The World’s Largest Open Access Agricultural & Applied Economics Digital Library

This document is discoverable and free to researchers across the

globe due to the work of AgEcon Search.

Help ensure our sustainability.

Give to AgEcon Search

AgEcon Search
http://ageconsearch.umn.edu
[email protected]

Papers downloaded from AgEcon Search may be used for non-commercial purposes and personal study only.
No other use, including posting to another Internet site, is permitted without permission from the copyright
owner (not AgEcon Search), or as allowed under the provisions of Fair Use, U.S. Copyright Act, Title 17 U.S.C.
Journal of Agricultural and Resource Economics 45(3):410–427 ISSN: 1068-5502 (Print); 2327-8285 (Online)
Copyright 2020 the authors doi: 10.22004/ag.econ.302463

The On-Farm and Near-Farm Effects of Wind

Turbines on Agricultural Land Values
Gabriel S. Sampson, Edward D. Perry, and Mykel R. Taylor

We estimate the effects of utility-scale wind turbines on agricultural land values in Kansas using
parcel-level transaction data from 2001 to 2017 in a hedonic price model. By matching transaction
data and wind turbine data at the common land units scale, we are able to ascertain on-farm effects
as well as near-farm effects. Across all our analyses, the preponderance of results suggests that
wind turbines do not affect agricultural property values, either on-farm or nearby, in a statistically
significant way. Thus, our results cannot confirm that wind turbines will increase land values when
installed on a parcel.

Key words: farmland, hedonic, land values, turbines, valuation, wind

Introduction

Wind power constitutes an important component of renewable power portfolios. From 2000 to
2013, wind power capacity in the United States grew from 2.5 gigawatts to over 60 gigawatts. The
U.S. Department of Energy projects total wind capacity to expand to over 220 gigawatts by 2030.
Despite the alleged benefits of wind power for mitigating carbon emissions and reducing reliance
on nonrenewable fuel sources, there are a number of unresolved controversies over the placement
of wind turbines and their potential influence on property values (e.g., Khatari, 2004; Groothuis,
Groothuis, and Whitehead, 2008; Heintzelman and Tuttle, 2012; Sunak and Madlener, 2017). The
projected future growth of the wind energy industry is likely to exacerbate these controversies.
This paper combines a rich set of agricultural land sale transaction data with location-specific
data on utility-scale wind turbines in Kansas to analyze the effects of wind turbines on agricultural
land values. By matching land sale transaction data and wind turbine data at the common land units
(CLU) scale, we are able to estimate two types of effects: (i) the effect of having one or more turbines
on the value of a parcel (i.e., the on-parcel impact) and (ii) the effect of having one or more turbines
on the value of nearby parcels (i.e., off-parcel impacts).
The literature examining the impacts of wind turbines on property values is still relatively new
and has produced mixed conclusions. Wind turbines have been characterized as having disamenity
effects on property values (Vyn and McCullough, 2014, review the literature). These effects include
concerns over noise generation, possible health effects, bird deaths, ice throw, and negative visual
effects. However, despite widespread concerns that negative public perception about wind turbines
is capitalized into property values (Ladenburg and Dubgaard, 2007; Krueger, Parsons, and Firestone,
2011; Heintzelman and Tuttle, 2012; Sunak and Madlener, 2017; Jensen et al., 2018), a number of
studies suggest that proximity to wind turbines has no impact on land values (Laposa and Mueller,
2010; Hoen et al., 2011, 2013; Lang, Opaluch, and Sfinarolakis, 2014; Vyn and McCullough, 2014;

Affiliations: Gabriel S. Sampson is an assistant professor, Edward D. Perry is an assistant professor, and Mykel R. Taylor is
an associate professor in the Department of Agricultural Economics at Kansas State University.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Review coordinated by Richard T. Woodward.
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 411

