Developers, Wall Street, and the Taxmen:

A Theory of Real Estate Development∗

Marta Faias Jaime Luque

Universidade NOVA de Lisboa ESCP Europe

Abstract

We build a general equilibrium model of commercial real estate (CRE) development (aka,

CRE asset creation), for an economy with several jurisdictions, and segmented commercial

good and equity markets. The CRE assets’ capital structures and cash flows, the prices of

equity, non-recourse mortgage debt and commercial goods, and the jurisdictions’ property

taxes and land use policies are all endogenous. We show that an equilibrium exists for this

economy. Our equilibrium characterization shows that CRE construction booms in jurisdic-

tions where developers hold a larger share of debt and equity (possibly, due to lower mortgage

default risk), and contracts otherwise. Regional negative shocks to developers’ funding debt

capacity result in leverage shifting toward other jurisdictions. Leverage also increases in ju-

risdictions that pursue a high property tax fiscal policy. Finally, we highlight how global real

estate markets can influence local jurisdictions’ fiscal policies.

JEL Classification: D52, D53, H73, R13, R33, R38

Keywords: commercial real estate (CRE) development; global real estate investors; local

households; capital structure; property taxes; CRE asset selection

∗
We would like to thank for comments and suggestions Anthony DeFusco, Guilles Duranton, Douglas Gale, Mike
Harrigan, Tom McCahill, Antonio Mello, Salvador Ortigueira, Erwan Quintin, Peter Ritz, Tim Riddiough, Manuel
Santos, Sheridan Titman, Matt Watcher, Abdullah Yavas, Jiro Yoshida, and the participants at the conferences of
SAET, EWGET, PET, the V Workshop on Equilibrium Analysis, the ASSA-AREUEA annual meetings at Chicago
(2017) and Atlanta (2019), and seminars at the University of Wisconsin-Madison and University of Miami. Email
addresses: [email protected] (M. Faias) and [email protected] (J. Luque). Corresponding author: Jaime Luque;
Calle Arroyofresno 1, 28035, Madrid, Spain; Phone: (+34) 911719017.

1
1 Introduction

Commercial real estate (CRE) development plays an integral role in the economy. It creates
spaces for jobs and provides a permanent source of revenue to investors. In the U.S. alone, there
is about $5 trillion worth of commercial real estate. The construction of these assets contributes
approximately 5% percent of the U.S. Gross Domestic Product (GDP), which in 2016 terms
amounts to a contribution of $861 billion to the U.S.’s economic output. In terms of employment,
commercial real estate development supports 6.25 million American jobs. But not all cities expe-
rience the same patterns in construction. Some cities grow and others shrink. Developers’ access
to debt and equity capital is crucial for a city to succeed.
In the last decades, regional and local commerical real estate markets have rapidly changed in
part because of the dominant presence of Wall Street financial intermediaries in real estate devel-
opment deals. The most common types of real estate investors are Real Estate Investment Trusts
(REITs), pension funds, insurance companies, and, to some extent, foreign investors. Only the
REIT sector owns more than $3 trillion in gross assets, supports an estimated 1.8 million full-time
workers, and contributes more than $56 billion in new construction and capital expenditures.1
Since the seminal contributions of Mills (1993) and Dinsmore (1998) on the prevalence of
REITs and other types of real estate investors in the development and acquisition of commercial
real estate, there has been a spirited debate on the ability of cities to attract capital investments.
Economic strains such as Brexit and the cooling of the Chinese economy are now driving inter-
national capital into commercial real estate in the U.S. in search for higher returns and stability.
Not only traditional gateway cities, such as New York, Los Angeles and Miami, are the focus
of investors’ attention, but also other secondary markets such as Denver, Phoenix and Nashville.
Local authorities are aware of the importance for developers to raise debt and equity, and compete
accordingly to make their cities attractive destinations for investors.
However, the presence of Wall Street – rather than Main Street investors – puts in question
1
REITs are corporate entities that can be privately or publicly held. In the United States, there are more than 200
stock exchange-listed REITs with a total equity market capitalization of approximately $1 trillion.

2
the role that global real estate investors have in urban and metropolitan economic development.
Real estate investors have a clear self-interest in reducing some of the costs decided by local
municipalities, including local property taxes, zoning and land use controls, and utility rates.
Depending on the maginitude of the real estate investors’ portfolios, stakeholders maybe be able
to influence whether and how a local community deals with a particular issue. An alarming
example is the case of Texas, where REITs own more than 29,600 properties, including office
buildings, luxury hotels, urban apartments, regional and strip malls, storage, and warehouse.
These important practical questions deserve a careful economic analysis and a general equi-
librium model happens to be the natural way to analyze these issues. The financing aspects of
a real estate development project cannot be analyzed without taking into consideration the local
fiscal and land use policies, nor can they be analyzed without a proper evaluation of the city’s
economic fundamentals. Local households are the final buyers of the goods produced and sold
by the commercial real estate assets and their prices determine the cash flows used to evaluate
the equity price of a commercial real estate asset. Neither we can ignore a developer’s access to
funding in the debt market because debt collateral constraints may determine a developer’s need
for equity. Also, a developer’s choice of capital structure has general equilibrium implications
in an economy with segmented equity markets. A more leveraged development in a jurisdiction
may be the result of real estate equity funds flowing to other more attractive jurisdictions. More-
over, because interests on debt are tax exempt but not the equity dividends, a local jurisdiction
fiscal policy (in the form of property taxation) may influence not only the capital structure of a
local development project, but also the capital structures of other projects outside the jurisdiction
(a general equilibrium effect).
We propose a model of real estate development that captures these considerations. We focus
on equity investments and commercial (income-producing) real estate properties, such as office
buildings, industrial space (e.g., heavy manufacturing, light assembly, and warehouses), retail
space (e.g., strip centers and regional malls), multifamily buildings (e.g., student housing and
mid-rise apartments), land (e.g., greenfield land), and hotels. We consider an economy with mul-

3
tiple jurisdictions and pay special attention to the financing aspects of the real estate development
deals, as well as their interaction with the local governments’ fiscal and land use policies. For
each development project in each jurisdiction, the capital structure, composed by the developer’s
debt, the developer’s (or entrepreneurial partner’s) equity, and the investors’ (or capital partners’)
equity, is endogenously determined in our model and driven by factors such as the developer’s
funding capacity, the jurisdiction’s property taxes, and the performance of the income-producing
real estate asset. Other important elements such as real estate equity cash flows and development
construction costs are also endogenous in our model.
We build our economy in a two-period setting. Uncertainty enters into the model because we
consider several states of nature in the second period. Our notion of equilibrium is in the tradition
of general competitive equilibrium models with incomplete financial markets and restricted par-
ticipation, where agents do not choose their location (jurisdiction), but are exogenously assigned
to jurisdictions. There are multiple jurisdictions and each jurisdiction has both local households
and a representative local developer. Global real estate equity investors can invest in all jurisdic-
tions, an important feature to model interdependent segmented commercial equity markets.
Because CRE development projects are capital intensive, developers need to obtain financing
by issuing non-recourse mortgage debt in a global market and selling equity to investors in the
first period. Developers then use raised the capital to buy construction inputs, including land.
Given the jurisdiction’s choice of the type of a CRE that a developer is allowed to construct,
inputs are transformed into a CRE asset using a Cobb-Douglas production technology. The CRE
asset produces (multiple) commercial goods in the second period, possibly in different amounts
across states of nature. Local households purchase these commercial goods. This interpretation
of commercial goods produced by CRE assets and sold to local households is in the spirit of
Debreu (1959)’s classical book Theory of Value. For example, for the case of hotels, the con-
sumption of a “hotel good” should be interpreted as the purchase of a certain number of nights
in a hotel room with specific characteristics in a given location. Similar interpretations apply
for the consumption of office space or student housing. Because households may have different

4
preferences for the consumption of commercial goods, the price of similar commercial goods can
differ across jurisdictions. This in turn implies that similar CRE assets in different jurisdictions
may generate different cash flows in equilibrium.
Given the jurisdictions’ fiscal and land use policies, we identify mild conditions under which
a competitive equilibrium exists for our economy. This result is not trivial because of the ad-
ditional economic structure that our model requires in order to accommodate the development
(aka creation) of CRE assets and the segmentation of commercial good and CRE equity mar-
kets. One particular difficulty is to guarantee that a developer’s budget set correspondence takes
convex values. This is particularly challenging in a setting where the market value of a CRE
asset depends on the developer’s choice of materials and the equity return in the second period
depends on the (endogenous) type of CRE asset chosen in the first period. Another subtlety of
our existence proof is the property of lower semicontinuity of the budget correspondence (the
standard approach of embeding all prices in the simplex does not guarantee the existence of an
interior point with segmented real estate equity and commercial good markets). We circumvent
this problem by finding an endogenous upper bound for real estate equity prices (previous results
in the literature of market segmentation with a fixed point theory approach are not useful in our
setting).
Having established this general result, we provide a characterization of equilibrium in a model
with only two jurisdictions and two states of nature in the second period. This setting allows us to
understand the impact of several economic shocks and public policies on the feasibility of CRE
development projects and capital structures.
First, we analyze the determinants of default risk on the valuation of CRE equity and mortgage
debt. We show that a higher risk of mortgage debt default decreases the equilibrium debt market
price. The expected default risk is also the main driver of the shadow value of the equity collateral
constraint that a developer must satisfy in order to get a non-recourse loan. This shadow value
in turn increases the equity price for a CRE asset facing a higher mortgage default risk. Roughly
speaking, mortgage default increases market pressure on equity through the corresponding equity

5
price.
Second, we investigate the determinants of differences in construction patters between juris-
dictions. We find that a higher mortgage default loss in a jurisdiction induces real estate investors
to rebalance their equity holdings toward the high default risk jurisdiction, crowding out devel-
oper’s equity in the high default risk jurisdiction and expanding the developer’s equity in the low
default risk jurisdiction. As a result, construction booms in the low default risk jurisdiction and
contracts in the high default risk jurisdiction.
Third, we show that global real estate equity investors increase their equity exposure in those
CRE projects whose developers become more debt constrained. In addition, we find an equity-
debt substitution effect not only within the capital structure of a real esatate development project,
but also across jurisdictions: levarage in the real estate construction sector shifts from one ju-
risdiction to another when access to the debt market changes for developers in one of the two
jurisdictions.
Fourth, we consider a situation where a jurisdiction decides to increase its property tax rate
to improve the provision of local public goods. Given the differences in tax treatment between
equity and debt, the developer in the higher property tax jurisdiction increases leverage. Because
the price of equity becomes relatively less expensive than debt, the developer in the low tax
jurisdiction rebalances its portfolio toward equity, so leverage decreases in that jurisdiction.
This last result ignores the possibility that jurisdictions strategically choose property taxes
in a non-cooperative way to increase profits. To understand the implications of dropping this
assumption, we extend our two-period model to incorporate a Nash game in a pre-stage of the
economy where strategic jurisdiction authorities choose their property taxes and the types of CRE
developments allowed in their jurisdictions. These decisions influence the type of competitive
equilibrium (e.g., the composition of the different capital structures and the commercial good and
equity prices). High property taxes are not always optimal for jurisdictions that seek to maximize
profits. Low taxes may generate higher profits by drawing more real estate equity investments to
the jurisdictions, increasing equity tax revenues as a result. We provide an example that illustrates

6
this, which resembles the “race to the bottom” concept in the literature of financial competition
among political jurisdictions (Cary 1974, Drezner 2001, and Carruthers and Lamoreaux 2016).
Our result extends to commercial real estate equity investments and highlights the impact that
global real estate markets have on local jurisdictions. Rather than having fiscal policy driving
financial markets, our economy captures the dominant role of Wall Street on the fiscal policies of
“small” municipalities.
Further contribution to the literature
The question of what determines the capital structure of a firm has been extensively re-
searched in the last decades since the seminal work of Modigliani and Miller (1958, 1963).
Significant contributions are Miller (1977) on the impact of taxes, Jensen and Meckling (1976)
on agency costs, Kraus and Litzenberger (1973) on bankruptcy costs, and DeAngelo and Masulis
(1980) on non-debt tax shields. Titman and Wessels (1988) analyze the explanatory power of
these and other related theories of optimal capital structure. The literature has expanded and re-
cently the focus has been on financial intermediation and the implications of certain regulations
on the banks’ optimal choice of capital structure (Gale 2004, DeAngelo and Stulz 2015, Gornall
and Strebulaev 2015, Allen, Carletti and Marquez 2015, Gale and Gottardi 2015, and Amaral,
Corbae, and Quintin 2017). However, capital structure theory is not as well developed for real
estate development and investments.
While there is some work that has highlighted the importance of capital constraints, taxes on
equity dividends, and default risk on the capital structure of a real estate asset (Gau and Wang
1990, Giambona, Mello, and Riddiough 2016), it is fair to say that most research is empirical.
Existing theoretical papers rely either on a “real options” pricing partial equilibrium model (see
e.g. Titman 1985, Capozza and Helsley 1990, Capozza and Sick 1990, Williams 1991, Childs,
Riddiough, and Triantis 1996, and Grenadier 1995a,b) or on a general equilibrium model with
entrepreneurial developers and a housing/land market (see e.g. Henderson 1974, Helsley and
Strange 1997, Konishi 2013, and Anglin, Dale-Johnson, Gao and Zhu 2014).2 Our paper pro-
2
Henderson (1974) studies the evolution of cities when intracity Marshallian externalities in production are
present and discusses the role of entrepreneurship. Helsley and Strange (1997) consider the case of (city) devel-

7
poses a different approach than these two literatures by considering a general equilibrium model
that focuses on the financing aspects of commercial real estate development when jurisdictions
compete to attract real estate equity investments.
We incorporate and endogenize for the first time and in a consistent way several important
financial variables of real estate development and investment decisions, such as the developer’s
capital structure – composed by debt, common equity of the capital partners, and common equity
of the developer partner –, the property taxes chosen by jurisdiction managers, the cash flows of
the real estate asset, and the types of development projects chosen by developers and jurisdiction
managers. Our results that illustrate an equity-debt substitution effect not only within the capital
structure of a real esatate development project, but also across jurisdictions contributes to the
recent theoretical literature on risk shifting and asset substitution. See the seminar paper of
Stiglitz and Weiss (1981), and Martinez-Miera and Repullo (2010) for a recent contribution.
In addition to having developers choosing their optimal capital structure, our paper also de-
parts from the literature in that it captures how global capital markets influence local fiscal pol-
icy. Our model exploits the trade-off between equity taxes, default risk on debt and leverage, but
leaves aside other considerations such as agency costs.
Our equilibrium model is also related to the literature of competitive market economies with
incomplete markets, developed by Diamond (1967), Radner (1974), and Grossman and Hart
(1979), among others. The main departure from this literature is our focus on income-producing
real estate assets. This particular type of asset generates cash flows that are dependent on the
market price of the goods sold in local segmented markets. Thus, real estate cash flows are en-
dogenous because commercial good prices are determined by market clearing given the demand
of households at the jurisdiction level and the construction of CRE assets that produce and sell
these commercial goods. Even when the size and type of real estate assets is the same in two
jurisdictions, cash flows may differ if the fundamentals of their respective local economies are
opers that provide local public goods with limited power and an explicit geographical structure of the city. Konishi
(2013) considers an economy with a large number of atomless land developers who can enter the market freely in an
idealized version of Tiebout (1956).

8
different.
In our setting, developers are subject to collateralized debt constraints and markets can be
incomplete. In addition, we depart from the standard one good economy because we allow for
multiple construction inputs in the development phase and also for multiple consumption goods
that are sold in the jurisdiction where the commercial real estate asset was constructed. Our
setting comes at a cost because with multiple goods and incomplete markets, we are not able
to establish the efficiency property of a decentralized competitive economy (Geanakoplos and
Polemarchakis 1986). In this sense, our work departs from Gale and Gottardi (2017), who follow
a similar approach than Makowski (1983) and Hart (1979) by considering an alternative general
equilibrium model of financial intermediation with one good and complete markets where the
efficiency property holds.
Our work is also closely related to the the literature on financial innovation – also referred to
as the security design literature – pioneered by Allen and Gale (1991) (see Allen and Gale 1994
and Duffie and Rahi 1995 for surveys). This literature has focused on the design of financial
securities, such as bonds, stocks, mortgages, and mortgage-backed securities, and the relationship
with economic development.3 Up to our knowledge, we are the first paper to uncover the specific
case of real estate development (aka creation) in a model with segmented equity and commercial
good markets.
Our paper also relates to Rahi and Zigrand (2009), who study financial innovation and wel-
fare in a two-stage equilibrium model with segmented markets. These markets (“jurisdictions”
in our terminology) are characterized by differential marginal valuations and can list the same
assets. Investors cannot trade assets in more than one market (jurisdiction). The ability to trade
across markets is left for arbitrageurs, who turn out to be the issuers of assets in the first stage
of their model. However, as Rahi and Zigrand (2009) recognize, this interpretation of issuance
and implied listing is rather specific (when a company lists its shares in an jurisdiction, arbi-
trage possibilities are not the main reason to go public). In our theory of real estate development
3
See, for example, Amaral, Corbae, and Quintin (2017) for a recent study of the relationship between financial
engineering (repackaging) and development.

