Practical Real Estate Lawyer
January: 7–24.
Small Business Liability Relief and Brownfields Revitalization Act of 2002 SBLRBRA, 42 USC.
§§ 9604-05, 9607, 9622, 9628.
Speiss, D. M. 2008. Public participation in brownfields cleanup and redevelopment: The role of
community organizations. http://deepblue.lib.umich.edu/handle/2027.42/60850
US Conference of Mayors. 2010. Recycling America’s land: A national report on brown-
fields redevelopment (1993–2010). http://www.usmayors.org/pressreleases/uploads/
November2010BFreport.pdf
U.S. Environmental Protection Agency. 2005. Investing in partnership, possibility, and
people: A report to stakeholders (Moving Forward). www.epa.gov/brownfields/
overview/05Stakeholder/StakeholderReport_MovingForward.pdf
U.S. Environmental Protection Agency. 2005. Standards and practices for all appropriate inqui-
ries, 40 C.F.R. § 312. http://www.epa.gov/brownfields/aai/index.htm
U.S. Environmental Protection Agency. 2011a. A guide to federal tax incentives for brownfields
redevelopment. www.epa.gov/brownfields/tax/tax_guide.pdf
U.S. Environmental Protection Agency. 2011b. Brownfields and land revitalization. http://epa.
gov/brownfields/about.htm
U.S. Environmental Protection Agency. 2011c. State brownfields and voluntary response pro
grams: An update from the states. http://www.epa.gov/brownfields/state_tribal/update2011/
bf_states_report_2011.pdf
van Hook, D. E., J. A. Shaw, and K. J. Kloo. 2003. The challenge of brownfield clus-
ters: Implementing a multi-site approach for brownfield remediation and reuse. New York
University Environmental Law Journal 12: 111–152.
Wernstedt, K., L. Heberle, A. Alberini, and P. Meyer. 2004. The brownfields phenomenon: Much
ado about something or the timing of the shrewd? Resources for the Future, Discussion
Paper 04-46. http://www.rff.org/Documents/RFF-DP-04-46.pdf
Wernstedt, K., A. Blackman, T. P. Lyon, and K. Novak. 2010. Voluntary environmental pro-
grams at contaminated properties. Resources for the Future, Discussion Paper 10-18. http://
www.rff.org/RFF/Documents/RFF-DP-10-18.pdf
C HA P T E R 2 5
R E G U L AT O RY TA K I N G S
T HOM AS J. M IC E L I A N D KAT H L E E N SE G E R S ON
The regulatory takings issue potentially arises whenever a government regulation

restricts the use of private property without actually seizing title to it. Examples include
zoning, environmental and safety regulations, historic landmark designation, rules
requiring equal accommodation for the disabled, and so on.1 From an economic per-
spective, regulations that reduce the value of private property are not fundamentally
different from outright takings (i.e., seizures of property); the difference is one of degree
rather than of kind. Thus, any deprivation or restriction of a particular right reduces the
value of the property proportionately. A physical taking, which deprives the owner of all
rights, is simply one end of a continuum.
Given this analytical equivalence between seizures and regulations, a separate treatment
of the two types of actions does not appear to be warranted by economic theory. Yet the
fact remains that courts have treated them quite differently: whereas compensation is vir-
tually always required for seizures, it is rarely awarded for regulations. Indeed, courts have
historically granted the government broad police powers to enact regulations in the public
interest without the need to compensate property owners for lost value. Still, in some cases,
courts have ruled that if a regulation goes “too far” in restricting private property, it will be
ruled a “regulatory taking” and compensation will be due. The question, therefore, is where
the dividing line is (or should be) between compensable and noncompensable regulations.
There exists a considerable body of case law and legal scholarship aimed at answering this
“compensation question.” A review of the various tests that have emerged from this inves-
tigation illustrates the range of perspectives that have been brought to bear on this debate
and also reflects the apparent lack of consensus on an adequate answer. The goal of this
chapter is to survey the contribution economic theory has made to this debate, highlighting
1
See Miceli and Segerson (1996), Meltz et al. (1999), and Miceli (2011, Chapter 5) for detailed
examinations of several of these regulations within the context of takings law. Note that, although the
existing case law describes the range of government actions that have so far been challenged under
takings law, it does not limit those things that can be challenged in the future. In principle, landowners
can challenge any action that reduces the value of their land as constituting a taking.
Regulatory Takings 669
the most significant insights and results. We do this using a simple model of takings or land
use regulation that provides a unifying framework for discussing the economics of regula-
tory takings. We use the model to illustrate both basic economic principles related to com-
pensation and a number of extensions that have been considered in the literature.
Much of this literature is normative in the sense that it proposes compensation rules
aimed at achieving efficient regulatory and land use decisions and hence does not pur-
port to provide a positive theory of the case law (which is viewed by many as incoherent)
or to account for competing values like distributive justice. We will discuss one rule,
however, that we believe goes a long way toward unifying the various tests that courts
and legal scholars have proposed.
As a prelude to the economic analysis, we provide brief reviews of the case law and
legal literature in this area.
1. Overview of the Case Law
1.1 The Physical Invasion Test

Nearly all courts have agreed that any government action that involves some sort of
physical invasion of a landowner’s property, even when it does not literally seize title,
constitutes a compensable taking. For example, in Loretto v. Teleprompter,2 the Court
held that a state law allowing cable television providers to install wires and other
equipment on a private building was a taking. Although the physical invasion test is
well-established in takings law, it is of limited usefulness because it offers no guidance
for the vast majority of government actions, like zoning and environmental regulations,
that involve no invasion. The remaining legal tests concern these sorts of cases.
1.2 The Noxious Use Doctrine

An important early test was established in the case of Mugler v. Kansas,3 which con-
cerned a law passed by the state of Kansas, pursuant to a prohibition amendment to
the Kansas constitution, forbidding the operation of breweries. The owner of a brewery
sued for compensation on the grounds that the law constituted a taking of his property,
but the US Supreme Court denied the claim based on the state’s right to regulate, with-
out compensation, those activities that are deemed “to be injurious to the health, mor-
als, or safety of the community,”4 so-called “noxious uses.” This ruling, referred to as the
2 458 U.S. 419 (1982).

3 123 U.S. 623 (1887).
4 Ibid., 668.
670 Thomas J. Miceli and Kathleen Segerson
noxious use doctrine, recognized that the government has broad regulatory powers to
prevent land uses seen as potentially harmful to the public.
Zoning ordinances provide the most common illustration of this principle, and courts
have routinely upheld them as valid exercises of the government’s right to regulate land
use in the public interest. The first case to reach such a conclusion was Village of Euclid
v. Ambler Realty,5 which upheld a town ordinance zoning a portion of the plaintiff ’s land
for residential use. The Court maintained that the ordinance fell within the municipal-
ity’s inherent right, under the police power, to protect public health, safety, morals, and
general welfare (Meltz et al. 1999, 214).
1.3 The Diminution of Value Test

Under the noxious use doctrine, the impact of a regulation on the landowner’s value
apparently had no bearing on the compensation question. As long as the landowner
retained title, the government had broad regulatory authority to restrict his or her land.
That view changed, however, when, in 1922, the Supreme Court decided the famous
case of Pennsylvania Coal v. Mahon.6 The case concerned a law passed by the State of
Pennsylvania aimed at protecting the safety of surface owners against the risk of cave-ins
(or subsidence) by requiring that coal companies leave enough coal in the ground to
support the surface. The Pennsylvania Coal Company brought suit seeking compen-
sation on the grounds that the regulation was a taking of its legal right to mine all of
the coal under the surface. (Under a common legal arrangement, the mining company
had sold the surface rights but had retained the mineral rights to the subsurface coal.)
Although the case seemed to be an easy one under the noxious use doctrine, given that
the law clearly met the standard of protecting the safety of the surface owners, the Court
ruled that compensation was due.
Writing for the majority, Justice Oliver Wendell Holmes argued that, apart from the nox-
ious use doctrine, there must be a limit to the government’s power to regulate private prop-
erty. That limit, he said, is embodied in the impact of the regulation on the landowner:
One fact for consideration in determining such limits is the extent of the diminu-
tion [in the landowner’s value]. When it reaches a certain magnitude, in most if not
all cases there must be an exercise of eminent domain and compensation to sustain
the act.7
This argument forms the basis for the diminution of value test for compensation, which
says that compensation is due if the loss to the landowner as a result of a regulation
5 272 U.S. 365 (1926)

6 260 U.S. 393 (1922). Also see Friedman (1986), who characterized the Penn Coal decision as a
“watershed” in takings law.
7 Id., 413.
is sufficiently large. Of course, this raises the question of what amount of loss is large
enough to meet the compensation threshold; Holmes only said that “if regulation goes
too far it will be recognized as a taking.”8 He therefore left it to future courts to decide on
a case-by-case basis what constitutes “too far.”
More than six decades later, the Supreme Court confronted a case with an almost
identical factual scenario as in Pennsylvania Coal. The issue again was whether an
antisubsidence statute passed by the state legislature was a taking of the coal compa-
ny’s rights, but, in apparent contradiction of its earlier ruling, the Court in Keystone
Bituminous Coal Assn. v. DeBenedictus (1987) ruled that it was not. In endeavoring to
distinguish the two cases, the Court argued that the statute at issue in Keystone protected
a broader public interest, whereas the earlier statute had been aimed at protecting only a
few private parties. Although the distinction between public and private in this context
is not strictly valid from an economic perspective, we will argue in Section 4.2.2 (see
footnote 30) that changing values of both the surface and mining rights over the inter-
vening time period can provide a legitimate basis for distinguishing the two cases.
1.4 The Penn Central Balancing Test

The need to balance the factors raised in these earlier cases—namely, the intent of the
regulation and its impact on the landowner’s property value—was made explicit in the
case of Penn Central Transportation Co. v. City of New York.9 This case arose out of the
city’s decision to designate Grand Central Terminal as a historic landmark, thereby lim-
iting the sort of alterations that the owners could make.10 Thus, when the Landmark
Preservation Commission turned down a proposal by Penn Central to build a multi-
story office building above the terminal, the owners sued, claiming a taking of their right
to develop. In deciding against compensation, the Supreme Court advanced a three-part
test for determining whether a compensable taking has occurred. The relevant factors
were (1) the character of the government action, (2) whether or not the regulation inter-
fered with “investment-backed expectations,” and (3) the extent of the diminution of
value. The first and third of these factors clearly identified the importance of both the
noxious use doctrine and the diminution of value test, although, once again, without
offering explicit guidance on how to balance one against the other.
The second factor, emphasizing the importance of investment-backed expectations,
captures the idea that any loss suffered by the landowner must have been based on rea-
sonable expectations, backed up by actual investments (Mandelker 1987; Fischel 1995,
50). In other words, an owner could not claim to have been denied uses that he never
would have contemplated or that would not have been allowed by law in the absence of
8 Id., 415.
9 438 U.S. 104 (1978).
10 On landmark designation and takings law, see Gold (1976).
the regulation. Thus, as a necessary condition for compensation, a claimant would have
to show evidence that he had in fact planned to undertake the prohibited development.
In his dissenting opinion to the Penn Central ruling, Justice William Rehnquist added a
further consideration when he stated that “a taking does not take place if the prohibition
applies across a broad cross section of land and thereby ‘secure[s]‌an average reciprocity
of advantage.’ ”11 The phrase “average reciprocity of advantage,” first used by Holmes in his
Pennsylvania Coal opinion, suggests that monetary compensation need not be paid if a reg-
ulation restricts all landowners equally, thereby spreading both the benefits and the costs
of the regulation. The Supreme Court employed similar logic in the case of Agins v. Tiburon
when it held that a landowner subject to a zoning restriction “will share with other owners
the benefits and burdens of the city’s exercise of its police power. In assessing the fairness of
zoning ordinances, these benefits must be considered along with any diminution in mar-
ket value that the appellants might suffer.”12 We will return to this logic in Section 5.
1.5 The Nuisance Exception

An important extension of takings law emerged from the case of Lucas v. South Carolina
Coastal Council (1992).13 The case involved a land developer who had purchased two
beachfront lots in South Carolina with the intention of developing them for residential
use. Such a use seemed reasonable at the time of purchase since several similarly situated
neighboring lots had already been developed. However, after the developer’s purchase
but before he began development, the South Carolina legislature passed a law prohibit-
ing further beachfront development in the area in an effort to control coastal erosion.
The developer sued, claiming a taking because the lots were rendered valueless by the
regulation. A trial court found in his favor and awarded full compensation. However,
the South Carolina Supreme Court reversed the ruling, despite the trial court’s finding
of a nearly complete diminution of value, relying instead on the regulation’s stated pur-
pose of preventing harm to the public—the old noxious use doctrine.
The U.S. Supreme Court, in turn, reversed the South Carolina Supreme Court and
said that compensation was due based on the fact that the regulation denied “all ben-
eficial or productive use of land.”14 Still, the Court left open the possibility that the state
could avoid paying compensation, despite the total loss, if it could show that the land use
prevented by the regulation constituted a nuisance under the state’s common law. This
standard, known as the nuisance exception, provides an objective basis, founded in the
common law, for determining what constitutes a noxious use.
One question not clarified by the Lucas decision, however, was the extent of dimi-
nution necessary to trigger automatic compensation. The regulation at issue in Lucas
11
Id., 147.
12 157 Cal.Rptr. 373, 1979; affirmed 447 U.S. 255, 262 (1980).
13 505 U.S. 1003 (1992).
14 Id., 1015.
clearly met any standard the Court could have applied because it caused a virtual total
loss (the nuisance exception aside). The question therefore remained whether some-
thing short of full diminution would also qualify. The Supreme Court revisited this issue
in Palazzolo v. Rhode Island,15 which concerned a landowner who sought compensa-
tion when he was denied permission to develop waterfront property under a wetlands
preservation law passed by the state of Rhode Island. The landowner claimed that the
regulation met the requirement for compensation under Lucas because it denied him
“all economically beneficial use” of the land, but the Court found that the regulation, in
fact, left the owner with developable land worth $200,000, compared to his claimed loss
of $3.15 million (a 94% diminution). Apparently, therefore, a diminution of at least 95%
is required to constitute a “total” deprivation under Lucas.
The extent of the diminution arose in a different way in the case of Tahoe-Sierra
Preservation Council v. Tahoe Regional Planning Agency,16 which concerned a temporary
moratorium on development that deprived a group of developers of the entire value of
their holdings for a 32-month period. The plaintiffs claimed that this action constituted
a taking under the Lucas rule, requiring compensation for regulations that deprive own-
ers of all productive uses of their land, albeit for a limited period of time. The Supreme
Court disagreed, however, arguing that the diminution was only partial in relation to
the “parcel as a whole,” accounting not only for its spatial dimension but for its tempo-
ral dimension as well. Recognizing that the value of the parcel would be restored once
the moratorium was lifted, the Court argued that the Lucas rule did not apply, and that,
under the Penn Central balancing test, no compensation was due.
1.6 Wetlands and Endangered Species Protection

Much recent litigation in the area of regulatory takings law has arisen in the context of
regulations aimed at protecting wetlands and endangered species. Wetlands represent
a natural resource that has only recently been recognized as providing important social
benefits, including providing a habitat for wildlife, flood control, water quality mainte-
nance, and both recreational and commercial use (Hartmann and Goldstein 1994). The
recent recognition of these values has led to the enactment of government regulations at
both the state and federal level aimed at preserving wetlands. However, since wetlands
predominantly exist on privately owned land, and because their primary value to the
owner is usually for future development, these regulations have naturally generated a
large volume of takings claims.17
Efficiency dictates that conversion of wetlands to development should occur to the
point at which the marginal social value of land in development equals its marginal
15
533 U.S. 606 (2001).
16 535 U.S. 302 (2002).
17 According to Meltz et al. (1999, 366, note 5), there were about 400 cases involving wetlands
regulations between 1960 and 1990, of which about half raised the takings issue.
social value if left in an undeveloped state. Thus, although it is probably efficient to

convert some wetlands to alternative uses, especially in early stages of economic devel-
opment, private landowners almost certainly would go beyond that efficient point if
unrestrained because they would not internalize the full social value of the resource.
Regulation is therefore necessary to achieve the efficient balance. Still, the question
remains whether landowners are entitled to compensation for their resulting loss.
Generally, courts have held that the denial of a permit to develop a wetland does not
constitute a compensable taking of the owner’s property. One argument in support
of this position has been to claim that the proposed use of the land would represent a
nuisance. However, in Florida Rock Industries v. United States, the U.S. Claims Court
held that “the assertion that a proposed activity would be a nuisance merely because
Congress chose to restrict, regulate, or prohibit it for the public benefit indicates circular
reasoning that would yield the destruction of the fifth amendment.”18 As a result, the
Court found that a taking had occurred. Although this argument is consistent with the
Lucas nuisance exception in its reliance on the common law for defining a nuisance,
it fails to recognize the possible efficiency benefits that a departure from nuisance law
might allow. Although nuisance law will often provide a useful approximation for effi-
cient regulation, this will not always be the case. Indeed, the fact that wetlands were once
themselves thought to be nuisances worthy of removal, but are now highly valued as an
important natural resource by many, illustrates the point (Meltz et al. 1999, 365).
Nuisance law is not the only applicable legal doctrine in wetlands cases; the “public
trust doctrine” is also relevant. The public trust is an ancient doctrine that grants own-
ership of navigable waterways, shorelines, and the open sea to the public. According
to this doctrine, landowners do not have a right to impair these resources (Lueck and
Miceli 2007, 237). In Just v. Marinette County, for example, the court held that “[a]‌n
owner of land has no absolute and unlimited right to change the essential character of
his land so as to use it for a purpose for which it was unsuited in its natural state and
which injures the rights of others.”19 Based on this logic, the court found that a regula-
tion preventing the landowner from filling a wetland was not a taking, even though the
proposed use would not have constituted a nuisance.
Like resource preservation, the protection of endangered species, especially those
endangered by human activity, has become an important objective of government pol-
icy (Boyle and Bishop 1987). The most important legislative action in this regard was
the passage of the Endangered Species Act (ESA) in 1973.20 Under this Act, the Fish and
Wildlife Service (FWS) was authorized to “list” a species as endangered or threatened
and to designate the “critical habitat” of that species for special protection or manage-
ment. The Act further stipulated that the criterion for listing a species is to be based on
“the best scientific and commercial data, without reference to economic costs or private
18 21 Cl.Ct. 161, 168 (1990).

19 201 N.W.2d 761, 768 (Wisc. 1972).
20 16 U.S.C. §§ 1531–1544.
property impacts.” In contrast, habitat designation is to be based on both scientific data

and “economic impact and any other relevant impact,” thus theoretically allowing con-
sideration of landowners’ interests (Meltz et al. 1999, 392).
As with wetlands, the preservation of endangered species warrants government inter-
vention because of the externalities involved (Harrington 1981). However, the takings
issue also arises because of the loss suffered by landowners as a result of the various
restrictions on their allowed activities. For example, owners are prohibited from “tak-
ing” members of a listed species unless it is done in a good faith attempt to protect a
person, where a “take” is defined to include, among other things, harassing, harming,
pursuing, or hunting a listed animal. More ominously for landowners, the FWS has
defined “harm” to include significant habitat modification or degradation (Meltz et al.
1999, 393), and, in 1995, the Supreme Court upheld that interpretation as reasonable.21
Conversely, land use restrictions under the ESA have generally produced relatively
modest impacts on landowners’ value, which, based on the prevailing legal standard
requiring a landowner to show a virtual total loss in value, does not bode well for the
success of taking claims.22 Of course there are exceptions to this, such as when the owner
of a stand of timber is prevented from harvesting it. The risk here, as will be discussed
later, is that the threat of an uncompensated regulation can result in perverse (and
costly) landowner incentives, such as a decision to clear cut the stand early to avoid its
being declared a habitat or to conceal the fact that an endangered species might reside in
a certain locale.
2. Other Proposed Tests for

Compensation
This section describes several tests for compensation that have been proposed in the
scholarly literature on takings. As will be seen, these tests vary in their economic content
and logical consistency. A test first proposed by Sax (1964) asserts that the government
owes compensation when it acquires property rights for use in its enterprise capac-
ity, as when it provides a public good, but it does not owe compensation when it acts
as a disinterested arbitrator in a private dispute, as when it prevents one private party
from imposing external costs on other private parties.23 Rubenfeld (1993) elaborates on
this test by arguing that a taking occurs when the government takes property for some
21
Sweet Home Chapter of Communities for a Great Oregon v. Babbitt, 515 U.S. 687 (1995).
22 Meltz et al. (1999, 396) note that, as of 1999, not a single court decision finding a taking under the
ESA had been reported.
23 Rose (1983) makes a similar argument.
productive use—a so-called “using”—as opposed to merely depriving the owner of its
use. Rose-Ackerman and Rossi (2000) propose a similar standard.
In a second article, Sax (1971) argues that the government does not owe compensa-
tion for any actions that it undertakes to regulate external costs. Daniel Bromley adopts
a similar perspective in arguing that paying compensation for such regulations would
represent “indemnification for an inability to continue to impose unwanted costs on
others” (Bromley 1993, 677). According to this view, which echoes the noxious use doc-
trine, the law does not (and should not) protect the right of landowners to engage in
activities that impose harm on others. However, the difficulty with this test, as noted by
Fischel (1985, 153), is that it offers “no workable distinction. . . between land uses that
create spillovers and those that do not. Every economic activity can be argued to affect
someone else” [emphasis in original].
A similar delineation of property rights underlies the harm-benefit rule, which says
that no compensation is due for regulations that prevent a landowner from imposing a
harm on others (e.g., a regulation against pollution), but compensation is due for regula-
tions that compel the landowner to confer a benefit on the public (e.g., a ban on devel-
opment to preserve open space). Although this rule has some intuitive appeal, it, too,
is unsupported by economic theory in the sense that a prevented harm can always be
defined as a benefit, and a forgone benefit can be defined as a harm (Fischel 1985, 158).
Based on this logic, Justice Scalia, in his Lucas opinion, dismissed the harm-benefit rule
as lacking a coherent legal basis for deciding the compensability of regulations.24
What underlies the failure of the harm-benefit rule, or any nuisance-based approach
to the compensation question, is Coase’s insight that all harms are reciprocal in nature,
meaning that both an injurer and victim must be present for an accident to occur (Coase
1960). Thus, regulations that prevent harms or confer benefits are indistinguishable in
terms of the cost to the injurer and the gain to the victim. What is lacking in tests is
a benchmark reflecting neutral conduct, which would serve as the basis for deciding
when compensation should and should not be paid. Fischel (1985, 158–160) offers such
a benchmark in the form of his normal behavior standard, which is based on arguments
first made by Ellickson (1973, 1977). According to this standard, no compensation is
due for regulations that prevent landowners from engaging in “subnormal” behavior,
but compensation is due for regulations that compel them to undertake “above-normal”
behavior, where “normal” behavior is defined by community standards based on what
landowners can reasonably expect to be able to do with their land. This “reasonableness
standard” therefore replaces the arbitrary distinction between harms and benefits in the
24
“When it is understood that ‘prevention of harmful use’ was merely our early formulation of the
police power justification necessary to sustain (without compensation) any regulatory diminution of
value; and that the distinction between regulation that ‘prevents harmful use’ and that which ‘confers
benefit’ is difficult, if not impossible, to discern an objective, value free basis; it becomes self evident
that noxious use language cannot serve as a touchstone to distinguish regulatory takings—which
require compensation—from regulatory deprivations that do not require compensation” (Lucas v. South
Carolina Coastal Council, 505 U.S. 1003, 1026, 1992).
harm-benefit rule. What makes this an economic standard (rather than being another
arbitrary distinction) is that it economizes on the transaction costs of achieving an effi-
cient land use pattern. Specifically, by setting the “zero compensation point” at normal
behavior, the costs of compliance will be minimized because most landowners will
engage in normal behavior automatically (i.e., without the need for government action).
Wittman (1984) proposes a similar compensation rule that is based on the behavior
of the government rather than of landowners. Specifically, he argues that the transac-
tion costs of paying compensation will be minimized if compensation is limited to cases
in which the government acts inefficiently, based on the presumption that “we would
expect the government to act efficiently more often than not” (Wittman 1984, 74). An
important drawback of both the Fischel and Wittman standards, however, is that they fail
to account for the role of the compensation rule in creating the proper incentives for land-
owners and/or the government to act efficiently. The next section deals with this issue.
In a very influential article, Michelman (1967) proposed a standard that is based on
a comparison of the settlement (or transaction) costs associated with paying compen-
sation and the demoralization costs of not paying compensation, where the latter are
defined to be those costs incurred by landowners and their sympathizers once they
realize that they will not be compensated for their losses. According to Michelman’s
standard, if the settlement costs are lower, compensation should be paid, whereas if
demoralization costs are lower, compensation should not be paid.25
Finally, Richard Epstein’s view on the compensation question is based on the Lockean
notion that the government should not stand in a preferred position compared to private
citizens (Epstein 1985, C
hapter 2). In this perspective, the government has no more rights
in its interactions with private citizens than does any other private citizen, inasmuch as
the government is merely an agent of those citizens when they act collectively. Thus,
when a government action wrongfully deprives a private citizen of valuable property, it
should have to pay compensation, just as a private citizen would have to pay for imposing
similar harm under nuisance (tort) law. In contrast, when a government action prevents
a private citizen from imposing harm on other citizens—as when it prevents that citizen
from creating a nuisance (or noxious use)—it should not have to pay, based on the right
that private citizens have to be free from nuisances caused by fellow citizens.26 Epstein’s
view thus closely corresponds to the nuisance exception established in the Lucas case.
3. Role of Economics in the Takings Debate
Before turning to the economic models, we provide some comments on the role that
economics can, and more importantly cannot, play in resolving issues related to the
25 See Fischel and Shapiro (1988), who discuss Michelman’s test in light of more recent theories.
26 See Epstein (1985, 36) and Epstein (1995, 133).
takings debate. Fundamentally, debates about the scope of the government’s power of
eminent domain concern the relationship between the state and its citizens. In par-
ticular, “What can the state demand of the individual citizens whom it represents and
governs?” (Epstein 1985, 3). As noted, under the Lockean conception of property, the
government derives its power from the consent of the governed and therefore cannot
infringe on individual property rights except insofar as that is necessary to prevent
property owners from interfering with one another’s rights. In contrast, a Benthamite
view of property sees the government’s role as defining property rights in such a way
as to achieve the greatest good for the greatest number. It follows that infringement on
the rights of some citizens is acceptable if it is part of an overall policy that results in an
increase in social welfare (however that is measured).
An economic approach to the takings issue is potentially compatible with both views,
depending on the concept of efficiency one employs. The conclusions one would reach
regarding the compensation question, however, would likely be quite different. Under
traditional Pareto optimality, government actions are only judged to be efficient if no
one is made worse off by the action and some are made strictly better off. This crite-
rion, which reflects the Lockean view regarding the protection of property rights, would
clearly require full compensation to be paid for all government actions, regardless of any
realized benefits. It would therefore be satisfied whenever the government regulation
yields aggregate gains that exceed the losses, and compensation is paid. In contrast, the
criterion of potential Pareto optimality (or Kaldor-Hicks efficiency) would not neces-
sarily require actual compensation of losers but only that compensation be “possible.”
This approach is clearly more congenial to the Benthamite view of property. It would
be satisfied whenever the government regulation yields aggregate gains that exceed the
losses, regardless of whether compensation is paid. Thus, either notion of efficiency is
consistent with an economic perspective. As a result, an economic approach does not
offer a clear resolution of the fundamental question about property rights that is a criti-
cal dimension of the takings issue. It does, however, allow a consideration of how com-
pensation decisions affect resource allocation and incentives for various parties, and
thus, ultimately, the magnitude of aggregate welfare. This perspective and its implica-
tions constitute the primary contribution of economic models of regulatory takings.
4. Economic Models of Land Use and