Center for Economic Development and Business Research, 2019) or may have differential impacts
depending on whether communities support or oppose wind energy (Vyn, 2018; Boyle et al., 2019).
Previous studies have largely focused on how turbines impact residential properties rather than
agricultural properties (two exceptions are Vyn and McCullough, 2014; Shultz, Hall, and Strager,
2015). Residential land studies provide useful information, but many wind farms are located in
agricultural regions and there are reasons that wind farms may have different effects on agricultural
land values. For example, recent research has shown that wind farms can alter local temperature and
precipitation (Li et al., 2018) as well as increase nearby crop yields (Chen, 2019; Kaffine, 2019).1
Thus, in contrast to residential property values, there is the possibility that agricultural land values
may actually increase in response to placement of nearby wind turbines if the crop yield effects are
large.
An additional consideration specific to agricultural properties is the on-farm effects of having
a wind farm. Specifically, for agricultural landowners that contract with wind energy companies to
lease land use rights, a natural question that arises is what the presence of turbines implies for the
value of their land. In promoting their Wind Powering America program, the U.S. Department of
Energy (2004) estimated that wind energy would provide $1.2 billion in new income to agricultural
and rural landowners. There is some recent evidence that wind lease payments can raise the value of
one’s land (Myrna, Odening, and Ritter, 2019). However, turbines can negatively impact irrigability
and farmability factors such as equipment maneuverability and drainage (Baker et al., 2018).
Kansas is a leading state in agricultural production and also ranks highly in wind energy
potential. However, stakeholder opinion is starkly divided over the prospect of wind energy
expansion (Dodge, 2019; Lefler, 2019; Shorman, 2019). Analyzing the impacts of wind energy
projects on agricultural land values can therefore assist policymakers in conducting cost–benefit
analysis of wind energy expansion. Moreover, understanding how on-farm land values are affected
by wind turbine installations can provide information to landowners interested in diversifying their
farm incomes and wealth portfolios.
A unique aspect of our research is the use of parcel-level sales data of every agricultural land
transaction in Kansas from 2001 to 2017. In addition, we have data on all 2,506 utility-scale wind
turbines constructed in Kansas between 2001 and 2017. We spatially match the turbine data layer
to 14,196 total agricultural land transactions occurring in counties with at least one wind turbine for
the years 2001 through 2017. Of these 14,196 transactions, 1,530 parcels were sold at least twice.
Methodologically, this paper takes the hedonic price model approach to estimate the effect of
proximity to wind turbines on land values. The treatment groups are defined by various measures
of proximity, including the inverse distance of a parcel to the nearest turbine, a set of dummies
representing whether a parcel belongs to a set of concentric rings about a turbine, and a dummy
for whether a parcel has turbines directly on it. To control for the possibility that placement of
wind turbines is correlated with omitted variables, we include a rich set of spatial dummies up to
the resolution of township-level (461 total). Year and month dummies are also included to control
for idiosyncratic temporal factors influencing land sales (e.g., commodity price fluctuation, interest
rates). Lastly, we are able to exploit a subsample of 1,530 repeat sales in a parcel-level fixed effects
framework (i.e., the finest level of controls possible).
While we do find positive land value effects in certain specifications, the preponderance of our
hedonic estimates provides little to no evidence of statistically significant impacts of wind turbines
on agricultural land values.2 Thus, our results cannot confirm that wind turbines will increase
land values when installed on a parcel. One interpretation of this result is that the lease payments
negotiated between wind energy companies and land owners are on average approximately equal to

1 Producers in Harper County, Kansas, have reported more rainfall and less hail measured on weather gauges that are

closer to wind turbines (Davis, 2018).

2 The positive effects are restricted to the inverse distance treatment group. However, the positive effect goes away when

parcel-level fixed effects are included.