9
and investments, we depart from these assumptions and allow global investors to buy real estate
equity in multiple jurisdictions. In addition, besides the obvious difference in motivations, our
model also departs from Rahi and Zigrand (2009) in that financial innovation is not driven by
arbitrageurs, but by the sequential actions of local jurisdiction authorities and developers. The
former choose the type of commercial real estate asset that developers are allowed to construct
in their respective jurisdictions (e.g., hotel, shopping mall, etc). The latter choose, for a given
type of real estate asset, the combination of construction inputs that determines the size of the
development.
The remainder of this paper is structured as follows. In Section 2, we present the model, the
equilibrium concept, and the result of equilibrium existence. Section 3 proposes an extension
to the model in which the jurisdiction’s fiscal and land use policies become endogenous vari-
ables. Section 4 builds a simplified economy with two jurisdictions and provides an equilibrium
characterization analysis. Section 5 concludes. The Appendix is devoted to the proofs.

2 The model

Consider a multi-jurisdictional economy with K > 0 jurisdictions. In each jurisdiction k, there

is a representative local developer dk that finances new developments by issuing debt and equity.
In addition, there are Hk ≥ 1 households, whose consumption of commercial goods is restricted
to jurisdiction k.4 We denote the set of households in jurisdiction k by Hk = {1, ..., hk , ..., Hk }.
The set of developers in the economy is D = {d1 , ..., dk , ..., dK }. For simplicity, we ignore banks
as financial intermediaries and assume that households buy debt directly from developers. In
addition, in this economy there are I ≥ 1 global real estate investors that can buy equity in all
new developments across jurisdictions. The set of investors is I = {1, ..., i, ..., I}. We write
4
Extending the model to allow for multiple developers in each jurisdiction is feasible; all that is needed is to
rewrite the market clearing equations of debt and equity to accommodate the additional CRE assets of the same
type constructed by the developers in a jurisdiction. However, this extension would complicate the analysis without
adding much additional economic insight, as it would introduce sharing rules for investors’ equity investments among
the different CRE assets within the same jurisdiction. Another extension could be to consider global developers, in
which case risk sharing across projects in different jurisdictions would be a possibility.

10
a ∈ A to denote an agent independently of its type, where A = I ∪ D ∪ {Hk }K
k=1 . Finally, we

write K = {1, ..., K} to denote the set of all jurisdictions.

Our focus is not on where agents choose to live, but on the implications of allowing global
real estate investors to do business in multiple jurisdictions. In this sense, our approach departs
from the Tiebout literature where consumers “vote with their feet” where to live and work (see
e.g. Tiebout 1956, Konishi 2008, and Luque 2013). Thus, we assume that the distribution of
agents into jurisdictions is exogenously given and satisfies the following conditions.5
In this economy, there are two periods, t = 1, 2, and S states of nature in the second period.
S denotes the set of states of the second period. Because we allow for the construction of CRE
assets, the space of commodities changes between periods 1 and 2. There is a consumption good,
indexed by l = 0 and called the “numeraire”. This good can be any good that facilitates trade,
e.g., cash. We assume that all agents can trade the numeraire good in the global market in both
periods. In addition, in period 1, there are also L1 inputs used for construction. We denote
the set of construction inputs by L1 = {1, ..., L1 }. We denote an agent a’s endowments of the
numeraire good and construction inputs in the first period by ω a1 ∈ R1+L1
++ . The set of construction

inputs L1 includes the land available for new development in each jurisdiction.6 For example, if
input l ∈ L1 is land in jurisdiction k, then the total amount of land available for development in
jurisdiction k is a∈A ω al1 . Because ω al1 < ∞ for all a ∈ A and all l ∈ L1 , this upper bound
P

limits the supply of land in each jurisdiction.

In the second period, agents are endowed with the numeraire good and each CRE asset pro-
duces L2 commercial goods, with corresponding set L2 = {1, 2, ..., L2 }. We denote agent a’s
endowment of the numeraire good at state s of the second period by ω a0 (s) ∈ R++ .
We refer to the numeraire, construction inputs, and commercial goods as “commodities”.
The vector of commodities purchased by an agent a is xa = (xa1 , (xa (s), s = 1, ..., S)) ∈
S(1+KL2 )
R1+L
+
1
×R+ . Notice that the consumption possibilities in the second period at a state s ∈ S
5
See Berliant and ten Raa (1994) for a discussion of the different methodologies in the fields of regional science,
regional economics, and urban economics.
6
For simplicity, we do not explicitly model the production of construction materials (e.g., the process of cutting
trees and using timber pieces to construct the frames of large structures) and take their supply as given.

11
are 1 + KL2 goods (the numeraire good and KL2 commercial goods). This is because there are
K jurisdictions in the economy and we assume that a jurisdiction’s CRE asset produces L2 com-
mercial goods. We introduce, however, restrictions on the consumption of some of these goods.
In particular, when an agent is a developer or a household in a jurisdiction k, the consumption of
a commercial good produced in another jurisdiction is zero, i.e., xalk0 (s) = 0 if a ∈ {dk } ∪ Hk
and l ∈ L2 (k 0 ) with k 0 6= k. Investors belong to all jurisdictions, so they are not restricted to
consume commercial goods in different jurisdictions.
We find convenient to denote the vector of all individual commodity purchases in the economy
by xA = (xa : a ∈ A). The vector of commodity prices is p = (p1 , (p(s), s = 1, ..., S)) ∈
S(1+KL2 )
R1+L
+
1
× R+ . We refer to the price of the numeraire good in the first period by p01 . For the
second period, we denote the price of a commercial good l ∈ L2 sold to households of jurisdiction
k at state s by plk (s). We write the price of the numeraire good at state s as p0 (s).
Agents derive utility from the consumption of the numeraire and the commercial goods avail-
able in the juridictions where they belong to. The consumption of construction inputs does not
(1+L1 )+S(1+KL2 )
provide agents any (direct) utility. We denote an agent a’s utility function by ua : R+ →
R.

Assumption 1: (i) For any agent a ∈ A, the utility function ua is continuous and strongly
quasi-concave;7 (ii) For all k ∈ K and for any agent a ∈ {dk } ∪ Hk , ua (x1 , (x(s), s =
1, ..., S)) = ua (x01 , (xk (s), s = 1, ..., S)), where xk (s) ∈ R1+L2 is the bundle of commercial
goods available at jurisdiction k, and, for any investor i ∈ I, ui (x1 , (x(s), s = 1, ..., S)) =
ui (x01 , (x(s), s = 1, ..., S)); that is, developers, households, and investors do not assign any
utility to the construction inputs; moreover, developers and households do not assign any utility
to commercial goods in jurisdiction where they do not belong to. (iii) For any agent a ∈ A, ua is
strictly increasing in all commodities to which agent a assigns utility.
Given a convex set X ⊂ Rn , a function f : X → R is strongly quasi-concave if f (λx + (1 − λ)y) >
7

min{f (x), f (y)}, for any (x, y) ∈ X × X such that f (x) 6= f (y). This property is weaker than strict quasi-
concavity, which requires f (λx + (1 − λ)y) > min{f (x), f (y)}, for any (x, y) ∈ X × X such that x 6= y.

12
2.1 CRE development projects

In the first period, developer dk can buy construction inputs to develop one (and only one) CRE
asset jk in jurisdiction k. These construction inputs are evaluated by the following Cobb-Douglas
production function:
yk = T F Pk · Πl∈L1 (xdl1k )αlk (1)

where

• T F Pk is a parameter that stands for the “total factor productivity” specific to jurisdiction k.
T F Pk may differ across jurisdictions because their natural resources, their agglomeration
economies, or the intensity of competition.8 T F Pk may also capture the infrastructure and
the amenities of a jurisdiction.9

• parameter αlk ∈ [0, 1] denotes the weight assigned to construction input l 6= 1 by jurisdic-
tion manager k.

• xdl1k is the amount of construction input l ∈ L1 that developer dk buys in period 1. Materials,
once employed for the construction of the CRE asset, cannot be further traded.

• The type of CRE asset is determined by the vector of production weights αk = (αlk )l∈L1 ,
i.e., different vectors αk leads to different types of CRE assets. For example, in a two-
jurisdiction economy with two construction inputs, we can associate the vector αk =
(α1k , α2k ) = (1/2, 1/2) with a hotel, and the vector αk0 = (α1k0 , α2k0 ) = (1/3, 2/3) with a
shopping mall. If the jurisdiction manager k chooses αk = (α1k , α2k ) = (1/2, 1/2) and if
the representative developer dk chooses xd1k = (2, 3), then the CRE development project jk
has a size equal to yk = 2.45. Different combinations of xd1k result in different type-specific
CRE asset sizes.
8
The fact that firms in large jurisdictions are more productive is a well-established fact in the empirical literature;
see, e.g., Rosenthal and Strange (2004) and Duraton et al. (2012).
9
In addition, we could make T F Pk a function of the jurisdiction’s fixed and variable costs, which in turn deter-
mine its public investments and depend on parameter αk (see Section 2.3). Here, for simplicity, we just take T F Pk
as a parameter specific of a jurisdiction k.

13
 If developer dk constructs a positive amount of CRE asset yk > 0, it is initially entitled to
all CRE equity in that asset and we write ēdkk = 1. If yk = 0, then ēdkk = 0. And because only
developers can engage in construction activities, we set ēak = 1 if a 6= dk . We find convenient to
write ēA = (ēa : a ∈ A) to denote the initial CRE equity ownership bundles of all agents in the
economy.
The production of commercial goods only depends on the type and size of the CRE asset. In
particular, at state s of the second period, the supply of consumption goods by CRE asset jk in
jurisdiction k is given by the following function:

fk (yk )(s) : [0, +∞) → RL+2

with fk (0)(s) = 0.10

For a CRE asset jk with size yk , the amount of commercial good l ∈ L2 supplied to local
households at state s is flk (yk )(s). The price of this commercial good is denoted by plk (s). Thus,
a CRE asset with size yk generates an endogenous price-dependent cash flow equal to

X
ck (yk ; p)(s) = plk (s)flk (yk )(s) at s ∈ S.
l∈L2

We impose the following assumptions on function fk :

Assumption 2: (i) fk (·)(s) is additively separable and homogeneous of degree 1; (ii) for
every k ∈ K and l ∈ L2 , flk is increasing and concave; (iii) and for all k ∈ K and s ∈ S,
fk (yk )(s) 6= 0 when yk 6= 0.

Assumption 2.i implies that, if yk = κθ1 + θ2 and θ1 , θ2 > 0, then fk (yk )(s) = κfk (θ1 )(s) +
10
For CRE assets such as hotels, office space, and student housing, consumption of those commercial goods
should be interpreted as the purchase of a particular type of space for a given period in a given location with specific
surrounding amenities. For CRE assets that involve the production and sale of physical goods, we differentiate
between industry and retail. The former produces some goods that are sold to retail owners. The latter buys those
goods from the industry and sells them to the individuals in the jurisdiction. For the sake of simplicity, our model
does not differentiate between the two. Instead, we assume that households purchase commercial goods directly
from the local CRE property. An extension of our model that differentiates between industry and retail real assets
could be done by considering an additional period (t = 3), where the goods purchased in t = 2 by retail owners are
sold to local households.

14
fk (θ1 )(s). This allows us to split the cash flows of a CRE asset among different equity owners.
We impose Assumption 2.ii to guarantee that the developer’s budget constraints are convex.11 We
will use Assumption 2.iii to prove that there is no excess of supply in the equity market.

2.2 Equity and debt

In this economy, only equity is subject to taxes (this model of taxation is consistent with the
so-called “pass-through taxation”, see discussion in Section 2.4). Issuing debt also requires a
developer to keep with some equity in the property that can be pledge to the lender in case of
default.
A developer dk can sell all, part, or none of its initial CRE equity ēdkk to investors. Investors
can buy equity in one or several CRE development projects and thus can be thought of as equity
REITs.12 We assume that households are not financially sophisticated in the sense that they do
not have access to the real estate equity market.13
With these considerations at hand, we proceed to introduce the following notation. Let Eka
denote an agent a’s equity positions on CRE project jk , where a = dk , i. We write E A ≡ (E a :
a ∈ D ∪ I) ∈ RK+IK
+ to denote the vector of equity positions corresponding to all developers
and investors of the economy, respectively (there are I investors and K developers – one per
jurisdiction – and we do not allow developers to buy CRE equity in jurisdictions other than their
own). Since households are just consumers and do not have access to the CRE equity market, we
set Ekh0k = 0 for all k, k 0 ∈ K.
An equity stake on a CRE asset is just a claim to the future payoffs generated by this asset.
11
Later, in section 4, we shall provide an example of a simplified economy, where an equilibrium can also exist in
a context where fljk is a linear increasing function.
12
Equity REITs are real estate companies that acquire commercial properties – such as office buildings, shop-
ping centers and apartment buildings – and lease the space in the structures to tenants. See “Guide to REITs”:
https://www.reit.com/investing/reit-basics/guide-equity-reits
13
This modelling choice requires an impatience assumption on the utility function (see below)

15
The market clearing condition for equity shares corresponding to CRE asset jk is as follows:

X
Ekdk + Eki = ēdkk . (2)
i∈I

When Ekdk = 0, the developer sells all the initial equity and does not keep any equity for himself.
If instead Ekdk ∈ (0, ēdkk ), the developer keeps part but not all of the ownership on the property.
When Ekdk = ēdkk , the developer owns all equity of the project, so he is entitled to all of the
property’s cash flows.
We normalize ēdkk = 1 to have Eka denoting both the agent a’s face value of equity on CRE
property jk in jurisdiction k, and the agent a’s equity share on jk . Thus, an agent a with Eka is
entitled to the CRE asset jk ’s cash flow Eka ck (yk ; p)(s). By market clearing equation (2), we can
write
X
ck (yk ; p)(s) = Ekdk ck (yk ; p)(s) + Eki ck (yk ; p)(s), for all s ∈ S.
i∈I

Roughly speaking, the endogenous price-dependent cash flow that a CRE asset with size yk gen-
erates at state s equals the sum of all payments received by the equity owners of this asset.
We denote the (endogenous) price of one unit of CRE equity in asset jk in period 1 by qk .
Then, an agent a that purchases an equity stake Eka > 0 on asset jk pays qk Eka in the first period
and receives Eka ck (yk ; p)(s) units of the numeraire good at state s of the second period.
We assume that debt is a nominal asset (we need this assumption for our proof of Theorem 1
below in order to guarantee the existence of an interior point in an agent’s budget constraint). An
a a
agent buying (lender) a face value of debt equal to D+ ≥ 0 pays τ D+ in the first period, where τ
is the (endogenous) discount price of debt.14 Similarly, an agent selling (borrower) a face value
a a
of debt equal to D− ≥ 0 receives τ D− in the first period. We allow households and developers
to trade debt in the global market. Howerver, global investors are not allowed to borrow, i.e.,
i
D− = 0 (we rule out modelling globar investors as mortgage REITs).15
14
Debt prices, which reflect the discounted cashflow of future debt payments, are endogenous in our model.
15
See Campello and Giambona (2013) and Cvijanović (2014) for insights on this possibility.