Regulation
Probably the most important contribution economists have made to the regulatory
takings debate has been the examination of the impact of compensation on the invest-
ment incentives of landowners whose property is at risk of being taken or regulated.
This line of research began with the seminal paper by Blume, Rubinfeld, and Shapiro
(1984) (hereafter, BRS), which showed that paying full compensation for takings creates
a moral hazard problem that causes landowners to overinvest in land that is targeted for
regulation. An implication of the BRS analysis is the so-called no-compensation result,
which demonstrated that zero compensation is efficient. The BRS model is actually
more subtle than this conclusion suggests, but the no-compensation result has naturally
received the most attention and has provided a stimulus for subsequent research, much
of it aimed at providing countervailing arguments.
The no-compensation result was controversial principally because of its perceived
unfairness and apparent inconsistency with the constitutional requirement of just com-
pensation (at least for physical takings). From an economic perspective, however, the
result is a simple consequence of the well-known moral hazard problem associated with
full insurance. This section presents a simple version of the BRS model in which the gov-
ernment’s decision to take or regulate the owner’s property is treated as exogenous. The
model provides a unifying framework for discussing the large literature on the econom-
ics of takings that has arisen since BRS. Consistent with this literature, the basic prin-
ciples derived from it are equally valid in the contexts of outright takings and regulation
of property. Subsequent sections then examine various extensions to the basic model.
4.1 Exogenous Probability of a Taking

The BRS model uses the following notation:
V(x) = market value of a piece of land after x dollars of improvements have been
made, where V′>0, and V″<0;
p = probability that the land will be taken for public use;
B = fixed value of the land in public use if taken;
C(x) = compensation paid to the owner in the event of a taking.
The timing of events is as follows: first the landowner decides how much to invest in
improving his land and then the taking/regulation decision occurs. The owner’s initial
investment is irreversible, so if the land is taken, its value in private use, V(x), as well as
the cost of the investment, x, are lost. Since the original BRS paper considered physical
takings, the interpretations of the variables above reflect this. However, as just noted,
the model and the results derived from it are equally valid in the context of regulatory
takings using the following interpretations of the variables. Under a regulatory taking,
V(x) represents the additional market value that the landowner would realize in the
absence of a regulation, or, equivalently, the loss in market value that would result from
the regulation; B represents the benefit of the regulation or, equivalently, the external
harm avoided by it; and p is the probability that the regulation is imposed.
Since the taking occurs randomly in this model, the only economic decision is the
owner’s choice of x. The socially optimal investment maximizes the expected social
value of the land:
pB + (1 − p)V (x ) − x. (1)
The resulting first-order condition is
(1 − p)V ′(x ) = 1, (2)
which defines the optimal investment, x*. Note that the amount of investment is
decreasing in p. Thus, as the probability of a taking increases, the landowner should
invest less, so as to reduce the loss in the event of a taking.
Now consider the actual choice of x by the landowner. His goal is to maximize his
expected private return from the land, which is given by
pC(x ) + (1 − p)V (x ) − x. (3)
Note that this expression differs from (1) by the first term. The first-order condition
defining the owner’s optimal investment is
pC ′(x ) + (1 − p)V ′(x ) = 1. (4)
Comparing (4) and (2) immediately shows that C′ = 0 is a sufficient condition for effi-
ciency; that is, lump sum compensation induces efficient investment. A special case of
lump sum compensation is C ≡ 0, or zero compensation. Intuitively, zero compensation
prevents the owner from overinvesting in his land because he internalizes the loss that
would result if the land is taken. This is the “no-compensation result” of BRS.
4.2 Endogenous Probability of a Taking

Several counterarguments have been advanced in favor of compensation, the most com-
mon being that compensation is needed to prevent the government from overregulating
(Johnson 1977). These models can be categorized based on the assumption that is made
about the government’s behavior.
4.2.1 Benevolent (Pigovian) Government

A benevolent government is defined as one that makes the taking decision to maximize
social welfare. Fischel and Shapiro (1989) refer to such a government as “Pigovian”
because it considers the social costs and benefits of its actions. To capture this formally,
let the value of the land in public use, B, now be a random variable whose value is only
learned after the landowner invests x. A benevolent government will only take the land
if it turns out to be worth more in public than in private use, given x. Thus, once B is real-
ized, a taking will occur if and only if B ≥ V(x). Let F(B) be the distribution function of
B, where F′(B) ≡ f(B) is the density. The landowner is assumed to know F(B), so that, at
the time he makes his investment decision, he knows that the probability of a taking is
equal to 1 − F(V(x)) for any x.
The socially optimal choice of x now maximizes
F (V (x ))V (x ) + [1 − F (V (x ))]E[B | B ≥ V (x )] − x
∞
= F (V (x ))V (x ) + ∫ BdF (B) − x. (5)
V (x )
The resulting first-order condition is
F (V (x ))V ′(x ) = 1, (6)
which has the same interpretation as (2), with F(V(x)) replacing 1 − p as the probabil-
ity that the land will not be taken. The expected private value of the land in this case is
given by
F (V (x ))V (x ) + [1 − F (V (x ))]C(x ) − x , (7)
and the resulting first-order condition for x is
F (V ( x ))V ′ ( x ) + [1 − F (V (x ))]C ′ ( x )
+ F ′ (V ( x )) V ′ ( x ) [V ( x ) − C ( x )] = 1.. (8)

Comparison of (8) and (6) shows that lump sum compensation (C′ = 0) is no longer

sufficient for efficiency. Instead, compensation must be equal to the full value of the land
at its efficient level of investment, or C = V(x*). Intuitively, full compensation is neces-
sary to prevent the landowner from either overinvesting or underinvesting to alter the
probability of a taking (Miceli 1991). Specifically, if C(x) < V(x) the final term on the
left-hand side of (8) is positive. Thus, the landowner will have an incentive to overinvest
in order to reduce the probability of a taking since he expects to be undercompensated.
Conversely, if C(x) > V(x), he will have an incentive to underinvest to increase the prob-
ability of a taking since he expects to be overcompensated. Only a rule of full compensa-
tion, or C(x) = V(x), will eliminate this incentive. Combining this result with the lump
sum requirement yields the efficient rule, C = V(x*).27
27
This result was anticipated by Cooter’s (1985) option approach, under which the government
acquires an option from the landowner that allows it to take the land at any point for a prespecified
price. If this approach were used, P would replace C(x) in (7), and the first-order condition in (8) would
become F(V(x))V′(x) + F′(V(x))V′(x)[V(x)−P] = 1. (Note that the C′ term drops out here because the
price is viewed as fixed with respect to the investment choice, x.) It follows that P = C(x*) for efficiency.
The preceding compensation rule is not the only one that induces efficient invest-
ment in this case. Hermalin (1995) showed that two other rules are also efficient. Under
the first, C = B; that is, the government must pay the landowner the full value of the
public project in the event of a taking (the gain-based compensation rule). In this case,
the landowner internalizes the social value of the land given in (5) and therefore makes
the efficient investment choice. Alternatively, suppose that compensation is zero in the
event of a taking, but the owner has the option to keep the land by paying the govern-
ment its social value, B. A rational owner will exercise this “buy-back” option if and only
if B < V(x). Thus, only efficient takings will go forward. (Note, therefore, that the gov-
ernment’s decision about when to initiate a taking is immaterial, as long as it truthfully
reveals B to the landowner.) Under this rule, the landowner’s expected return is equal to
F (V ( x ))E[V ( x ) − B | V ( x ) > B] + [1 − F (V ( x ))]. 0 − x

V (x)
= ∫ [V (x) − B]dF (B) − x. (9)
0
Maximizing (9) with respect to x yields the first-order condition in (6). Thus, the
landowner makes the efficient investment choice.
In this analysis, the landowner is able to affect the probability of a taking through
investments that increase the private value of the land, V(x). However, Innes (2000)
notes that a landowner might also be able to affect the probability of a taking through
investments that change the public value of the land. For example, he might be able
to affect the desirability of his property as habitat for an endangered species that the
government might seek to protect. This implies that the investment x shifts the distri-
bution function of B; that is, the distribution becomes F(B,x). However, this does not
change the fundamental result that, when the landowner can affect the probability of a
taking, zero compensation does not lead to efficient landowner investment. In this con-
text, Innes shows that efficiency can be restored by compensating the landowner for the
public value of the land if it is taken or by employing a “negligence compensation” rule
under which landowners receive compensation only if they have acted efficiently when
investing in the public value of their land.
4.2.2 Nonbenevolent (Majoritarian) Government

More realistic models of government behavior suppose that it acts in the interests of
the majority of landowners, subject to budgetary restrictions (Giammarino and Nosal
2005). Thus, suppose that the government makes its taking decision by comparing the
value of the public project to the amount of compensation that it must pay in the event
of a taking, rather than to the opportunity cost of the land. Such a government is said to
have “fiscal illusion,” in that it only considers the budgetary impacts of its actions (BRS
1984). In this case, the government will initiate a taking if and only if B ≥ C(x), which
implies that the probability of a taking is 1 − F(C(x)), given x.
An obvious way to induce the government to make the correct taking decision is to set
C = V(x) (full compensation), but this rule will revive the moral hazard problem. One solu-
tion is to set C = V(x*), which solves both the fiscal illusion problem (because compensation
is full) and the moral hazard problem (because compensation is lump sum). As an alterna-
tive, consider the gain-based rule that sets C = B. As one saw earlier, this rule solves the
moral hazard problem, but a nonbenevolent government will be indifferent between taking
the land and not taking it. The landowner, however, will only want the taking to occur if B ≥
V(x), which is the efficient condition (given x). Thus, if the government follows the wishes
of the landowner, the rule will be (weakly) efficient regarding the taking decision.
Consider next Hermalin’s buy-back rule. Again, the landowner will control the taking
decision in this case and will do so efficiently since he will buy back the land if and only if
B < V(x). Because we showed that this rule also solves the moral hazard problem, it will
achieve efficiency of both the land use and takings decisions.
The final rule we consider involves a “threshold test” for compensation as first pro-
posed by Miceli and Segerson (1994, 1996). The rule works as follows: if the government
acts inefficiently to take or regulate land, it will be required to pay full compensation,
but if it acts efficiently, it will not have to pay.28 Formally, the rule can be written as
V (x ), if B < V (x *)
C=
if B > V (x *). (10)
0,
The efficiency of this rule can be established as follows.29 First, assuming that the land-
owner invested efficiently, the government will take or regulate the land if and only if it
is efficient to do so because it wishes to avoid paying compensation, which would result
in a loss of B − V(x*) when the taking is inefficient (i.e., when B < V(x*)). As a result,
landowners will anticipate that only efficient takings (or regulations) will occur and that
compensation for these actions will be zero. Thus, they will choose x*. This logic estab-
lishes that the Nash equilibrium under rule (10) will be efficient regarding both the land
use and taking decisions.
As a positive matter, the rule in (10) has considerable appeal because it goes a long
way toward explaining actual legal doctrine in the area of regulatory takings. Most obvi-
ously, the rule resembles the diminution of value test from Pennsylvania Coal because it
establishes a threshold for when a regulation “goes too far.” Specifically, compensation
will be due when the regulation is inefficiently imposed. The threshold rule also pro-
vides a standard for applying the noxious use doctrine. Specifically, a noxious use can
be defined as an activity that is efficiently regulated by the government and for which
compensation is therefore not required. Note that, according to this interpretation, the
noxious use doctrine and the diminution of value test are two ways of saying the same
28
Miceli and Segerson (1994) also propose a threshold rule under which compensation hinges
on whether the landowner acted efficiently in investing in the property. This rule is similar to the
“negligence compensation” rule subsequently proposed by Innes (2000) in the context where the
landowner’s investment affects the public use value of the land.
29 For a more detailed proof, see Miceli and Segerson (1994).
thing: the noxious use doctrine emphasizes cases in which the government has acted
efficiently in imposing a regulation, and so compensation is not due (corresponding to
the second line of (10)), whereas the diminution of value test emphasizes cases in which
the government has not acted efficiently, and so compensation is due (corresponding to
the first line of (10)).30
Similar reasoning shows that the Lucas nuisance exception fits easily into this frame-
work. Recall that the nuisance exception allows the government to avoid paying com-
pensation when it regulates activities that would be judged a nuisance under the state’s
common law. But how is a nuisance defined by the common law? The usual standard
is reasonableness, which is defined by asking whether a reasonable person would con-
clude that the amount of harm caused by the activity in question outweighs the ben-
efit.31 In other words, it is based on a cost-benefit calculation. Thus, the threshold for
compensation implied by the nuisance exception is identical to that under the proposed
threshold rule.
Extending this logic shows that the threshold rule provides an alternative “neutral
conduct” point for applying the harm-benefit rule. Specifically, by setting neutral con-
duct equal to efficient conduct, a regulation can be said to “confer a benefit” (and hence
require compensation) when it imposes inefficient restrictions on landowners, whereas
it can be said to “prevent a harm” (and hence not require compensation) when it imposes
an efficient restriction. The threshold rule is also consistent with Fischel’s normal behav-
ior standard, which, recall, set normal behavior based on a landowner’s reasonable
expectations (i.e., based on community norms) about permissible land uses.
4.3 Constitutional Choice Models

A different class of land use models is based on the notion that the government is not a
distinct entity with motives of its own but is merely the vehicle by which the citizens in
a given jurisdiction act collectively to govern themselves, including deciding on land
use policies and how to finance the cost of any required compensation. These “consti-
tutional choice models” envision a process in which citizens initially hold a hypotheti-
cal constitutional convention to choose the compensation rule from behind a veil of
ignorance about which particular parcels will be taken or regulated (Rawls 1971). Then,
these same citizens choose the amount (but not the specific parcels) of land to be regu-
lated or taken. Given this, individual landowners then make their investment decisions
without knowing if their land will be subject to the taking. Finally, the actual takings
decisions are made. In this setting, landowners know that they are both potential targets
30
Based on this interpretation, the rule in (10) provides a means of reconciling the apparently
conflicting decisions in Mugler, Pennsylvania Coal, and Keystone Bituminous Coal Assn. See Miceli (2011,
chapter 5) for a detailed discussion of this aspect of the threshold rule.
31 See Landes and Posner (1987, C hapter 2) for an economic theory of nuisance law.
of regulation but also beneficiaries of that regulation, and any compensation awarded to
victims of regulation must be financed out of taxes levied on all citizens. In designing the
compensation rule at stage one, citizens will therefore presumably take account of both
sides of the public ledger and thus will not be overly generous or stingy with regard to
compensation (Fischel 1995, 211).
The formal model of this process, first developed by Fischel and Shapiro (1989),32 uses
the following notation:
n = total number of identical parcels subject to a taking risk;

s = number of parcels to be taken for public use, s ≤ n;
B(s) = social value of the taken land, B′ > 0, B″ < 0.
T = per-person tax liability to finance compensation.
All other variables are defined as above. The public good, B, is assumed to be pure in
the sense that it is enjoyed by all landowners, including those whose land is taken. The
tax is also assessed on all landowners.
In this model, citizen landowners, acting from behind a veil of ignorance, choose
the number of parcels to take. However, since the specific parcels that will be taken
are only revealed after landowners have made their investment decisions, in the ini-
tial state, each landowner assesses an equal probability, p, that his or her parcel will
be taken, where p = s/n. The probability that a parcel will not be taken is therefore 1−
p = (n − s)/n. The wealth of landowners in the “no-taking” and “taking” states, respec-
tively, are given by
w N = V (x ) − T + B(s) − x (11)

wT = C(x ) − T + B(s) − x. (12)
The expected wealth of each landowner is therefore
E(w ) = (1 − p)V (x ) + pC(x ) − T + B(s) − x. (13)
The public budget must be balanced, so nT = sC, or, using the definition of p,
T = pC. (14)
(This assumes that the tax is assessed solely to finance compensation for takings.)
32 Also see Nosal (2001).

As in the BRS model, landowners choose x to maximize their expected wealth, taking
the compensation rule as given. In the current model, they also take as given the amount
of land to be taken, s (or, equivalently, the probability of a taking). The new element here
is the tax payment, T. If landowners also treat T as fixed (i.e., a lump sum tax), then the
first-order condition emerging from (13) would be identical to that in (4), and the BRS
result would be obtained. (That is, C′ = 0 would be a sufficient condition for efficient
investment.) However, suppose, more realistically, that taxes are assessed proportion-
ately on property values. That is, let T = tV(x), where t is the property tax rate. Also,
suppose that compensation is defined as a proportion of land value, or C(x) = αV(x) for
some parameter α. Substituting these expressions for T and C(x) into (13) and taking the
derivative with respect to x yields the first-order condition
(1 − p)V ′(x ) + pαV ′(x ) − tV ′(x ) = 1,

or
(1 − p)V ′(x ) + ( pα − t )V ′(x ) = 1. (15)
Now observe that, according to the balanced budget condition in (14), tV(x) = pαV(x),
or t = αp. Thus, the second term on the left-hand side of (15) vanishes, yielding (2). The
landowner therefore makes the efficient investment choice for any value of α; that is, any
compensation amount. In other words, the compensation rule is irrelevant with respect
to the land use decision. The reason for this result is that the compensation and tax dis-
tortions exactly offset each other through the balanced budget condition (Miceli 2008).
Finally, consider the choice of s, or how much land to take. Landowners also make
this choice from behind a veil of ignorance to maximize (13), subject to the balanced
budget condition in (14). Note that, in making this choice, they recognize the fact that
p(s) = s/n. The resulting first-order condition for s, after canceling terms, is
nB ′(s) = V (x ), (16)
which is the Samuelson condition for a pure public good. That is, land should be devoted
to public use until the marginal benefit of the last unit taken equals its opportunity cost
in private use. Thus, landowners authorize the efficient amount of takings for any given
x. As was true of the land use decision, this result is independent of the form of the com-
pensation rule and for the same reason. Thus, any compensation rule, including zero
and full compensation, would yield efficient decisions under this model.
4.4 Dynamic Models of Development
The discussion of land use incentives to this point has been based on a static model of
land use in the sense that the timing of the landowner’s development decision was not an
issue. This section extends the model to address two dynamic land use issues. The first
concerns the timing of development, and the second concerns the impact of the land-
owner’s expectation regarding the threat of regulation on the purchase price of the land.
4.4.1 The Timing of Development

The timing of development is an important issue because landowners faced with the
threat of a regulation may be impelled to develop prematurely in order to reduce or
eliminate that threat. For example, a developer may fill a wetland in order to preempt an
impending ban on development. Similarly, a landowner may harvest the timber on his
land prematurely to reduce the likelihood that it would provide habitat for a protected
species (see, for example, Innes et al. 1998). In fact, Lueck and Michael (2003) find sta-
tistical evidence that timber plots with greater proximity to colonies of the endangered
red-cockaded woodpeckers are more likely to be harvested early. Preemptive habitat
destruction of this type could actually lead to an overall reduction in the population of a
species that the land use restrictions are intended to protect.
A number of authors have presented theoretical models that investigate the impact
of compensation on the timing decision (e.g., Miceli and Segerson 1996; Innes 1997;
Riddiough 1997; Turnbull 2002; Lueck and Michael 2003). The basic insight regarding
premature or preemptive development can be illustrated using the following two-period
model of the land use decision, based on Miceli and Segerson (1996, Chapter 8). Let
VN = present value of the land if developed now;

VL = present value of the land if developed later;
p = probability that there will be a social benefit from preventing development in the
future, given no development now;
B = the resulting social benefit from prohibiting development in the future (either in
the form of an explicit benefit or a foregone harm);
V0 = residual value of the land to the landowner if development is prohibited, where
0 ≤ V0 < VL;
C = compensation paid to landowners who are prohibited from developing in the
future.
Assume that development in the present period cannot be prevented and that, once
it goes forward, the social benefit from prohibiting development can never be realized.
Also assume that if the land is not developed in period one and B is not realized in period
two, then the optimal course of action is to develop the land (i.e., there is no chance that
B will be realized in some future period).
The key question in this setting is whether it is optimal for the landowner to develop
the land now or to wait. If he develops now, the social (= private) value of the land is
fixed at VN, whereas if he waits, the expected social value is p(B + V0) + (1 − p)VL. Thus,
waiting preserves the option to use the land for the public project. It is therefore socially
optimal to wait if and only if
p(B + V0 ) + (1 − p)VL > VN .

(17)
From the landowner’s perspective, if he develops now, his return is VN, whereas if he
waits, it is p(C + V0) + (1 − p)VL, which differs from the social value by the inclusion of C
rather than B in the first term. He will therefore choose to wait if and only if
p(C + V0 ) + (1 − p)VL > VN .