412 September 2020 Journal of Agricultural and Resource Economics

Figure 1. Location of Wind Turbines and Parcel Sales Data

land owners’ minimum willingness to accept (i.e., the compensation offsets any disamenity values).3
Additionally, our results suggest that wind turbines have not produced statistically significant
negative external effects for agricultural property values in Kansas.
This paper is most closely related to the studies by Vyn and McCullough (2014), who estimate
agricultural property value impacts using 1,590 farmland sales in southern Ontario, and Shultz,
Hall, and Strager (2015), who use a cross-section of 2,912 assessed agricultural land values in
Pennsylvania. Both studies find little evidence of wind farm impacts. The present study not only
uses a much larger dataset (over 14,000 sales) across a longer time horizon but is also the first to
consider both the on-farm and near-farm land value impacts in the United States using arms-length
transaction data (rather than assessed values).4

Background and Study Area

In 2017, Kansas ranked among the top five states in wind energy generation and in future wind
energy potential (U.S. Energy Information Administration, 2018). Over one-third of Kansas’
electricity was generated from wind power in 2017. As of 2018, total installed wind capacity
was over 5,500 megawatts (MW) generated by 2,996 wind turbines. An additional 1,600 MW
of wind capacity is under construction or in development (American Wind Energy Association,
2019). However, landowners are divided in their opinion over wind energy expansion in Kansas
(Dodge, 2019; Lefler, 2019; Shorman, 2019). One faction views wind energy as a boon due to job
creation, tax revenues, and lease payments to landowners. Another faction is concerned with possible
impacts to neighboring livestock, noise pollution, loss of enjoyable viewsheds, and depressed land
values. In fact, Sedgwick County (home of Wichita—the largest city in Kansas) recently banned
the development of large-scale wind turbines (Lefler, 2019), while commissioners in Reno County
rejected a permit for a new 220 MW wind farm (Shorman, 2019).
The wind power project data used in this study include the location and specification of all utility-
scale wind turbines installed and operational up through the year 2017 (Figure 1). We therefore limit

3 The most common compensation structure is a royalty or fixed-fee arrangement made to the landowner on a monthly

or yearly basis. Such structures tie the value of the wind project to the land. Lump-sum payments are the least common and
would result in the value of the wind project not being tied to the land (Windustry, 2009).
4 Ma and Swinton (2012) demonstrate that hedonic estimates can be downward biased when assessed values are used

instead of arms-length transactions.

Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 413

Table 1. Study Area Wind Facilities and Land Transactions

Corn and
2017 Average Soybean
Capacity 2017 Startup 2010 Parcel Sales Acres
County (MW) Turbines Year Population Sales ($/acre) (thousands)
Barber 183.2 92 2009 4,861 670 4,134 5.4
Butler 151.0 101 2005 65,880 1,299 3,142 80.0
Clark 429.0 208 2016 2,215 378 2,828 1.4
Cloud 201.4 70 2008 9,533 530 2,417 48.0
Coffey 199.0 95 2015 8,601 524 2,099 96.6
Elk 199.8 111 2011 2,882 571 3,246 14.2
Ellis 206.5 115 2013 28,452 524 1,948 3.5
Ellsworth 186.0 124 2008 6,497 460 1,890 3.6
Ford 417.2 235 2006 33,848 767 2,255 53.7
Grant 112.9 61 2013 7,829 705 3,013 49.1
Gray 507.8 342 2001 6,006 867 3,193 93.7
Harper 281.6 176 2012 6,034 745 1,930 5.7
Haskell 136.9 74 2013 4,256 622 4,557 108.4
Kingman 104.0 65 2012 7,858 586 2,071 19.0
Kiowa 116.6 76 2010 2,553 414 2,796 36.3
Lincoln 264.3 165 2008 3,241 632 2,168 12.3
Marshall 72.0 36 2016 10,117 605 3,194 168.0
Ness 168.3 94 2015 3,107 504 1,923 4.3
Pratt 208.3 121 2016 9,656 529 2,265 73.2
Rush 46.0 20 2015 3,307 453 1,148 9.0
Sumner 150.0 75 2015 24,132 1,153 2,170 68.0
Trego 30.4 17 2015 3,001 356 1,603 8.9
Wichita 99.0 33 2009 2,234 354 2,094 34.9