16
 CRE debt is non-recourse and must be collateralized with CRE equity. The collateral debt
constraint in period 1 for an agent a = dk , hk in jurisdiction k is as follows:

a
Eka ≥ σ k D− a
≡ υ k (D− ) (3)

where σ k > 0 stands for the equity collateral requirement for each unit of debt issued by an agent
a in jurisdiction k.16 See Giambona, Mello, and Riddiough (2016) for a discusison of covenants
imposed on secured (non-recourse) mortgage contracts.
We denote the interest rate at state s especified in the mortgage contract by r(s) − 1 and
for simplicity we set r(s) = r, for all s ∈ S. Because the commercial mortgage is assumed to
be non-recourse, the borrower’s delivery rate at state s per unit of debt purchased will not be the
face value r, but rather the minimum between the debt face value and the cash flow corresponding
to the equity collateral. Formally, a borrower a’s effective mortgage payoff at state s for CRE
project in jurisdiction k is
Qk (s) = min{r, σ k ck (yk ; p)(s)}

Thus, if r > σ k ck (yk ; p)(s), the borrower declares default and transfers the CRE asset payoff
a
υ k (D− )ck (yk ; p)(s) to the lender.
As in Geanakoplos and Zame (2014), we consider a “perfectly competitive world in which
lenders and borrowers meet in a large market, and not a world with a single lender and borrower
negotiating with each other”. There is an agency (or clearing house) that manages the mortgage
payments and determines the mortgage rate of return Φ1 ∈ [0, 1]S that is paid to the lenders at
the different states of nature of period 2. In particular, given the borrowers’ effective mortgage
16
Notice that collateral constraint (3) differs from the standard one in general equilibrium where collateral is
a durable consumption good (see e.g. Geanakoplos and Zame 2014). We instead require the borrower to keep
a minimum amount of CRE equity as collateral in case of default. Also notice that constraint (3) substitutes the
standard exogeneous short sale constraint of type Da ≥ −D̄a with D̄a > 0. This is because, once we take into
account the equity market clearing equation (2), we have Eka ≤ ēdkk and therefore Da ≥ −D̄a where D̄a = ēdkk /σ k
for our economy. Finally, notice also that in equilibrium we expect developers to be in the long side of the debt
market, and thus the short sale constraint would become Ddk ≥ −ēdkk /σ k . This constraint rules out the possibility
of indeterminacy in agents’ portfolios choices in our setting with an endogenous financial structure and possibly
redundant assets.

17
payments across jurisdictions, the agency delivers the following return at state s of period 2 for
each unit of mortgage purchased in the first period:

P P a
k∈K a∈{Dk ∪Hk } Qk (s)D− a
P P
if a∈{Dk ∪Hk } D− > 0

a
 P P
 r k∈K a∈{D ∪H } D− k∈K
k k


φs =



a
 P P
 1 if k∈K a∈{Dk ∪Hk } D− =0

2.3 The jurisdiction’s profit function

Jurisdictions incur in some costs when providing public goods. Examples of these costs are roads,
sewerage, fire protection, police, legal advisors, supervision, and accounting. Here we split the
costs for local authority k between fixed costs, λ(αk ) > 0, and a variable cost ε(αk ) > 0. Both
cost functions are defined in terms of the numeraire good. We assume that ε(0) = 0 and λ(0) = 0,
and also that ε(αk ) and λ(αk ) are, respectively, homogenous of degree 1 and 0 in αk . The latter
assumption guarantees that, without loss of generality, we can normalize the return vector of a
CRE asset. In terms of economic interpretation, making fixed and variable costs a function of αk
implies that the type of CRE asset dictates the associated cost of public infrastructure.
To finance these costs, the local authority imposes a tax on CRE property that is proportional
to the agent’s equity holdings. In particular, an agent holding Eka equity shares in asset jk must
pay γ ak Eka units of the numeraire good to the jurisdiction, where γ ak > 0 is the agent type specific
property tax rate. This model of taxation is consistent with the so-called “pass-through taxation”,
in which the owners of the CRE property are personally responsible for paying taxes and expenses
according to some pari-passu rule, which normally takes the form of a proportional sharing rule
with respect to the owners’ equity holdings. The “pass-through taxation” model includes Limited
Liability Companies (LLC), which are one of the most prevalent business forms in the United
States.
We take tax rate γ k as given in Sections 2 and 3. Later, in Section 4, we provide an extension
to the model that makes this variable endogenous.

18
 Assumption 3: For all k ∈ K and all a ∈ A, γ ak belongs to the compact set Γak = {γ ak ∈
R+ : ε(αk ) ≤ γ ak ≤ γ̄ k } where γ̄ k > ε(αk ).

Let us now define a jurisdiction by a triplet (k, αk , γ k ) that specifies the set of players k in
jurisdiction k, the type of CRE asset that developers can construct (determined by vector αk ), and
the vector of property taxes γ k = (γ hk k , γ dkk , γ ik ), respectively. Hereafter, we refer to K as the ju-
risdiction structure, i.e., K ≡ {(k, αk , γ k ), k = 1, ..., K}. By defining the set of all jurisdictions
in K that contain agent a by Ka ≡ {(k, αk , γ k ) ∈ K : a ∈ k}, we can restrict agents’ choices.
By abuse of notation, k designates de jurisdiction and also stands for the triplet (k, αk , γ k ). Later,
in Section 3 we introduce a pre-stage where we endogenize (αk , γ k ).
Given vector E A , the profits of a jurisdiction k in the first period are given by

X
πk ≡ γ ak Eka − (λ(αk ) + ε(αk )yk )
a∈k

Profits are redistributed among jurisdiction members. Consider an agent a ∈ k and let its
share of jurisdiction k’s profit be δ ak ∈ [0, 1]. By choosing a vector (δ ak )a∈k , such that a∈k δ ak =
P

1, the jurisdiction manager is effectively redistributing resources among agents in the jurisdiction.
Thus, CRE property taxes have a redistributive effect in our model because jurisdiction profits
revert to the agents that live and do business in the jurisdiction according to some weights.17
A theory of political economy could be elaborated by endogenizing these shares (we leave this
possibility for future research).
17
This assumption is standard in competitive financial economies with intermediation costs. See e.g. Markeprand
(2008) and Prechac (1996).

19
2.4 Optimization and equilibrium

The budget constraints of an agent a ∈ A in period 1 and state s ∈ S of period 2 are, respectively,

X X X
pl1 (xal1 − ω al1 ) − p01 δ ak π k + τ D+
a a
− τ D− + qk (Eka − ēak ) + p01 γ k Eka ≤ 0
l∈{0}∪L1 k∈Ka k∈Ka

(4)
X X X
p0 (s)(xa0 (s) − ω a0 (s)) + a
plk (s)xalk (s) ≤ φs D+ a
− Qk (s)D− + Eka ck (yk ; p)(s) (5)
k∈Ka l∈L2 k∈Ka

1+L01 S(1+L2 K)
Given equity prices q ∈ RK
+ , commodity prices p ∈ R+ × R+ , and the price of debt
τ ∈ R+ , an agent a’s optimization problem consists of choosing a vector

1+L01 S(1+L2 K)
(xa1 , xa (1), ..., xa (S), D−
a a
, D+ , E a ) ∈ R+ × R+ × R+ × R+ × RK
+

that maximizes his utility function ua , subject to his budget constraints of periods 1 and 2, the
i
collateralized debt constraint (3), and the sign constraints D− = 0, Ekh0k = 0 for all k, k 0 ∈ K,
and xalk0 (s) = 0 if a ∈ {dk } ∪ Hk and l ∈ L2 (k 0 ) with k 0 6= k. W
The formal definition of an equilibrium is as follows:

Definition 1: Given a jurisdiction structure K, a competitive equilibrium for this economy

consists of a system (xA , D−
A A
, D+ , E A , p, q, τ , (π k )k∈K ), such that:
(i) each agent a solves its optimization problem;
(ii) for each jurisdiction k, the profit function π k is

X
πk = γ k Eka − (λ(αk ) + ε(αk )yk )
a∈{dk }∪I

(iii) the
 return of mortgage debt at state s is
a
P P
k∈K a∈{Dk ∪Hk } Qk (s)D− a
P P
if a∈{Dk ∪Hk } D− > 0

a
 P P
 r k∈K a∈{D ∪H } D− k∈K
k k


φs =



a
 P P
 1 if k∈K a∈{Dk ∪Hk } D− =0

20
 (iv) the following market clearing conditions hold:

• global market clearing for the numeraire consumption good in period 1:

a
− ω a01 +
P P
a∈A (x01 k∈K (λ(αk ) + ε(αk )yk )) = 0

• global market clearing for the construction inputs in period 1:

a
− ω al1 ) = 0, ∀l ∈ L1
P
a∈A (xl1

• global market clearing for the numeraire consumption good at state s ∈ S of period 2:

(xa0 (s) − ω a0 (s)) = 0,

P
a∈A

• local market clearing for commercial goods:

xalk (s) − flk (yk )(s) = 0

P
∀k ∈ K, ∀l ∈ L2 , a∈k

• local market clearing for equity:

∀k ∈ K, Ekdk + Eki = ēdkk

P
i∈I

• global market clearing for debt:

a a
P
a∈A (D+ − D− )=0

In equilibrium we expect developers need to raise equity and issue non-recourse mortgage
debt to pay for construction materials in the first period. For this equilibirum configuration, the
capital structure of a CRE asset has the following structure:

Remark 1: The capital structure of a CRE asset at jurisdiction k is endogenous and composed
by the developer’s debt Dkdk , the common equity of the entrepreneurial (developer) partner Ekdk ,
and the common equity of the capital (investor) partners (Eki )i∈I .

2.4.1 Additional key assumptions and the existence result

A subtle condition in general equilibrium models is the lower semicontinuity property of the
agent’s budget constraint in the first period. The usual approach to guarantee this property is

21
assuming that all agents of the economy have positive endowments of all goods. In our model,
we cannot impose this assumption for the commercial goods because these goods are endoge-
nously produced. To guarantee the existence of an interior point in the budget constraint set for
each profile of prices, we consider the following assumptions on the endowments of construction
inputs (l ∈ L1 ) and the numeraire good (indexed by “0”):18

Assumption 4: For all agents a ∈ A, ω a01 > (λ(αk )+ε(αk )(1+K))K, ω al1 > 0, ∀ l ∈ L1 , and
ω a0 (s) > 0. Moreover, for all developers dk ∈ D, ω d01k > γ k ŷk , where ŷk ≡ T P Fk · Πl∈L1 (ω dl1k )αlk .

Another subtlety to guarantee the property of lower semicontinuity of the budget set corre-
spondence has to do with the presence of portfolio constraints, which prevent the usual normal-
ization of commodities and assets prices in the first period. Since we cannot impose additional
restrictions on the agent’s budget constraints and portfolio sets – as these are written to capture
our particular setting – we consider the following mild impatience assumption for an economy
with restricted participation:

Assumption 5: For any agent a ∈ A and any x ∈ X a , there exists a bundle %(θ, x) ∈ R1+L
+
1
,
given θ ∈ (0, 1), such that ua (x1 + %(θ, x), (θ(x(s))s=1,...,S ) > ua (x1 , (x(s))s=1,...,S ).

Assumption 5 says that we can always find a large consumption for an agent in period 1
such that this agent is better off with this extra consumption in period 1 but less consumption in
every state of period 2.19 This assumption is satisfied by many different types of utility functions
that are unbounded on the first period consumption, such as von-Neumann utility functions with
quasi-linear, Cobb-Douglas, or Leontieff kernels, e.g., Cobb-Douglas, CES, and CARA (see
Seghir and Torres-Martinez 2011). Also, notice that this type of utility function does not depend
on the representation of individuals’ preferences and does not require further assumptions on the
portfolio sets.
18
We leave for future research the relaxation of some of the elements in Assumption 4, namely, getting rid of the
interiority assumptions of agents’ endowments (see Rincon-Zapatero and Santos 2009 for similar issues). In Section
4, we provide a simplified version of our economy in which an equilibrium exists without requiring agents having
positive endowments of all commodities in the first period.
19
This way to reframe the standard impatience assumption of an agent preferring to consume today rather than
tomorrow is convenient to state in terms of the “primitives” of the model (see ).

22
 Theorem 1: Let Assumptions 1.i-iii, 2.i, 2.ii, 3, 4, and 5 hold. Then, given a jurisdiction
structure K ≡ {(k, αk , γ k ), k = 1, ..., K}, there exists a competitive equilibrium.

The proof of Theorem 1 is left for the Appendix. This proof relies on fixed point theory.

2.4.2 Subtleties of our existence proof

In the rest of this section, we discuss the implications of imposing Assumptions 2, 4 and 5, and
the technical subtleties of our existence proof in view of existing results in the literature of general
equilibrium with segmented markets.
First, it is not trivial that a developer’s budget set correspondence takes convex values. To see
this, notice that the market value of a CRE asset depends on the developer’s choice of materials.
Also notice that the equity return in the second period depends on the CRE asset throught func-
tions (fk )k∈K , and so is endogenous. Then, to guarantee the convexity of a developer’s budget
set, we have to impose Assumption 2.ii, namely, we require that fk is a concave and increasing
function for every k ∈ K.20
The main difficulty of our equilibrium existence proof has to do with the lower semicontinuity
property of the agents’ budget constraints. Since the CRE equity markets are segmented (CRE
equity not available to all agents in the economy), we cannot take the usual approach where an
auctioneer chooses both the commodity and security prices in the simplex. For if the auctioneer
chooses the price of one type of CRE equity equal to 1, the remaining commodity, debt, and
equity prices would be zero. But then, there would be jurisdictions with commodity and security
prices equal to zero and the budget constraints of agents with single jurisdiction memberships
would hold with equality. Lower semicontinuity of the budget correspondence would fail as a
result since we could not guarantee the existence of an interior point. To circumvent this problem,
we let the price auctioneer for the first period choose commodity prices in the simplex (a compact
set). For asset prices, we have to find an endogenous upper bound.
20
The convexity of a developer’s budget set is not compromised by the budget constraint in the first period because
the production function for the CRE asset is Cobb-Douglas and this function is assumed to be concave.

23
 Another issue related to the lower semi-continuity of the budget set correspondence has to
do with the existence of an interior point in the budget constraint. Even if agents have strictly
positive endowments of the numeraire good in the second period, the value of endowments may
not be strictly positive for some agents because we embed commodity prices into the simplex.
To overcome this difficulty, we rely on two assumptions: 1) endowments of the numeraire and
construction imputs are stritly positive for all agents in the first period, and 2) the nominal return
of risk-free debt is positive. With prices chosen in the simplex and strictly positive endowments
of all commodities in the first period, the value of the endowment in the first period is strictly
positive. The following bundle is an interior point of the budget constraints for all agents: all
commodity purchases equal to zero, CRE equity positions also equal to zero, and a small position
in risk-free debt.
Next, we argue that previous results in the literature of market segmentation with a fixed point
theory approach are not useful for finding endogenous upper bounds for CRE equity prices.
In one strand of this literature, authors consider exogenous trading constraints, but impose
financial survival assumptions or spanning conditions on the set of admissible portfolios (see
for instance Balasko, Cass, and Siconolfi 1990 for a seminal contribution, and, more recently,
Angeloni and Cornet 2006, Aouani and Cornet 2009, and Cornet and Gopalan 2010). These
assumptions are not suitable for our particular setting. For instance, households do not satisfy the
survival assumption because they do not trade equity. Also, the property tax on equity prevents
us from considering a spanning condition on the set of admissible portfolios.
Another strand of the literature applies fixed point theory but in a setting with endogenous
portfolio constraints. For instance, the paper by Cea-Echenique and Torres-Martinez (2016) imposes
a super-replication condition that allows payments associated with segmented contracts to be
super-replicated by durable goods and/or contracts that agents can short sell. This assumption
does not fit into our framework because of the following reasons: 1) we do not have durable
goods, 2) there is a collateral constraint on debt, and 3) CRE equity returns are endogenous
as they depend on the developers’ choices. In a similar context, Faias and Torres-Martinez

24
(2017) consider instead assumptions on the utility function – precisely, indiference curves through
individuals’ endowments do not intersect the consumption set boundary. However, the “essen-
tiality of commodities” assumption requires endowments to be strictly positive, which is not true
in our setting for the case of commercial goods in the second period. Finally, the paper by Seghir
and Torres Martinez (2011) considers an impatience assumption in the utility function, which
requires that small reductions in the consumption of the second period can be compensated by
an increase in the consumption of the first period. We adapt this assumption to our framework
by imposing Assumption 5. A key difference from Seghir and Torres Martinez (2011) is that
they assume that second period endowments are strictly positive, which is not true in our case.
We overcome this difficult by imposing Assumption 4, which says that developers have enough
endowment of the numeraire in the first period to pay the property tax associated with the equity
of the CRE asset that they could produce with their endowment of construction inputs. With this
trick, we make sure that developers can always transfer wealth from the first period to the second
period.

2.4.3 Financial innovation with segmented markets: the case of real estate development

We conclude this section with a remark that relates our equilibrium setting with Rahi and Zigrand
(2009) model of financial innovation with segmented markets.
Remark 2: Our Walrasian economy has segmented good and equity markets, and financial
markets can be incomplete. This setting is similar to Section 3 of Rahi and Zigrand (2009) in
the sense that we do not consider “arbitrageurs”. The main difference from our models is that
we allow for global investors that can buy and sell CRE equity in multiple jurisdictions. An
equilibrium exists for our economy and, therefore, we can rule out arbitrage opportunities even
when global investors can operate in multiple jurisdictions.