(18)
Comparing (17) and (18) reveals that the only compensation rule that guaran-
tees that the landowner will make the correct decision is C = B. Any lesser amount
of compensation, including zero compensation, runs the risk of causing premature
development.
In addition to affecting the timing of development, the compensation rule can also
affect landowners’ incentives to reveal information about the public (e.g., conserva-
tion) value of their land. For example, in the absence of compensation, landowners do
not have an incentive to cooperate with regulators seeking to collect information about
public values prior to regulation. Providing some form of compensation (perhaps con-
ditional on landowner behavior or coupled with other conditions) can encourage land-
owners to cooperate with the collection of information or to reveal private information
(Polasky et al. 1997; Polasky and Doremus 1998; Innes et al. 1998).
4.4.2 Capitalization and Compensation

In his highly influential article, Michelman (1967) argued that a landowner who bought
a piece of property under the threat of an impending regulation would have no claim for
compensation if the regulation is later enacted because the purchase price would have
appropriately discounted the cost of the regulation. In other words, the price would have
“capitalized” the taking risk. This is a persuasive argument that has found its way into
the case law. For example, in H.F.H. Ltd. v. Superior Court,33 the Court denied relief to
a landowner whose commercial property was rezoned as residential based on the argu-
ment that “the long settled state of zoning law renders the possibility of change in zoning
33 542 P.2d 237, 246 (1975).

clearly foreseeable to land speculators and other purchasers of property, who discount
their estimate of its value by the probability of such a change” (246).
To demonstrate the capitalization argument formally (Miceli and Segerson 1996,
chapter 6), let
V = market value of a piece of property if unregulated;

VR = market value of the property if regulated, where 0 ≤ VR < V;34
p = probability that a regulation will be imposed.
Suppose that the current owner wishes to sell the property after the regulatory threat
has become public knowledge. Assuming that both buyers and sellers are risk neutral,
the maximum amount a rational buyer would be willing to pay for the property would be
(1 − p)V + p(VR + C ), (19)

which reflects both the risk of the regulation and the expected compensation. In the case
of zero compensation, the buyer would only pay (1 − p)V + pVR < V. Thus, if the regula-
tion were subsequently imposed, he would not have a good argument for compensation
since the sale price was appropriately discounted.
Epstein (1985, 151–158) and Fischel and Shapiro (1988) both point out, however, that
the seller would have a good argument for compensation since, at the time the possibil-
ity of the regulation was first announced, he suffered a capital loss equal to the differ-
ence between the discounted sale price and V, the value of the land in the absence of a
regulatory threat. In particular, his loss would be V −[(1−p)V + pVR] = p(V − VR). The
compensation question thus reverts to the original owner.
One way to eliminate the original owner’s loss would be to pay full compensation, or
C = V − VR, to the buyer at the time the regulation is actually enacted. Note that substitut-
ing this amount into (19) yields V, which means that the seller suffers no loss at the time
of sale. Alternatively, suppose the original owner is given the right to assert a takings claim
at the time of sale based on the probability that the regulation will be enacted later. Stein
(2000) refers to this as a “sale ripened” claim. In that case, the buyer would pay a price
equal to (1 − p)V + pVR since he would have no takings claim later (i.e., C = 0), but the
seller would receive compensation equal to p(V−VR) at the time of sale, yielding him an
overall return of (1 − p)V + pVR + p(V − VR) = V. Again, his loss is eliminated. In theory,
therefore, both approaches to the problem of a sale in the face of a regulatory threat are
equivalent in the sense that the original owner is fully compensated. In practice, however,
the sale-ripened approach is probably inferior both because it would entail more frequent
litigation and because it involves the difficult informational burden of calculating the risk
of a future regulation.
34 It is also possible that a regulation could enhance the value of other properties, a point we return to
in Section 4. (See, especially, note 37 and the associated text; also see Fischel [1995, 81].)
4.5 Balancing Risk and Incentives

A final reason for paying compensation is to provide risk-averse landowners with
insurance against a taking or regulatory risk. In advancing this argument, Blume and
Rubinfeld (1984) contend that the government needs to provide this protection because
private insurance for takings risk is generally not available. (Also see Rose-Ackerman
[1992] and Kaplow [1986, [1992].) Further, compensation must be mandatory, as by
constitutional dictate, because a nonbenevolent government might otherwise refuse to
insure those parcels that it plans to take or regulate. As we have already seen, however,
full compensation for takings creates the risk of landowner moral hazard, so the optimal
compensation rule must balance risk-sharing and incentives.35
4.6 The Social Cost of Funds

Aside from landowner moral hazard, the preceding discussion focuses primarily on
alternative economic arguments for paying compensation. However, requiring com-
pensation for land use restrictions would impose substantial resource requirements on
regulatory bodies. For example, an early estimate by Goldstein and Watson (1997) sug-
gested that requiring compensation for restrictions on wetlands development could
cost regulatory agencies $350–400 billion in 1994 dollars, or roughly $500–560 bil-
lion in today’s (2013) dollars. The revenue to pay compensation for regulatory takings
would generally have to be raised through distortionary taxation, implying a poten-
tially significant deadweight loss. This loss constitutes a cost of paying compensa-
tion that would have to be weighed against any benefits. For this reason, Innes (2000)
argues for use of a compensation rule that provides efficient incentives with the low-
est possible cost to the government. Note, for example, that the threshold rule in (10)
would not entail any deadweight loss because, under this rule, no compensation is paid
in equilibrium.36
5. In-Kind Compensation
Government regulations are pervasive and in many cases impose substantial bur-
dens on property owners in terms of lost value. It does not follow, however, that
property owners as a whole are necessarily made worse off by the imposition of such
35
For a formal analysis of this tradeoff in a takings context, see Miceli and Segerson (2007, 49–50).
For more general discussions of the tradeoff, see Stiglitz (1974), Holmstrom (1979), and Shavell (1979).
36 See Innes et al. (1998) and Innes (2000) for a more detailed discussion of the implications of the
deadweight cost of government taxation for the design of compensation rules.

regulations37 or even that landowners directly subject to the regulatory restrictions are
necessarily uncompensated. The reason for this paradoxical assertion is that the con-
stitutional requirement of just compensation does not specify that compensation must
always be monetary; it can also be in-kind (Epstein 1985, Chapter 14).
To see what this means, note that in settings where regulations are widely imposed,
as in the case of zoning restrictions, all property owners are equally burdened by the
regulations, but they are also equally benefited by them. These benefits provide a form
of implicit or in-kind compensation to all affected landowners. This argument implies
that a compensable taking has not occurred when a regulation secures an “average reci-
procity of advantage” across all property owners.38 It also reflects Michelman’s (1967,
1223) assertion that “[a]‌decision not to compensate is not unfair as long as the disap-
pointed claimant ought to be able to appreciate how such decisions might fit into a con-
sistent practice which holds forth a lesser long run risk to people like him than would
any consistent practice which is naturally suggested by the opposite decision.”
5.1 Neighborhood Externalities and Compensation

The economics of this perspective is based on the problem of “neighborhood exter-
nalities,” which represent the spillover effects (costs or benefits) that neighboring prop-
erty owners impose on one another as a result of their land use decisions (Miceli 2011,
136–139). For example, the manner in which owners use or maintain their property
obviously affects their own values but also the values of neighboring owners. Because
owners generally ignore these spillover effects, they may engage in socially inefficient
practices. For example, they may skimp on maintenance or paint their houses unusual col-
ors (Davis and Whinston 1961).
Often, the problem of neighborhood externalities is solved privately by means of agree-
ments, explicit or implicit, among residents (e.g., Cannaday 1994; Hughes and Turnbull
1996). In some cases, however, transaction costs limit the ability of these sorts of pri-
vate responses to the problem of neighborhood externalities. This is especially true for
large-scale externalities, such as those created by business operations or in very dense
neighborhoods where residents are strangers. In these settings, it is in the interests of prop-
erty owners to allow the government, acting in their collective behalf, to impose regulations
that allow (or rather, “force”) them to achieve an efficient land use pattern.
Based on this logic, regulations aimed at achieving this outcome would not be compen-
sable takings because landowners as a group actually benefit from them. Thus, for example,
a zoning regulation that prevents a landowner from opening a gas station in a residential
neighborhood would not give rise to a taking claim because, although the claimant might
37
See, for example, Truesdell et al. (2006) for evidence that wetlands regulations resulted in both
“takings” and “givings”; i.e., reductions in property values for some landowners but increases in values
for others.
38 Pennsylvania Coal v. Mahon, 260 U.S. 393, 415 (1922).
be able to demonstrate a loss in value due to the restriction, this loss would only exist rela-
tive to a background in which all other landowners are prevented from engaging in such
use. In other words, the claimant’s “loss” is calculated based on his unilateral departure
from the efficient land use pattern. Thus, he would have no claim for compensation. Indeed,
if the regulation is efficiently structured, it would actually raise the claimant’s property value
relative to the situation in which no regulation is in place and all landowners are free to pur-
sue their private interests unimpeded (Schall 1976). It is in this sense that all landowners are
said to receive in-kind compensation for the restrictions imposed by broad (and efficient)
government actions.
At the start of this chapter we asserted that, from an economic perspective, regulatory
takings lie on a continuum with physical takings and therefore should, in principle, be
treated the same. The preceding argument, however, provides a possible economic basis
for the dissimilar treatment of the two types of cases. Specifically, the nearly universal pay-
ment of compensation for physical takings, which typically involves the acquisition of only
a few parcels, reflects the concentration of costs on those owners whose land is taken and
for which they receive little or no in-kind compensation. Thus, monetary compensation
is necessary to satisfy the just compensation requirement. In contrast, the denial of com-
pensation for most regulations reflects their broad impact across property owners, with its
promise of in-kind compensation through increased property value, as measured relative
to a world in which no regulations are imposed on individual land use decisions.
5.2 The Essential Nexus and Proportionality Requirements

A different sort of argument for in-kind compensation was evaluated by the Supreme
Court in the case of Nollan v. California Coastal Commission.39 The case concerned
the buyer of a beachfront cottage who wanted to build a larger house on the lot. The
California Coastal Commission granted permission for the expansion, but only on
the condition that Nollan would agree to allow public access to the adjoining beach.
Although beach access would clearly represent a physical invasion of the own-
er’s property, and hence would constitute a taking under ordinary circumstances,
the Commission’s logic was that the requisite compensation was implicit in the
Commission’s granting of the development right. Thus, it maintained, no further com-
pensation was due. The Supreme Court disagreed based on the argument that there
had to be an “essential nexus” between any conditions attached by the government to
the development permit and the impact of the proposed development. Since in the
Nollan case it found that no such nexus existed, the implicit transaction was not legally
acceptable.
It is important to note that the Court’s ruling did not invalidate the logic of the govern-
ment’s argument; rather, it suggested that the proposed transaction was not acceptable
39 483 U.S. 825 (1987).

based on the facts of the case. The Supreme Court further refined its position on this
issue seven years later in the case of Dolan v. Tigard,40 which involved a requirement
by the City of Tigard that the owner of a hardware store had to deed a portion of her
property to the City for use as a bike path and open space as a condition for its allow-
ing her to expand the store. The City’s argument in making this request was that the
open space and bike path would mitigate the costs to the community arising from the
expanded business operation. The Court in this case found, in contrast to Nollan, that
there did exist a nexus between the city’s demand and the proposed expansion since
the bike path and open space would in fact mitigate the resulting damage. However, it
also suggested that the costs imposed on the landowner by the demand were dispro-
portionate in comparison to the social benefits. In order to avoid the need for explicit
compensation, the government had to demonstrate a “rough proportionality” between
the social harm from the proposed development and the value of the property that was
being taken in exchange. In other words, the in-kind benefit received by the landowner
had to provide sufficient compensation for her losses in order to meet the requirement
of just compensation.
Note that the difference between the rulings in Nollan and Dolan is merely one of
degree. Whereas Nollan found no relation between the government’s demand and the
landowner’s proposed development, Dolan found an inadequate relation (Fischel 1995,
349). Been (1991) nevertheless criticized the Court’s awarding of compensation in
Nollan (and presumably would have likewise criticized the reasoning in Dolan) based
on the argument that the claimant was protected against what he deemed to be an
unreasonable government demand by his option to exit the jurisdiction (Ghosh 1997).
However, Fischel (1995, 345) notes that, in Nollan, the regulation in question was tied
to the particular location—namely, the beachfront—rather than to an activity that the
claimant could easily have resumed in a different location. Thus, exit did not provide an
adequate escape for Nollan. The exit argument applies better to the facts of Dolan, which
involved a business that the claimant presumably could have relocated without substan-
tially diminishing its value.41
6. Summary of Economic Research on

Takings Law and Future Directions
Among legal scholars, the prevailing view is that the case law on regulatory takings is mud-
dled at best and chaotic at worst. This has been true ever since the Supreme Court’s deci-
sion in Pennsylvania Coal, which erased the apparently bright line separating compensable
40 512 U.S. 374 (1994).

41 Of course, businesses often have location-specific goodwill that would be lost in the event of exit.
takings from mere regulations that had been established in Mugler. The current state of the
law, which epitomizes this confusion, is the multipronged balancing test from Penn Central.
In our view, the contribution of the economics literature on takings, especially since the
BRS article, has been to bring some order to the debate, first, by formalizing the fundamen-
tal tradeoff between the land use decisions of owners whose land is at risk of a regulation
and the regulatory decisions of the government; and second, by showing how compensa-
tion rules can provide incentives for both decisions to be made efficiently. Although, as we
have noted, much of the literature has been normative in the sense of prescribing optimal
rules along these lines, we also emphasize that some rules can be interpreted in a positive
light as rationalizing the balancing approach that has emerged from the case law.
Aside from incentives, economic theory has also pointed out the risk sharing features
of compensation. When landowners are risk averse, optimal risk sharing is an important
aspect of efficiency, especially because land represents the largest component of most
people’s wealth. The absence of private insurance for takings risk provides an important
efficiency rationale for compensation, although as we noted, this factor may conflict
with incentives for efficient land use in the face of that risk.
Although we believe that the economic approach to takings law has been exceedingly
fruitful, there are several issues that warrant further work. One concerns the informa-
tion requirements of the proposed rules, many of which depend on the efficiency of
either the landowner’s or the regulator’s decision. The question of how the court would
acquire such information needs to be answered before these rules can be used in prac-
tice. Related to this is the increasing need for courts, in assessing the efficiency of various
policies, to value noneconomic goods like the environment or endangered species, and,
on the other side of the ledger, to account for the nonmarket (but legitimately economic)
value that owners attach to their land.42 Another issue concerns the motivation of judges
who play a big role in the evolution of the common law (no matter what the area) but
whose objectives are not well understood.43
Finally, it is important to recognize the limitations of economic theory in evaluating
takings law. After all, the Fifth Amendment may never have been intended to advance an
economic theory of takings. A broader perspective therefore requires the allowance for
other values besides efficiency, like fairness or justice.44 The challenge for future research
on the takings issue is therefore to incorporate these competing values to develop a more
complete understanding of the case law in this area.
42 On this last point, see Plassmann and Tidemann (2008) and Shapiro and Pincus (2007).
43 As a result, most models of legal change ignore the role of judges or treat them in an ad hoc way.
44 See, for example, Tideman and Plassmann (2005) and Niemann and Shapiro (2008).
Acknowledgments
We acknowledge the helpful comments of Josh Duke on an earlier draft of this chapter.
References
Been, V. 1991. ‘Exit’ as a constraint on land use exactions: Rethinking the unconstitutional
conditions doctrine. Columbia Law Review 91: 473–545.
Blume, L. and D. Rubinfeld. 1984. Compensation for takings: An economic analysis. California
Law Review 72: 569–628.
Boyle, K., and R. Bishop. 1987. Valuing wildlife in benefit-cost analysis: A case study involving
endangered species. Water Resources Research 23: 943–950.
Bromley, D. 1993. Regulatory takings: Coherent concept or logical contradiction? Vermont Law
Review 17: 647–682.
Cannaday, R. 1994. Condominium covenants: Cats, yes; dogs, no. Journal of Urban Economics
35: 71–82.
Coase, R. 1960. The problem of social cost. Journal of Law and Economics 3: 1–44.
Cooter, R. 1985. Unity in tort, contract, and property: The model of precaution. California Law
Review 73: 1–51.
Davis, O., and A. Whinston. 1961. The economics of urban renewal. Journal of Contemporary
Problems 26: 105–117.
Ellickson, R. 1973. Alternatives to zoning: Covenants, nuisance rules, and fines as land use
controls. University of Chicago Law Review 40: 681–782.
Ellickson, R. 1977. Suburban growth controls: An economic and legal analysis. Yale Law Journal
86: 385–511.
Epstein, R. 1985. Takings: Private property and the power of eminent domain. Cambridge,
MA: Harvard University Press.
Epstein, R. 1995. Simple rules for a complex world. Cambridge, MA: Harvard University Press.
Fischel, W. 1985. The economics of zoning laws: A property rights approach to American land use
controls. Baltimore: Johns Hopkins University. Press.
Fischel, W. 1995. Regulatory takings: Law, economics, and politics . Cambridge, MA: Harvard
University Press.
Fischel, W., and P. Shapiro. 1988. Takings, insurance, and Michelman: Comments on economic
interpretations of ‘just compensation’ law. Journal of Legal Studies 17: 269–293.
Fischel, W., and P. Shapiro. 1989. A constitutional choice model of compensation for takings.
International Review of Law and Economics 9: 115–128.
Friedman, L. 1986. A search for seizure: Pennsylvania Coal v. Mahon in context. Law and
History Review 4: 1–22.
Ghosh, S. 1997. Takings, the exit option and just compensation. International Review of Law
and Economics 17: 157–176.
Giammarino, R., and E. Nosal. 2005. Loggers versus campers: Compensation for the taking of
property rights. Journal of Law, Economics, and Organization 21: 136–152.
Gold, A. 1976. The welfare economics of historic preservation. Connecticut Law Review
8: 348–369.
Goldstein, J. H., and W. D. Watson. 1997. Property rights, regulatory taking, and compensa-
tion: Implications for environmental protection. Contemporary Economic Policy 15: 32–42.
Harrington, W. 1981. The Endangered Species Act and the search for balance. Natural Resources
Journal 21: 71–92.
Hartmann, J., and J. Goldstein. 1994. The impact of federal programs on wetlands, Vol. II.
A report to congress by the Secretary of the Interior, Washington, D.C.
Hermalin, B. 1995. An economic analysis of takings. Journal of Law, Economics, and Organization
11: 64–86.
Holmstrom, B. 1979. Moral hazard and observability. Bell Journal of Economics 10: 74–91.
Hughes, W., and G. Turnbull. 1996. Restrictive land covenant. Journal of Real Estate Finance
and Economics 12: 9–21.
Innes, R. 1997. Takings, compensation and equal treatment for owners of developed and unde-
veloped property. Journal of Law and Economics 40: 403–432.
Innes, R. 2000. The economics of takings and compensation when land and its public use value
are in private hands. Land Economics 76: 195–212.
Innes, R., S. Polasky, and J. Tschirhart. 1998. Takings, compensation and endangered species
protection on private lands. Journal of Economic Perspectives 12: 35–52.
Johnson, M. 1977. Planning without prices: A discussion of land use regulation without com-
pensation. In Planning without prices, ed. B. Siegan, 63–111 Lexington, MA: Lexington
Books.
Kaplow, L. 1986. An economic analysis of legal transitions. Harvard Law Review 99: 509–617.
Kaplow, L. 1992. Government relief for risk associated with government action. Scandinavian
Journal of Economics 94: 525–541.
Landes, W., and R. Posner. 1987. The economic structure of tort law. Cambridge, MA: Harvard
Univ. Press.
Lueck, D., and T. Miceli. 2007. Property law. In Handbook of law and economics, eds. A.
M. Polinsky and S. Shavell, 183–257 Amsterdam: Elsevier.
Lueck, D., and J. Michael. 2003. Preemptive habitat destruction under the Endangered Species
Act. Journal of Law and Economics 46: 27–61.
Mandelker, D. 1987. Investment-backed expectations: Is there a taking? Journal of Urban and
Contemporary Problems 31: 3–43.
Meltz, R., D. Merriam, and R. Frank. 1999. The takings issue: Constitutional limits on land use
control and environmental regulation. Washington, DC: Island Press.
Miceli, T. 1991. Compensation for the taking of land under eminent domain. Journal of
Institutional and Theoretical Economics 147: 354–363.
Miceli, T. 2008. Public goods, taxes, and takings. International Review of Law and Economics
28: 287–293.
Miceli, T. 2011. The economic theory of eminent domain: Private property, public use. Cambridge,
MA: Cambridge University Press.
Miceli, T., and K. Segerson. 1994. Regulatory takings: When should compensation be paid?
Journal of Legal Studies 23: 749–776.
Miceli, T., and K. Segerson. 1996. Regulatory takings: An economic analysis with applications.
Greenwich, CT: JAI Press.
Miceli, T., and K. Segerson. 2007. The economics of eminent domain: Private property, public
use, and just compensation. In Foundations and trends in microeconomics, Vol. 3. ed. W. Kip
Viscusi. Boston: Now Publishers.
Michelman, F. 1967. Property, utility, and fairness: Comments on the ethical foundations of
‘just compensation’ law. Harvard Law Review 80: 1165–1258.
Niemann, P., and P. Shapiro. 2008. Efficiency and fairness: Compensation for takings.
International Review of Law and Economics 28: 157–165.
Nosal, E. 2001. The taking of land: Market value compensation should be paid. Journal of Public
Plassmann, F., and T. N. Tideman. 2008. Accurate valuation in the absence of markets. Public
Finance Review 36: 334–358.
Polasky, S., and H. Doremus. 1998. When the truth hurts: Endangered species policy on pri-
vate land with imperfect information. Journal of Environmental Economics and Management
35: 22–47.
Polasky, S., H. Doremus, and B. Rettig. 1997. Endangered species conservation on private land.
Contemporary Economic Policy 15: 66–76.
Rawls, J. 1971. A theory of justice. Cambridge, MA: Belknap Press.
Riddiough, T. 1997. The economic consequences of regulatory taking risk on land value and
development. Journal of Urban Economics 41: 56–77.
Rose, C. 1983. Planning and dealing: Piecemeal land controls as a problem of local legitimacy.
California Law Review 71: 837–912.
Rose-Ackerman, S. 1992. Regulatory takings: Policy analysis and democratic principles. In
Taking property and just compensation: Law and economic perspectives of the takings issue,
ed. N. Mercuro, 25–44 Boston: Kluwer Academic.
Rose-Ackerman, S., and J. Rossi. 2000. Disentangling deregulatory takings. Virginia Law
Review 86: 1435–1495.
Rubenfeld, J. 1993. Usings. Yale Law Journal 102: 1077–1163.
Sax, J. 1964. Takings and the police power. Yale Law Journal 74: 36–76.
Sax, J. 1971. Takings, private property, and public rights. Yale Law Journal 81: 149–186.
Schall, L. 1976. Urban renewal policy and economic efficiency. American Economic Review
66: 612–628.
Shapiro, P., and J. Pincus. 2007. Efficiency and equity in the assemblage of land: The L2H2 auc-
tion. Working paper, Department of Economics, University of California, Santa Barbara.
Shavell, S. 1979. On moral hazard and insurance. Quarterly Journal of Economics 93: 541–562.
Stein, G. 2000. Who gets the taking claim? Changes in law use law, pre-enactment owners, and
post-enactment buyers. Ohio State Law Journal 61: 89–165.
Stiglitz, J. 1974. Incentives and risk sharing in sharecropping. Review of Economic Studies
79: 578–595.
Tideman, T., and F. Plassmann. 2005. Fair and efficient compensation for taking property under
uncertainty. Journal of Public Economic Theory 7: 471–495.
Truesdell, M. K., J. C. Bergstrom, and J. H. Dorfman. 2006. Regulatory takings and the dimi-
nution of value: An empirical analysis of takings and givings. Journal of Agricultural and
Applied Economics 38: 585–595.
Turnbull, G. 2002. Land development under the threat of taking. Southern Economic Journal
69: 468–501.
Wittman, D. 1984. Liability for harm or restitution for benefit? Journal of Legal Studies 13: 57–80.
C HA P T E R 2 6
EMINENT D OMAIN AND THE