the analysis to land sale transaction data for the years up to 2017. In total, 23 counties had active
wind energy projects in 2017 (see Table 1), including 2,506 active wind turbines, ranging in size
from 0.05 MW to 3.0 MW, with a total wind energy capacity of 4,471 MW. Projects range in size
from 0.07 MW to 419 MW, with an average capacity of 124 MW.
Table 1 summarizes the number and average value of the land transactions across the 23 counties
with wind energy projects, along with 2010 census populations and thousands of acres planted to
corn or soybeans, obtained from the USDA National Agricultural Statistics Service. Average land
values range from about $1,148/acre in Rush County to over $4,500/acre in Haskell County. Average
land values are generally larger for counties having more corn- and soybean-planted acres. The
notable exceptions are Barber and Elk Counties, which are likely influenced by their proximity to
Wichita.

Data

The data used in the analysis are taken from a variety of sources, at the finest resolution possible.

Land Transactions

To conduct the analysis, we leverage parcel-level sales data for every agricultural land transaction
of at least 40 acres in size in Kansas from 2001 to 2017. The data were obtained from the
Property Valuation Division (PVD) of the Kansas Department of Revenue (Figure 1). In order to be
characterized as a farmland transaction, a parcel must be at least 75% cropland by area. We restrict
414 September 2020 Journal of Agricultural and Resource Economics

Table 2. Summary Statistics

Variable (units) Mean Std. Dev. Min. Max.
Price per acre ($/acre) 2, 604.9 2, 734.2 139.3 16, 854.0
Total agricultural acres 172.2 128.1 40.0 1, 429.8
Percentage of parcel irrigated (%) 8.2 23.8 0.0 100.0
Commute time to 10,000 population (hrs) 0.8 0.4 0.0 2.0
Commute time to 40,000 population (hrs) 1.7 0.9 0.4 3.8
Root zone available water storage (mm) 232.2 58.8 50.0 330.0
Soil organic carbon (kg/m2 ) 9.7 3.0 1.8 25.4
Percentage of parcel with acidic soils (%) 0.8 6.3 0.0 100.0
Percentage of parcel with basic soils (%) 50.3 43.0 0.0 100.0
Slope (%) 3.5 2.5 0.0 21.5
Elevation (ft) 605.6 228.8 205.0 1, 448.0
Growing season precipitation (inches) 19.8 4.1 13.1 29.0
Evapotranspiration (inches) 34.8 1.6 30.8 38.1
Degree days between 10◦ and 32◦ Celsius 2, 074.8 104.8 1, 798.7 2, 332.1
(degrees × days)
Degree days over 32◦ Celsius (degrees × days) 46.0 10.3 20.3 80.5

the analysis to arms-length transactions to ensure accurate reflections of fair market values and to
the 23 counties having at least one turbine by 2017. We drop parcels having multiple sales within
the same year because these are unlikely to represent separate competitive transactions (about 470
transactions dropped). In total, we have data on 14,196 transactions. Our PVD sales data include
information on total amount of sale, estimates of dollar amount improvements to land, and acres
of the parcel that are dryland or irrigated.5 We exclude the value of improvements from the price
because this usually reflects the value of storage barns and outbuildings (but not turbines).6,7 All
prices are converted to 2017 dollars using the Consumer Price Index. Table 2 provides summary
statistics for land transactions and characteristics included in our analysis.
Two main concerns when working with agricultural land sales data are sparseness of transactions
reflecting fair market value and whether the decision to list or purchase land is endogenous to
land characteristics. The former concern is not an issue in this setting because we have over
14,000 arms-length transactions covering 23 counties over a 17-year period. The latter concern
can be characterized as a sample-selection problem and has been empirically documented in cases
where rural land has competing agricultural and non-agricultural uses (Koundouri and Pashardes,
2003). Sample selection is unlikely to be a problem in this setting because there is little exurban
development pressure in Kansas, which would drive wholly competing uses of the land (White,
Morzillo, and Alig, 2009). Additionally, we are able to control for nonagricultural development
pressure by including variables on commute times to cities of various size (described below) and by
including a rich set of spatial dummies (up to the township level).