25
3 Optimal property tax and selection of CRE assets

The presence of global investors makes public policy an important tool for local and national
governments as they seek to attract capital for commercial real estate development. Recurrent
policies used by local governments are fiscal and land use policies. The most important fiscal
instrument for local governments is property taxes (e.g., property taxes contribute more than 60
percent of the City of Madison’s revenues in Wisconsin). There is also a wide spectrum of land
use policies, but one of the most important and effective one for urban design is the restriction
that a jurisdiction may impose on the type of a real estate development, e.g., size and asset class.
Local governments’ public policies are crucial in attracting CRE investments. If well executed,
they create value for the jurisdiction, not only in terms of tax revenues, but also in terms of job
creation and amenities for its citizens.
So far, we have taken the property tax γ k and the Cobb-Douglas exponents αk for all juris-
dictions k ∈ K as parameters and proved that the equilibrium set is non-empty.21 We denote
by E(α, γ) the set of competitive equilibria for a given profile of types of CRE assets and corre-
sponding taxes, (α, γ) = ((α1 , γ 1 ), ..., (αK , γ K )). In this section, we propose an extension to our
model of Section 2 in which strategic jurisdiction managers choose (α, γ) in a non-cooperative
game. More precisely, we consider a strategic game in which jurisdiction authorities decide their
respective types of CRE projects and property taxes. For this, each of these jurisdiction author-
ities evaluates a profile of strategies by replacing the equilibrium (or combination of equilibria)
associated with the profile of policy variables into their corresponding profit function.
Let each jurisdiction authority consider a discrete set Λk of possible types of CRE develop-
ment projects (that is, αk ∈ Λk for all k ∈ K) and a discrete set Γk = Γhk k × Γdkk × Γik of possible
fiscal policies for all k ∈ K.

Assumption 7: The spaces Λk and Γk are discrete for all k ∈ K.

21
Recall that developers in a jurisdiction can only construct one type of CRE asset, which is chosen by the jurisdic-
tion manager. However, it is possible that developers in the same jurisdiciton would choose a different combination
of materials and thus develop the same CRE asset but with different sizes. Notice that allowing for the construction
of multiple types of CRE assets would complicate our analysis because that would require having the jurisdiction
manager choose multiple vectors of Cobb-Douglas production weights.

26
 Examples of available CRE development projects in Λk are shopping centers, retail spaces,
offices, and hotels. Examples of property taxes in Γk are a non-invasive fiscal policy with low
property taxes and a redistributive fiscal policy with high property taxes. The jurisdiction manager
chooses pairs (αk , γ k ) ∈ Λk × Γk that satisfy certain criteria. An example of such criteria could
be to consider only those CRE development projects that are compatible with a low tax scheme,
i.e., those CRE development projects with low fixed and variable costs, i.e., low λ(αk ) and ε(αk ),
respectively.
The aim of each jurisdiction manager is to maximize its profits, which consist of the difference
between the property taxes the jurisdiction receives minus the costs it incurs to provide a CRE
project. In particular, given a profile (α, γ) of CRE projects and taxes, let (xA , DA , E A , p, q, τ , (π k )k∈K )
be an equilibrium vector of the economy; then, for this equilibrium, the profit of a jurisdiction k
is
X
πk = γ k Eka − (λ(αk ) + ε(αk )yk ) .
a∈{dk }∪I

As discussed by Allen and Gale (1989), it would be difficult to provide general conditions that
guarantee uniqueness in this environment with endogenous security design. Here, we face the
same difficulty because, for each profile (α, γ), the set of equilibria may not be single. We
circumvent this problem by assuming that jurisdiction k’s manager evaluates its payoff using an
index Ψk (E(α, γ)) that depends on the equilibrium set of profits that emerge given the profile of
parameters (α, γ).
Let Πk (α, γ) = {π k : (xA , D−
A A
, D+ , E A , p, q, τ , (π k )k∈K ) ∈ E(α, γ)}, then, Ψk (E(α, γ)) =
Ψk (Πk (α, γ)) . When the equilibrium set E(α, γ) is unique, then for each profile (α, γ) ∈ Λk ×
Γk , Πk (α, γ) = {π k } is also unique and thus Ψk (E(α, γ)) = π k . When the equilibrium set for
a profile (α, γ) is not single, we define index Ψk (E(α, γ)) using a measurable selector of the
equilibrium correspondence E(α, γ). In particular, let (x̃A , D̃−
A A
, D̃+ , Ẽ A , p̃, q̃, τ̃ , (π̃ k )k∈K )(α, γ)
be the measurable selector of correspondence E(α, γ). Then, Ψk (E(α, γ)) = π̃ k . This index is
well defined if such measurable selector exists.22
22
The notion of an equilibrium selector is well-known and has been used in different strands of the literature; see,

27
 Lemma 7: There exists a measurable selector (x̃A , D̃−
A A
, D̃+ , Ẽ A , p̃, q̃, τ̃ , (π̃ k )k∈K ) for the
equilibrium correspondence E.

The proof of this lemma follows by applying the Kuratowski-Ryll-Nardzewski measurable

selection theorem (see Aliprantis and Border 2006, p. 600). To see this, notice that in our
model, the set of available CRE development projects and tax profiles is finite and, therefore, the
equilibrium correspondence E(α, λ) is trivially a weak measurable correspondence.
Considering equilibrium selections has implications for economic policy. To see this, notice
that the index function Ψk (E(α, γ)) can be defined in many different ways. For example, if
the jurisdiction manager is risk averse and has prudent behavior, then it would be reasonable
to consider an index that corresponds to the minimal profit generated among the set of possible
equilibria, i.e.,
Ψk (α, γ) = min{π k : π k ∈ Πk (α, γ)}.

Because for each profile (α, γ) the set of equilibrium profits belongs to a compact set, the proof
of existence of this minimum is trivial.
Local authorities play a strategic game G = {(Λk × Υk , Ψk )k∈K }, where Λk × Υk and Ψk are,
respectively, the strategy set and the payoff function of local authority k ∈ K. A Nash equilibrium
in mixed strategies for this game consists of a probability measure over the set of property taxes
and CRE projects.
The game G pins down the local authorities’ equilibrium strategies.

Definition 2: An equilibrium for the economy is a profile (α, γ) ∈ (Λ1 × Υ1 ) × · · · (ΛK × ΥK )

and a Walrasian equilibrium system (xA , D−
A A
, D+ , E A , p, q, τ , (π k )k∈K ) ∈ E(α, γ), such that

(i) (α, γ) is a Nash equilibrium for the game G = {(Λk × Υk , Ψk (α, γ))k∈K }, and
for example, Miao (2006) in recursive macroeconomics, Berliant and Page (2001) in public economics, Simon and
Zame (1990) in game theory, Faias, Moreno, and Pascoa (2002) and Luque and Faias (2017) in financial economics,
and Stahn (1999) for a general equilibrium model with monopolistic behavior. In these models, in general, a profile
of actions gives rise to a set of equilibrim outcomes. Then, to obtain an equilibrium existence result with well-defined
payoff functions, which themselves depend on these profiles, authors use equilibrium selections. For example, this
is the case of Cournot-Walras models with a continuous space of actions, where continuous random selections are
used.

28
 (ii) for each k ∈ K, π k = Ψk (ε(α, γ)).

Theorem 2: Let Assumption 7 hold. Then, there exists an equilibrium for the first stage of
the economy.

Theorem 2 guarantees the existence of an equilibrium, possibly in mixed strategies (due to

the discreteness of sets Λ and Γ).

4 Equilibrium characterization

The seminar paper by Modigliani and Miller (1958) shows that the capital structure of a financial
asset is irrelevant. However, this result ignores important institutional features, such as limits on
debt issuance, different tax treatment between equity and debt, and default risk – see Admati and
Hellwig (2013) for a recent paper that discusses these possibilities in the rather different context
of the banking sector, Gau and Wang (1990) for a discussion of usual restrictions in real estate
investments and their relationship with the capital structure of an income-producing property,
and Sun, Titman, and Twite (2015) for evidence of the impact of the recent financial crisis on the
limits to debt capacity for commercial real estate assets. In our model, the capital structure of a
real estate development is relevant because we consider property taxes, default on non-recourse
collateralized mortgage debt, and potentially incomplete markets. Moreover, capital structures
of CRE development projects in different jurisdictions are interdependent in our setting with
segmented markets because investors can buy equity in multiple jurisdictions.
In this section, we perform an equilibrium analysis of these possibilities. For simplicity,
we focus on an economy with two jurisdictions (k = 1, 2), two states of nature in the second
period (s = s1 , s2 ), and only one construction material (l = 1) used for development in both
jurisdictions. We normalize to 1 the prices of the numeraire good (l = 0) in period 1 and also at
states s1 and s2 in period 2. Function fk (yk )(s) : yk → xk (s) ∈ R+ , which maps the CRE asset
size into an amount of a single commercial good at state s, is assumed to be linear. In particular,
fk (yk )(s) = Yk (s)yk , where Yk (s) ≥ 0 for s = s1 , s2 and k = k1 , k2 . For simplicity, we consider

29
only one type of CRE asset and accordingly we equate the coefficient αk in production function
1 between jurisdictions, i.e., α1 = α2 . The promised mortgage payoff r is set equal to 1.
Each jurisdiction k has two households and one developer, denoted by hk , Hk , and dk , re-
spectively. The real estate equity investor i belongs to both jurisdictions. The utility functional
form of an agent a ∈ A is

X X
ua (xa ) = θal ln xal + (θa0 (s) ln xa0 (s) + θa1 (s) ln xa1 (s) + θa2 (s) ln xa2 (s))
l={0,1} s={1,2}

where the θ-parameters denote agent a’s weights assigned to the different available goods.23 We
set all θ-parameters are zero, except for θH h1 h2 d1 d2
0 > 0, θ 1 (s) > 0, θ 2 (s) > 0, θ 0 (s) > 0, θ 0 (s) > 0,
1

θi0 (s) > 0, for s = s1 , s2 . Households hk and Hk differ by when they prefer to consume. We think
of hk as a young household who enjoys the consumption of his respective commercial good at
both states of the second period, while Hk is an old household who enjoys the consumption of the
numeraire good in the first period. Both developers and the investor only enjoy the consumption
of the numeraire good in the second period.
Endowments are as follows. Old household Hk owns ω H
1
k
> 0 units of the construction
material. Young household hk is endowed with ω h0 k > 0 units of the numeraire good (we assume
ω h0 1 = ω h0 2 ) and ω h0 k (s) > 0 units of the numeraire good at state s = 1, 2. The real estate equity
investor i is endowed with ω i0 (s) > 0 units of the numeraire good at state s = 1, 2 of the second
period. Developers have no commodity endowments whatsoever, so debt and equity are the only
means to transfer wealth from the first to the second period.
We think of the real estate investor as an “equity REIT”, whose business only consists of
i
buying CRE equity; thus, we exclude this investor from borrowing in the debt market; thus, D− =
0. In addition, we restrict households from investing in CRE equity and leave this possibility only
to the real estate equity investor and the developers.
23
The consumption of commercial goods is restricted to only those households that belong to the jurisdiction in
question. For example, the consumption of the commercial good produced in the second jurisdiction, denoted by
xa2 (s), can only be positive in this simplified economy if a ∈ k2 .

30
 Finally, to avoid dealing with the sharing of jurisdiction profits, we set a policy that assigns
jurisdiction profits to the corresponding old household, i.e., δ Hk = 1 for k = k1 , k2 .

4.1 Determinants of mortgage debt and equity market prices

For this economy, an equilibrium as specified by Definition 1 satisfies the following properties.

Proposition 1 (Mortgage recovery rate): φs , the equilibrium mortgage recovery rate at

state s, can be expressed as a weighted sum of effective delivery rates Q1 (s) and Q2 (s), where
weights are the corresponding developer’s share of debt with respect to total debt in the economy.
Delivery rate φs also increases with the young households endowments of the numeraire good at
state s. Formally,

d1 d2
X D− D−
φs = ω h0 k (s) + d1 Q (s) + d1
d2 1 d2
Q2 (s) (6)
k
D− + D− D− + D−

We define mortgage yield at state s as φs /τ , mortgage default loss of the CRE project in
jurisdiction k as 1 − Qk (s), and high-risk default jurisdiction as the jurisdiction facing a negative
shock to its effective mortgage delivery rate Qk (s). With this definitions in mind, we work out
expression (6) to assert the following:

Corollary 1 (Mortgage yield): The mortgage yield at a given state s decreases with mort-
gage default losses at that state, the developer’s debt face value in the high-risk default jurisdic-
tion, and the scarcity of young households’ resources at state s with respect to their resources in
the first period. Formally,

P hk d1 d2
φs k ω 0 (s) D− D−
= P hk
+ Q
hk 1
(s) + hk
Q2 (s) (7)
τ
P P
k ω0 k ω0 k ω0

Next, let us define φ̂s = (λhs k /λh1 k )φs as the household hk ’s discounted personalized recovery
rate using household hk ’s shadow values λhs k and λh1 k for his budget constraints in period 1 and
state s, respectively. Similarly, we define Q̂k (s) = (λds k /λd1k )Qk (s) as the developer dk ’s dis-

31
counted personalized effective delivery rate using developer dk ’s shadow values λds k and λd1k for
his budget constraints at state s and period 1, respectively.

Proposition 2 (Binding collateral constraint and expected default): The shadow value of
dk
collateral constraint (3) corresponding to mortgage debt D− and denoted by ξ dk is a function of
the expected default losses of the CRE project in jurisdiction k. Formally,

X  ξ dk σ
k
φ̂s − Q̂k (s) = dk (8)
s
λ1

Corollary 2 (Mortgage debt valuation): The negative effect of a higher default loss 1 −
Qk (s) on the mortgage debt equilibrium price τ more than offsets the positive impact of a higher
default loss 1 − Qk (s) on τ through the developer dk ’s collateral constraint shadow price ξ dk .

Proposition 3 (Equity valuation): The equity equilibrium price qk accrued of the CRE tax
γ k increases with the developer dk ’s discounted value of CRE cash flows and the expected default
losses of mortgage debt in jurisdiction k (through shadow value ξ dk ). Formally,

X λdk
qk + γ k = s
c (y ; p)(s)
dk k k
+ ξ dk (9)
s
λ1

where ck (yk ; p)(s) = pk (s)Yk (s)yk .

Equilibrium condition (8) captures how a higher risk of mortgage debt default in jurisdiction
k’s CRE project increases market pressure on equity through price qk . Equilibrium condition
(8) also shows how fiscal policy can be used to offset the positive impact of higher default risk
(through higher ξ dk ) on qk ; namely, a decrease in γ k may (partially) offset an increase in ξ dk .

4.2 CRE construction patters and capital markets

We now turn our attention to differences in construction patterns between jurisdictions. We as-
sume that both jurisdictions have the same space available for construction, but may differ in

32
the amount of construction material employed for CRE development, here denoted by xd1k for
developer dk in jurisdiction k. We define the height of commercial real estate in jurisdiction k
by Heightk = xd1k . For our next result, we also find convenient to define the developer dk ’s
expected market price of a CRE space unit (square foot) in his development project jk , net of the
productivity factor T F Pk , by
X λdk
s
mk ≡ pk (s)Yk (s)
s
λd1k

Proposition 4 (Differences in construction patterns between jurisdictions): CRE height in

jurisdiction k1 increases with respect to jurisdiction k2 if:

1. Developer d1 ’s “equity capital” E1d1 in CRE project j1 increases with respect to developer
d2 ’s “equity capital” E2d2 .

2. The CRE project j1 ’s productivity parameter T F P1 increases with respect to T F P2 ;

3. m1 , the expected market price of a CRE space unit (square foot) net of productivity gains
increases in CRE project j1 , increases with respect to m2 .

Formally, the following equilibrium equation captures items 1, 2 and 3 of Proposition 4:

1/1−α
E1d1 T F P1 m1

Height1
= (10)
Height2 E2d2 T F P2 m2

dk
The next corollary uses the collateral constraint (3) on mortgage debt D− to provide an
alternative way of looking at item 1 in Proposition 4 in terms of collateral requirements σ 1 and
σ2.