L A N D A S S E M B LY P R O B L E M
JO SH UA M . DU K E
In Kelo v. New London, 545 US 469 (2005), the US Supreme Court spurred a renewed
political, legal, and economic focus on eminent domain. The case concerned confis-
cations for urban redevelopment and the interpretation of “public use” from the Fifth
Amendment of the US Constitution. The decision precipitated widespread outrage, and
some states fashioned legislation to tie the hands of local governments in using emi-
nent domain. Kelo therefore simultaneously expanded eminent domain and created a
backlash. Most outrage focused on the perceived unfairness of eminent domain, and the
facts of Kelo do indeed suggest significant burdens borne by those whose property was
confiscated. Economic research is especially well positioned to inform the contentious
discourse that underlies the efficiency-derived normative arguments used to support
the legal precedent on eminent domain. An economic assessment of land market fail-
ures from eminent domain helps to distill institutional meaning for a post-Kelo world
and to predict the impact of the decision on land markets.
The Fifth Amendment reads, in part, “nor shall private property be taken for public
use without just compensation.” This is the “eminent domain” clause. Legal scholarship
and case law focus on two aspects of the clause. The first concerns “just compensation”
questions, such as when is compensation due, and what is “just”? The second analyzes
what constitutes “public use.” Since the 1980s, two principal approaches emerged to
analyze the economics of land market problems related to eminent domain, and these
mirror the two legal aspects.
Economic research investigates when the payment of just compensation is efficient,
and Blume, Rubinfeld, and Shapiro (1984) present the seminal model. Their article
and ones that extend it offer many results, but the main finding is that full compensa-
tion may introduce a moral hazard incentive problem. In Chapter 25 of this handbook,
Miceli and Segerson thoroughly review their own and other extensions of the Blume,
Rubinfeld, and Shapiro (1984) model. These models largely focus on compensation,
especially in the context of the efficiency of regulatory takings law. Although regulatory
Eminent Domain and the Land Assembly Problem 699
takings indisputably emerged from eminent domain law, regulatory takings law has
since evolved into a complex set of bright-line and ad hoc tests, and these tests do not
generally apply to cases of physical appropriation. Thus, regulatory takings and eminent
domain are largely distinct in current law. In law, regulatory takings scholarship tends
to focus on whether compensation should be paid for a governmental action, whereas
eminent domain scholarship assesses on the constitutionality (or permissibility) of the
action. Because just compensation is not the central focus of eminent domain law, eco-
nomic modeling must also examine questions about the performance of the underlying
land market.
This chapter focuses on the second area of law—what is “public use”—and a sec-
ond economic problem associated with land market performance—land assembly.
Economic land assembly models can help inform questions about the permissibility of
eminent domain actions. The underlying efficiency problem is the inability of markets
to deliver optimally sized redevelopment, which warrants some exercises of eminent
domain. In land assembly problems, a developer attempts to buy and combine a set of
contiguous urban parcels, which time has seen partitioned to an extent that private rede-
velopment faces difficulties in capturing economies of scale. Why can a developer and
sellers not write socially optimal assembly contracts? The conventional economic story
involves the holdout problem; sellers act as local monopolists, holding their perfectly
heterogeneous product out of the market to gain market power and thereby capture a
share of the redevelopment rents (Posner 1992, 56). Despite the plausibility of the story,
surprisingly few economists have replicated this simple intuition with formal models.
This chapter begins with a brief summary of the eminent domain jurisprudence.
Next, the main economic approaches to analyzing eminent domain are reviewed
(readers are directed to Miceli and Segerson’s chapter in this handbook for an in-depth
assessment of the just compensation models). Eminent domain potentially corrects
two inefficiencies in land markets: (1) inefficiency from holdouts and (2) the under-
supply of public goods from urban redevelopment. Yet, eminent domain will not be
efficient in all settings.
This chapter then develops an original model of land assembly that is able to explain
inefficiencies in land assembly that derive from informational asymmetry. The model
identifies conditions under which eminent domain is likely to be more efficient than pri-
vate assembly and vice versa. One key finding is that eminent domain affects the perfor-
mance of the land assembly market, and thus market failure must be assessed in light of
holdout, public good, and the option for eminent domain. Beyond efficiency, the results
suggest that information asymmetry is the predominant reason for failure in the land
assembly market—it is here that the warrant for eminent domain resides. This model
helps assess recent controversies arising from eminent domain jurisprudence, and it
offers a new opportunity for further economic investigation of urban land market fail-
ures. The model shows that the untoward intentions associated with the popular debate
about Kelo have been given too much credit for market assembly and eminent domain
problems. Instead, it is information asymmetry that drives much of these conflicts.
Simply, private information prevents socially optimal contracts. A better understanding
700 Joshua M. Duke
of the conditions under which land markets fail to deliver assembly will then lead to
a better understanding of how new legislation and judicial rules will deliver improved
outcomes. A final section concludes.
1. The Law of Eminent Domain
Although the eminent domain clause of the Fifth Amendment is simple and brief, the
past century saw protracted academic disputes and legal cases that shaped the permis-
sible scope of eminent domain. Controversy was particularly acute with respect to the
definition of “public use.” In what has become known as the “narrow” view of public use,
the public must actually use the confiscated property. For instance, under the narrow
view, the public must actually use land confiscated from a farmer, say, as a public road.
The “narrow” view likely frames most examples in newspapers, public discourse, and
classrooms, so it likely captures the public perception of eminent domain.
However, case law tends to favor the “broad” view of public use, which allows for pri-
vate use of confiscated property so long as the use is to public advantage. The broad
view of public use evolved from a series of early cases (see Strickley v. Highland Boy Gold
Mining Company, 200 US 527, 1906) and, most recently, in Kelo. Berman v. Parker, 348
US 26 (1954) involved large-scale, urban redevelopment in Washington, DC, which
sought to confiscate some non-“blighted” properties along with blighted ones. The US
Supreme Court found the confiscation to be permissible, even though private parties
would end up with the redeveloped property because, for public advantage, the “area
must be planned as a whole,” and large-scale redevelopment cannot be accomplished on
a piecemeal basis. The Court in Berman explicitly recognized the land assembly prob-
lem and seemed to define the public interest as preventing local monopolies: “If owner
after owner were permitted to resist these redevelopment programs on the ground that
his particular property was not being used against the public interest, integrated plans
for redevelopment would suffer greatly” (348 US 26, at 35). Subsequent cases reinforced
this broad view. This includes validating as public use eminent domain actions that
prevent a land oligopoly (Hawaii Housing Authority v. Midkiff 467 US 229, 1984) and
enhance market competition (Ruckelshaus v. Monsanto Co., 467 US 986, 1984).
In the Kelo case, a five-justice majority validated the New London Development
Commission’s (NLDC) use of delegated eminent domain power to condemn 15
non-blighted houses for a loosely conceptualized, large redevelopment project. A rule
emerged that “public use” will be even more broadly interpreted. Specifically, blight was
no longer necessary for using eminent domain in redevelopment, and a new sufficient
condition became the prospect of increasing a jurisdiction’s tax base. Specific examples
of synergies were described, including jobs, tax revenue, and “build(ing) momentum
for the revitalization of downtown New London” (Kelo 545 US 469, 2005). Following the
case, many legislatures drafted statutes to restrain public agencies from using eminent
domain except under a narrow set of circumstances.
Several features of the Kelo case drive the renewed focus on the use of eminent
domain. The lower court opinions and other source material suggest that this redevel-
opment project was not profitable for a private developer (i.e., the project had net pri-
vate costs without subsidies), nor was it thought providential for the city of New London
(i.e., net social costs). The houses were in a well-maintained neighborhood, suggesting
the absence of that standard justification—removing blight. Moreover, the proposed
redevelopment plan was generally vague about what would be created—a parking lot,
retail opportunities, “support the existing park.” In short, the costs of eminent domain
on existing owners seem large, whereas the benefits of an assembled (large) parcel seem
ambiguous at best. Adding to the tenuous nature of the project, it turns out that the
NLDC was prepared to bear a range of costs of redevelopment—relying on its ability
to mobilize public resources for the task. The web-based accounts of the conflict sug-
gested that the developers hired by the NLDC might be handed assembled parcels at an
agreeable discount over what would have been required had eminent domain not been
available.
A flawed legal context was not aided by the inflammatory language deployed by sev-
eral justices in their struggle with the facts at hand. Justice O’Connor complained that
such practices, if allowed to persist unchecked, would find cities condemning cheap
hotels to allow for the appearance of elegant and expensive ones. The heated dissents
then contributed to the public outrage mentioned earlier.
The Kelo decision not only focused public attention on eminent domain law and
practice, it also suggested an expanded “public use” doctrine. A host of economic impli-
cations are associated with this new legal dispensation. Holdouts may hinder socially
beneficial land assemblies (i.e., positive externalities of redevelopment), whereas the
deployment of eminent domain will solve the problem of holdouts but captures some
subjective values from those in a position of being a monopoly supplier of needed
parcels (the holdout). Existing economic models—particularly, Miceli and Segerson
(2007)—offer insights into the gains and losses associated with both approaches to the
development challenge. But some issues have remained unexamined. Unlike previous
models predicated on certainty, there is abiding uncertainty on the part of developers
and landowners about whether eminent domain can and will be used to the developer’s
advantage.
2. Recent Economic Models of

Land Assembly
Existing models explore the holdout problem in at least four ways. One approach
focuses on delay, whereby sellers threaten to hold up a socially valuable project
(Menezes and Pitchford 2003); then, Nash bargaining occurs to divide the rents (Miceli
and Segerson 2007). A second approach is to conceptualize redevelopment as delivering
702 Joshua M. Duke
positive pecuniary externalities to adjacent owners. Then, a developer cannot fully

internalize pecuniary benefits through markets (O’Flaherty 1994). The problem also
may be viewed as one of anticommons, in which any seller can veto a socially optimal
transaction and thereby prevent efficient projects from occurring unless they are suffi-
ciently valuable (Heller 1998; Buchanan and Yoon 2000). A fourth approach shows that
asymmetric information may prevent optimal contracting (Eckart 1985; Strange 1995).
Unfortunately, little synthesis exists to reconcile these four approaches.
This chapter contends that there are really two distinct underlying efficiency prob-
lems: When should eminent domain be used, and when does the private market for land
assembly fail? The first problem may be termed the “eminent domain problem,” whereas
the second is the “land assembly problem.” These problems are related because eminent
domain is rationalized to overcome land assembly market failures, including monopoly,
externalities, and information asymmetry. However, strictly interpreted, the problems
are separate, and land assembly inefficiencies with and without eminent domain ought
to be assessed. In the next section, I offer a model that does this with a focus on land
assembly problems arising from information asymmetry.
One paper was identified that addresses both the land assembly problem (arising
from monopoly) and eminent domain as a solution to problem (Miceli and Segerson
2007). The justification for the existence of eminent domain is to overcome monopoly
by preventing delay—forcing the timing of the transaction—or by preventing sellers
from securing such high prices that redevelopments are inefficiently small, thereby forc-
ing sellers to accept the land-market price as just compensation. Miceli and Segerson
(2007) model the assembly market as a two-period game in which sellers engage in
cooperative Nash bargaining with the developer. Their model is particularly effective
in explicitly demonstrating the holdout phenomenon, and it also demonstrates how
eminent domain mitigates the incentive problem, although at a possible cost to sell-
ers. Miceli and Segerson (2007) conclude that eminent domain may lead to inefficiently
large redevelopments.
Eminent domain may also internalize the positive redevelopment externality by forc-
ing projects to be the optimal size. O’Flaherty (1994) discusses this possibility, but not
explicitly, in the context of his land assembly model. Overall, O’Flaherty seems pessi-
mistic about the corrective capacity of eminent domain because he believes local gov-
ernments face political constraints that result in inefficiently small redevelopments.
The role of information in assessing eminent domain has received scant attention.
Munch (1976) offered the closest approximation, considering the performance of
eminent domain relative to land assembly in light of transaction costs. Although not
modeled as an information asymmetry, the analysis of transaction costs may hold impli-
cations about the role of information in comparing the performance of market assem-
bly compared to public assembly. Munch (1976) concludes that eminent domain is less
efficient than decentralized market-driven land assembly. A major shortcoming in this
analysis is that Munch (1976) ignores externalities, endogenous behavior, and does not
first demonstrate the land assembly problem with the model.
This review of existing economic models suggests a gap in explaining the conditions
under which eminent domain outperforms market assembly in allocating scarce resources.
Eminent domain performance must be assessed relative to market assembly under the
same conditions and with explicit consideration of the market failures of monopoly (i.e.,
holdouts), externalities, and information asymmetries. As will be shown, information
asymmetry offers a great deal of explanatory power for land assembly market failure.
3. Application: A Land Assembly Model of

Eminent Domain
As the preceding section clarifies, only a small set of existing models explains the land
market inefficiency warranting eminent domain. In contrast, well-developed legal
scholarship exists on eminent domain. The model presented here combines key facets
of existing models (such as delay, positive redevelopment externalities, and compen-
sation) with information asymmetry. The application seeks to offer a more complete
explanation of market failure in settings where eminent domain may or may not offer
efficiency-enhancing institutional change. The original model is also designed to cap-
ture the key features of modern eminent domain law, such as those found in the Kelo
case. The features of case law include the private and social efficiency of assembly, the
role of delay costs for the developer introduced by uncertainty, and a new concept of
delay benefits to sellers.
The model reveals that the option of eminent domain in the future helps explain the
failure of market-based land assembly in the present. In addition, the model illustrates
the effects of the imposition of different decision rules on the public body exercising
eminent domain. Eminent domain proceedings can be used to force a transaction with
the developer, but, as one sees in Kelo, eminent domain can also be used to force and
subsidize land-assembly transaction. Recent legislative limits on eminent domain in
response to Kelo alter the behavior in private assembly markets, but they will not prevent
all opportunities for redevelopment. This section develops the general model structure
and identifies key assumptions. Then, in turn, the model is evaluated under two possible
information conditions: complete and incomplete information. The model is evaluated
and questions are posed about possible policy impacts.
3.1 Model Structure

Developer, D, seeks to assemble parcels owned by A and B into a redevelopment project.
These three parties constitute society, initially, although subsequent assumptions allow
possible positive redevelopment externalities (following O’Flaherty 1994). The model
(Figure 26.1) shows a two-period market assembly interaction and, later, an eminent
704 Joshua M. Duke
D
P A1, P B1 Indicates market
assembly outcomes
A
Period Accept Reject
1
B
Reject Accept Accept Reject
D V-P A1-P B1 D D
P A1 P A2, P B2
P B2 P B1 P A2
A
Period B Reject Accept
A
2
B
Reject Accept Reject Accept
Reject Accept Reject Accept
1-PA1-d1 V-PA1-PB2-d1 A2 B1
1-PB1-d1 V-P -P -d1 -d1 1-PB2-d1 1-PA2-d1 V-PA2-PB2-d1
PA1 PA1 v PA2 v v PA2 PA2
αv PB1 PB1 PB1 αv P B2 αv PB2
FIGURE 26.1 Market assembly subgame.
domain subgame (Figure 26.2) that will be imposed on the outcomes if market assembly
fails. The initial period when nature selects the player type variables is omitted from the
game in Figures 26.1 and 26.2:
• V is the value to D of the assembly measured in units of m, where m is the market

value of one unassembled parcel;
• d1, d2 are the costs of project delay incurred by D (delay cost is explained below);
• v is the subjective value of A’s parcel to A; and
• α is the subjective value parameter s.t. αv is the subjective value of B’s parcel to B.
Nature also selects the high-type and low-type party. In period 1, D makes take-it-
or-leave-it offers PA1 and PB1 to sellers, A and B, who then accept or decline. If one or
both sellers decline, the developer bears delay cost, d1, and makes new take-it-or-leave-it
offers PA2 and PB2 to the remaining seller or sellers. This section develops a series of
assumptions about parcel values and states of the world, which will guide the modeling
sections.
3.2 Parcel Values
Let both parcels be homogeneous in market value, but let owners hold heteroge-
neous “subjective” values (following the term from Posner 1992, 57). Owners have
Indicates pubic assembly outcome
d2
G
eminent domain@Pr(π) No eminent domain@Pr(1-π)
V-2-d1-d2 -d1-d2
1 v
1 αv
FIGURE 26.2 Eminent domain subgame.
heterogeneous values for many goods, such as housing, because individual prefer-
ences vary over the many amenities and disamenities that constitute the housing good.
Heterogeneity for a given type of house manifests as owners’ minima in their willingness
to accept (WTA) compensation. Owners with WTA at or below current market prices
(adjusted for transactions costs) are actively marketing their houses. Owners who are
not in the market, therefore, reveal that they hold a subjective value in excess of the mar-
ket value. As Miceli (2011, 57–58) explains, many scholars anticipate that the difference
between subjective value and market value increases over time for homeowners as they
gain idiosyncratic ties to their houses. For this model, simplifying assumptions are made
to study owners with subjective values that would be most likely to lead to holdouts.
Assumption 1. Sellers own homogeneous parcels with market value, m = 1.

Assumption 2. Sellers hold heterogeneous subjective values: v for A, αv for B, where
α є (0,1).
The actual price level is immaterial, so a numéraire scaling is imposed such that all other
values modeled are interpreted as relative to a single-parcel market value. The homoge-
neity assumption simplifies the model by controlling for a parcel characteristic that does
not contribute to explaining the occurrence of the assembly problem. Subjective-value
heterogeneity, however, does help explain why one seller might require a higher payment
than another and, equally, captures seller willingness to hold out for a certain offer and to
bear risk. In addition, heterogeneity allows for imperfect information on the part of D to
have a more realistic complicating effect on assembly. A standard assumption in assembly
models is that owners are not willing to sell at current market prices.
Assumption 3. Sellers are not currently marginal sellers in the market: v > αv > 1.
This is consistent with the economic concept of upward sloping supply. The intu-
ition that subjective value exceeds market value is well recognized in economics and
706 Joshua M. Duke
law,1 although it does not automatically follow that just compensation for eminent
domain should exceed m = 1.2
Much of the controversy involving eminent domain in law and economics centers
on the payment of m = 1, as “just compensation,” to owners with higher, but unobserv-
able, values. Posner (1992, 57) describes the quantity v − 1 as a tax on subjective value.
This is also the distinction in institutional economics between value in use and value in
exchange. From the perspective of a planner, representing the rest of society, this same
quantity can be seen as a fiscal savings. However, this fiscal savings may lead to some
projects that are inefficiently large.
What does it mean to be a “holdout”? This model will operationalize two forms of
holdout behavior.
Definition 4. A seller is a weak holdout when the seller demands or is paid more than
the market value, but less than the seller’s reservation value: v > PA > m; αv>PB>m.
Definition 5. A seller is a strong holdout when the seller demands or is paid more than
the seller’s reservation value: PA > v; PB > αv.
This chapter introduces the strong versus weak holdout distinction to capture dif-
ferent distributional outcomes and potential unobservable resource allocation inef-
ficiencies affected by eminent domain. A strong holdout seeks to extract rents from
the redevelopment project, whereas a weak holdout is taxed by eminent domain.
Because of the unobservability of subjective values, planners and other observers may
perceive holdouts to be any seller demanding or getting paid more than the market
value. The model focuses on the strong holdout phenomenon because it is most likely
to trigger inefficiently small redevelopments, and it captures the monopoly-type rent
extraction best.
3.3 Social Value of Assembly

Two sets of assumptions describe possible states of the world regarding the social value
of redevelopment relative to that of the unassembled parcels and include possible
positive redevelopment externalities. The following assumptions measure the relative
value of V.
1
Posner (1992, 56) captures this perspective well: “The familiar argument that the eminent domain
power is necessary to overcome the stubbornness of people who refuse to sell at a ‘reasonable’ (that is, the
market) price is bad economics. If I refuse to sell for less than $250,000 a house that no one else would
pay more than $100,000 for, it does not follow that I am irrational, even if no ‘objective’ factors such as
moving expenses justify my insisting on such a premium. It follows only that I value the house more than
other people. This extra value has the same status in economic analysis as any other value.”
2 Some do not view the subjective value as a legitimate measure of value for conducting efficiency
analyses.
Project
v(1+α)
not socially
efficient
Project socially
optimal
N.A.
Assumption
3
0
0 2 V
FIGURE 26.3 Social optimality of redevelopment.
Assumption 6. Assembly is a potential Pareto improvement (PPI): V ≥ v(1 + α).

Assumption 7. Assembly is not a PPI, but is privately efficient for D: v(1 + α)>V ≥ 2.
A Pareto improvement (assumption 6) occurs if gains from the project are distributed
to all three parties so that no one is made worse off. If the social value of the project is less
than the subjective values but more than the market values, then assumption 7 holds.
In other words, assembly appears financially efficient but is not economically efficient
because it consumes more resources than it creates. Under eminent domain, developers
and planners may perceive such projects to be “efficient” because the benefits of devel-
opment exceed just compensation, which are v and αv when awarded at market values.
These values are compared in Figure 26.3.
Redevelopment projects are often rationalized with positive externalities. For exam-
ple, the redevelopment in Kelo was rationalized as a way to capture “synergies” with a
recently sited, neighboring large firm. Following O’Flaherty (1994), let the value of the
positive externality be Y measured in units of m.
Assumption 8. Y ≥ 0, where the addition of Y is large enough to make Assumption 7

into Assumption 6.
If assumption 6 and 8 hold, then the PPI is enhanced by the externality but social
efficiency is not affected: V + Y ≥ v(1 + α) > 2. If assumptions 7 and 8 hold, then the
PPI is directly affected by the externality: V + Y ≥ v(1 + α) > 2. Although many possible
impacts of externalities on the social value assumptions are possible, the model focuses
only on these two. Figure 26.4 displays these relationships.
708 Joshua M. Duke
v(1+α) Project
not socially
efficient Project socially
efficient, but not
privately efficient
Project
socially
optimal
N.A.
Assumption
3
0
0 2 V
FIGURE 26.4 Social optimality of redevelopment with externalities.
3.4 Time and the Costs of Delay

The game consists of two periods of market assembly interactions and a third period of
public assembly (eminent domain). Note that the game can run for 1, 2, or 3 periods. Delay
costs, d1, are imposed on developers unable to secure assembly during negotiation period
1. If period 2 negotiations also fail, then the wait for possible eminent domain imposes a
second delay cost, d2.
Delay does not affect sellers’ decision making in market assembly. Specifically, any given
price in period 1 is valued equivalently to that same price in the future because the discount-
ing cost perfectly offsets the benefit of owning the parcel during the delay. This assumption
is likely reasonable for most cases; otherwise, disequilibrium prevails and owners would
have some incentive to enter the market in the present.
Consider the five instances of failed market assembly in the subgame in Figure 26.1
(occurring when one or both sellers reject the offer in period 2). Then, a legislative or
quasi-judicial body (the “public authority”) may use eminent domain. As in Kelo, the use
of eminent domain for redevelopment may involve the public authority pursuing emi-
nent domain and then turning over parcels to developers in exchange for market value,
m = 1, or even less when the parcels are reallocated as a discount. The period of time wait-
ing for public action at future time, T, imposes yearly eminent domain delay costs, ρ,3 on
3 In developing a land redevelopment plan involving assembly, D bears a host of project planning
costs. Some of these costs are sunk, but others incur as yearly costs, ρ. The yearly costs, ρ, might include
the developer such that d2 = ρ(1 − δT)/r, where r < 1 is the discount rate and δt = 1/(1 − r)t
is the discount factor that raises nominal eminent domain subgame payoffs to those at
the market assembly time.
Sellers view the delay of eminent domain differently than does the developer. At
time T, eminent domain occurs and, in expectation, sellers will anticipate nominal
gains that are less than their nominal subjective value (this is shown below). However,
at all periods leading up to T, sellers enjoy their full subjective values (in annual-
ized, nominal terms). In effect, the longer eminent domain is put off (i.e., as T → ∞),
the more subjective value under eminent domain approaches that of subjective value
with no eminent domain. Consider an example. A seller facing eminent domain at
time 10 derives greater utility from his or her parcel than if he or she faced eminent
domain at time 3 because that seller will be able to enjoy full subjective value during
time 3 to 10. This becomes a delay benefit for the seller. However, it is not valued at
an annualized v; rather, the benefit becomes an avoided opportunity cost of the dif-
ference between residing in the parcel and moving. The next best option is best mea-
sured by the market value, m = 1, and thus the delay benefit is: d3 = (v − 1)(1 − δT)/r
and d4 = (αv − 1)(1 − δT)/r. The author was unable to find the delay-benefit concept in
the existing literature.
After T years, the public authority renders a decision—one unknown prior to
T. The unknowable aspect of this decision pertains to (1) whether or not the proj-
ect will proceed, (2) the scope of the project, or (3) whether a developer’s desired
parcels will be included in a broader project. Redevelopment occurs when assem-
bly is mandated via eminent domain. The outcome is treated as an exogenous,
common-knowledge parameter where the probability of redevelopment is distrib-
uted uniformly πє(0,1).4
Let impacts on the developer and sellers incurred in market assembly increase with the
discount rate, such that V, d1, m, v, and α, are equivalent at any time. Then, the developer
payoff in the public assembly subgame is V − 2 − d1 − d2 if urban redevelopment occurs, and
− d1 − d2 otherwise—assuming the participation constraint is satisfied. Just compensation
is a market-value measure, m = 1. The public assembly payoff assumes that the developer
must pay for the parcels instead of a public body confiscating, paying m = 1, and then turn-
ing the parcels over to the developer at zero cost. The payoff to each seller is m = 1 if urban
redevelopment occurs and is the subjective value otherwise. The expected payoff for all par-
ties, with risk neutrality assumed, will simply weight the payoffs by the probabilities of their
occurrence.
the costs of sustaining the development plan if delayed. For instance, the development requires access to
capital and hedging against future increases in capital costs requires “lock-in” fees. A planning staff may
need to be maintained during the delay period. Also, delay involves the recurrent costs of negotiation
and gathering new information on an evolving local economy and land market.
4 The options of no eminent domain (π = 0) and certain eminent domain (π = 1) are not allowed. The
former is equivalent to the market assembly game. Both can potentially complicate the algebra.
710 Joshua M. Duke
3.5 Information
Asymmetric information exacerbates assembly conflicts, and much of the acrimony
over eminent domain arises from fairness concerns: the presumption that subjective
values exceed just compensation and the loss of autonomy in participating in a mar-
ket. D has private information about V, d1, and d2, whereas m is common knowledge,
as is the game structure. One limitation of the model (and an opportunity for future
research) is that D cannot invest in gaining information and cannot alter the structure of
the game to sort sellers. Furthermore, the model does not include a bargaining interac-
tion, which, when coupled with the sequential structure of the game, implies that seller
choice is not affected by knowledge of V. This contrasts with Eckart (1985) and Strange
(1995) who produced land assembly problems from sellers’ lack of knowledge about V,
although these articles proposed different game structures.
For A and B, information is imperfect because they move simultaneously, and it is
incomplete because they do not know the other’s subjective value. The game structure
and information availability thereby appear advantageous to D, who can make take-it-
or-leave-it offers to sellers. This advantage probably captures reality better than that of
market-savvy sellers, and the apparent structural advantage to D will be shown to bal-
ance with sellers’ monopolist power.
This chapter conceptualizes the complications of information on assembly in two
ways. First, governments do not exercise eminent domain with certainty. The process
of eminent domain involves competing claims in the political process, where a devel-
oper may lobby for a project and sellers lobby against. Exogenous factors, such as the
goals of political leaders and financial pressures within the community, also affect
the decision to use eminent domain. In the model, eminent domain occurs with the
common-knowledge probability of π. From D’s perspective, the risk that eminent
domain will not be used, 1 − π, becomes a cost. This parameterization extends the Miceli
and Segerson (2007) model, where eminent domain occurs with certainty and immedi-
ately at the developer’s behest.
The second complication is an information asymmetry, which is modeled only as it
affects D. The information in the game is complete when D knows v and α and is other-
wise incomplete.
Assumption 9. Complete information: D knows v and α.