Soils

Soil characteristics likely to affect rents from agriculture are obtained from the SSURGO soil survey
on the website of the USDA Natural Resource Conservation Service (NRCS). The PVD data contain
information on the acres of the parcel represented by each soil type. We link these soil types to the
SSURGO data, which provide information on the characteristics of each soil type and aggregate
5 Additional details of the PVD transaction data are described in Tsoodle, Golden, and Featherstone (2006) and Sampson,
Hendricks, and Taylor (2019).
6 According to conversations with land appraisers, wind turbines are not included in the value of improvements. This is

because wind turbine capital is owned by the wind energy utility rather than the landowner.
7 In later subsections, we investigate differential impacts across parcels that have residential value and those that do not.

In short, we find no evidence of statistically significant differential impacts.

Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 415

the characteristics to the parcel level. The following soil characteristics are used as controls in
regressions: percentage of parcel with pH less than 6 (acidic soils), percentage of parcel with pH
greater than 7.5 (basic soils), plant available water storage, and soil organic carbon in the top 150
cm of the soil horizon.

Climate

Daily gridded weather data are obtained from PRISM and are linked to sections of the Public
Land Survey System, which are then merged to the parcels. We construct four climate variables
(1981–2012 average) for each section: average growing season precipitation, average number of
annual degree days between 10◦ and 32◦ Celsius, average number of annual degree days greater
than 32◦ Celsius (i.e., heat levels detrimental to crop growth, Schlenker, Hanemann, and Fisher,
2006), and average reference growing season evapotranspiration. The climate variables are expected
to capture average climate conditions related to agriculture at the section level.

Urban Influence

We control for urban influence by using data on the commute time to a city with a population of
10,000 or more and commute time to a city with a population of 40,000 or more. Commute times
from each parcel are calculated using Google Maps.

Wind Turbines

We obtain the locations and technical specification of utility-scale wind turbines in Kansas from the
U.S. Wind Turbine Database, jointly operated by the U.S. Geological Survey, Lawrence Berkeley
National Laboratory, and American Wind Energy Association. Turbine locations are obtained from
high-resolution aerial imagery. Technical specifications for the turbines are obtained from the make
and model, as provided by the manufacturer. Utility-scale turbines are characterized as turbines
capable of generating power to feed into the grid to supply a utility. Turbine locations are judged to
be within a 10-meter error tolerance. In total, we obtain the location and specification of all 2,506
utility-scale turbines in Kansas that became active up to the year 2017. Table 1 provides turbine
summary statistics.
The locations of each turbine and land parcel were matched using QGIS. To determine whether
a parcel contained a wind turbine, we spatially merged the turbine and parcel locations to CLU files
for each of the 23 study counties. A CLU is an individual contiguous farming parcel, defined as
the smallest unit of land having a permanent, contiguous boundary and is used by the USDA Farm
Service Agency when linking farm records to maps or images. We overlay the parcel and turbine
coordinates to the CLU files and label any CLU sharing a parcel coordinate and turbine coordinate
as being a parcel with a turbine on it. Figure 2 provides an example.