Corollary 3: When developers’ collateral constraints bind in both jurisdictions, a laxer col-
lateral requirement in jurisdiction k1 with respect to jurisdiction k2 (a higher σ 1 with respect
to σ 2 ) increases CRE height in k1 with respect to CRE height in k2 . Formally, we can rewrite
expression (10) as follows:

 d1 1/1−α
Height1 σ 1 D− T F P 1 m1
= d2 T F P m
(11)
Height2 σ 2 D− 2 2

33
 Now we discuss the general equilibrium effects of a higher expected default risk on equity
and debt prices and the capital structures of CRE development projects in the two jurisdictions.
Without loss of generality, let jurisdiction k2 experience a higher mortgage default loss at state
s2 , i.e., 1 − Q2 (2) increases. This shock may be due to a decrease in production capacity Y2 (2)
or a lower productivity parameter T F P2 .

Proposition 5: A higher mortgage default loss at state s2 in jurisdiction k2 induces the real
estate investor i to rebalance its equity investments toward the high-risk jurisdiction, crowding
d2
out developer d2 ’s equity E2d2 and debt D− , while increasing developer d1 ’s equity E1d1 and debt
d1
D− . As a result, construction booms in the low-risk jurisdiction k1 and contracts in the high-risk
jurisdiction k2 , in turn increasing the CRE height in jurisdiction k1 with respect to jurisdiction
k2 .

4.3 Interdependent capital structures without mortgage default

Mortgage default and competition for construction materials are not necessary conditions for
the real estate investor to rebalance its equity portfolio between jurisdictions. To illustrate this,
we slightly modify our previous economy by assuming recourse mortgage debt where payoff
is always r = 1 (debt is risk-free). Also, there are now two construction materials. CRE de-
velopment uses a different material in each jurisdiction; in particular, y1d1 = T F P1 (xd111 )α1 and
y2d2 = T F P2 (xd212 )α2 , where material l = 1 is used for CRE development in jurisdiction k1 and
material l = 2 is used for CRE development in jurisdiction k2 . We ignore default and the collat-
eral constraint (3) on mortgage debt, but impose the following exogenous short sale constraint on
a
recourse debt to guarantee equilibrium existence (see Hart 1979): D− ≤ D̄a , where D̄a > 0. We
refeer to D̄a as the agent a’s debt capacity. Finally, we let the old household Hk be the sole owner
of material l = k, i.e., ω H H1 H2 H2
1 > 0, ω 2 = 0, ω 2 > 0, and ω 1 = 0. The rest of specifications are
1

similar to our previous economy. The following proposition highlights the particular equilibrium
properties for this alternative economy.

Proposition 6: For this economy with risk-free debt and no developers’ competition for con-

34
struction materials, an equilibrium satisfies the following properties:

1. Commercial good price differentials across states of nature in the second period are driven
by the CRE asset’s commercial good production capacity (i.e., fk (yk )(s1 ) versus fk (yk )(s2 ));

2. The CRE equity price accrued of the property tax in a jurisdiction is driven by the CRE
asset’s cash flows (the default term is absent in this specification);

3. Construction input price differentials across jurisdictions are driven by the difference in
marginal productivity of the jurisdiction-specific CRE assets.

We finish this section with some numerical examples that illustrate how the CRE capital
structures in different jurisdictions may change after a shock to a developer’s funding capacity or
a different local fiscal policy.

Example 1 (benchmark): Let us consider the following parameter values: ω H

1
1
= ωH
2
2
=
ω h0 1 = ω h0 2 = ω i0 = 1, ω H H2 H1 H2
0 (1) + ω 0 (1) = ω 0 (2) + ω 0 (2) = 1, r = 1, D̄
1 d1
= D̄d2 = 1,
T F P1 = T F P2 = 1, α1 = α2 = 0.6, Y1 (1) = Y2 (1) = 1, Y1 (2) = Y2 (2) = 0.5, ε(α1 ) = α1 /6,
ε(α2 ) = α2 /6, λ(α1 ) = λ̂1 α1 with λ̂1 = 0.333, λ(α2 ) = λ̂2 α2 with λ̂2 = 0.333, δ H
1
1
= 0,
δ h1 1 = δ h2 2 = 0.7, δ d11 = δ d22 = 0.2, δ i1 = δ i2 = 0.1, and γ a11 = γ 1 = 0.5 and γ a22 = γ 2 = 0.5
for a1 = d1 , i and a2 = d2 , i. All θ-parameters are equal to zero, except for θH
0
1
= θH
0
2
= 3,
θh1 1 (1) = θh2 2 (1) = 1, θh1 1 (2) = θh2 2 (2) = 0.5, θd01 (s) = θd02 (s) = 0.24, and θi0 (s) = 0.52. For
these parameter values, we obtain a unique equilibrium solution where q1 = q2 = 2.500, τ =
1.440, p1 = p2 = 1.500, p1 (1) = p2 (1) = 1.500, p1 (2) = p2 (2) = 3.000, and E1i = E2i = 0.173.

For the above parameters, Figure 1 illustrates the equilibrium capital structures of CRE assets
j1 and j2 in an economy where the developer’s debt, the developer’s equity, and the investor’s
equity are the same in both jurisdictions (for the sake of brevity, we leave the details of parameter
and equilibrium values for the Appendix).

35
Figure 1: This figure illustrates the different capital structure components of CRE assets j1 and
j2 , given the parameter values of Example 1 in the Appendix. Quantities are expressed in real
terms, i.e., nominal amount times the corresponding price.

Example 2 (Shocks to developers’ funding capacities):24 Let us modify our benchmark

example by increasing the developer d1 ’s debt capacity to D̄d1 = 1.1, while decreasing the
developer d2 ’s debt capacity to D̄d2 = 0.9. Roughly speaking, the access to funding for the
developer in the second jurisdiction (k2 ) worsens with respect to developer in the first jurisdiction
(k1 ).
Figure 2 illustrates the equilibrium values of the different capital structure components of
CRE assets j1 and j2 for this case. There we see that, compared to our benchmark example in
Figure 1, the developer with worse funding capacity (d2 ) decreases its absolute real exposure to
CRE debt and equity, while the investor increases its equity exposure. The opposite happens
in the CRE development project of the developer with better funding capacity. Thus, when a
developer has poorer access to debt financing, the equity investor finds optimal to fill the gap.
Leverage (debt-to-equity ratio) decreases (increases) in the jurisdiction where the developer has
worse (better) funding capacity. Thus, leverage shifts from the jurisdiction with shortage of debt
to the jurisdiction with better debt funding capacity.
24
See Sun, Titman, and Twite (2015) for evidence of the impact of the recent financial crisis on the limits to debt
capacity for commercial real estate assets.

36
Figure 2: This figure illustrates the different capital structure components of CRE assets j1 and
j2 , given the parameter values of Example 2 in the Appendix. Quantities are expressed in real
terms.

Example 3 (The property tax rate increases in jurisdiction k2 to finance an increase in

public spending): We next consider a situation where jurisdiction k2 ’s fixed cost λ̂2 increases
from 0.2. to 0.5 (public spending increases) and finances this by increasing its tax rate γ 2 from
0.5 to 0.8. Figure 3 illustrates how the CRE capital structures in both jurisdictions change as a
result. The increase in γ 2 decreases the developer d2 ’s equity compared to the equilibrium value
in Figure 1 due to the developer’s substitution of equity for tax free debt. The developer d2 ’s
debt and the investor’s equity contributions increase as a result. Roughly speaking, we find that a
jurisdiction that increases local government spending on public infrastructure and finances these
additional spending with higher taxes ends up with more levered developers, while jurisdictions
with more conservative fiscal policies can effectively reduce developers’ leverage.

37
Figure 3: This figure illustrates the different capital structure components of CRE assets j1 and
j2 , given the parameter values of Example 3 in the Appendix. Quantities are expressed in real
terms.

Example 4 (Competition between jurisdiction managers): Finally, we illustrate by means

of a numerical example how jurisdiction competition in property taxation may result in a “race to
the bottom”. For this, we consider a setting where the two jurisdictions compete through taxation
and land regulation to attract equity investments. For this, we consider again the parameter values
used to construct our benchmark example in Figure 1, with the exception that now jurisdiction
k’s set of strategies is Λk × Γk = {(αhigh
k , γ high
k ), (αlow low
k , γ k )} = {(0.80, 0.53), (0.60, 0.50)}, for

both k = k1 , k2 . The economic interpretation is that a high Cobb-Douglas exponential parameter

α is associated with a high variable cost for the jurisdiction, which in turn requires a high property
tax. Jurisdiction manager k = 1, 2 chooses (αk , γ k ) in Λk ×Γk in order to maximize the following
profit function:25

 
αk 
πk = 1− τ (α, γ)D̄dk + (qk (α, γ) + γ 1 )Eki (α, γ) − pk (α, γ)ω H
k
k
− λ̂k αk (12)
6γ k

where (α, γ) = (α1 , α2 , γ 1 , γ 2 ). For simplicity, we assume that δ dkk = 0 and δ ik = 0, for k = 1, 2.
25
We obtain function (12) by  first writing the jurisdiction k’s profit  function as
αk α k
dk dk
πk = (γ k − αk /6)T F Pk x1k − λ̂1 αk , and then replacing T F Pk x1k with
dk


τ (α, γ)D̄dk i I
+ (qk (α, γ) + γ 1 )Ek (α, γ) − p1k (α, γ)ω 1k /γ k using (27a) and (27b), and taking δ k = 0.

38
After computing the equilibrium associated with each pair of strategies and the corresponding
profit functions for each jurisdiction manager, we obtain the following payoffs:

Jurisdiction k = 2 (αhigh
2 , γ high
2 ) (αlow low
2 , γ2 )

Jurisdiction k = 1

(αhigh
1 , γ high
1 ) (0.13, 0.13) (0.13, 0.20)

(αlow low
1 , γ1 ) (0.20, 0.13) (0.20, 0.20)

(αlow low
k , γ k ) is a dominant strategy for both k = 1, 2 in this game. Thus, the Nash equilibrium

consists of the pair of strategies ((αlow low low low

1 , γ 1 ), (α2 , γ 2 )). This result resembles the “race to the

bottom” concept in literature of competition among political jurisdictions (Cary 1974, Drezner
2001, and Carruthers and Lamoreaux 2016). In particular, we find that low property taxes are
optimal for profit maximizing jurisdictions that compete to attract global real estate equity in-
vestments.

5 Conclusions

In this paper we build a general equilibrium model of CRE development. The cash flows of a CRE
asset depend on the amount of commercial goods sold to households in the jurisdiction where the
asset is located. Global real estate investors compete to buy equity stakes on these assets. Be-
cause investors can buy CRE equity in different jurisdictions, the economy does not consist of
isolated markets. Market interdependence means that the investors’ decisions to buy more CRE
equity in a jurisdiction affect the capital structure of CRE assets located in other jurisdictions.
We identify mild conditions that guarantee the existence of equilibrium for this economy. This
result contributes to the literatures of optimal security design and market segmentation in general
equilibrium (Allen and Gale’s 1991, Allen and Gale 1994, Duffie and Rahi 1995, and Rahi and
Zigrand 2009) and also to the literature pioneered by Modigliani and Miller (1958) and Bradley,
Jarrel, and Kim (1894) on the existence of an optimal capital structure in equilibrium by consid-
ering the case of commercial real estate assets.

39
 Our model endogenizes many important variables, such as the capital structure of a CRE
asset, the CRE cash flows, the construction costs, the prices of commercial goods, the property
taxes, and the type of CRE assets that a jurisdiction selects. To motivate this aspect of our the-
ory, we propose a simplified version of our general equilibrium model that illustrates the market
mechanism of different shocks on the economy. In addition, we provide several numerical ex-
amples. For instance, we see that a negative shock to a developer’s funding capacity increases
the equity-to-total-equity ratio of the capital partner and decreases the debt-to-equity and the de-
veloper’s equity-to-total-equity ratios. Because the capital structures of CRE assets in different
jurisdictions are interconnected, this shock increases the leverage ratio in those CRE assets be-
longing to jurisdictions that are not hit by the shock. Moreover, developers in those jurisdictions
must increase their equity contributions in order to offset the investors’ decrease of CRE equity
purchases in their jurisdictions. We also explore other shocks to the economy, such as a decrease
in the production capacity of a CRE asset and an increase in the property tax in one jurisdiction.
These examples also offer interesting insights regarding the inflation of commercial goods and
the changes in the capital structures of CRE assets.
There are many other issues that can be explored in future research under the lens of our
model. For example, our economy could be extended to accommodate the difference between
industrial real estate and retail properties. This extension is briefly discussed in Section 2.
General equilibrium models with collateral constraints seem to be the correct approach for
this extension (see e.g. Geanakoplos and Zame 2014, Gale and Gottardi 2015, and Fostel and
Geanakoplos 2016). An interesting question that could be addressed following this approach
would be to quantify the importance of collateral versus taxes for the capital structures of different
CRE assets (see Li, Whited, and Wu 2016 for a similar question in the context of corporations).
Also it would be interesting to characterize the relationship between the CRE asset collateral and
the developer’s funding capacity (see Campello and Giambona 2013 and Cvijanović 2014 for
empirical work on this issue).
Another avenue for future research is to understand the role that transfers of property tax

40
revenue to the agents that live and do business in a jurisdiction have in the economy. When
the jurisdiction’s profits revert to local households, taxes can be seen as a standard redistributive
device from CRE developers and investors to households. When the jurisdiction profits revert to
developers, transfers can be seen as Tax Incremental Financing. And if the jurisdiction transfers
tax revenues to investors, the subsidies can be seen as tax credits (see Minnassian 2016). As in
the classical theory of general equilibrium, all that we would require is that the profit sharing
weights sum up to one across the agents of the jurisdiction (Luque 2018).
Our model can also guide empirical evaluations of the role that CRE property taxes and debt
collateral requirements have on attracting real estate equity investments in a globalized economy
in which jurisdictions compete to attract global equity investors.

References

[1] A. Admati and M. Hellwig (2013), The bankers’ new clothes: What’s wrong with banking
and what to do about it. Princeton University Press: Princeton and Oxford.

[2] C. D. Aliprantis and Border, K. C. (2006), Infinite Dimensional Analysis, 3rd Ed., Springer,
Berlin.

[3] F. Allen, E. Carletti, and R. Marquez (2015), “Deposits and Bank Capital Structure”, J
Finan Econ 118, 601-619.

[4] F. Allen and D. Gale (1991): “Arbitrage, Short Sales and Financial Innovation,” Economet-
rica 59, 1041–1068.

[5] F. Allen and D. Gale (1994): Financial Innovation and Risk Sharing. Cambridge, MA: MIT
Press.

[6] P. Amaral, D. Corbae, and E. Quintin (2017), “Financial Engineering and Economic Devel-
opment”, mimeo.

41
 [7] L. Angeloni and B. Cornet (2006), “Existence of Financial Equilibria in a Multi-period
Stochastic Economy”, Advances in Math Econ 8, 933-955.

[8] Anglin, P. M., D. Dale-Johnson, Y. Gao and G. Zhu (2014), “Patterns of Growth in Chinese
Cities: Implications of the Land Lease”, J Urban Econ 83, 87-107.

[9] Z. Aouani and B. Cornet (2009), “Existence of Financial Equilibria with Restricted Partici-
pation”, J Math Econ 45, 772-786.

[10] Y. Balasko, D. Cass, and P. Siconolfi (1990), “The Structure of Financial Equilibrium with
Exogenous Yields: The Case of Restricted Participation”, J Math Econ 19, 195-216.

[11] M. Berliant and T. ten Raa (1994), “Regional Science: The State of the Art”, Reg Sci and
Urban Econ 24, 631-647.

[12] M. Berliant and F. Page (2001), “Income Taxes and the Provision of Public Goods: Exis-
tence of an Optimum”, Econometrica 69, 771-784.

[13] M. Bradley, G. Jarrel, and E. Kim (1894), “On the Existence of an Optimal Capital Struc-
ture: Theory and Evidece”, J Finan 39, 857-878.

[14] M. Campello and E. Giambona (2013), “Real Assets and Capital Stucture”, J Finan Quant
Analysis 48, 1333-1370.

[15] D. R. Capozza and R.W. Helsley (1990), “The Stochastic City”, J Urban Econ 28, 187-203.

[16] D. R. Capozza and G. A. Sick (1990), “Risk and Return in Land Markets”, mimeo, Univer-
sity of Michigan.

[17] W. Cary (1974), “Federalism and Corporate Law: Reflections Upon Delaware”, Yale Law
Journal 83, 663-705.

[18] B. G. Carruthers and N. R. Lamoreaux (2016), “Regulatory Races: The Effects of Jurisdic-
tional Competition on Regulatory Standards”, J Econ Lit, 54 (1): 52-97.