Assumption 10. Incomplete information: D does not know v and α, but knows
assumption 3 (i.e., that A and B are not marginal sellers).
Incomplete information also implies D cannot identify the low type. In reality, a
developer would likely have some information about sellers’ subjective values. Future
empirical and theoretical efforts may choose to explore these information assumptions
and their impact on behavior and the efficiency of eminent domain.
4. Application: Land Assembly with

Complete Information
Although complete information (D knows v, α) may not completely characterize real-

ity, such an assumption is useful for establishing a baseline for market assembly perfor-
mance. Proponents of market assembly (and opponents of public assembly) argue that
markets produce superior results to those of eminent domain. However, perceptions of
market superiority in assembly are likely driven by subtle assumptions about the insig-
nificance of prevailing market imperfections, such as relatively low market power, no
positive externalities, and low transaction costs. This is not surprising. Standard eco-
nomic models show that perfect markets will automatically align private decisions with
socially optimal allocations of resources. In part, market assembly advocates also may
be driven by the Pareto improvement fairness characteristic of voluntary market trans-
actions. In contrast, public assembly advocates might be driven by the same criteria
of fairness and efficiency, seeing instead massive imperfections in market assemblies
and monopolists holding up socially beneficial projects to extract undeserved holdout
rents.
Assumptions about information lie at the heart of disputes about eminent domain.
These assumptions are largely driven by one’s worldview, and reconciling evidence is
difficult to come by. However, theory can offer insight. The ability of markets to deliver
optimal results without information problems must be assessed—and this assessment
must be made relative to eminent domain under similar conditions. This section evalu-
ates relative performance under assumptions of complete information. Then, in the next
section, the relative ability of market and public assembly under a limited information
condition is assessed.
4.1 Behavior in Market Assembly Without Eminent Domain

Under Complete Information
For now, let eminent domain be unavailable if D fails to purchase both parcels during
the market assembly period (only the Figure 26. 1 subgame). Seller payoffs are straight-
forward because, as argued earlier, delay does not affect seller decision making. If a seller
accepts an offer, then the seller receives PIt for I = A,B, t = 1,2. If a seller rejects the offer,
A receives v, and B receives αv. For D, market assembly payoffs are V − PA1 − PB1 if in
period 1 and V − PA2 − PB2 − d1 if in period 2. If no seller accepts an offer, then D incurs
delay cost, −d1. If one seller accepts and the other rejects, then D may resell the pur-
chased parcel for m = 1, but bears the costs of one purchase PIt for I = A,B, t = 1,2, and the
delay, −d1.
712 Joshua M. Duke
Will D participate in market assembly? Although Figure 26.1 does not allow explicitly
for D to opt out of the interactions in periods 1 and 2, D would never participate in any
market assembly game when it would be privately inefficient.
Participation Constraint D. D participates in period 1 if V − PA1 − PB1 ≥ 0, and in

period 2 if V − PA2− PB2− d1 ≥ 0, where PIt for I = A,B, t = 1,2, are optimal offers.
Characterizing equilibrium is straightforward, with backward induction identifying

optimal strategies. Consider the three second-period subgames. Optimal seller choice
is independent of the other seller’s choice, and sellers cannot credibly reject any take-it-
or-leave-it offer that makes them better off or (for simplicity) leaves them indifferent. A
will accept only if PAt ≥ v, and B will accept only if PBt ≥ αv, for t = 1, 2. Otherwise, each
rejects. D deduces sellers’ simple acceptance rule and will drive A and B to indifference
in period 2: PA2 = v and PB2 = αv. D prefers to avoid d1 and would thus also drive sellers
to indifference in period 1. If the participation constraint holds, in equilibrium, sellers
accept the subjective value offer in both periods.
One envisions three possible stories to support seller behavior. Sellers do not act strate-
gically beyond the structure of this game. These sellers accept a price that drives them to
indifference. In order to focus on other model aspects of interest, this story is maintained.
However, two competing perspectives exist. Sellers might require an epsilon bonus in the
period 1 to break the indifference within period 1 and between periods 1 and 2. This epsi-
lon could be modeled, but it would needlessly complicate the presentation. A third possi-
bility is that sellers would strategically seek to bargain over the delay costs, which they can
impose unilaterally by rejecting the period 1 offer. Such bargaining has been thoroughly
modeled by Miceli and Segerson (2007). If the epsilon bonus and strategic delay cost com-
prise part of the sellers’ reservation prices, the two competing perspectives become ana-
lytically identical to the maintained assumption that sellers accept the indifference price.
Table 26.1 compares equilibria of this complete information game under the states-of-
the-world assumptions about externalities and social value. Without externalities
(assumption 8 does not hold), equilibria depend only on the assumption about the social
value of the project, V, but, in both cases, market assembly is socially optimal. If the proj-
ect is a PPI (assumption 6), then market assembly occurs and, if otherwise (assumption
7), D does not participate. Also, if the PPI assumption holds, then market assembly pro-
duces a Pareto improvement because sellers receive payments equivalent to their subjec-
tive values and D keeps all gains from pursuing the project. This analysis is not surprising;
when no market imperfections exist, markets yield socially optimal outcomes.
If externalities exist, then social optimality depends on the social value assumption. If
the project is already privately efficient (assumption 6), then the addition of externalities
(assumption 8) does not alter market performance. However, if the project is not a PPI
(assumption 7, 8), then the project does not occur despite its social optimality. The ben-
eficiaries of the externality do not gain. This represents the prototypical public-goods
type market failure and captures the argument made by planners and politicians in favor
of using eminent domain for redevelopment.
Table 26.1 Equilibria in market assembly games of complete information

Value of project
Not PPI,
privately Privately
PPI efficient inefficient
Equilibrium Complete info Complete Info
conditions No externality Externality No externality Externality
D participates Yes or no? Yes Yes No No

D offers
P A1 v v
P B1 αv αv
P A2 v v
P B2 αv αv
A plays
Period 1 Accept Accept
B plays
Payoffs
D V−v−αv V−v−αv
A v v
B αv αv
3rd parties N.A. Y N.A. 0
Analysis
Assembly Occurs? Market Market No No
Is assembly Yes Yes No Yes
optimal?
Social Welfare Δ V−v(1 + α) V−v(1 + α) + Y 0 0
Maximizes? Yes Yes Yes No*
PI? Yes Yes N.A. No
* Fails to gain Y. PI, Pareto improvement; PPI, potential PI.
Overall, market assembly under complete information produces results that are intu-
itive and straightforward. Market assembly can achieve social optimality when markets
lack imperfections and when redevelopment is privately efficient. However, market
assembly does not always produce socially optimal results, even under complete infor-
mation. Public goods associated with redevelopment may lead to an undersupply of
assembly (in previous literature, inefficiently small redevelopments). In addition, there
exists no holdout problem in this game because sellers always receive their subjective
values, at minimum.
Two competing, but difficult to reconcile, visions of redevelopment likely dictate how
severely one evaluates the ability of market assembly to generate social optimality. One
vision views redevelopment externalities as substantive and pervasive. It is for exactly
714 Joshua M. Duke
these reasons—revitalization of moribund urban economies—that redevelopment

is proposed. This camp worries about the ability of market assembly alone to deliver
optimal redevelopments even under the ideal conditions of perfect assembly. A second
vision would likely argue that these externalities themselves produce the appropriate
market incentives for developers to internalize them by expanding the scale of their
redevelopments. If one truly believes that information is complete, then no obstacles
in this model (except the two-seller structure) would stop developers from achieving a
socially optimal scale of redevelopment.5
Collectively, this assessment warrants questions of how market assembly will perform
under conditions of incomplete information. Yet, this model does not fully character-
ize the performance of assembly markets because eminent domain is not modeled. In
Section 4.2, the assumption of complete information is maintained so the model can
inform the performance of assembly markets when eminent domain becomes available.
As one may suspect, under complete information, eminent domain will benefit D and
harm A and B relative to the land assembly market alone.
4.2 Behavior in Market Assembly with Eminent Domain

Under Complete Information
Eminent domain is rationalized as a way to achieve socially valuable land assembly when
market assembly fails. Following the preceding subsection, the assembly market’s failure
under complete information refers to a public-goods market failure rather than holdouts
from monopoly or delay. This subsection will show that, although eminent domain can
overcome this market failure, the option of public assembly distorts the incentive for a
developer to negotiate in the market. This option means that sellers can never do as well
as they might in market assembly without eminent domain. The equilibrium described
here shows that sellers accept offers below their subjective values, and this might lead to
inefficiently large redevelopments (under assumption 7). The developer will do better
with eminent domain. These results correspond to the public concerns in Kelo and other
cases about the “unfairness” of eminent domain. Indeed, eminent domain does have
unfavorable distributional impacts on sellers (under this model structure and assump-
tions). However, the model results will also show that greater uncertainty about eminent
domain and longer delays attenuate the developer’s advantage in eminent domain.
Eminent domain changes the strategies of D, A, and B. Sellers now trade off offers
with expected payoffs under the risk of confiscation rather than subjective values. For
now, assume there is no delay benefit to sellers. Sellers’ expected eminent domain payoff
is π + (1 − π)v for A and π + (1 − π)αv for B. So, in period 2, sellers will accept if D’s offer
exceeds expected eminent domain payoff: A accepts if PA2 ≥ π + (1 − π)v, and B accepts
5 Obstacles outside the model might include capital and bargaining issues with holdouts (see Miceli
and Segerson 2007).
if PBt ≥ π + (1 − π)αv. Otherwise, they each reject. Knowing this, D lowers the period 2
offers to force indifference. A similar assessment can be made in period 1; sellers must
accept the expected indifference payoff from eminent domain.
Equilibria can now be described. D’s participation constraint remains the same as in
market assembly alone, except now the optimal offers are less (see proposition 1 below).
This expands the set of interactions in which D will participate (see proposition 2),
creating a difference between social optimality and private efficiency in the absence of
externalities. D offers PA1 = PA2 = π + (1 − π)v, PB1 = PB2 = π + (1 − π)αv. A and B “accept”
in both periods. There are several implications.
Proposition 1. Under complete information, sellers receive and accept lower market assem-
bly offers when eminent domain is available than when it was not. The developer does
better.
Proof. It is sufficient to show for A that the acceptable offer under eminent domain is less
than the acceptable offer without eminent domain: π + (1 − π)v < v. This implies π(1 − v) +
v < v ==> π(1 − v)<0. The LHS is negative because of assumptions: πє(0,1) and v > 1. Also,
because D pays less to A and B, D is better off.
Proposition 2. With the eminent domain option, D will participate under more conditions
than under market assembly alone. Let this set of land market conditions belong to set Z.
Proof. Let points in set Z be project-seller value pairs: {v(1 + α),V}. Consider only
period 1, because assembly occurs in equilibrium in period 1. Under market assem-
bly, the participation constraint showed that D participated in all market conditions
when V ≥ v(1 + α). Let these conditions be set, X. Proposition 2 is proved if X is a per-
fect subset of Z and X≠Z. This holds if the market assembly participation constraint
without eminent domain, V ≥ v(1 + α), differs (and produces more possible pairs)
from the one with eminent domain, V ≥ π + (1 − π)v + π + (1 − π)αv, or if v(1 + α) > π +
(1 − π)v + π + (1 −π)αv. Simplifying, v + vα > π + v−vπ + π + αv−αvπ ==> 0> 1−v + 1−αv ==>
v + αv > 2, which is true by assumption 3.
Proposition 3. When no externalities exist, eminent domain expands the set of market
assemblies, and each of these new assemblies is socially inefficient.
Proof. From proposition 2, conflicts in Z but not in X satisfy: V ≥ π + (1 − π)v + π +

(1 − π)αv, but not V ≥ v(1 + α). Thus, the additional market assemblies conflicts created by
eminent domain are those where v(1 + α)>V ≥ π + (1 − π)v + π + (1 − π)αv. This condition
matches the definition of social inefficiency. These conflicts are displayed in Figure 26.5.
These conditions show that merely having the possibility of eminent domain changes
many aspects of market assembly. This is because complete information allows D to set-
tle each privately efficient conflict in market assembly and at a lower cost to D because
716 Joshua M. Duke
v(1+α)
Project
SIn, PIn
Project
SEf=Socially Efficient
SIn, PEf
SIn=Socially Inefficient
PEf=Privately Efficient
Project PIn=Privately Inefficient
SEF
N.A.
Assumption
3
0
0 2 V
FIGURE 26.5 Social optimality of redevelopment with eminent domain (with and without
externalities).
of the threat of eminent domain. Without externalities, eminent domain produces

socially suboptimal outcomes. Market assembly had been optimal—because of the lack
of imperfections—and now eminent domain leads to lower payments to sellers, which
represents an apparent unfairness because wealth is transferred directly from sellers to
a developer. In addition, eminent domain has allowed some socially inefficient market
assemblies to proceed. Graphically, one sees the inefficiency by a parallel shifting up and
to the left in the participation constraint line guiding D.
However, if redevelopment externalities exist, the results are not as clear about the
shortcomings of eminent domain. Here the “distortionary” line in Figure 26.5 actually
moves in the same direction as the externality line, in effect potentially improving the
alignment of social optimality and private efficiency. In other words, the distortion cre-
ated by eminent domain may attenuate the distortion created by the market failure. This
is not to say that internalization occurs because of eminent domain. On the contrary,
the two distortions would only align by chance. Nevertheless, the implication is clear.
Eminent domain can improve efficiency when externalities exist. The following propo-
sition formalizes these results.
Proposition 4. Given externalities, eminent domain improves the efficiency of market

assembly if π(v + αv − 2)≤Y.
Proof. From proposition 3, eminent domain distorts efficiency with all conflicts where v(1
+ α)>V ≥ π + (1 − π)v + π + (1 − π)αv. Redevelopment externalities, Y, are added to V, which
shrinks the number of conflicts satisfying the left inequality. The potential eminent domain
distortion to efficiency corresponds to the possible values of V in the preceding inequality,
which has magnitude: v(1 + α) − [π + (1 − π)v + π + (1 − π)αv]. This magnitude can be simplified
to π(v + αv − 2). This quantity can be compared to Y, the efficiency loss from not using eminent
domain for these conflicts. For π(v + αv − 2) < Y, eminent domain in effect internalizes some
of the externality, and at equality perfect internalization occurs. Thus, for π(v + αv − 2)≤Y,
eminent domain improves efficiency. For π(v + αv − 2) > Y, the internalization has occurred
and an efficiency loss begins to mount for the eminent domain distortion alone: Y − π(v +
αv − 2). Thus, optimally, society should balance the benefits of eminent domain Y with the
costs Y − π(v + αv − 2).
Proposition 4 is the principal efficiency result of this analysis. It also implies that when
externalities are small, eminent domain is less likely to provide an efficiency gain, all else
equal. If externalities are sufficiently small, there may not be any land market assembly
characteristics such that eminent domain will enhance social efficiency. Another impli-
cation is that eminent domain is more likely to enhance social efficiency when the exter-
nality is large relative to the eminent domain distortion.
5. Application: Land Assembly with

Incomplete Information
Incomplete information is a severe case, contrasting starkly with the state of the world
developed earlier because, strictly, an incomplete information world means that D
knows absolutely nothing about v and α. Mathematically, D would not know if the maxi-
mum of v is infinite. Then, all attempts at achieving market assembly fail. In general,
market assembly fails in many cases when information is incomplete—this is as to be
expected because the information asymmetry manifests itself as potentially infinite
transaction costs. Eminent domain corrects this failure in many cases, although it can
potentially overcorrect, as seen earlier.
5.1 Behavior in Market Assembly Without Eminent Domain

Under Incomplete Information
D does not know sellers’ valuations other than that they exceed m = 1. Sellers’ participa-
tion constraints (above) still guide their behavior. However, D’s behavior is complicated
by the information asymmetry. D no longer knows whether sellers will accept any offer
and thus faces two risks: (1) bearing the delay cost, d1, and (2) buying only from one
seller and having to scrap the project for salvage value, m = 1. The first risk is known to
D and could be optimally balanced with the expected benefit, V − PA1 − PB1, to form a
participation constraint. However, the second risk of salvaging PIt − 1, I = A,B, t = 1,2, is
pure uncertainty and explodes with the infinite possible maximum of v. One must then
posits a possible story about D’s strategic thinking.
718 Joshua M. Duke
Assume that D believes that the v maximum might approach infinity, as the pure
uncertainty condition suggests. Here, D will not participate in the assembly market at all
because the expected benefit cannot mathematically outweigh the salvaging risk. (The
salvaging risk is that D loses the difference between potentially infinite v and known
market value m = 1, which means the potential loss also is potentially infinite.) Thus, D’s
participation constraint cannot be satisfied, and land assembly never occurs.
If the project would otherwise satisfy the social efficiency condition, then market
assembly can be seen as inefficient. Redevelopment externalities exacerbate this ineffi-
ciency. If the project was not socially efficient, then market assembly is socially optimal.
However, externalities might expand the set of socially efficient projects, pushing some
from the socially inefficient to socially efficient category. The inefficiency result leads
logically to calls for eminent domain.
5.2 Behavior in Market Assembly with Eminent Domain

Under Incomplete Information
Under incomplete information, eminent domain would tend to generate more
socially optimal outcomes than would market assembly. However, this improve-
ment in social efficiency will sometimes require a transfer from sellers to D. Eminent
domain would occur if D offered prices of 0 or 1 in the market and then simply waited
for public assembly. Of course, D’s participation constraint would need to be satis-
fied. Figure 26.6 shows that, with eminent domain, many socially optimal conflicts
will now generate public assembly. Eminent domain improves social efficiency for
all conflicts in the lower right area (i.e., all socially efficient conflicts)—subject to the
preceding result that no assembly occurs without eminent domain when informa-
tion is incomplete. However, eminent domain will lead to some socially inefficient
assemblies. This region varies with π, as explained in Section 4 on complete informa-
tion. Furthermore, as with complete information, externalities would exacerbate the
V(1+α) V(1+α) No Project

Occurs: SEf
No Project
outcome Project
Occurs: SIn, PEf
SEf outcome
SEf=Socially Efficient
No Project Project SIn=Socially Inefficient
Occurs: SEf PEf=Privately Efficient
N.A. SIn outcome N.A. PIn=Privately Inefficient
Assumption Assumption
3 3
0 0
0 2 V 0 2 V
FIGURE 26.6 Social optimality of redevelopment with and without eminent domain (no
externalities; incomplete information). (Left) Market assembly; (right) public assembly.
shortcomings of market assembly and would be attenuated for many conflicts under
eminent domain.
6. Application Assessment
The analysis demonstrates that market assembly is relatively superior to public

assembly, but only if one believes in a state of the world in which information is com-
plete and if redevelopment externalities are small or can be internalized by larger
private developments.6 In other words, the complete-information state of the world
suggests social optimality is more likely to occur when eminent domain (1) is not
available, (2) has severe restrictions on its use, or (3) is used infrequently. If, however,
one views the world as one in which information is incomplete, and if redevelopment
externalities are substantive and cannot be internalized in private developments,
then the analysis suggests that market assembly will be relatively worse than public
assembly.
Although these two pure states of the world are somewhat unrealistic, it is exactly
these sorts of assumptions that would be required to sustain absolute positions such as
that the market (or eminent domain) is always superior. In contrast, much political dis-
course seems to claim knowledge that eminent domain either is or is not socially advan-
tageous. The economist wonders if such claims have any relevance for social efficiency,
and the model presented here suggests that it is information in the land use conflict set-
ting that may determine efficiency.
Future research may find analytical traction in the case of partially incomplete
information. The results of the model under partially incomplete information can
then be compared to the two extreme benchmarks—thereby establishing the rela-
tive performance of market and public assembly. Unfortunately, current evidence
does not suggest what type of information may be available to developers. The model
allows several familiar issues to be examined, each of which will have different
impacts on developers and sellers and that therefore may exacerbate or ameliorate
conflict. Several possible approaches to analyzing land use policy are suggested in
Section 6.1.
6 This result comes from the model, but, intuitively, other conditions outside the model might lead
to the same claim. For instance, if bargaining or negotiation leads to a sufficient attenuation of the
information asymmetry, then market assembly is relatively superior.
720 Joshua M. Duke
6.1 Research Possibilities: Evaluating Eminent Domain and

Institutional Change
The precedent set by Kelo seems to increase the likelihood of eminent domain because
the set of assemblies qualifying for “public use” seems to expand. In the model, this
means that π increases. The model suggests this increases assembly, which may be effi-
cient if there are externalities and is inefficient otherwise. It is not clear how this will
affect the behavior of developers if information is “partially” incomplete. One hypothe-
sizes that it will increase D’s use of low-ball offers because public assembly is more likely.
This, in turn, would lower the likelihood of market assemblies.
Following Kelo, some states enacted legislation to curb the use of eminent domain.
For some types of conflicts, such as redevelopment without blight, eminent domain
would be prohibited. This would reduce the π for urban dwellers and would likely pro-
vide an incentive to D to make fewer low-ball offers and thus increase the use of market
assembly.
Moratoria are a land use policy option used when land market behavior outstrips the
planning process’s ability to adapt. This strategy is one plausible response to the change
emanating from courts and legislatures. Local governments might suspend the use of
eminent domain for a period of time, thus increasing T. If T is large, then the payoffs D
anticipates from eminent domain become small. This will result in a tendency for devel-
opers to abandon hopes for eminent domain and compel them to seek market assem-
bly. This is not necessarily efficient because the land assembly market fails under many
conditions.
Some authors (as described in Epstein 1985, 184) suggest that just compensation for
eminent domain should be set at market value plus a bonus for reservation value, say
10%—an approach traditionally followed in England. This would render the new pub-
lic assembly payment to be m′ = 1.1m. Future work might explore how this institution
might affect efficiency, and one anticipates that it may increase the likelihood of market
assembly because eminent domain would become more costly to D—it is more costly
because D must now pay 1.1m for each confiscated parcel, rather than m.
Free development represents an odd case, but one which may be common in rede-
velopment conflicts. One of the controversies in Kelo concerned the perception that the
developers exaggerated the benefits of redevelopment because of the prospect of gaining
“free” land through the eminent domain process. Indeed, the promise of redevelopment
may lead public agencies to transfer the condemned land to developers at a very low
(perhaps zero) cost. In such cases, assembly need not be privately efficient, yet D may
still pursue a project of this type if D does not fully bear the costs of eminent domain.
Ultimately, this is not necessarily inefficient. Local governments may subsidize devel-
opers because they perceive very large positive redevelopment externalities. But such a
strategy would affect the performance of the land assembly market.
7. Conclusion
Eminent domain provides a mechanism to correct two substantive inefficiencies in