42
[19] S. Cea-Echenique and J.P. Torres-Martinez (2016), “Credit Segmentation in General Equi-
librium”, J Math Econ 62, 19-27

[20] P. D. Childs, T. Riddiough, and A. J. Triantis (1996), “Mixed-Uses and the Redevelopment
Option”, Real Estate Econ 24, 317-339.

[21] D. Cvijanović (2014), “Real Estate Prices and Firm Capital Structure”, Rev Fin Stud 27,
2690-2735.

[22] B. Cornet and R. Gopalan (2010), “Arbitrage and equilibrium with portfolio constraints”,
Econ. Theory 45, 227-252.

[23] H. DeAngelo and R. Masulis (1980), “Optimal Capital Structure Under Corporate and Per-
sonal Taxation”, J Finan Econ 8, 3-29.

[24] H. DeAngelo and R. Stulz (2015), “Liquidity-claim Production, Risk Management, and
Bank Capital Structure: Why High Leverage is Optimal for Banks”, J Finan Econ 116,
219-236.

[25] G. Debreu (1959), “Theory of Value: An Axiomatic Analysis of Economic Equilibrium”,

Cowles Foundation Monographs Series, Yale University Press.

[26] D. Diamond (1967), “The Role of a Stock Market in a General Equilibrium Model with
Technological Uncertainty”, American Economic Review 57, 759-776.

[27] C. Dismore (1998), “The Impact of Public Capital Markets on Urban Real Estate”, discus-
sion paper prepared for the Brookings Institution Center on Urban and Metropolitan Policy.

[28] D. W. Drezner (2001), “Globalization and Policy Convergence”, Int Stud Rev 3, 53-78.

[29] D. Duffie and R. Rahi (1995): “Financial Market Innovation and Security Design: An In-
troduction,” J Econ Theory, 65 (1), 1-42.

43
[30] G. Duranton, P. Philippe-Combes, L. Gobillon, D. Puga, and S. Roux (2012), “The Pro-
ductivity Advantages of Large Cities: Distinguishing Agglomeration from Firm Selection”,
Econometrica 80, 2543-2594.

[31] M. Faias (2008), “Approximate Equilibrium in Pure Strategies for a Two-Stage Game of
Asset Creation”, Decisions in Economics and Finance 31, 117-136.

[32] M. Faias, E. Moreno, and M. Pascoa (2002), “Real Indeterminacy of Equilibria and Manip-
ulability”, J Math Econ 37, 325-340.

[33] M. Faias and J. Luque (2017), “Endogenous Formation of Security Exchanges”, Econ The-
ory, forthcoming. DOI: 10.1007/s00199-016-0989-9.

[34] M. Faias and J.P. Torres-Martinez (2017), “Credit Market Segmentation, Essentiality of
Commodities, and Supermodularity”, J Math Econ 70, 115-122.

[35] A. Fostel and J. Geanakoplos (2016), “Financial Innovation, Collateral and Investment”,
Amer Econ J Macro 8, 242-284.

[36] D. Gale (2004), “Notes on Optimal Capital Regulation”, in: St-Amant, P., C. Wilkins
(Eds.), The Evolving Financial System and Public Policy. Bank of Canada, Ottawa, On-
tario, Canada, pp. 225-253.

[37] D. Gale and P. Gottardi (2015), “A General Equilibrium Theory of Capital Structure”,
mimeo.

[38] D. Gale and P. Gottardi (2015), “Capital Structure, Investment and Fire Sales”, Rev Finan
Stud 28, 2502-2533.

[39] G. W. Gau and K. Wang (1990), “Capital Structure Decisions in Real Estate Investment”,
AREUEA Journal 18, 501-521.

[40] J. Geanakoplos and W.R. Zame (2014), “Collateral Equilibrium, I: A Basic Framework”,
Econ Theory 56, 443-492.

44
[41] J. Geanakoplos and H. Polemarchakis (1986), “Existence, Regularity, and Constrained Sub-
optimality of Competitive Allocations When the Asset Market Is Incomplete”, Cowles
Foundation Working Paper 652.

[42] E. Giambona, A. Mello, and T. Riddiough (2016), “Real Assets, Collateral, and the Limits
of Debt Capacity”, Real Estate Econ, forthcoming.

[43] W. Gornal and I. A. Strebulaev (2016), “Financing as a Supply Chain: The Capital Structure
of Banks and Borrowers”, J Finan Econ, forthcoming.

[44] S. R. Grenadier (1995a), “Valuing Lease Contracts: A Real-Options Approach”, J Finan

Econ 38, 297-331.

[45] S. Grenadier (1995b), “The Persistence of Real Estate Cycles”, Real Estate Finan and Econ
10, 95-119.

[46] S. J. Grossman and O. Hart (1979), “A Theory of Competitive Equilibrium in Stock Market
Economies”, Econometrica 47, 293-329.

[47] O. Hart (1979), “On Shareholder Unanimity in Stock Market Economies”, Econometrica
47, 1057-1083.

[48] R. W. Helsley and W. C. Strange (1997), “Limited Developers”, Canadian J Econ 30, 329-
348.

[49] J.V. Henderson (1974), “The Sizes and Types of Cities’, Amer Econ Rev 64, 640-656.

[50] M. C. Jensen and W. Meckling (1976), “Theory of the Firm: Managerial Behavior, Agency
Cost and Capital Structure”, J. Finan Econ 3, 11-25.

[51] A. Krausand and R. Litzenberger (1973), “A State Preference Model of Optimal Financial
Leverage”, J Finan 28, 911-922.

45
[52] H. Konishi (2013), “Entrepreneurial Land Developers: Joint Production, Local Externalities
and Mixed Housing Developments”, J Urban Econ 75, 68-79.

[53] H. Konishi (2008), “Tiebout’s Tale in Spatial Economies: Entrepreneurship, Self-Selection,

and Efficiency”, Reg Sci and Urban Econ 38, 461-471.

[54] S. Li, T. M. Whited, and Y. Wu (2016), “Collateral, Taxes, and Leverage”, Rev Finan Studi
29, 1453-1500.

[55] J. Luque (2018), “Assessing the Role of TIF and LIHTC in an Equilibrium Model
of Affordable Housing Development”, Reg Sci and Urban Econ, forthcoming DOI:
10.1016/j.regsciurbeco.2018.06.005

[56] J. Luque (2013), “Heterogeneous Tiebout Communities with Private Production and Anony-
mous Crowding”, Reg Sci and Urban Econ 43, 117-123.

[57] L. Makowski (1983), “Competitive Stock Markets”, Rev Econ Stud 50, 305-330.

[58] T. Markeprand (2008), “On Financial Equilibrium with Intermediation Costs”, J Math Econ
44, 148 - 156.

[59] D. Martinez-Miera and R. Repullo (2010), “Does Competition Reduce the Risk of Bank
Failure?”, Rev Finan Stud 23, 3638-3663.

[60] A. Mas-Collel and J. Nachbar (1991), “On the Finiteness of the Number of Critical Equi-
libria, with Application to the Random Selections”, J Math Econ 20, 397 - 409.

[61] J. Miao (2006), “Competitive Equilibria of Economies with a Continuum of Consumers and
Aggregate Shocks”, J Econ Theory 128, 274-298.

[62] E. S. Mills (1993), “The Future of Office Markets”, Working Paper, Kellogg Graduate
School of Business, Northwestern University.

[63] M. H. Miller (1977), “Debt and Taxes” J Finan 32, 261-275.

46
[64] M. Minassian (2016), “Tax Deductions and Credits for Own-
ers of Commercial Real Estate in Massachusetts”, Minassian CPA.
http://www.minassiancpa.com/articles/realestate/masstaxincentives.html

[65] F. F. Modigliani and M. H. Miller (1958), “The Cost of Capital, Corporation Finance, and
the Theory of Investment”, Amer Econ Rev 48, 261-297.

[66] F. F. Modigliani and M. H. Miller (1963), “Corporate Income Taxes and the Cost of Capital”,
Amer Econ Rev 48, 261-297.

[67] C. Prechac (1996), “Existence of Equilibrium in Incomplete Markets with Intermediation

Costs”, J Math Econ, 25 (3), 373-380.

[68] R. Radner (1974), “A Note on the Unanimity of Stockholders Preferences Among Alter-
native Production Plans: A Reformulation of the Ekren-Wilson Model”, Bell Journal of
Economics and Management Science 5, 181-184.

[69] R. Rahi and J.P. Zigrand (2009), “Strategic Financial Innovation in Segmented Markets”,
Rev Finan Stud 22, 2941-2971.

[70] J. P. Rincon-Zapatero and M. Santos (2009), “Differentiability of the Value Function With-
out Interiority Assumptions”, J Econ Theory 144, 1948-1964.

[71] S. Rosenthal and W. Strange (2004), “Evidence on the Nature and Sources of Agglom-
eration Economies,” in Handbook of Regional and Urban Economics, Vol. 4, ed. by V.
Henderson and J.-F. Thisse. Amsterdam: North-Holland, 2119–2171.

[72] A. Seghir and J.P. Torres-Martinez (2011), “On Equilibrium Existence with Endogenous
Restricted Financial Participation”, J Math Econ 47, 37-42.

[73] L. Simon and W. R. Zame (1990), “Discontinuous Games and Endogenous Sharing Rules”,
Econometrica 58, 861-872 .

47
[74] H. Stahn (1999), “Monopolistic Behaviors and General Equilibrium: A Generalization of
Nikaido’s Work”, J Math Econ 32, 87-112.

[75] L. Sun, S. D. Titman, and G. J. Twite (2015), “REIT and Commercial Real Estate Returns:
A Postmortem of the Financial Crisis”, Real Estate Econ 43, 8-36.

[76] C. M. Tiebout (1956), “A Pure Theory of Local Expenditures”, J Polit Economy 64, 416-
424.

[77] S. Titman (1985), “Urban Land Prices Under Uncertainty”, Amer Econ Rev 75, 505-514.

[78] S. Titman and R. Wessels (1988), “The Determinants of Capital Structure Choice”, J Finan
43, 1-19.

[79] S. Van den Heuvel (2008), “The Welfare Cost of Bank Capital Requirements”, J Mon Econ
55, 298-320.

[80] J. T. Williams (1991), “Real Estate Development as an Option”, J Real Estate Finance Econ
4, 191-208.

A Appendix

In this Appendix we provide the proof of our equilibrium existence theorem 1 for the case where
mortgage contracts are recourse (no default is allowed). In addition, we report the proofs of the
characterization results for our simplified economy of Section 4 and the computation details of
Examples 1 to 3.

A.1 Proof of Theorem 1

Proof of Theorem 1: Our approach is to first construct a generalized game of the economy intro-
duced in Section 2, then prove that the set of equilibria for our generalized game is non-empty,

48
and then show that an equilibrium of the generalized game is in fact a competitive equilibrium
that satisfies all conditions in Definition 1.
To this aim, let us first define the following thresholds:

ω a01 denotes the aggregate endowment of the numeraire good in the first
P
• W01 ≡ a∈A

period. We shall use this threshold to bound consumption of the numeraire in the first
period.

• We shall use threshold Ê ≡ maxk=1,...,K (1/γ k )(W01 ) to find an upper bound on equity
purchases. Notice that we chose threshold Ê in such a way that the property tax pay-
ment, denominated in terms of the numeraire good, cannot exceed the total amount of the
numeraire good in the economy.

• We shall use threshold

X X
W ≡ max{W01 , max ω al1 , max ω a0 (s), max flk (s)(Y )}
l∈L1 l∈S (s,l,k)∈S×L2 ×K
a∈A a∈A

to bound consumption in the second period, where Y = maxk=1,...,K ȳk , where ȳk =
Y
T F Pk (ω al )αlk .
l∈L1

• n ∈ N+ is a parameter that we shall use for bounding consumption of commodities in the

first period. The subtlety is that in this economy the prices of debt and equity stakes can be
very large and, if this happens, the consumption bundles of commodities in the first period
can also be very large. However, we shall find an upper bound for these variables that
depend only on the primitives of the economy and thus we shall conclude that this constant
n will not be binding.

The set of players in this generalized game consists of the set of agents A = I∪{dk }K K
k=1 ∪{Hk }k=1

and the following auctioneers: an auctioneer that chooses period 1 prices, an auctioneer that
chooses period 2 prices, and an auctioneer per jurisdiction that chooses profits.

49
 Next, let us truncate the set of admissible consumption bundles and financial positions corre-
sponding to households, developers, and investors.

• Each household h ∈ {Hk }K h h h h

k=1 chooses a vector (x , D+ , D− , ) in the compact set Ω (n) =
(1+L2 )S
[0, n]1+L1 × [0, 2W ](1+L2 )S × [0, 2(#A)D]2 ⊂ R1+L
+
1
× R+ × R.

• Each developer d ∈ D ≡{dk }K d d d d

k=1 chooses a vector (x , D+ , D− , E ) in the compact set
(1+L2 )S
Ωd (n) = [0, n]1+L1 × [0, 2W ](1+L2 )S × [0, 2(#A)D]2 × [0, 2Ê] ⊂ R1+L
+
1
× R+ ×R×
R+ , where #A denotes the number of agents in the economy.

• Each investor i ∈ I chooses a plan (xi , D+

i
, E i ) in the compact set Ωi (n) = [0, n]1+L1 ×
S(1+L2 K)
[0, 2W ]S(1+L2 K) × [0, 2(#A)D] × [0, 2Ê]K ⊂ R1+L
+
1
× R+ × R × RK
+.

The goal of households, developers, and investors is to maximize their utility function by
choosing a bundle in their respective compact sets

(xh , D+
h h
)h∈H , (xd , D+
d d
, E d )d∈D , (xi , D+
i i
, E i )i∈I ∈ (Ωh (n))HK ×(Ωd (n))K ×(Ωi (n))I


, D− , D− , D−

1+L1 +S(1+L2 )
which satisfies their respective budget constraints. The consumption sets are X a = R+
1+L1 +S(1+KL2 )
if a ∈ D ∪ H, and X i = R+ if i ∈ I.
We now define the auctioneers’ feasible sets.

• The price-auctioneers in periods 1 and 2 choose prices in the following simplexes: ∆L1 =
P1+L1 (1+L2 K)S P(1+L2 K)S
{p ∈ R1+L
+
1
: l=1 p l = 1} and ∆ (1+L2 K)S−1
= {p ∈ R + : l=1 pl = 1},
respectively. These sets are non-empty, convex, and compact.

 
• The profit-auctioneer chooses profits π k in the compact set −λ(αk ) − ε(αk )(1 + I), a∈A ω a01 .
P

The lower bound −λ(αk ) − ε(αk )(1 + I) on the feasible set for profit π k follows from the
fact that parameter γ k must be such that ε(αk ) ≤ γ k , for all k = 1, ..., K. The upper
bound on π k , a∈A ω a01 , is given by the aggregate of the numeraire good in the first period
P

since profits are denominated in units of the numeraire. Notice that we fix an exogenous

50
 upper bound for the sake of simplicity; in fact, we could obtain the upper bound on profits
endogenously.

Auctioneers solve the following optimization problems:

• The price auctioneer in period 1 chooses (p1 , τ , q) ∈ ∆L1 × [0, m]1+K , with m ∈ N+ , in
order to maximize the following function:

X XX X X X
p01 (xa01 − ω a01 ) − p01 δ ak π k + pl1 (xal1 − ω al1 ) + τ a
(D+ a
− D− )+
a∈A a∈A k∈K l∈L1 a∈A a∈A
 
X X X X
+ p01 γ k Eka + qk  Eka − ēak 
k∈K a∈{dk }∪I k∈K a∈{dk }∪I

• The price auctioneer in period 2 chooses (p(s), s = 1, ..., S) ∈ ∆(1+L2 K)S−1 in order to
maximize the following function:
 
X X X X
p0 (s) (xa0 (s) − ω a0 (s)) + plk (s)  xalk (s) − flk (yk )(s) .
s∈S a∈A (s,l,k)∈S×L2 ×K a∈Hk ∪I∪{dk }

• For all k = 1, ..., K, the profit-auctioneer in jurisdiction k chooses π k in the closed set
 
−λ(αk ) − ε(αk )(1 + I), a∈A ω a01 in order to minimize the following function:
P

 2
X
π k − γ k Eka + λ(αk ) + ε(αk )yk 
a∈{dk }∪I

• For all s ∈ S, a house clearing auctioneer chooses the delivery rate φs ∈ [0, 1] to minimize

P P a
!2
k∈K a∈{Dk ∪Hk } Qk (s)D−
φs − P P a
r k∈K a∈{Dk ∪Hk } D−

a
> 0 and minimize (φs − 1)2 if a
P P P P
if k∈K a∈{Dk ∪Hk } D− k∈K a∈{Dk ∪Hk } D− = 0.

51
 We refer to the above generalized game as G(n, m). Next, we verify that the player’s best
response correspondences for this game satisfy the conditions of Kakutani’s fixed point theorem.
First, the objective functions of households, developers, and investors are continuous and
strongly quasi-concave as stated in Assumptions 1.i-iii, and their choice sets are non-empty,
convex, and compact.
Second, for each vector (p1 , (p(s), s = 1, ..., S), τ , q, (π k )k∈K , (φs )s∈S ) of prices, profits and
delivery rates, the choice set of each agent a ∈ A has an interior point. According to Assumptions
4 and 5, the endowments of the numeraire good and construction inputs are strictly positive for
every agent. Moreover, for every agent a ∈ A, p1 ω a1 + p01 k∈K δ ak π k > 0 (this follows because
P

ω a01 > (λ(αk ) + ε(αk )(1 + K))K and p1 ∈ ∆1+L1 ). Thus, xa = 0 and E a = 0, together with a Da
a
< p1 · ω a1 + p01 K a a
P
satisfying inequalities τ D+ k=1 δ k π k and φs D+ > 0 for all s ∈ S, is an interior

point of the budget correspondence. This guarantees the lower hemicontinuity property of the
agent’s budget set correspondence. Since upper hemicontinuity also holds in our setting, we can
use Berge’s Maximum Theorem to claim that, for these players, the best response correspondence
is upper-hemicontinuous with non-empty and compact values. The best response correspondence
also takes convex values – this follows from the convexity of the budget set correspondence
and strongly quasi-convavity of the objective function. For developers, this is also true, but
the convexity property is not immediate. That property follows from the fact that the production
function, which transforms construction inputs into a CRE asset, is concave, and also the fact that
production functions (flk )l∈L1 ,k∈K that assign CRE assets into commercial goods are increasing
and concave by Assumption 2.ii.
Third, the objective function of the profit-auctioneer in each jurisdiction is continuous and
convex and its choice set is non-empty, convex, and compact. In addition, the price-auctioneers’
objective functions are linear and, therefore, continuous and strictly quasi-concave in their choice
variable. Moreover, their choice sets are non-empty, convex, and compact. Thus, for each of these
auctioneers, its best response correspondence is also upper-hemicontinous an takes non-empty,
compact and convex values. Kakutani’s fixed point theorem guarantees that the generalized game

52
G(n, m) has a Nash equilibrium, which is the fixed point of the product of the best response
correspondences.

Lemma A.1: Suppose that Assumptions 4 and 5 hold. Then, if x̄a01 < W01 for all a ∈ A,
there exists a threshold m̄ ∈ N for an equilibrium (x̄A , D̄A , Ē A , p̄, q̄, τ̄ , (π̄)k∈K ) of the generalized
game G(n, m), such that max{q̄k1 , q̄k2 , ..., q̄kK , τ } < m̄.
Proof of Lemma A.1: Let ỹk ≡ T P Fk ·Πl∈L1 (ω dl1k )αlk and (f˜k (s), s = 1, ..., S) ≡ (fk (ỹk )(s), s =
1, ..., S) (here f˜k (s) is the bundle of commercial goods that can be produced in period 2 with the
CRE assets developed by developer dk using his endowment of construction inputs). According
to Assumption 4, a developer dk has enough endowment ω d01k of the numeraire good in the first
period to buy equity of the CRE asset that he develops using his own endowment of construc-
tion inputs. Thus, a developer can always transfer at least an amount of wealth p(s)f˜k (s) from
period 1 to each state of period 2. We conclude that the bundle (ω d01k − gjk (ỹk ), 0, (ω 0 (s), f˜k (s)),
s = 1, ..., S)) is always feasible for developer dk .
According to Assumption 5, given θ ∈ (0, 1), there exists % ∈ R1+L
+
1
such that

udk (W01 , (2W (1, ..., 1), s ∈ S)) < udk (W01 + %, (2θW (1, ..., 1), s ∈ S)),

where % = %(W01 , (2W (1, ..., 1), s = 1, ..., S)). If we take θ ∈ (0, 1), such that 2W (1, ..., 1)θ <
0.7f˜k (s) and 2W θ < 0.7ω d0k (s), for all s ∈ S, then

udk (W01 , (2W (1, ..., 1), s ∈ S)) < udk (W01 + %, (0.7ω d0k (s), 0.7f˜k (s)), s ∈ S))

or

udk (W01 , (2W (1, ..., 1), s ∈ S)) < udk (ω d01k − 0.7γ k ỹk + %̃, (0.7ω d0k (s), 0.7f˜k (s)), s ∈ S)),

where %̃ = W01 − ω d01k + 0.7γ k ỹk + % (notice that parameter %̃ only depends on the fundamen-
tals of the economy). Since x̄a01 < W01 , for all a ∈ A, by monotonicity it is also true that

53
udk (x̄d0,1
k
, (x̄dk (s), s ∈ S)) < udk (W01 , (2W (1, ..., 1), s ∈ S)), which by transitivity implies that

udk (x̄d01k , (x̄dk (s), s = 1, ..., S)) < udk (ω d01k − 0.7γ k ỹk + %̃, (0.7ω d0k (s), 0.7f˜k (s)), s ∈ S)).

This means that developer dk cannot buy %̃ units of numeraire in period 1 with the resources that
he saves when buying only part of the CRE equity, which is p0,1 gk (0.3yk ) + qk 0.3yk . That is,

p̄0,1 %̃ − 0.3p0,1 γ k ỹk

p01 0.3γ k ỹk + qk 0.3ỹk < p̄0,1 %̃ ⇔ qk < m̄m = .
0.3ỹk

Finally, set m̄E = maxk∈K m̄k (observe that, for every k ∈ K, threshold m̄m depends only
on the primitive parameters of the economy). Inequality udk (W01 , (2W (1, ..., 1), s ∈ S)) <
udk (W01 + %, (0.7ω d0k (s), 0.7f˜k (s)), s ∈ S)) also implies

udk (W01 , (2W (1, ..., 1), s = 1, ..., S)) < udk (ω d01k − γ k ỹk + %̂, (0.7ω d0k (s), 0.7f˜k (s)), s ∈ S)),

where %̂ = W01 −ω d01k +γ k ỹk +% (notice that %̂ only depends on the fundamentals of the economy).
Then, developer dk cannot afford the bundle (ω d01k −γ k ỹk +%̂, (0.7ω d0k (s), 0.7f˜k (s)), s = 1, ..., S));
in particular, dk cannot buy the bundle %̂ ∈ R in period 1 using the debt payment that he would
receive by selling the bundle (0.3ω d0k (s), 0.3f˜k (s)), s = 1, ..., S).
Finally, let Ddk be such that rDdk < min{0.3ω d0k (s), 0.3f˜k,1 (s), ..., 0.3f˜k,1 (s)}. Then, −τ Dh <
p̄01 ρ̂, that is, τ < m̄D = ρ̂/(−Ddk )). It just remains to set m̄ ≡ max{m̄E , m̄D } and n̄ = 2W + m̄.
This concludes the proof of Lemma A.1. 

Lemma A.2: An equilibrium (x̄A , D̄A , Ē A , p̄, q̄, τ̄ , (π̄)k∈K ) of the generalized game G(n, m)
for (n, m) > (n̄, m̄) is a competitive equilibrium as defined in Definition 1.
Proof of Lemma A.2: By adding the first period budget constraints of all agents in the

54
economy, we obtain

X X X X
p01 (x̄a01 − ω a01 ) + pl1 (x̄al1 − ω al1 ) + τ D̄a +
a∈A l∈L1 a∈A a∈A
 
X X X X
p01 γ k Ēka + qk  Ēka − ēak  ≤ 0.
k∈K a∈{dk }∪I k∈K a∈{dk }∪I

Then, taking into account the problem of the price-auctioneer in period 1, we conclude that
there is no excess of demand for commodities and assets; that is,

a
− ω a01 − δ ak π k ) + γ k Ēka ≤ 0
P P P P
1. a∈A (x̄01 k∈K k∈K a∈{dk }∪I

a
− ω al1 ) ≤ 0, for all l ∈ L1
P
2. a∈A (x̄l1

D̄a ≤ 0
P
3. a∈A

Ēka ≤ ēdkk , for all k ∈ K ≤ 0

P
4. a∈{dk }∪I

a
− ω a01 − δ ak π k ) + γ k Ēka > 0, then the price auctioneer
P P P P
If a∈A (x̄01 k∈K k∈K a∈{dk }∪I

would choose p01 = 1 and a price equal to zero for the other commodities and assets. This
allows this auctioneer to obtain a positive value for its objective function. However, this is
in contradiction with the aggregation of the budget constraints. The same argument allows us
to conclude that aggregate debt and equity holdings is less than or equal to zero. Finally, if
P a dk
a∈{dk }∪I Ēk − ēk > 0, for some asset jk , then the auctioneer would choose q̄k = m̄ (notice

that x̄a01 < W01 ), which would contradict Lemma A.1 for n > n̄. The same argument applies for
debt.
The inequality in the first numbered list item is equivalent to

X X XX X X
x̄a01 ≤ ω a01 + δ ak π k − γ k Ēka ,
a∈A a∈A k∈K a∈A k∈K a∈{dk }∪I

which in turn is equivalent to

55
 X X X X X
x̄a01 ≤ ω a01 + πk − γ k Ēka
a∈A a∈A k∈K k∈K a∈{dk }∪I

δ ak = 1. Given that π k = γ k Ēka − λ(αk ) + ε(αk )ȳk , we get

P P
because a∈A a∈{dk }∪I

X X X
x̄a01 ≤ ω a01 − (λ(αk ) + ε(αk )ȳk ).
a∈A a∈A k∈K

By adding the budget constraints of all agents over all states of nature in period 2, we obtain

 
X X X X
p̄0 (s) (x̄a0 (s) − ω a0 (s)) + p̄lk (s)  x̄al (s) ≤
s∈S a∈A (s,l,k)∈S×L2 ×K a∈{dk }∪Hk ∪I
X XX X
≤ Sr D̄a + Ēka ck (ȳk ; p̄(s))
a∈A s∈S k∈K a∈{dk }∪I
X XX
≤ Sr D̄a + ck (ȳk )
a∈A s∈S k∈K

D̄a ≤ 0, we can rewrite the above

P
Using the definition of ck (yk ) and the fact that Sr a∈A

inequality as follows:

 
X X X X
p̄0 (s) (x̄a0 (s) − ω a0 (s)) + p̄lk (s)  x̄al (s) − flk (ȳk )(s) ≤ 0.
s∈S a∈A (s,l,k)∈S×L2 ×K a∈{dk }∪Hk ∪I

Then, given the problem of the auctioneer of period 2, we conclude that there is no excess of
demand for commodities at date 2, i.e.,

a
− ω a0 (s)) ≤ 0, for all s ∈ S;
P
5. a∈A (x̄0 (s)

x̄a (s) − fk (yk )(s) ≤ 0 for all s ∈ S and k ∈ K.

P
6. a∈{dk }∪Hk ∪I

56
 In equilibrium, we also have that, for every jurisdiction k ∈ K, profits are

X
πk = γ k Ēka − (λ(αk ) + ε(αk )ȳk ) .
a∈{dk }∪I

Observe that this value for the profit function belongs to the profit-auctioneer’s strategy set; more-
over, this is the value that minimizes that auctioneer’s objective function.
Next, we prove that there is no excess of supply of commodities other than construction
inputs. For if there were excess supply of one of the commodities in either period 1 or period
2, then the respective price-auctioneer would choose the price of that commodity equal to zero.
However, this would be in contradiction with the existence of an optimal plan for the agents of
this economy, since utility functions are increasing.
If there is excess supply of a construction input, then the price-auctioneer would choose its
price equal to zero and this would be in contradiction with the existence of an optimal plan for
developers. To see this, notice that by increasing the purchased amount of construction inputs, the
developer could increase his income in period 1 to spend it on the consumption of the numeraire
good 01 (recall that the utility function is increasing in the consumption of the numeraire good).
In addition, there is no excess supply of debt or equity. Again, we prove this by contradiction.
If there were excess supply of debt or equity, the price-auctioneer would choose the price of that
particular asset equal to zero and this would contradict the existence of an optimal plan because,
by the monotonicity of the preferences, debt pays strictly positive returns in every state of nature
and equity also pays strictly positive returns in all states of nature if pk (s)  0 and Assumption
2.iii hold. Observe that, in a Nash equilibrium of the generalized game, pk (s)  0 for all s ∈ S,
by monotonicity of preferences in period 2.
Finally, we establish the optimality of consumption plans. For this, first notice that, for each
agent a ∈ A, (x̄a , D̄a , Ē a ) satisfies the budget constraint given prices and profits (p̄, q̄, τ̄ , (π̄)k∈K ).
Also, (x̄a , D̄a , Ē a ) belongs to the interior of Ωa (n). Therefore, by the convexity of the budget sets
and the strongly quasi-concavity of the utility functions, we have that the bundle (x̄a , D̄a , Ē a ) is

57
optimal in the budget set with prices and profits (p̄, q̄, τ̄ , (π̄)k∈K ).

A.2 Equilibrium characterization proofs

Proof of Proposition 1: Proposition 1 follows from the market clearing equation for the nu-
meraire good at state s ( k pk (s)Yk (s)yk = ω h0 k (s) + Q1 (s)D−
d1 d2
P
+ Q2 (s)D− ) and equation
P h1 h2
k pk (s)Yk (s)yk = φs (D+ + D+ ) (using the young households budget constraints and the mar-

ket clearing equations for the numeraire good, the commercial goods at state s, and mortgage
debt). 

Proof of Corollary 1: Corollary 1 follows from expression (6), the young households budget
constraints, and the market clearing equations for mortgage debt. 

Proof of Proposition 2: Proposition 2 follows from young household hk ’s optimality condi-

hk
(τ = s (λhs k /λh1 k )φs ) and the developer dk ’s optimality condition with
P
tion with respect to D+
dk
(τ = s (λds k /λd1k )Qk (s) + ξ dk σ k /λd1k ). 
P
respect to D−

Proof of Corollary 2: A higher 1 − Qk (s) decreases φs . Necessary optimality condition

P hk hk hk
τ = s (λs /λ1 )φs for D+ > 0 then implies that τ must decrease when default loss 1 −

Qk (s) increases. Statement in Corollary 2 then follows by this fact and the necessary optimality
condition τ = s (λds k /λd1k )Qk (s) + ξ dk σ k /λd1k for D−
dk
P
> 0. 

Proof of Proposition 3: Proposition 3 follows from developer dk ’s necessary optimality con-

P  
dition for Ekdk > 0 and Proposition 2’s equilibrium condition (8), where ξ dk = (λd1k /σ k ) s φ̂s − Q̂k (s) .


Proof of Proposition 4: Proposition 4 follows from the developer dk ’s necessary optimality

condition with respect to xd1k > 0,

X  
dk dk dk dk αk −1
p1 = (λs /λ1 ) Ek pk (s)Yk (s)T F Pk αk (x1 ) (13)
s

Then, equating (13) for developers d1 and d2 and using α1 = α2 , we obtain after some algebra

58
the equilibrium property (10). 

Proof of Corollary 3: Equilibrium property (11) follows from expression (10) and the as-
sumption of binding collateral constraints (ξ d1 > 0 and ξ d2 > 0). 

Proof of Proposition 5: A higher 1 − Q2 (2) translates into a lower mortgage delivery rate
at state s2 (φ2 ), in turn decreasing the mortgage discount market price τ (τ = s (λhs k /λh1 k )φs ,
P

hk
young household hk ’s optimality condition with respect to D+ > 0) and the developer d2 ’s
d2
amount of debt D− (market price auctioneer and debt market clearing).
P  
The initial shock also increases expected default risk s φ̂s − Q̂2 (s) in jurisdiction k = 2,
making the developer d2 ’s collateral constraint more binding (ξ d2 increases due to equilibrium
P  
condition s φ̂s − Q̂k (s) = ξ dk σ k /λd1k ; see proof of Proposition 2), in turn increasing the
CRE asset j2 ’s equity market price q2 (qk + γ k = s (λds k /λd1k )ck (yk ; p)(s) + ξ dk , see proof of
P

Proposition 3) and the investor i’s equity E2i in project j2 (market price auctioneer and equity
market clearing). This crowds out developer d2 ’s equity share in project j2 (i.e., E2d2 decreases
since E2d2 + E2i = 1).
The investor’s equity porfolio rebalancing toward jurisdiction k2 decreases investor i’s equity
E1i and the equity market price q1 of CRE asset j1 (following the investor’s budget constraint ω i0 =
(q1 + γ 1 )E1i + (q2 + γ 2 )E2i ). As a result, developer d1 holds the remaining equity in CRE project
j1 (i.e., E1d1 increases since E1d1 + E1i = 1). This additional constituted equity collateral increases
d1
developer d1 ’s debt when his collateral constraint is binding (i.e., D− = σ 1 E1d1 increases).
Finally, notice that expression (10) in Proposition 4 then implies that the CRE height in juris-
diction k1 with respect to jurisdiction k2 . 