urban land markets. One form of inefficiency arises from the holdout problem, or local
monopolies associated with a heterogeneous land market. A second problem is that
assembly is a part of efforts to revitalize areas through urban redevelopment, which in
turn supplies public good benefits. A private assembly market fails to overcome these
two failures. Eminent domain law can be supported by these economic rationales.
Yet eminent domain will not be efficient in all settings. The application presented in
this chapter shows that there are conditions under which eminent domain is likely to
be more efficient than private assembly and vice versa. One key finding is that eminent
domain affects the performance of the land assembly market, and thus market failure
must be assessed in light of holdout, public good, and the option for eminent domain.
Other economic models have investigated efficiency implications of eminent domain,
including delay and just compensation.
Beyond efficiency, this chapter offers results suggesting that information asymmetry
is the predominant reason for failure in the land assembly market—it is here that the
warrant for eminent domain resides. Economists should devote increasing attention to
information and move the debate beyond issues of holdouts as monopoly, where perni-
cious sellers extract rents in markets and pernicious developers then capture rents by
aligning with power-hungry planners and incompetent local governments. Such inten-
tions have been ascribed particularly to developers, planners, and governments in the
wake of Kelo. The model shows that the ascribed intentions, whether true or untrue,
have been given too much credit for market—and public—assembly problems. Rather,
it is a cognitive failing that drives these conflicts—private information prevents socially
optimal contracts. A better understanding of the conditions under which land markets
fail to deliver assembly will then lead to a better understanding of how new legislation
and judicial rules will deliver improved outcomes.
Surprisingly, the land economic literature offers few economic studies of assembly and
eminent domain. There are several potential reasons for this lack of literature. First, it is
difficult to identify a behavioral reason why inefficiency exists—even though the assem-
bly market failures of monopoly and public goods are readily understood. When one
seeks to model behavior systematically, it becomes challenging to locate the precise failure
leading to the inefficient outcome. As this chapter clarifies, several successful models have
focused on delay and externalities. The application in this chapter focuses on information.
A second aspect of the literature has likely had a calming effect—specifically, the
Blume, Rubinfeld, and Shapiro (1984) result of efficient zero compensation. Blume,
Rubinfeld, and Shapiro (1984) offered an early and, some might say, definitive result on
eminent domain. By finding that zero compensation is efficient because of the moral
hazard problem, some economists may have been dissuaded from further study of emi-
nent domain by the profundity and persuasiveness of the result. Yet, policy makers and
722 Joshua M. Duke
the general public may have made little use of this result because it runs so counter to
notions of fairness—notions that moreover are ingrained in the US Constitution.
As the literature review and the application show, there are other ways than just com-
pensation to examine eminent domain. The “public use” part of the Fifth Amendment is
also important to politicians and judges, but it has received little attention from econo-
mists. Economic research on eminent domain is not settled. Collectively, this lack of
economic attention limited economists’ impact on the public debate following Kelo.
Ideally, future research will continue to investigate land assembly market failures.
Acknowledgments
Any errors are attributable to the author. Challenging discussions and feedback from
Dan Bromley helped shape and improve this work. The author is also grateful for
insightful feedback from Peter Schwarz and JunJie Wu, seminar participants at the
University of Delaware Legal Studies Program, and workshop participants at the Society
for Environmental Law and Economics.
References
Buchanan, J. M., and Y. J. Yoon. 2000. Symmetric tragedies: Commons and anticommons.
Journal of Law and Economics 43(1): 1–13.
Eckart, W. 1985. On the land assembly problem. Journal of Urban Economics 18: 364–378.
Epstein, R. A. 1985. Takings: Private property and the power of eminent domain. Cambridge,
MA: Harvard University Press.
Heller, M. A. 1998. The tragedy of the anticommons: Property in the transition from Marx to
markets. Harvard Law Review 111: 621–688.
Menezes, F., and R. Pitchford. 2003. The land assembly problem revisited. Regional Science and
Urban Economics 34(2): 155–162.
Miceli, T. J. 2011. The economic theory of eminent domain. New York: Cambridge University Press.
Miceli, T. J., and K. Segerson. 2007. A bargaining model of holdouts and takings. American Law
and Economics Review 9(1): 160–174.
Miceli, T.J. and K. Segerson. 2014. Regulatory takings. In The Oxford handbook of land econom-
ics, ed. J.M. Duke and J Wu. New York: Oxford University Press.
Munch, P. 1976. An economic analysis of eminent domain. The Journal of Political Economy
84(3): 473–497.
O’Flaherty, B. 1994. Land assembly and urban renewal. Regional Science and Urban Economics
24: 287–300.
Posner, R. A. 1992. Economic analysis of law, 4th ed. Boston: Little, Brown and Company.
Strange, W. C. 1995. Information, holdouts, and land assembly. Journal of Urban Economics
38: 317–332.
C HA P T E R 2 7
FUTURE RESEARCH
DIRECTIONS IN LAND
ECONOMICS
JO SH UA M . DU K E A N D J U N J I E W U
Land is special in many ways. In the foreword, Bromley argues that our idea of “place”
conflates with land, giving land primacy. Wars are fought over it. To the research econo-
mist, the commodity “land” poses special modeling challenges, not the least of which is
that many do not see land as a commodity at all. As Bromley notes, land is different from
other commodities that economists might model, like toothpaste. This special differ-
ence is not about market value. Land economists have long recognized the challenge of
putting land on the quantity axis of a market model. Land has describable but also inef-
fable qualities (Bromley, foreword); or, as some see it, land is “extremely heterogeneous”
(Irwin and Wrenn, Chapter 13). Irwin and Wrenn also argue that land modeling creates
special challenges in that decisions about its uses are affected by market and nonmar-
ket feedbacks, and it has important dynamic characteristics because past choices may
accumulate.
Put differently, the special nature of land makes all owners interdependent monopo-
lists foisting and bearing innumerable external costs and benefits, both spatial and tem-
poral. Massive and complex challenges stand in the way of markets effectively allocating
land among competing uses. And yet land markets function and not always poorly.
Why? How well do they function? Can policy improve outcomes? Land economics has a
long history of providing insights. As this handbook shows, land economists are poised
to offer innovative, powerful answers in the near future.
Future research directions in land economics will expand the study of optimal land
allocation, market failures preventing optimality, and policy impacts. The chapters in
this handbook make clear that new data exist to better describe the spatial aspects of
land, and new techniques allow economists to ask new questions or approach older
questions looking for new insights. Together, these forces offer rich opportunities for
economists to explain land use behavior and improve land market outcomes.
Although the handbook describes many advances, this chapter presents five promi-
nent trends that will occupy a large share of future research. The ability to develop more
sophisticated models that describe land outcomes is one of the most promising recent
developments. These counterfactual analyses exemplify a key advantage of rigorous eco-
nomic analysis. Another trend is toward more sophisticated models that explain how
people and firms sort on the landscape. Integrated feedback modeling of human and
ecological processes offers another rich field of study. Path-breaking articles from past
decade have laid out these methods, and land economists are poised to extend and apply
these techniques to a host of new problems.
These trends will help provide the answers to questions that society asks of econo-
mists. What changes would have happened to land uses in region R if clean water policy
U had not been implemented? How will land use change in the future if location V has
carbon policy W? How will incentive-based conservation policy X interact with zoning
Y and, in turn, affect the supply of ecosystem services Z? How will provision Z affect
the land choices made under X and Y? Land economists are better than ever poised to
answer these questions.
The purpose of this chapter is to identify broad research priorities and directions for
land economists, distilling lessons from recommendations in the preceding 26 chapters.
Each chapter comprehensively covers one area of land economics, and this final chapter
synthesizes five key directions for the field. The first two directions involve improved
modeling capacity: spatially explicit structural modeling and integrated economic and
ecological modeling. The third direction focuses on advancing methods to understand
and uncover agents’ behavior in settings involving land decisions. The fourth direction
explores how to use abundant yet incomplete or inconsistent data. The fifth direction
involves overcoming information challenges in policy design.
The focus on five directions is necessarily selective. For instance, most chapters
discuss specific applications and topic-based directions and policy needs. These are
difficult to synthesize, so we focus on broader trends. Another decision involved
determining what exactly constitutes a “future research direction.” Some chapters
identify where the authors anticipate the literature to be moving, either in the short-
or long-term. This positive approach differs from a normative one (i.e., where the
authors think the literature should be going). We try to include both perspectives in
this review.
1. Future Research Directions
An overarching trend that drives future research directions in land economics is the
integrated approach that involves both integrated economic and ecological modeling
and cross-fertilization among land-related economics fields. Several recent advance-
ments in economics, including the emergence of the new economic geography and eco-
logical economics, drive the integration among land-related economics fields. These
Future Research Directions in Land Economics 725
advancements have led to the increasing recognition that land use patterns, economic
growth, and the spatial distribution of economic activities and environmental impacts
are highly interdependent. Recent advancements in information technology has also
propelled integrated research and made it possible. Further development of integrated
research will require additional theoretic, empirical, and methodological advances. Five
future research directions are discussed here.
1.1 Spatially Explicit Structural Modeling

The increasing recognition of interdependency between land use patterns and eco-
nomic growth highlights the need for spatially explicit structural modeling. Partridge
and Rickman (Chapter 1) argue that land use and economic development research
should be modeled, jointly, as complex systems; otherwise, “piecemeal” analyses lead
to inconsistent lessons about development. In particular, Partridge and Rickman sug-
gest that land economists should explain the location behavior of households and firms,
whereas development economists should model the role of land in determining where
economic activity occurs. Parker (Chapter 16) sees opportunities for agent-based mod-
els to explore the relationship between employment centers and residential locations.
These studies suggest that a spatially explicit structural modeling approach would bet-
ter explain economic performance, the distribution of economic activity in a region,
impact of shocks (such as energy shocks and housing market bubbles), and poverty
(Partridge and Rickman, Chapter 1). Although economists are accustomed to structural
models, a great deal of the current empirical work relies on reduced-form models. Irwin
et al. (2009) call for structural modeling to better identify the potential causal linkages
among the many interdependent processes that affect urban-rural growth.
Economic models simplify reality, and economists keenly understand that their
models are built on assumptions. Some noneconomists reject economic results, point-
ing to assumptions that they perceive as invalid. Objections are frequently framed as
criticisms that economic models have oversimplified real-world complexities (Parker,
Chapter 16). Many economists would contest the validity and applicability of such
charges. They might argue that structural modeling is not an effort to address the long-
standing “oversimplification” objection but to gain additional insights by capturing the
essential linkages of the processes that shape economic and environmental outcomes.
Integrated research offers ways for economists to increase complexity, including link-
ing quantitative modeling efforts in different economic fields and in noneconomic disci-
plines. Integration can take many forms, as explained in Khanna, Zilberman, and Crago
(Chapter 4). They view the challenge of integration from the perspective of multiple
models that all examine similar phenomena. How can these models be linked? Khanna,
Zilberman, and Crago (Chapter 4) argue that a simple form of integration is triangula-
tion, or examining a distribution of estimates provided by different models. A nesting or
“off-line” linking of models provides a higher level of integration. Modularity offers even
higher levels of integration in which the “subroutines” of one model become incorpo-
rated in another.
Irwin and Wrenn (Chapter 13) identify three general opportunities for integrating
economic models. First, they see possibilities for simplifying difficult structural econo-
metric modeling problems by selectively combining these models with reduced-form
parameters estimates. A second opportunity identified by Irwin and Wrenn is to explore
complementarities between structural spatial equilibrium models and agent-based
models. Third, they believe spatial land use models can be linked to equilibrium models
from other sectors.
1.2 Integrated Economic and Ecological Modeling

A great need also exists for integrated economic and ecological modeling. Economic
activities affect ecosystems, and changes in ecosystems in turn affect economic perfor-
mance. Thus, there is a need for integrated economic and ecological modeling, regard-
less of whether the focus of the study is on a natural resource-based economic system
or a human-affected ecological system. Several chapters promote integrating economic
models of land use and ecosystem services models, especially using ecological mod-
els from outside economics (Lewis and Nelson, Chapter 7; Johnston et al., Chapter 8;
Attavanich et al. Chapter 10).
The chapters that use integrated models make clear that creating ecosystem service
linkages is challenging. Even descriptions on a single dimension of ecological pro-
cesses (such as habitat in one location for one species) require relatively large teams
with significant time investments. Beyond time and effort, insufficient model integra-
tion was cited as a shortcoming (Attavanich et al., Chapter 10). Furthermore, integrated
research should be careful not to proceed where the science cannot support its conclu-
sions. Johnston, Swallow, Bauer, Uchida, and Anderson (Chapter 8) explain and offer
examples about how truly integrated models of ecosystem services and land use can be
constructed. But they also emphasize that our current ability to model these interac-
tions accurately is limited. If “methodological sacrifices” are made to provide an inte-
grated model, then the validity of the measures and, more broadly, the approach will be
jeopardized.
Several chapters describe topical opportunities for integrated research. For example,
McCarl et al. (Chapter 9) argue that land use research on climate has insufficiently mod-
eled the interactions between adaptation and mitigation. They argue that this can be
conceptualized as an optimal portfolio problem. Montgomery (Chapter 11) calls for
feedback modeling between decisions to address fire fuel accumulation and fire sup-
pression decisions. Barbier (Chapter 6) finds too few studies offer detailed, integrated
modeling of land use change and economic development. Zilberman et al. (Chapter 2)
argue that the role of technology adoption in vertical integration requires additional
research and that this relationship should also model agricultural land use behavior.
This host of new methods offers better opportunities to examine complex linkages
between policy choice and land use behavior. Gnedenko and Heffley (Chapter 20) advo-
cate this form of integration, arguing for enhanced modeling of community-level deci-
sions and of the decisions of nearby communities or higher level governments. They
focus on decisions about fiscal policies for open space, but their point applies to other
public goods. Traditionally, the link is established through an open-city condition, but
Gnedenko and Heffley (Chapter 20) see opportunities for models that better capture
linkages and potential spillovers arising from fiscal policies.
1.3 Methods to Uncover Agents’ Behavior

The methods chapters cover spatial, econometric, simulation, and experimental
approaches. Most of these methods have been developed to better understand selection
issues. In other words, the methods allow economists to understand agents’ interac-
tions and decisions without necessarily observing in a given location a real-world policy
or a real-world market. Inferences can be made without bias. Collectively, economists
are better able to understand land behavior and phenomena, but the opportunities for
applying these techniques have only just begun to be realized.
Irwin and Wrenn (Chapter 13) offer a comprehensive and synthetic review of land
modeling, so this summary section on future directions will necessarily be brief. The
modeling chapters also offer details about the future directions of various empiri-
cal, simulation, and experimental methods. Klaiber and Kuminoff (Chapter 14)
explain equilibrium sorting models. The latest simulation methods are described in
econometric-based settings (Plantinga and Lewis, Chapter 15) and agent-based settings
(Parker, Chapter 16). Cho, Kim, and Roberts (Chapter 17) explain recent developments
in spatial econometric modeling (see also Plantinga and Lewis, Chapter 15; Brady and
Irwin 2011). Experimental methods, both quasi-experimental (Towe, Lewis, and Lynch,
Chapter 18) and lab/field experiments (Messer, Duke, and Lynch, Chapter 19), are
detailed. Many, but not all, of these methods involve some form of hedonic modeling.
The unifying message is that new spatial, econometric, simulation, and experimental
methods exist so that economists no longer need to wait for actual institutional change
to discern its likely impact on a landscape. The results of even untested land policies
can be informed by sophisticated methods. The chapters do an excellent job of explain-
ing how empirical evidence can be used to draw inference directly, to apply inferen-
tial results from one location to another, to motivate behavior rules in simulation, or to
guide experimental design. Even when no empirical insight can be found, the chapters
on lab/field experiments (Messer et al., Chapter 19) and agent-based modeling (Parker,
Chapter 16) explain how economists can still analyze land policy.
The topical and applied chapters also highlight exactly how these methods will
improve economic research. Nickerson and Zhang (Chapter 5), Barbier (Chapter 6),
and Cho et al. (Chapter 17) all call for expanded temporal studies of spatially explicit
land phenomena. However, Irwin and Wrenn (Chapter 13) note that these efforts require
further work on overcoming the curse of dimensionality introduced by dynamics in

empirical land use models.
Nickerson and Zhang (Chapter 5) offer recommendations for using new spatial
techniques, regression discontinuity design, and matching methods to improve farm-
land value research. Ferris and Lynch (Chapter 21) argue that these new techniques are
poised to make breakthroughs on our understanding of endogeneity of land policy and
outcomes. Methods to addresses observed endogeneity and other inferential challenges
are common themes in the handbook. In particular, authors recognize a need for better
modeling of empirical outcomes from regulations (Lewis and Nelson, Chapter 7; Stone
and Wu, Chapter 12). This relates to another important direction: research that uncov-
ers structural parameters. Readers will find Irwin and Wrenn (chapter 13) and Klaiber
and Kuminoff (Chapter 14) directly address these approaches. But Irwin and Wrenn
also caution that the literature requires further empirical tests of the assumptions main-
tained to drive these models.
Some chapters highlighted methodological challenges that current theoretical and
empirical models cannot address. In the case of estimating the sources and changes in
farmland values, Nickerson and Zhang (Chapter 5) argue for a better merging of two
largely distinct approaches—behavioral versus data-driven. They recommend building
dynamic structural models of behavior that explain how landowner expectations are
formed and that employ recent time series techniques to extend foundational models of
value, such as Just and Miranowski (1993). Such models, Nickerson and Zhang suggest,
would offer evidence on the speed with which farmland values change in response to
changes in drivers.
Another example concerns the manner in which objective functions are speci-
fied. The land policies of governments include many articulated goals, but these goals
are not always measured commensurately and benefit-cost data are often unavail-
able. A specific example might be farmland preservation policy, which seeks to
achieve acreage goals while also delivering a series of amenities and other services
such as food security, environmental services, and wildlife habitat (Nickerson and
Hellerstein 2003).
Can economists contribute insight to these situations without a wholesale repudiation
of the manner in which the objective is framed? Ferris and Lynch (Chapter 21) argue
that, in these cases, programs will require multiple techniques to satisfy multiple goals.
But current economic theory offers a poor understanding of the interactions among
these techniques, and economists need theoretical advances and empirical insights. Fire
hazard reduction and fire suppression response provides another example of a policy
setting with complex goals. Although existing analytical approaches in dynamic pro-
gramming do not readily address such problems, Montgomery (Chapter 11) sees poten-
tial for recent advances in approximate dynamic programming to be adapted to solve
spatial and dynamic optimization problems. Other chapters in this handbook suggest
similar ways that new methods can contribute insight to nagging challenges, even when
the problem is specified in a complex manner.
1.4 How to Use Abundant Yet Incomplete or Inconsistent Data

Many causal mechanisms that influence land use and location patterns differ depending
on the spatial scale of analysis (Irwin et al. 2009). Consequently, data are often collected
at different spatial scales, depending on the issue of interest. What are the implications
of scale-dependent processes for modeling and policy analysis? How should abundant,
yet scale-incompatible data be used? These are important issues for future research. As
Irwin et al. (2009) point out, we know relatively little about how microlevel processes
and heterogeneity affect the current aggregate outcomes. Therefore, efforts are needed
to develop models that capture decisions at the microlevel, that allow interactions at
multiple scales, and that are able to predict aggregate outcomes.
Spatially explicit data on land prices, uses, and services are rich and increasingly avail-
able. Several chapters detail the datasets available for researchers to study specific areas
of land economics (see, for instance, Klaiber and Kuminoff, Chapter 14, and Claassen
et al., Chapter 23). Many chapters suggest that newly abundant data, especially spatially
explicit data, offer tremendous opportunities for economists to create new insights on
markets and policy performance. Some chapters also identify data needs for currently
unanswerable questions.
Among the many chapters highlighting the need for better data, the unifying theme
is space. The future importance of spatial data surprises no land economist. Economists
eagerly anticipate new and better spatially explicit data, but they also identify specific
data needs. For example, Nickerson and Zhang (Chapter 5) call for more spatially disag-
gregated data and spatially ordered transaction data in the study of farmland markets.
Claassen et al. (Chapter 23) advocate for measures of transaction costs in landowners’
supply of agri-environmental services. Barbier (Chapter 6) identifies the need for tem-
porally explicit data on the geographical location of rural poor in the developing world,
especially with reference to vulnerable ecological sites.
Several other chapters anticipate the need for better measures of land use with specific
applications in mind. For instance, Stone and Wu (Chapter 12) call for improved mea-
sures of endogenous social amenities, or the intangible desirability of neighborhoods.
Lewis and Nelson (Chapter 7) seek better measures on the benefits of conservation.
Many application areas would benefit from better measures for dealing with counter-
factuals (see, e.g., the discussion of conservation policy intervention from Lewis and
Nelson, Chapter 7). More and better data from land markets can lead to resolution of
persistent econometric issues, such as omitted variable bias, which is a nagging econo-
metric challenge in many settings. Nickerson and Zhang (Chapter 5) explicitly cite this
issue with respect to farmland value studies, but it applies broadly to most empirical
land use models—especially the workhorse revealed preference models.
Other authors suggest that some spatial data exist but that economists have only
just begun to take advantage of the opportunities for analysis. Partridge and Rickman
(Chapter 1) recommend an expansion in the use of geo-coded firm-level data, espe-
cially when used in an integrated model with micro geo-coded housing data. Gnedenko
and Heffley (Chapter 20) argue for further and better matching of land cover data and
parcel-specific data on zoning and other land use controls. Yet existing data are not
always readily useable in spatial formats. Ferris and Lynch (Chapter 21) suggest that
farmland owners and other households’ opinion, income, and preference survey data
exist but that confidentiality rules prohibit most researchers from connecting these ana-
lytical units of observation with spatial data sources.
Several authors focused on making better efforts to measure unobserved costs
(in supplying ecosystem services and other settings) at the farm or parcel level.
Hodge (Chapter 22) identifies opportunities to use these cost data to better target
agri-environmental policy at the farm level. Several other chapters (chapters 21 and 23)
cite the importance of conservation targeting, but also the need to develop better target-
ing techniques; these techniques will require better measures of space, benefits, costs,
and the like.
1.5 Overcoming Information Challenges in Policy Design

Economists have a long history of examining the burdens of information costs on
resource allocation efficiency. Economists design contracts and institutions to improve
allocation in light of information asymmetry and use terms such as “second best” to
describe a situation in which one does as well as one can, given the information chal-
lenges (e.g., see Smith 1995; Wu and Babcock 1996). Many chapters in the handbook
identify areas where future work can continue to create cost-effective solutions despite
information asymmetry. For instance, these problems were mentioned in settings
such as habitat conservation (Chapter 7), empirical modeling of landowner returns
(Chapter 15), conservation auctions (Chapter 19), and urban landowner information
on reservation value (Chapter 26).
Solutions to information problems can be derived from within existing markets,
which means property rights are assigned and buyers/sellers are defined (Schmid 1999).
This also means that the market provides some structure to the information prob-
lems facing participants and policy makers who seek to adjust market outcomes. Yet
the chapters also make clear that land economists envision a future in which econo-
mists increasingly contribute to the design of policies, prior to the assignment of rights.
Recent interest by policy makers in incentive problems such as additionality, baselines,
leakage/slippage, and stacking seems to have energized land economists. Although
economists have recognized these incentive problems for years, the needs and funding
of policy is redirecting economic efforts (see, e.g., Claassen et al. Chapter 23).
Policy makers seek innovative solutions out of a desire for fiscal efficiency (i.e., obtain-
ing a given level of agri-environmental services for the least cost) or fiscal illusion (i.e.,
creating demand for services by capping sources of emissions or development). But land
economists recognize that the underlying problem arises from information asymmetry
in the censoring of landowners’ opportunity costs, which complicates optimal solutions
(Lewis, Plantinga, and Wu 2009). Several chapters point out emerging trends where
land economists can assess the performance of competing institutions and thus affect
policy before the assignment of rights. For instance, recent literature examines market
instrument comparisons in land conservation (consider Arnold et al.’s 2013 comparison
of auctions, contracts, and taxes in government procurement of ecosystem services) and
in work covering activities related to land (consider alternate baselines for incentivizing
best management practices in water quality trading in Ghosh et al. 2011).
The handbook chapters offer many specific directions in the study of information and
land economics. Fundamentally, the increasingly availability of information and inter-
connectedness of people affects and is affected by the way land is used. Mills (Chapter 3)
examines the economic performance of differently sized metropolitan areas and, in the
process, sketches a research agenda for anticipating future growth patterns. Mills pre-
dicts that suburban land use will continue to rapidly expand and thus poses a hypothesis
for future research.
Many chapters consider future research directions arising from information asym-
metry with respect to supply curve of land use. The chapters highlight the challenges of
uncovering these opportunity costs in a number of settings, including the nonpecuniary
benefits of agricultural land use and the role in technology adoption (Zilberman et al.,
Chapter 2). Several chapters call for studies of the unobservable costs of delivering con-
servation and other agri-environmental services (Lewis and Nelson, Chapter 7; Hodge,
Chapter 22). Plantinga and Lewis (Chapter 15) discuss how to model landowners’ pri-
vate information.
In Chapter 22, Hodge offers some sobering thoughts about a desire to create even
second-best contracts—issues that future research must come to terms with lest this
work be unproductive. Hodge starts with a relatively uncontroversial position (that
some asymmetric information is unavoidable), but notes that all agri-environmental
efforts rely to some extent on an ability to write enforceable contracts. Hodge wonders
if one also can define those agricultural practices that optimally deliver a given environ-
mental benefit, for a given farm, and with cognizance of permanence. In other words,
what happens when the contract ends? Eisen (Chapter 24) considers the effect of vol-
untary cleanup programs that lead to brownfield sites that are not contaminant free
many years following a cleanup. In this situation, what is the implication for efficient
“reopener” policy and land markets (i.e., policies that allow governments to require
additional clean up in the future)?
What about additionality? Will increased contracting for agri-environmental services
“crowd out” the norm among some land managers of delivering environmental services
for free (Hodge Chapter 22)? In the case of unreclaimed brownfields, residual contam-
ination identified at a later date would have occurred anyway. How should voluntary
cleanup programs be evaluated in light of this contamination (Eisen, Chapter 24)?
These are two examples of land economics studies in very different settings,
although the concerns are remarkably similar. Despite economists’ often very clever
work in market design, perhaps the remaining information problems of market fail-
ures are too severe. Even if one begins to understand fully how landowners respond
to incentives—say, contracts for agri-environmental services under information
asymmetry—Claassen et al. (Chapter 23) cite a need to understand the effect of differ-

ent funding mechanisms, such as federal cost-share. Certainly, one future direction will
be to seek better rights-based solutions. But another direction will likely be to devote
greater attention to comparing the relative performance of rights-based solutions under
alternate institutions.
Another quite different type of information problem that complicates land decisions
and policy involves uncertainty. For instance, eminent domain (Duke, Chapter 26), reg-
ulatory diminutions (Miceli and Segerson, Chapter 25), and brownfield policy (Eisen,
Chapter 24) are made under uncertain conditions about the future. How do future local
economic conditions and uncertain future contamination affect optimal urban redevel-
opment policy in the present? Perhaps the regulatory takings literature has advanced the
furthest in dealing with these issues because of robust work on efficient compensation
for eminent domain following Blume, Rubinfeld, and Shapiro (1984).
Miceli and Segerson (Chapter 25) point out that the regulatory takings results
depend on information about the impacts of a landowner’s or regulator’s decision,
but it is unclear to what extent courts could have access to this information. One
anticipates, nevertheless, that nonmarket valuation research, especially on ecosystem
services (see Johnston et al., Chapter 8), would have a great deal of impact in provid-
ing this information in the future. In the eminent domain case, Duke (Chapter 26)
argues that economists need more evidence on alternate rules about when confisca-
tions are authorized because these not only impact efficiency (after the fact through
compensation), but the expectations of outcomes in eminent domain also affect land
assembly market performance. Some future work should explore broadly applicable
rules for efficient urban redevelopment and efficient compensation for regulatory
diminutions.
Nevertheless, one challenge is that the two key literatures mentioned in this section
seem increasingly to have grown apart. Incentive-based agri-environmental policy now
seems quite distant from the economics of land use law (often regulatory standards).
The former focuses on incentivizing behavior, whereas the latter focuses on the implica-
tions of involuntary land use standards. In general, the regulatory law and economics
research establishes a set of baseline market conditions from which marginal changes in
the status quo occur through incentive-based policies. Towe et al. (Chapter 18) term this
a “shift” from regulatory to incentive-based regulation. Future work may want to reex-
amine these approaches for opportunities to enhance efficiency through linked models.
One opportunity might be to recognize that the constitutionality of regulatory poli-
cies determines the price of incentive policies. For example, if regulatory takings law
affords little protection to landowners in the form of no compensation for small to mod-
erate diminutions in value, then a conservation easement should be less (fiscally) expen-
sive. Future work may therefore seek to better understand the effectiveness of regulatory
policies in terms of the direct benefits and costs of control and in terms of the impacts on
future incentive-based efforts.
2. Reflections on Future Directions in