Proof of Proposition 6: The statements in Proposition 6 corresponding to Lemmas 2, 5, and

7 below.
Households h1 and h2 are excluded from the CRE equity market and, therefore, the only
financial instrument that they can use to transfer wealth from the first to the second period is
risk-free debt.26 Because they prefer to consume tomorrow, they will sell their numeraire good
26
Household’s debt could be seen as a deposit of this household in a bank account. For the sake of simplicity, we

59
endowment and purchase as much risk-free debt as possible (risk-free debt pays 1 unit of the
numeraire good in both states of the second period). Thus, in equilibrium we expect27

1 h1
Dh1 = ω >0 (14)
τ 0
1
Dh2 = ω h0 2 > 0 (15)
τ

Households purchase the commercial good produced in their respective jurisdictions and,
therefore, the market clearing of the commercial good occurs at the jurisdictional (local) level.
This implies that xh1 1 (s) = Y1 (s)T F P1 (xd11 )α1 and xh2 2 (s) = Y2 (s)T F P2 (xd22 )α2 , for s = s1 , s2 .
Because investor i prefers consumption of the numeraire good in the second period, we expect
him to sell his endowment of good 0 and buy CRE equity in one or both jurisdictions. The
following condition follows from the investor’s budget constraint in the first period:

(q1 + γ 1 )E1i + (q2 + γ 2 )E2i = ω i01 (16)

Next, we prove Proposition 1 with a series of lemmas.

Lemma 1: The aggregate costs of private development and local public good provision equal
the total amount of numeraire good available in the economy in the first period, i.e.,

X   α1
p ω H1 + p ω H2 + (λ(αk ) + ε(αk )T F P1 xd11k ) = ω h011 + ω h012 + ω i01 (17)
| 1 1 {z 2 2 }
cost of construction inputs k∈K
| {z }
cost of local public good provision

Proof: Equation (17) follows from the old households’ budget constraint in period 1 and the
market clearing equation for the numeraire good. 

Lemma 2: Scarcity of a commercial good drives the price differential of this good between
states of nature. In particular, the price of a commercial good becomes more expensive at state
ignore banks as potential financial intermediaries in this economy.
27
Conditions (14) and (15) are useful to obtain the following Lemmas 2 to 5 in the proof of Proposition 1.

60
s relative to state s0 6= s if the amount of the commercial good produced and sold at s is smaller
than at state s0 . In particular,

p1 (1)/p1 (2) = Y1 (2)/Y1 (1) (18)

p2 (1)/p2 (2) = Y2 (2)/Y2 (1) (19)

Moreover, if the value of one unit of the commercial good produced in a jurisdiction is the same at
both states, consumption of the numeraire good is the same at both states for the two developers
and the investor, i.e., xd1 d1 d2 d2 i i
0 (1) = x0 (2), x0 (1) = x0 (2), and x0 (1) = x0 (2).

Proof: The budget constraint of households h1 and h2 at states s1 and s2 are such that

α1
rDh1 = p1 (s)Y1 (s)T F P1 xd11 , for s = s1 , s2 (20)

α2
rDh2 = p2 (s)Y2 (s)T F P2 xd22 , for s = s1 , s2 (21)

Conditions (18) and (19) follow from conditions (20) and (21), respectively (i.e., dividing the
corresponding expression for state s1 by the corresponding expression for state s2 ). Moreover,
conditions (18) and (19), and the developers’ budget constraints in the second period imply that
xd01 (1) = x0d1 (2) and xd02 (1) = xd02 (2). These equalities, together with the market clearing con-
ditions of the numeraire good at states s1 and s2 , and assumption ω H H2
0 (1) = ω 0 (2), imply that
1

xi0 (1) = xi0 (2). 

Lemma 3: At each state of the second period, the sum of CRE cash flows and debt promises
equals the total amount of the numeraire good in the economy, i.e.,

X   αk
pk (s)Yk (s)T F Pk xdkk + r(Dd1 + Dd2 ) = ω H
0 (s) , for s = s1 , s2
1
(22)
k∈{1,2}

Proof: We obtain equation (22) using equalities xd01 (1) = xd01 (2) and xd02 (1) = xd02 (2), to-
gether with the investor and developers’ budget constraints at states s1 and s2 , and the market

61
clearing equation for the numeraire good in the second period. 

Lemma 4: The relative value of commercial goods produced in jurisdictions k1 and k2 is

driven by the relative amounts of the numeraire good endowments that households h1 and h2
have in the first period. In particular,

α1
p11 (s)Y1 (s)T F P1 xd11 ω h0 1

=
d2 α2
(23)
p21 (s)Y2 (s)T F P2 x2 ω h0 2

Proof: Condition (23) follows by equating the price of debt (τ ) that results from households
h1 and h2 ’ budget constraints of period 0 and state s1 . 

Roughly speaking, differences in wealth between households in different jurisdictions deter-

mine the difference in credit that these households extend to developers. Because households use
their loan payments in the second period to purchase the commercial good of their respective ju-
risdiction, differences in loan amounts determine the differences in commercial goods valuation
and also the differences in (endogenous) cash flows generated by the different CRE assets.

Lemma 5: The CRE equity price accrued of the property tax in a jurisdiction is driven by the
value of the commercial good produced and sold in that jurisdiction. In particular,

r (q1 + γ 1 ) = 2(p1 (1)Y1 (1)) (24)

r (q2 + γ 2 ) = 2(p2 (1)Y2 (1)) (25)

Proof: Conditions

λi1 (q1 + γ 1 ) = 2(λi (s1 ) + λi (s2 ))(p1 (1)Y1 (1))

λi1 (q2 + γ 2 ) = 2(λi (s1 ) + λi (s2 ))(p2 (1)Y2 (1))

follow from investor i’s first order conditions with respect to CRE equity E1i and E2i , respectively,
where λi1 and λi (s) are the shadow values of investor’s budget constraints in period 1 and state

62
s of period 2, respectively. These equations take into account conditions (18) and (19).28 With
risk-free debt, we have that λi1 /(λi (s1 ) + λi (s2 )) = r and, therefore, conditions (24) and (25)
follow accordingly. 

Lemma 6: When developers borrow at their maximum capacity, the total amount of debt in
the economy equals the resources that households have in the first period, i.e.,

X
τ (D̄d1 + D̄d2 ) = ω h0 k (26)
k={k1 ,k2 }

Moreover, the equilibrium price of debt τ must satisfy the following equations:


 d1
d1 α1
τ = p11 ω H
11
1
− (q 1 + γ 1 )E i
1 + γ 1 T F P1 x 1 /D̄ (27a)

 d2
d2 α2
τ = p21 ω H
21
1
− (q 2 + γ 2 )E i
2 + γ 2 T F P2 x 2 /D̄ (27b)

Proof: Equations (26), (27a), and (27b) follow from conditions (14) and (15), together with
the market clearing equation for debt, the market clearing condition for equity (at the jurisdiction
level), and the developers’ budget constraint in the first period. 

Developers are the only agents in this economy who can create CRE assets. For that, they
need to buy construction inputs. The respective market clearing conditions imply that these
purchases must be such that xd11 = ω H d2 H2
1 and x2 = ω 2 . In addition, the equilibrium must satisfy
1

the following condition:

Lemma 7: The difference in prices of construction inputs in different jurisdictions is driven

by the difference in marginal productivity of the CRE assets. In particular,

2 α2 −1 1 α1 −1
p1 − p2 = α2 T F P2 (ω H
2 ) p2 (1)Y2 (1) − α1 T F P1 (ω H
1 ) p11 (1)Y11 (1) (28)

Proof: (28) follows from the first order optimality conditions of developers d1 and d2 with
28
Notice that (24) and (25) assume that the shadow values of sign constraints Ejdkk ≥ 0 and Ejdkk ≤ yjdkk are zero
(below, we will verify that our equilibrium is such that these constraints are non-binding for both k = k1 , k2 ).

63
respect to construction inputs 11 and 21, respectively, and conditions (18) and (19). 

A.3 Numerical examples

In this subsection, we provide all details regarding the computation of examples corresponding
to Figures 1, 2, and 3 (Examples 1, 2, and 3, respectively). In addition, here we also discuss how
poor CRE asset performance may affect price inflation of commercial goods in a jurisdiction
(Example 4).

Example 1 (benchmark): For the parameter values in Example 1, we obtain a unique equilib-
rium solution where q1 = q2 = 2.500, τ = 1.440, p11 = p21 = 1.500, p11 (1) = p21 (1) = 1.500,
p11 (2) = p21 (2) = 3.000, and E1i = E2i = 0.173.29 Figure 1 illustrates the equilibrium values of
the different capital structure components of CRE assets j1 and j2 , namely, the developer’s debt,
the developer’s equity, and the investor’s equity. These quantities are expressed in real terms, i.e.,
nominal amount times the corresponding price.
When analyzing the capital structure of a CRE development project, analysts look at financial
ratios, such as the debt-to-equity ratio or the ratio of an investor’s equity-to-total-equity. A debt-
to-equity ratio higher than 0.500 indicates that the CRE capital structure has a greater proportion
of its capital funding from lenders rather than equity investors. An “investor’s equity-to-total-
equity” ratio higher than 0.500 indicates that the investor owns more than 50 percent of the
equity of a CRE asset. Our theory obtains these ratios endogenously determined in equilibrium.
For example, the ratios corresponding to our benchmark example in Figure 1 are 0.576 for the
“developer’s debt-to-total-equity” ratio, 0.827 for the “developer’s equity-to-total-equity” ratio,
and 0.173 for the “investor’s equity-to-total-equity” ratio. The respective ratios are the same in
both jurisdictions. 

Example 2 (funding capacity): We modify our benchmark example by increasing the debt
29
We first obtain the value of equilibrium variables q1 , q2 , τ , p11 , p12 , p11 (1), p12 (1), p11 (2), p12 (2), E1i , and E2i
by solving the following system of equations: (17), (18), (19), (22), (23), (16), (24), (25), (26), (27a), (27b), and
(??). These equilibrium values, in turn, allow us to solve for the rest of the equilibrium variables by using the agents’
budget constraints and market clearing equations.

64
limit for developer d1 to D̄d1 = 1.1, while decreasing the debt limit for developer d2 to D̄d2 =
0.9.30 Roughly speaking, access to funding for developers in the first jurisdiction improves, while
it worsens for those in the second jurisdiction. The new equilibrium is such that q1 = q2 = 2.500,
τ = 1.440, p11 = p21 = 1.500, p11 (1) = p21 (1) = 1.500, p11 (2) = p21 (2) = 3.000, E1i = 0.125,
and E2i = 0.221. Figure 2 illustrates the equilibrium values of the different capital structure
components of CRE assets j1 and j2 for the new parameter values. As mentioned in Section 3,
we see that, compared to our benchmark example, the developer with worse funding capacity
(d2 ) decreases his absolute real exposure to CRE debt and equity, while the investor increases his
equity exposure. The opposite happens in the CRE development project of the developer with
better funding capacity.
We can get further insights into the composition of the two capital structures by looking at and
comparing financial ratios. For the parameter values of example 2, we find a “developer’s debt-
to-total-equity” ratio equal to 0.634 in CRE asset j1 and 0.518 in CRE asset j2 ; a “developer’s
equity-to-total-equity” ratio equal to 0.875 in CRE asset j1 and 0.779 in CRE asset j2 ; and an
“investor’s equity-to-total-equity” ratio equal to 0.125 in CRE asset j1 and 0.221 in CRE asset j2 .
We summarize the equilibrium values of the financial ratios under consideration for Examples 1
and 2 in table 1.

Example 1 Example 2
(benchmark example) (funding capacity)
k1 k2 k1 k2
developer’s debt to total CRE equity 0.576 0.576 0.634 0.518
developer’s equity to total CRE equity 0.827 0.827 0.875 0.779
investor’s equity to total CRE equity 0.173 0.173 0.125 0.221
Table 1: This table reports the financial ratios of the equilibria corresponding to Examples 1 and 2.

Compared to our benchmark example, we conclude that both the developer’s debt-to-equity
ratio and equity-to-total-equity ratio decrease (increase) for the CRE capital structure correspond-
30
By offsetting the decreasing in D̄d2 with an increase in D̄d1 , we are able to keep the remaining components of
equations (22) and (26) with the same equilibrium value as in the benchmark example.

65
ing to the developer with worse (better) funding capacity, while the investor’s equity-to-total-
equity ratio increases (decreases, respectively). 

Example 3: Now suppose that jurisdiction 2 experiences an increase in its fixed costs λ̂2 from
0.2. to 0.5, and that the jurisdiction’s manager responds by increasing its tax rate γ 2 from 0.5 to
0.8.31 The other parameter values are as in Example 1. In this case, q1 = 2.500, q2 = 2.200,
τ = 1.590, p1 = 1.572, p2 = 1.428, p1 (1) = p2 (1) = 1.500, p1 (2) = p2 (2) = 3.000, E1i = 0.147,
and E2i = 0.199.
Because the property tax γ 2 is paid by both the developer d2 and the investor i, we expect
that this tax increment has an impact on how these two agents allocate their resources. Figure
3 illustrates this. The increase in γ 2 decreases the developer d2 ’s equity compared to the equi-
librium value of our benchmark Example 1. As explained in Section 3, this change is due to
the developer’s substitution of equity for tax free debt. In particular, developer’s debt increases
because τ jumped from 1.440 (in Example 1) to 1.590 (in Example 3), while Dd2 remained equal
to D̄d2 = −1. The developer d2 ’s debt and the investor’s equity contributions increase as a result.
In the other jurisdiction, the investor’s equity contribution decreases. Interestingly, the in-
crement in E2i and the decrease in E1i respond to the change in the equity price q2 . Compared
to Example 1, q2 has decreased from 2.500 to 2.200, while q1 has remained the same (2.500).
Roughly speaking, equity in CRE asset j2 has become relatively cheaper. In CRE asset j1 , we
also see that the developer’s debt (τ Dd1 ) and equity contributions (q1 E1d1 ) increase.

Example 1 Example 3
(benchmark) (Tax policy γ 2 )
k1 k2 k1 k2
developer’s debt to total CRE equity 0.576 0.576 0.636 0.723
developer’s equity to total CRE equity 0.827 0.827 0.853 0.801
investor’s equity to total CRE equity 0.173 0.173 0.147 0.199
Table 2: This table reports the financial ratios of the equilibria corresponding to Examples 1 and 4.

31
Notice that these changes are such that neither jurisdiction k2 ’s profits nor the profit components of equilibrium
equations described in the above simplified economy change.

66
 Table 2 reports the equilibrium values for the three financial ratios under consideration. It
provides a complementary perspective on the comparison between the capital structures (in real
absolute amounts) in the two CRE assets. There we see that the debt-to-equity ratio increases for
j2 , while the developer’s equity-to-total-equity ratio E2d2 /(E2i + E2d2 ) decreases. This responds to
the developer d2 ’s substitution effect between equity and tax free debt. The investor’s equity-to-
total-equity ratio E2i /(E2i + E2d2 ) makes up the difference. 

Example 5: Let us consider again the same parameter values as in the benchmark example,
except that now we modify the amount of the commercial good produced at state s2 by CRE asset
j2 ; in particular, let Y2 (2) decrease from 0.5 to 0.1. Possible reasons are a natural disaster that
negatively impacts the CRE asset’s production capacity, a reduction in the supply of intermediate
inputs captured by f2 (y2d2 )(2), or even political reasons, such as a policy of expropriation of
resources. In this new equilibrium, only the the price of the commercial good 21 at state s2
changes (p2 (2) = 15.000). The values of other equilibrium variables, including the financial
ratios under consideration, do not change with respect to the benchmark example.32 

32
In particular, q1 = q2 = 2.500, τ = 1.440, p1 = p2 = 1.500, p1 (1) = p2 (1) = 1.500, p1 (2) = 3.000,
p2 (2) = 15.000, and E1i = E2i = 0.171.

67