Integrated Research and Policy
This handbook has been framed with the claim that land economics requires an inte-
grated approach. The preceding sections suggest directions for integration, and this
section provides some concluding thoughts about why this approach is needed. First,
partial equilibrium analysis is not always adequate to examine the questions soci-
ety needs answered. In these settings, all else is not equal. Land use decisions impact
ecosystem services, economic development, and outcomes in other markets. There
are feedback effects. As Klaiber and Kuminoff (Chapter 14) explain, institutions cause
nonmarginal changes in public good provision, and partial equilibrium results will not
apply. People will resort themselves on the landscape.
Second, land economic problem settings are often too fluid to warrant the simplifica-
tion economists seek to derive tight and tractable results, ready lab experiments, and
empirically testable theoretic results. For instance, the urban-rural land use interface is
remarkably mutable. Although it is easier to model an urban market and a distinct agri-
cultural land market, actual market behaviors stretch such economic simplifications.
Why do some parcels at the fringe remain in farming despite higher apparent returns
to conversion? Why are there farm parcels inside the fringe, exhibiting the hallmark of
sustainable agriculture? What drives exurban residential development? Why do some
farmers make preservation decisions that seem to make themselves worse off? Many
economists have begun to explain these phenomena, and future work will likely need
integrated models to develop higher level explanations. Similar needs with respect to
these urban-rural modeling challenges could be claimed for explaining human-nature
interactions.
Third, integrated work may help prevent unexpected suboptimal recommenda-
tions. Economic studies often focus on narrow policy applications for technical reasons.
Social phenomena are largely uncontrollable, at least by researchers. Narrow foci allow
for thoroughness of coverage. But we also live in a vast, complex, and integrated world.
If we fix one failure, we may trigger others, and there may be perverse incentives.
Beyond our application areas, policy requires that we do not look at land use in isola-
tion. We need to understand how agricultural land use affects ecosystems and urban
uses. If economists explain these many linkages, policy is apt to follow. However, cur-
rent land policy, especially in the United States, is notoriously divided. Local govern-
ments have direct control over most land decisions, but state and federal governments
indirectly influence many land outcomes via major legislation and judicial review.
Historical legacies led to our current incoherent land use control organization. That
said, there is no clear answer on the best way to reorganize land policy.
The “extreme heterogeneity” of land seems to suggest that policy should be made
locally, where information is best. Local communities also have the greatest incentives
to solve local problems. But land also delivers positive and negative externalities beyond
localities, suggesting that higher level policy is warranted. Are the costs of voting with
one’s feet less expensive than policy coordination or reinvention? What about the voice-
less future? Many economists model only cross-sections or dynamics. Integrated mod-
els of space and time are difficult, but there is a cost to holding one dimension constant.
If economists expand the size and scope of their questions, fewer questions will be
answered.
Irwin and Wrenn (Chapter 13) call policy usefulness “one of the ultimate tests of any
modeling approach.” There is limited, unambiguous evidence that some work by land
economists has had direct impacts on policy (Irwin and Wrenn, Chapter 13; Banzhaf
2010). But the pace of economic research diverges from policy needs. Irwin and Wrenn
also argue that policy makers demand “real-time policy” and “quick approximate
answers.” Economic research largely does not match these requirements, and the inte-
grated modeling called for in this handbook would slow economics down further.
Although these challenges are well-known to economists, they miss the role
of agenda-setting research. Land economics has a strong record here. Consider
agri-environmental policy. The shift from regulatory land use to incentive-based poli-
cies was most likely due to pioneering work by economists. As early incentive-based
programs experimented with fixed-price and reverse auctions, economists weighed
in with results. Today’s economic research results will shape land policy in the coming
decade. In this framing, integrated research can play a significant role.
The research agenda for the future has not been finalized, and this handbook offers
many avenues for future work. Economists better understand how to compare insti-
tutional outcomes, but many applications are needed, and economists do not conduct
enough replications (Irwin and Wrenn, Chapter 13; Messer et al., Chapter 19). Given
the multiple sources of market failure in land, economists do not pay enough continuing
attention to the normative content of the efficiency concept employed (Bromley 1990)
or to problems with the general theory of second best (Lipsey and Lancaster 1956).
Given that land policy delivers multiple outcomes, what is the marginal rate of substi-
tution among the benefits (Claassen et al. Chapter 23)? Can society do better by turn-
ing agri-environmental payments around and focusing on the environmental results
delivered rather than the costs of delivering them (Hodge, Chapter 22)? Given that land
and land uses are highly inelastic in supply, economists should continue to explore the
policy implications of capitalization—that is, to what extent do improvements flow as
rents to landowners?
The integrated approach involves both a cross-fertilization across land-related eco-
nomics fields and also an integration with models outside economics. The increasing
recognition that land use patterns, economic growth, and the spatial distribution of
economic activities and environmental impacts are highly interdependent has led to a
convergence of interest among “land economists” working in several fields of econom-
ics, including agricultural economics, natural resource economics, environmental eco-
nomics, regional science, and urban economics. This has made the potential gains from
collaboration much greater. The purpose of this handbook has been to stimulate further
integration and collaboration in land economics research.
References
Arnold, M., J. M. Duke, and K. Messer. 2013. Adverse selection in reverse auctions for ecosys-
tem services. Land Economics 89(3):387–412
Attavanich, W., B. S. Rashford, R. M. Adams, and B. A. McCarl. 2014. Land use, climate change,
and ecosystem services. In The Oxford handbook of land economics, eds. J. M. Duke and J.
Wu, 255–280. New York: Oxford University Press.
Banzhaf, H. S. 2010. Economics at the fringe: Non-market valuation studies and their role in
land use plans in the United States. Journal of Environmental Management 91: 592–602.
Barbier, E. B. 2014. Land use and sustainable economic development: developing world. In The
Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 139–159. New York: Oxford
University Press.
Brady, M., and E. Irwin. 2011. Accounting for spatial effects in economic models of land
use: Recent developments and challenges ahead. Environmental and Resource Economics
48(3): 487–509.
Bromley, D. W. 1990. The ideology of efficiency: Searching for a theory of policy analysis.
Journal of Environmental Economics and Management 19(1): 86–107.
Cho, S-H., S. G. Kim, and R. K. Roberts. 2014. Spatial econometric modeling of land use change.
Claassen, R., J. Cooper, C. Salvioni, and M. Veronesi. 2014. Agri-environmental policies: A
comparison of US and EU experiences. In The Oxford handbook of land economics, eds. J. M.
Duke and J. Wu, 648–667. New York: Oxford University Press.
Duke, J. M. 2014. Eminent domain and the land assembly problem. In The Oxford handbook
Duke, J. M., and J. Wu. 2014. Future research directions in land economics. In The Oxford hand-
book of land economics, eds. J. M. Duke and J. Wu. New York: Oxford University Press.
Eisen, J. B. 2014. Stigmatized sites and urban brownfield redevelopment. In The Oxford handbook
Ferris, J., and L. Lynch. 2014. Land conservation in the United States. In The Oxford handbook
Ghosh, G., M. Ribaudo, and J. Shortle. 2011. Baseline requirements can hinder trades in water
quality trading programs: Evidence from the Conestoga watershed. Journal of Environmental
Gnedenko, E., and D. Heffley. 2014. Open space preservation: Direct controls and fiscal incen-
tives. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 547–582. New
Hodge, I. 2014. European agri-environmental policy: The conservation and re-creation of cul-
tural landscapes. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu,
Irwin, E., K. P. Bell, N. E. Bockstael, D. Newburn, M. D. Partridge, and J. Wu. 2009. The econom-
ics of urban-rural space. Annual Review of Resource Economics 1(October): 1–26.
Johnston, R. J., S. K. Swallow, D. M. Bauer, E. Uchida, and C. M. Anderson. 2014. Connecting

ecosystem services to land use: Implications for valuation and policy. In The Oxford handbook
Just, R. E., and J. A. Miranowski. 1993. Understanding farmland price changes. American
tion. American Journal of Agricultural Economics 91:1080–1096.
Lipsey, R. G., and K. Lancaster. 1956. The general theory of second best. The Review of Economic
Studies 24(1): 11–32.
information and auction efficiency in ecosystem service markets. In The Oxford handbook
Miceli, T. J., and K. Segerson. 2014. Regulatory takings. In The Oxford handbook of land econom-
ics, eds. J. M. Duke and J. Wu, 698–722. New York: Oxford University Press.
Mills, E. S. Are large metropolitan areas still viable? 2014. In The Oxford handbook of land eco-
nomics, eds. J. M. Duke and J. Wu, 74–84. New York: Oxford University Press.
Montgomery, C. A. 2014. Fire: An agent and a consequence of land use change. In The Oxford hand-
book of land economics, eds. J. M. Duke and J. Wu, 281–301. New York: Oxford University Press.
Nickerson, C. J., and D. Hellerstein. 2003. Protecting rural amenities through farmland preser-
vation programs. Agricultural and Resource Economics Review 32(1): 129–144.
Nickerson, C. J., and W. Zhang. 2014. Modeling the determinants of farmland values in the
United States. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 111–
Parker, D. C. 2014. An economic perspective on agent-based models of land use and land cover
Partridge, M. D., and D. S. Rickman. 2014. Integrating economic development analysis and
land use economics. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu,
Plantinga, A. J., and D. J. Lewis. 2014. Landscape simulations with econometric-based land use
models. In The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 380–401. New
Schmid, A. A. 1999. Government, property, markets. . . In that order. . . Not government versus
markets. In The fundamental interrelationships between government and property, eds. N.
Mercuro and W. J. Samuels, 237–242. Greenwich, CT: JAI Press.
nism. American Journal of Agricultural Economics 77(1995): 93–105.
Stone, E., and J. Wu. 2014. Land use and municipal profiles. In The Oxford handbook of land
economics, eds. J. M. Duke and J. Wu, 302–324. New York: Oxford University Press.
Zilberman, D., M. Khanna, S. Kaplan, and E. Kim. 2014. Technology adoption and land use. In
The Oxford handbook of land economics, eds. J. M. Duke and J. Wu, 52–73. New York: Oxford
University Press.
Wu, JunJie, and B. Babcock. 1996. Contract design for the purchase of environmental goods
from agriculture. American Journal of Agricultural Economics 78(1996): 935–945.
Subject Index
Adaptation, climate change 6, 226, 239–244, Approximate Dynamic Programming

255–258, 726 (ADP) 295, 728
Additionality 7, 17, 237, 239, 600–602, 631–633, Asymmetric information 7, 14, 15, 60,
639, 730–731 186–190, 487, 495, 565, 595, 597, 604,
Adverse selection 7, 485, 495, 595, 596 650, 699, 702, 703, 710, 717, 719, 721,
Afforestation 190, 233, 234, 236, 242, 260, 389, 730, 731
588 Auctions
Agent-based models 2, 9, 10, 328–347, in general 344, 411, 485, 487, 490, 564, 597,
402–424, 430, 496, 567, 602, 725–727 599, 602
Agglomeration 25, 27, 29, 30, 31, 32, 40, conservation auctions 7, 190, 217, 218, 482,
185–186, 189–190, 331, 413, 490, 513 485, 490–504, 597, 730, 731
AGLINK-COSIMO model 96, 104–105 efficiency 11, 491, 493, 495–497, 502, 564
Agricultural Adjustment Act of 1933 116 reverse auction 11, 217, 218, 494, 495, 504,
Agricultural commodity 103, 200, 416, 417, 734
593, 595, 628
Agricultural Sector Model (ASM) 262–269, Balancing test methods 461, 468
271, 272 Behavioral economics 482, 490
Agricultural policy 15, 128, 197, 198, 583, 590, Bellman equation 339, 384
593, 604, 618, 633, 636 Berman v. Parker, 348 U.S. 26 (1954) 700
Agri-environmental programs and Best management practices (see also Land
policies 15, 17, 127, 583–605, 612–640, conservation) 66, 616, 626, 630, 731
729–732, 734 Biodiversity 5, 6, 178, 189, 197, 199, 210, 258,
Agro-Ecological Zones (AEZs) 97, 241 302, 327, 337, 380, 381, 383, 388, 389, 423,
Alonso's urban land market theory (see Urban 491, 496, 554, 558, 559, 589, 590, 594, 595,
economies) 600, 601, 612, 617, 620–624, 632, 634
Amenities Bioeconomics 7, 196, 208–214, 293
in general 2, 24, 25, 27, 28, 30, 31, 33–36, 41, Bioenergy/Biofuel 4, 52, 85–107, 121, 129,
294, 304–310, 319, 321, 322, 330, 335–339, 234–236, 402, 423, 629, 630
341, 343, 344, 352, 354–359, 362–376, 381, Biofuel and Environmental Policy Analysis
431, 466, 467, 537, 560, 571, 572, 590 Model (BEPAM) 94–96, 99–102, 105
endogenous amenities 306, 318–319, 343, Biomass 94, 95, 102, 104–107, 234, 282, 283
355, 358, 367, 369–374, 729 Biophysical and biogeophysical processes
environmental, agricultural, and natural 8, and outputs 86, 89, 90, 93–95, 103, 106,
27, 28, 32, 33, 35, 120, 121, 124, 198, 285–288, 184–186, 199, 200, 203, 229, 329, 410, 421,
291, 303, 310, 311, 319, 321, 336, 352, 516, 423
612, 617, 623, 639, 728 Boom and bust cycles 4, 142, 150, 154, 155, 261
open space 303, 309, 310, 315, 338, 343, 435, Bounded rationality 416
444, 513, 516, 519, 520, 524, 526, 528, 532, 540 Brownfields and contaminated sites 15,
urban 24, 25, 352, 651, 705 648–664, 731, 732
738 subject INDEX
Cap and trade 99, 236, 423, 551 Conservation Reserve Enhancement Program
Capitalization 33, 112–114, 124, 127, 336, 354, (CREP), USA 553, 562, 563, 564, 566, 615,
369, 375, 376, 519, 688, 689, 734 616, 634
Carbon density 234 Conservation Reserve Program (CRP),
Carbon sequestration (see also Greenhouse USA 7, 13, 183, 184, 218, 490, 491, 493, 495,
gases) 197, 199, 215, 234, 235, 237, 239, 260, 502, 519, 548, 552, 553, 557, 558, 562–567,
291, 423, 554, 558, 562, 563, 628, 629 593, 615, 616, 625–630, 634–636
Cash crops 58, 143 Conservation Stewardship Program
Central Place Theory (CPT) 24, 27–31 (CSP) 553, 616, 626, 635, 639
Clean Development Mechanism (CDM) 238 Constant Elasticity of Transformation (CET)
Clean Water Act (CWA) 165, 375, 514, frontier 97, 101
614, 634 Constitutional choice models 684
Cleanup standards 653, 655, 657, 659, 660 Coordination of Information on the
Climate change 6, 85, 87, 88, 90, 97, 163, 175, Environment (CORINE) 637
178, 226–245, 255–276, 283, 327, 446, 603, Core-periphery patterns 31
618, 620, 624, 625, 639, 649, 662 Counterfactuals and counterfactual policy
Cobb-Douglas utility function 341, 343, 344, analysis/simulation 40, 92, 103, 104, 176,
364, 525–527, 531 327–329, 336, 337, 345, 453, 458–460, 465,
Command-and-control (see also Land use 468, 475, 558, 567, 600–602, 633, 724, 729
policy) 165, 166, 175, 176 Coupled economic-ecological modeling 394,
Common Agricultural Policy (CAP), 403, 420, 423
European Union 15, 583, 584, 586, 587, Coupled Global Climate Model (CGCM) 263,
591, 603, 618, 619, 622–624, 634, 639 264, 269, 270, 272, 273
Common Monitoring and Evaluation Credit subsidies 64
Framework (CMEF), European Crop insurance 57, 63, 238, 614, 628
Union 631 Crop prices 4, 85, 90, 91, 95, 106, 255, 257, 258,
Compensation for takings 15, 165, 618, 619, 626, 629
668–694, 698, 699, 703, 706, 707, 709, Cropland 90–92, 94, 97, 98, 100, 101, 104, 105,
710, 720–722, 732 107, 139, 140, 227, 230, 232, 235, 236, 241,
Complementary technology in production 58 257, 258, 261, 266, 268, 269, 275, 282, 302,
Comprehensive Environmental Response, 381, 383, 552, 557, 600, 616, 626, 628, 629
Compensation, and Liability Act Cross-compliance 15, 552, 613, 614, 618, 619,
(CERCLA), Superfund 376, 649, 621, 623, 633, 634
655–660 Crowding out effect, the 26, 629
Computable General Equilibrium (CGE) Curse of dimensionality 292, 339, 346, 728
model (see General equilibrium model)
Conditional independence, assumption of Decentralization 36, 332, 515
(CIA) 458, 459, 465, Decoupling and decoupled income
Congestion 4, 7, 8, 26, 28, 34–36, 78, 81, 303, support 63, 64, 117, 586, 603, 618
309, 327, 331, 516, 540 Deforestation 5, 69, 104, 105, 107, 140, 150,
Conservation (see Land conservation and 153–155, 199, 227, 234, 237, 258, 259, 383
policy) Delaware Agricultural Lands Preservation
Conservation easement 12, 13, 121, 126, 164, Foundation 496, 499
197, 457, 496, 518–520, 548–550, 553–557, Developing economies/developing
566, 568, 573–575, 615–617, 636, 732 countries 4, 42, 52, 64, 66, 68, 88, 101,
Conservation Effects Assessment Project 139–155, 229, 230, 232, 233, 242, 244, 488,
(CEAP) 636 729
subject INDEX 739
Development rights (see Conservation cataloguing 204, 562

easements) definition of 196, 200
Difference-in-Difference (DID) approach 39, disservices 197
40, 127, 130, 320, 375, 453, 457, 462, 468, ecosystem fragmentation 34, 332
472, 571, 601 marginality 206–207
Diminution in value 670–673, 676, 683, 684, models 209–212, 262–268, 292, 577, 726
732 Eminent domain 13, 15, 550, 670, 678,
Disamenities 308, 309, 612, 613, 639, 705 698–722, 732
Discrete choice models of land use and land Emissions (see Environmental impacts and
cover change 131, 332, 339, 445 Greenhouse gases)
Dispersal matrix 210–212, 214 Empirical methods (see Econometric and
Distributed lag model 114 empirical analysis)
Dolan v. Tigard, 512 U.S. 374 (1994) 693 Endangered Species Act (ESA), USA 163–176,
Dualistic frontier economy 140–155 296, 380, 514, 551, 558, 674, 675
Dynamic processes and modeling 4, 13, 35, Endogeneity 28, 29, 37, 38, 40, 90, 94–98, 119,
53–55, 61, 67, 69, 86, 87, 92, 94, 96–98, 125, 127, 170, 216, 263, 295, 306–309, 312,
112, 113, 116, 117, 128, 129, 131, 141, 144, 314, 316–322, 329–331, 334, 335, 337, 339,
145, 164, 179, 183, 204, 205, 209, 234, 242, 343, 355, 357–359, 362, 367, 369–374, 388,
259, 261–263, 286, 292, 293, 295, 331, 332, 392, 397, 409, 414, 418, 419, 430, 439, 452,
334, 339–342, 345–347, 383, 384, 402, 403, 454–456, 464, 476, 487, 498, 504, 521–523,
406–412, 416–419, 424, 440, 444–446, 525, 527–529, 531, 534, 539–541, 568, 572,
557, 558, 687–689, 723, 728, 734 577, 680, 702, 728, 729
Enhanced Safety from Wildfire Act 291
Easements (see Conservation easement) Environmental Benefit Index (EBI) 564, 565,
Econometric and empirical analysis 2, 4, 7, 626
9–11, 16, 17, 24, 25, 30, 31, 33, 38, 39, 41, 43, Environmental contamination (see
53, 54, 61, 63, 87, 111, 113, 115–125, 128–131, Environmental impacts and Brownfields
151, 152, 165, 176, 179, 186, 190, 197–199, 202, and contaminated sites)
203, 206, 213, 217, 219, 255, 256, 266, 268, Environmental contracts (see Payment for
276, 287, 289, 304, 309, 311, 320, 322, 327– ecosystem services)
347, 353, 355, 358–368, 374, 376, 380–397, Environmental impacts 1, 4–6, 8, 14, 34–37,
404,407, 413, 414, 418, 419, 421, 422, 424, 56, 59, 65, 78, 86, 88, 90, 93–95, 98, 107,
430–446, 452–455, 458, 461, 464, 475, 481, 120, 226, 227, 233–238, 244, 255, 256, 263,
489, 495, 519, 521–523, 525, 527, 531, 532, 534, 267, 302, 309, 355, 423, 485, 490–491, 516,
537, 539, 541, 542, 569, 577, 583, 594, 597, 551, 552, 566, 584–586, 600, 603, 612, 627,
599, 625–632, 640, 710, 725–730, 733 628–638, 649, 650, 653–657, 659, 662, 663,
Ecosystem and ecosystem services (see also 676, 725, 730, 734
Payment for econsystem services) Environmental market (see Payment for
in general 2, 5, 6, 11, 16, 35, 36, 86, 88, 91, ecosystem services and Auctions)
163, 183, 196–220, 233, 235, 239, 241, 242, Environmental policy integrated climate
256, 261, 276, 282, 285, 287, 292, 294–297, model (EPIC) 95, 263, 565, 627, 629
302, 333, 337, 380, 402, 423, 475, 482, 490, Environmental Quality Incentives Program
493–495, 503, 504, 547, 548, 550, 556–558, (EQIP) 553, 557, 616, 618, 626, 635, 639
563–565, 567, 583, 590, 601, 603, 604, 624, Environmental stewardship 591, 592, 596, 602,
627, 724, 726, 730–733 604
benefits 6, 35, 86, 88, 197, 198, 200, 201, 203, Environmentally Sensitive Areas (ESAs) 584,
204–206, 292, 548, 556, 557 591, 592, 595, 598
740 subject INDEX
Equilibrium sorting model 10, 23, 32, 36, 41, Fire

303, 334–339, 352–376, 560, 727 in general 281–297
European Agricultural Fund for Rural aboriginal 281, 282
Development (EAFRD) 587–589, 619, anthropogenic 282–285, 294
623 risk 281, 286–292, 294, 295
European Network for Rural Development suppression 284–287, 289, 290, 293–295
(ENRD) 637 Flight from blight 303
Event sequencing mechanisms 407, 411, Florida Rock Industries v. United States, 21
414–416, 420 Cl.Ct. 161 (1990) 674
Experimental economics Forest land 3–5, 14, 28, 34, 37, 89–95, 97, 98,
in general 9, 11, 16, 215, 216, 288, 481–504, 101, 104–106, 139–143, 152–154, 163, 196,
507–510, 597, 727 209, 214, 227, 229, 231, 233–237, 240–242,
bias 484, 486 257–260, 282, 284, 285, 288–297, 302, 318,
experiment design 484–490, 497–499 319, 381–383, 387–389, 396, 430, 431, 466,
field experiments 4, 340, 483 513, 532–534, 548–550, 554, 561, 562, 568,
laboratory experiments 190, 340, 481, 569, 572, 587, 620, 623, 628, 633
483–485, 491, 495, 503, 733 Forest Fires Emergency Act, USA 284, 285
participant instructions 507–510 Forestry and Agricultural Sector Optimization
survey 483, 484, 486–488 Model (FASOM) 94, 95, 99, 105, 106
Externalities 6, 8, 11, 12, 37, 54, 62, 64, 88, 94, Fossil fuels 94, 99, 142, 234–236
260, 281, 286–289, 294, 302, 303, 307, Fragile lands (see also Marginal lands) 140,
308, 322, 331, 397, 414, 475, 513, 515, 517, 141, 145, 152, 155, 627
525, 547, 594, 620, 639, 652, 662, 675, Free-riding 215, 216, 218, 287, 515
691–692, 701–703, 706–708, 711–721, Frontier 4, 101, 105, 140–155
733
General equilibrium models (see also
Farming (see Agriculture) Equilibrium sorting model) 2, 26, 38, 86,
Farm Accountancy Data Network (FADN), 92, 96–98, 204, 308, 334, 337, 338, 346, 352,
European Union 637, 638 358, 369, 370, 372, 375, 418, 423
Farm and Ranch Land Protection Program General Linear Model (GLM), spatial 431, 445
(FRLPP), USA 554, 617, 635 Generalized Propensity Score (GPS) 127
Farm Bill, USA 116, 117, 552 Generalized Method of Moments
Farm Structure Survey (FSS), USA 637, 638 (GMM) 128, 366–368, 431, 432, 437, 439,
Farmland values 4, 111–131, 232, 519, 564, 728, 445
729 Geo-coded data 43, 729
Federal Insecticide, Fungicide, and Geographic Information System (GIS) 30, 43,
Rodenticide Act (FIFRA), USA 614 129, 199, 207, 304, 310–312, 314, 315, 320,
Federal Tort Claims Act, USA 291 322, 333, 383, 387, 388, 392, 403, 407, 408,
Feedback effects and loops 2, 3, 27, 28, 33, 34, 412, 438, 444, 575, 577, 633
38, 89, 90, 93, 95, 96, 98, 99, 106, 131, 200, Geographically Weighted Regression
260, 289, 294, 329, 331, 334, 335, 336, 337, (GWR) 41,125, 431, 432, 445
342, 343, 346, 347, 352, 370, 388, 403, 421, Global Biomass Optimization Model
423, 494, 723, 724, 726, 733 (GLOBIOM) 94, 95
Feedstocks 85, 86, 88–90, 94–96, 98, 100, 102, Global Trade Analysis Project (GTAP) 97, 98,
104–106, 234, 236, 629 101, 102
Field experiments (see Experimental Grassland 97, 98, 163, 227, 229, 230, 233, 235,
economics) 256, 258, 261, 262, 269, 534, 552, 562–564,
subject INDEX 741
590, 591, 594, 595, 600, 617, 621, 623, 624, Integrated Global System Model (IGSM) 97,
629, 632, 633, 634 98, 102
Grassland Reserve Program (GRP), USA 554, Integrated Valuation of Ecosystem Services
617, 634 and Tradeoffs (InVEST) 205
Green Revolution 52–54, 57, 58, 60 Integrated World Model (IWM) 102
Greenfield 12, 652, 657, 661–663 Intergovernmental Panel on Climate Change
Greenhouse gases (GHG) 5, 37, 85, 90, 91, (IPCC) 227–230, 232, 244, 255, 256, 258,
93–95, 98–100, 107, 197, 199, 215, 226–229, 262, 263, 267, 268
234–239, 242, 244, 255, 256, 258, 260, 283, Intermediate services, ecosystem 200, 203
291, 327, 423, 491, 547, 554, 558, 562, 563,
566, 567, 624–625, 628, 629, 649, 662 Joint production, ecosystem and
household 204, 594
Habitat 1, 2, 5–7, 10, 14, 140, 163–191, 197, 199, Just-Pope production function 57
205, 209–216, 256, 258, 260–262, 265, Just v. Marinette County, 201 N.W.2d 761
272, 273, 275, 291, 292, 296, 380–383, 386, (Wisc. 1972) 674
388, 389, 394, 475, 491, 493, 547, 551, 554,
557–559, 584, 585, 590, 591, 594, 602, 616, Kelo v. New London, 545 U.S. 469 (2005) 698–
619–621, 626, 630, 632, 633, 673–675, 682, 701, 703, 707, 708, 714, 720–722
687, 726, 728, 730 Keystone Bituminous Coal Assn. v.
Habitat Conservation Plan (HCP) 165, DeBenedictus, 480 U.S. 470 (1987) 671
167–171, 551
Hawaii Housing Authority v. Midkiff, 467 U.S. Laboratory experiments (see Experimental
986 (1984) 700 economics)
Healthy Forests Reserve Program (HFRP) 554 Land assembly 7, 15, 698–722, 732
Healthy Forests Restoration Act (HFRA) 290 Land Capability Class (LCC) rating 384, 385,
Heckman selection model 122, 125–127, 130 387
Hedonic model 4, 10, 30, 35, 112, 118–131, 288, Land conservation and policy (see also Land
308, 309, 315, 327, 334–336, 353, 355, 358, use policy, Conservation easement,
364, 373, 374–376, 381, 383, 385, 414, 419, Auctions, and Best management
519, 561, 663, 727 practices) 9, 10, 13, 14, 209, 210, 214, 453,
H.F.H. Ltd. v. Superior Court, 542 P.2d 237 456–458, 476, 482, 485, 491, 496, 498, 515,
(1975) 688 547–577, 616–617, 623, 626, 628–631, 636,
Holdouts 15, 698–722 637, 639, 727, 728, 731, 733, 734
Household location choice 9, 23–25, 31, Land-labor ratio 147–151
303–311, 316, 317, 320–322, 346, 353, 433, Land management 7, 184, 211, 216, 227, 234,
Hydrological processes 227, 232, 233, 261 284, 287, 288, 408, 410, 585, 589, 593, 597,
599, 604, 616, 620, 621, 622, 625
Incentive payments and programs (see Land Land markets 1, 3, 4, 11, 12, 26, 94, 112, 118, 120,
use policy) 121, 122, 126, 129, 131, 143, 331, 340, 340,
Inductive price expectation models 416 342, 343, 345, 346, 481, 484, 498, 503, 517,
Information (see also Asymmetric 547, 698, 699, 700, 721, 723, 729, 731
information) Land preservation policy (see Zoning and
complete 595, 710–717, 719 Conservation easement)
incomplete 381, 416, 458, 703, 710, 717–719 Land price 16, 25, 31, 33, 52, 56, 94, 111–131, 240,
In-migration 26, 27, 32–34, 411 327, 341, 521, 523, 527, 531, 729
Integrated Assessment Models (IAMs) 243, Land retirement 548, 552, 553, 575, 615, 616,
244, 259, 423 626–629, 636, 639
742 subject INDEX
Land trusts 486, 518, 548, 550, 551, 629 Market failure 11, 14, 197, 218, 380, 395, 515,
Land use intensity 210, 233, 235, 256, 302, 332, 516, 548, 698, 699, 702, 703, 712, 714, 716,
333, 381, 382, 384, 386, 387, 390–396, 452, 721–723, 731, 734
454, 584, 586, 592, 595, 620, 621 Maryland Agricultural Lands Preservation
Land use patterns 2, 3, 9, 13, 14, 16, 58, 87, 88, Foundation 496, 575
94, 163, 232, 256, 259, 262, 302–304, 327, Mathematical Programming (MP) 263, 267,
340, 341, 383, 514, 534, 563, 564, 566, 725, 418, 594, 595, 630
734 Maximum lot coverage 432–444
Land use policy Maximum Likelihood Estimation (MLE)
in general 12, 28, 35, 43, 198, 202, 204, 296, method 39, 318, 360, 361, 431, 432, 445
327, 333, 338, 370, 380, 389, 395, 444, 445, Metapopulation 208–214
453, 454, 475, 515, 517, 540, 541, 636, 684, Metropolitan Area 3, 4, 28, 30–36, 42, 74–83,
719, 720 304, 305, 310, 311, 320, 322, 731
incentive-based policy 6, 7, 12, 13, 15, Migration 4, 8, 24, 26, 27, 30, 32–34, 88, 141,
186–190, 303, 381, 385, 389, 454, 475, 476, 211, 231, 259, 264, 265, 267, 269, 281, 294,
516, 548, 549, 552–556, 569, 573–575, 613, 369, 376, 409, 411, 434, 651
621, 622, 628, 636, 724, 732, 734 Mining 140, 141, 143, 153, 670, 671
regulation (see also Zoning) 6, 13–15, 34, Mitigation, climate change 6, 94, 163, 226, 227,
36, 64–65, 163–165, 171, 176, 190, 209, 233–239, 242–245, 260, 275, 620, 625, 726
210, 286, 292, 296, 297, 303, 310, 312, 322, Modeling International Relationships
339, 340, 381, 431, 435, 442–444, 464, 487, in Applied General Equilibrium
547, 548, 551, 566, 569, 572, 573, 577, 605, (MIRAGE) 97, 98, 101, 102, 104, 105
613–615, 636, 657, 659, 668–694, 728, 732 Modularity, in multiple models 103, 725
Landscape simulations 10, 380–397 Monocentric city model (see Urban Economics)
Leakage (see also Slippage) 17, 236, 237, 239, Monte Carlo method 125, 295, 387, 389, 421
243, 730 Moral hazard 294, 595, 596, 679, 683, 690, 698,
Leapfrog development 331, 342, 344, 432, 520, 721
568, 570 Model Representing Potential Objects
Least-cost-plus-loss model 289 That Appear in The Ontology of
Livestock 59, 65, 89, 92–94, 98, 107, 140, 153, Human-Environmental Actions &
215, 227–230, 234–236, 240, 241, 256, 257, Decisions (MR POTATOHEAD) 406,
282, 552, 584, 586, 590, 615–617, 621, 633, 407, 410, 420
635, 638 Mugler v. Kansas, 123 U.S. 623 (1887) 669
Local government and public finance 513– Multifunctionality and multifunctional
520, 554 agriculture 199, 585, 590, 594
Local Indicators of Spatial Association Municipal profile 9, 302–322
(LISA) 438
Locally Weighted Regression (LWR) National Agricultural Statistics Service
(see also Geographically weighted (NASS), USA 120, 265, 628, 636
regression) 41 Natural amenity (see Amenities)
Loretto v. Teleprompter, 458 U.S. 419 Nature reserves (see Reserve-Site Selection)
(1982) 669 Negligence 290, 682, 683
Lucas v. South Carolina Coastal Council, 505 Nelder-Mead algorithm 367
U.S. 1003 (1992) 672–674, 676, 677, 684 New Economic Geography (NEG) 2, 3, 24, 26,
27, 30, 31, 33, 38, 724
Marginal land 63, 89, 94, 101, 102, 104, 106, Nollan v. California Coastal Commission, 483
140–155 U.S. 825 (1987) 692, 693
subject INDEX 743
Nonmarginal change 41, 206, 207, 328, 329, Permanence, in carbon sequestration 236,
334, 335, 337, 374, 375, 733 237, 239
Nonmarket goods and services 6, 66, 168, 169, Piguovian subsidy 186
171, 190, 196–201, 206, 215, 288, 291, 329, Policy shock 93, 96, 99, 101, 102, 375
331, 335, 338, 364, 376, 381, 410, 517, 561, Pollution (see Environmental impacts)
694, 723, 732 Prairie Pothole Region of North America
Nonmarket valuation, in general 6, 197–209, (PPR) 256, 261, 262, 264, 265, 267–269,
291, 337, 338, 414, 561, 602, 732 271–273, 275, 563
Nonpoint source pollution (see also Priority Funding Area (PFA), Maryland,
Environmental impacts) 485, 490, 492 USA 454–478, 569, 571
Noxious use, in regulatory takings 669–672, Private property 165, 166, 290, 294, 668, 670,
676, 677, 683, 684 698
Nuisance exception 672–674, 677, 684 Propensity Score Matching (PSM) 11, 40, 126,
127, 130, 320, 321, 452–476, 572, 573, 601,
Omitted variable bias 37, 41, 121, 122, 127–129, 631
354, 373, 374, 729 Property rights 12, 143, 236, 239, 497, 518, 520,
One-dimensional root-finding problem 371, 525, 548, 550, 554, 556, 573, 585, 586, 616,
372 617, 636, 668, 670, 671, 674–678, 698–701,
Open space 14, 25, 27, 28, 34, 88, 120, 197, 210, 730
303, 308–316, 330, 336–338, 343, 344, Property taxes 312, 316, 318–321, 354, 519–522,
354–356, 363, 370, 372–374, 432–436, 444, 526–529, 532, 534, 537–540, 556, 686
452, 455, 513–542, 561, 569, 572, 574, 576, Public goods 6, 15, 163, 165, 190, 216, 239, 287,
612, 617, 649, 676, 693, 727 303, 309, 310, 312, 322, 335, 336, 352, 354, 357,
Open-city model 14, 434, 521, 523–525, 528, 372, 515, 516, 520, 526, 539, 540, 547, 583,
540, 541, 727 585, 594, 595, 597, 599, 603, 620, 622, 624,
Organization for Economic Co-operation and 675, 685, 686, 699, 712–714, 721, 727, 733
Development (OECD) 55, 96, 585 Public lands 34, 286, 287, 288, 291, 294, 387,
Ownership 64, 88, 120, 287, 294, 380, 381, 387, 513, 514, 549, 550, 558, 559
388, 395, 396, 498–501, 507–510, 549, 555, Public Trust Doctrine 674
604, 605, 616, 674 Public use, in Constitutional law 679, 680,
683, 685, 686, 698–701, 720, 722
Palazzolo v. Rhode Island, 533 U.S. 606 Purchase of Agricultural Conservation
(2001) 673 Easements (PACE) (See Conservation
Park land 27, 308–12, 315–16,319, 321, 374, easement)
439–441, 466, 513–515, 539, 550, 560–562, Purchase of Development Rights (PDR) (See
651, 660 Conservation easement)
Partial equilibrium models 17, 86, 92, 93–96, Pure Characteristics Model (PCM) 337, 353,
99, 207, 338, 346, 369, 501, 733 357, 358, 363–372
Participants, experiment 11, 216–218, 481–504
Payment for ecosystem services 166, 183, 184, Quality of life 24, 27–30, 32–36, 81, 302, 322,
190, 214–218, 492, 552–554, 565, 583, 603, 514, 560, 587, 620
604, 613, 616, 627, 640 Quasi-Experimental (QE) approaches 2, 4, 11,
Penn Central Transportation Co. v. City of 25, 38–40, 129–131, 330, 353, 373–375, 452,
New York, 438 U.S. 104 (1978) 671–673, 453, 456, 457, 475, 627, 633, 727
694
Pennsylvania Coal v. Mahon, 260 U.S. 393 Ranchland and ranching 140–143, 150, 153,
(1922) 670, 683, 691, 693 154, 554
744 subject INDEX
Random utility model 338, 353, 358–363, 365, configuration 184, 286, 292, 306, 332, 380,
368–374, 381, 385–388 594, 602
Real options theory 332, 333 context 307–309
Redevelopment 1, 15, 168, 306, 648–664, dependence 4, 112, 122–125, 129, 184–190,
698–722, 732 431, 432, 436, 444, 577
Reduced-form model 9, 10, 16, 38, 93, 131, 319, externalities 6, 281, 286–288, 294, 397, 414
321, 327–345, 353, 373, 376, 385, 390, 393, heterogeneity 4, 94, 95, 123–125, 129, 275,
476, 527, 528, 531, 534, 725, 726 306, 332, 336, 340, 342, 344, 423, 602
Reduction in emissions from deforestation interactions 123, 125, 295, 381, 444
and degradation (REDD) 234 Spatial models
Regional economic development 3, 23–43 spatial econometric modeling 123–125, 130,
Regression discontinuity 130, 453, 456, 728 397, 430–446, 727
Regulatory takings 13, 15, 668–694, 698, 699, spatial equilibrium model 27, 28, 340–342,
732 344, 346, 726
Remediation 15, 648, 649, 653–664 spatial error model 123–125, 129
Renewable Energy Directive (RED), European spatial general linear model (GLM) 431,
Union 100, 104, 105 445
Renewable Fuel Standard (RFS) 100, 104, 106, spatial lag model 123, 125, 130
127, 629 spatial panel data model 446
Reserve-Site Selection (RSS) 164, 176, 177–183 spatial probit model 432, 437, 439, 443, 445
Return-on-Investment (ROI) 164, 176, Spatial simulation models (see also
179–183 Agent-based models) 332, 340, 341, 346,
Reverse auctions (see Auctions) 407
Ricardo model and Ricardian rent 3, 87, 146, Species Area Relationship (SAR) 177, 178
212, 410 Spillover 29, 36, 40, 124, 130, 308, 330, 337, 431,
Riparian land management 260, 576 515, 519, 523, 524, 541, 577, 676, 691, 727
Risk 3, 6, 53, 57–69, 88, 113, 115, 116, 163, Sprawl, urban and suburban 1, 3, 8, 12, 23,
167, 172, 179, 227, 231, 240, 257, 259, 281, 32, 34–37, 42, 303, 309, 327, 332, 333, 343,
286–297, 409, 413, 424, 433, 490, 596, 604, 344, 380, 402, 432, 433, 437, 438, 441, 443,
624, 654, 659, 663, 685–694, 709, 710, 714, 444, 454, 568, 569, 572, 573, 576, 649,
717, 718 661–663
Ruckelshaus v. Monsanto Co., 467 U.S. 986 Stacking 627, 730
(1984) 700 Stated Preference (SP) 198, 203, 210, 291, 602,
Rural Development Policy (RDP), European 630
Union 588, 619, 620, 622–624, 632, 637 Strickley v. Highland Boy Gold Mining
Company, 200 U.S. 527 (1906) 700
Salient incentives 481, 482, 487, 489 Structural models 4, 9, 10, 16, 25, 28, 38, 41,
Salt-and-pepper effect 395–397 42, 94, 115, 116, 128–131, 317–319, 327–347,
Slippage (see also Leakage) 17, 626, 730 352–376, 384, 407, 418, 422, 724–726, 728
Smart growth 11, 37, 454, 463–465, 476, 477, Suburbanization 7, 8, 79, 82, 83, 303–306, 322,
569, 575, 661, 662 327
Smart Growth Areas Act, Maryland, Survey on Agricultural Production Methods
USA 463, 476 (SAPM), European Union 637, 638
Soil conservation 552, 565, 622, 640, 666 Sustainability and sustainable
Soil Survey Geographic Database development 35, 37, 66, 139–155, 239, 540,
(SURGO) 466 553, 554, 587, 620–622, 624, 649, 661, 662,
Spatial analysis 664, 733
subject INDEX 745
Systematic Conservation Planning (SCP) 164, Vertical integration 65–66, 726

176, 383 Vertical preference structure 360, 365
Village of Euclid v. Ambler Realty, 272 U.S. 365
Tahoe-Sierra Preservation Council v. Tahoe (1926) 670
Regional Planning Agency, 535 U.S. 302 Voluntary cleanup programs (VCP) 658–664,
(2002) 673 731
Takings clause (see Eminent domain and Von Thunen model 3, 26, 28, 58, 86, 87, 112,
Regulatory takings) 410, 419, 430, 577
Technology adoption and diffusion 3, 52–69, Vulnerability, climate change 6, 226–227, 233,
87, 88, 415, 726, 731 244, 245,
The Nature Conservancy (TNC) 13, 164
Tiebout model 23, 32, 36, 303, 307–309, 314, Water Framework Directive, European
335, 352, 357, 515 Union 618, 620, 621, 635
Tillage 233, 235, 260, 594, 618, 627, 631, 638 Water filtration services of land 203
Timber and timberland 34, 140, 145, 153–155, Water pollution (see Environmental impacts)
164, 231, 237, 286, 287, 292, 293, 381, 675, Water-holding capacity of soil 54, 57
687 Welfare biases 203
Total Factor Productivity (TFP) 40 Wetland Reserve Enhancement Program
Tradable permits (see Cap and trade) (WREP) 615
Transaction cost 15, 30, 60, 113, 115, 116, 167, Wetland Reserve Program (WRP) 13, 553, 557,
236, 238, 239, 306, 493, 498, 515, 589, 562–565, 615, 616, 629, 634, 636
597–600, 602, 603, 630–632, 639, 640, Wetland and Waterfowl Model (WWM) 262–
677, 691, 702, 705, 711, 717, 729 265, 269–271
Tyranny of distance 30 Wetlands 34, 140, 196, 199, 207–210, 212–214,
256, 260–265, 267–271, 275, 491, 532, 533,
United Nations Framework Convention on 549, 552, 553, 562–565, 591, 615–618, 634,
Climate Change 234, 242 638, 673–675, 687, 690, 691
Urban economics Wildfire (see Fire)
in general 2, 23, 24, 86, 118, 303, 304, 309, Wildland 285
322, 341, 343, 541, 734 Wildland-Urban Interface (WUI) 285, 287,
Alonso's urban land market theory 31, 303, 288, 290, 294
430 Wildlife
Monocentric model 31, 303–306, 335, 341, in general 34, 163–191, 214, 282, 283, 285,
541 287, 292, 591, 594, 621
Urban Growth Boundary (UGB) 454, 551, 567, conservation 6, 163–165, 175, 176, 183–191,
569, 570 540, 550, 551, 563, 564, 585, 590
Urban land 3, 27, 35, 335, 381, 395–397, 404, damage 120
408, 409, 413, 430, 552, 556, 662, 699, 721, habitat 5, 14, 163, 183, 190, 197, 256, 260, 265,
730 275, 292, 491, 551, 565, 616, 626, 634, 638,
Urban revitalization (see also Redevelopment) 673, 728
12, 303, 463, 477, 650, 652, 655, population 163, 164, 200, 265, 556
700, 714 viewing and recreation 120, 260
Urban sprawl (see Sprawl) Wildlife Habitat Incentives Program,
Urbanization 3, 5, 7–9, 13, 14, 27, 283, 372, 383, USA 164, 616, 634
397, 513 Willingness to accept (WTA) (see also
Urban-rural boundary and growth 435, 519, Valuation) 7, 164, 165, 184–187, 189, 190,
725, 733 343, 386, 414, 415, 493, 495, 500, 630, 705
746 subject INDEX
Willingness to pay (WTP) (see also Valuation) Zoning (see also Land use policies) 4,
33, 82, 119, 131, 216, 217, 260, 261, 336, 338, 12–14, 23, 24, 32, 34, 88, 120, 125, 126, 130,
339, 343, 344, 352, 356, 369, 373–375, 414, 209, 212, 213, 303, 312, 314, 329, 331, 332,
484, 521, 547, 560–562, 602, 630 343, 368, 381, 385, 392–395, 408, 413,
Working land conservation 548, 549, 552–554, 434, 436, 439, 440, 444, 454, 457, 466,
557, 575, 615, 616, 625–630, 639 513, 516–528, 540–542, 551, 567–572,
576, 615, 668–670, 672, 688, 691,
Zero-truncated Poisson distribution 391 724, 730

The Oxford Handbook of Land Economics

Uploaded by

Copyright:

Available Formats

The Oxford Handbook of Land Economics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Oxford Handbook of Land Economics

Uploaded by

Copyright:

Available Formats

T H E OX F OR D HA N DB O OK OF

Oxford is a registered trademark of Oxford University Press

Published in the United States of America by

© Oxford University Press 2014

All rights reserved. No part of this publication may be reproduced, stored in

You must not circulate this work in any other form

Library of Congress Cataloging-in-Publication Data

I ntroduction: Land as an Integrating Theme in Economics 1

17. Spatial Econometric Modeling of Land Use Change 430

18. Using Quasi-Experimental Methods to Evaluate Land

19. Applying Experiments to Land Economics: Public Information and

20. Open Space Preservation: Direct Controls and Fiscal Incentives 513

21. Land Conservation in the United States 547

22. European Agri-Environmental Policy: The Conservation and

23. Agri-Environmental Policies: A Comparison of US and EU

24. Stigmatized Sites and Urban Brownfield Redevelopment 648

25. Regulatory Takings 668

26. Eminent Domain and the Land Assembly Problem 698

27. Future Research Directions in Land Economics 723

Subject Index 737

1 Karl Polanyi, The Great Transformation, Boston: Beacon Press, 2001.

The inconvenience of land as a fictitious commodity arises because economic mod-

Joshua M. Duke, Newark, Delaware

Richard M. Adams is Professor Emeritus in the Department of Applied Economics at

Ekaterina Gnedenko is Lecturer in the Department of Economics at Tufts University.

Jianhong E. Mu is Postdoc Scholar in the Department of Applied Economics at Oregon

Marcella Veronesi is Assistant Professor in the Department of Economics at the

1. Land Markets and Welfare

2. Determinants and Drivers of Land Use

Numerous societal changes, including economic development, technological progress,

3. Environmental and Socioeconomic

Land use changes, such as deforestation, urbanization, intensification of agriculture,

3.1 The Land Use and Environment Nexus

3.2 Socioeconomic Impacts of Land Use Change

Urbanization also presents opportunities to farmers. The emergence of a new cus-

4. Methodological Developments

The increasing importance of the environmental and socioeconomic issues associ-

4.1 Spatial Econometric Methods in Land Use

4.2 Simulation Methods in Land Use

4.3 Experimental Methods in Land Use

5. The Economics of Land Use

the effectiveness of each approach. Hodge, in Chapter 22, conducts an assessment of

6. Future Research Directions

Bromley, D. W. 1989. Economic interests and institutions: The conceptual foundations of public

FIGURE 1.1 Model of regional economic development with land use.

centers. Economic activity across space is strongly affected by access to agglomeration

1. Land Use in Regional Economic

Computable general equilibrium (CGE) models incorporate factor supply con-

2. Economic Development: Distance

2.1 Distance and Regional Economic Development

2 See Mulligan (1984) for a review of the CPT literature.

2.2 Land Economics and Intrametropolitan-Area Economic

3. Land Use, Quality of Life, and

4. Fiscal Federalism, Land Use, and

So-called “smart-growth” policies have been widely enacted to promote sustainable

5. Empirically Assessing Economic

(IV)/quasi-experimental approaches; and structural models.7 CGE models have already

in which e is the residual. A key assumption is Cov(e, X)= 0, or there is no endogeneity

Greenstone, Hornbeck, and Moretti (2010) (GHM) is an example of the advantages

Another econometric problem is unobservable variables. This is especially problem-