Methodology

This paper estimates the direct effect of wind turbines on the value of an agricultural parcel and the
indirect effects on the value of nearby parcels. Fundamentally, we are interested in estimating the
treatment effect of either (i) having a wind turbine on the parcel or (ii) having a turbine in nearby
proximity to the parcel (but not on the parcel).
There are a number of challenges in measuring the effect of turbines. The first is the date at
which the turbine begins to exist. We assume that the relevant existence date is the year that the
turbine becomes operational (we explore sensitivity to this assumption in a later section). The second
challenge is how to measure the effects of wind turbines. Because we match parcel locations and
416 September 2020 Journal of Agricultural and Resource Economics

Figure 2. Sample Parcel and Turbine Locations Together with CLUs

Notes: Parcel locations are for any parcel having a sales transaction during the period of analysis and not necessarily subsequent to a turbine
installation.

turbine locations at the CLU level, we can reasonably ascertain whether a parcel had a turbine on it
when it sold.
There are multiple reasons why a turbine might affect land value when it is installed on a
parcel. Foremost, the landowner obtains lease payments from granting the wind energy company
the rights to install and operate the turbine(s). Typically, the terms of the lease are 20–30 years,
with payments most often being made to the landowner over scheduled intervals (e.g., annually).
A recent survey of lease contracts revealed compensation payments ranging from $1,500 to $9,500
per turbine per year (Windustry, 2009).8 Interviews with farmers have demonstrated that a leading
motivator for pursuing on-farm wind turbines is diversification of income (Sutherland and Holstead,
2014). However, turbines also have a number of potentially adverse effects, such as compromising
viewsheds, annoyance from noise, bird deaths from the rotary blades, landscape impacts such
as compression of soils and soil erosion, annoyance to livestock, taking land out of production
for access roads and turbine pads, and disruption of farm equipment (e.g., harvesters, irrigation
structures) (Khatari, 2004; Heintzelman and Tuttle, 2012; Vyn and McCullough, 2014).
In theory, the terms of the lease agreement and the amount of compensation will emerge from
a bargaining process between the landowner and wind company (Coase, 1960; Libecap, 1993). The
landowner’s minimum willingness to accept is ex ante expected to be the amount that approximately
covers the opportunity cost of turbine installation (i.e., any disamenity cost and revenues forgone
from taking land out of production). For the wind company, the maximum offer will depend on the
power purchase agreement with the power plant. According to the Lawrence Berkeley National
Laboratory, recent wind power purchase agreements in the Midwest range from about $20 per
megawatt-hour (MWh) to $40/MWh (Berkeley Lab, 2019). Assuming an average annual electricity
generation of 2,000 MWh/MW and evaluating at our sample average turbine size of 1.8 MW implies
average annual gross revenue of $72,000–$144,000 per turbine to the wind company.

8 For parcels having wind energy installments, the average number of turbines is 1.3, with an average capacity of 2.4 MW.
Using the lease payment ranges above and assuming a 20-year lease term and 5% discount rate, the average capitalized value
would range from about $25,000 to $161,000. This represents about 6%–40% of the value of the average parcel in our sales
data.
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 417

Table 3. Summary Statistics for Wind Turbine Variables

Turbine Proximity Measure Mean Std. Dev. Max./Count
Average distance to nearest turbine (km) 97.0 100.0 438.7
Turbine on parcel 0.003 0.056 44
Turbine 0–2 km away 0.013 0.111 178
Turbine 2–4 km away 0.014 0.116 193
Turbine 4–6 km away 0.022 0.147 314

Thus, whether land values increase in response to a new stream of wind lease payments
will depend on the extent to which those lease payments exceed a landowner’s ex ante expected
opportunity costs. It is also worth noting that even if the lease payments exceed such costs, if similar
properties without turbines retain the option of readily and costlessly adding turbines at some future
date, then prospective buyers would not necessarily place greater value on properties with turbines
compared to those without. More specifically, in a competitive bidding market, if turbines were
not completely exclusive to the properties on which they reside, we would not expect to observe a
significant premium for such properties.9
For parcels having a turbine nearby, potential adverse effects include the loss of enjoyable
viewsheds (Sunak and Madlener, 2017), noise, disturbing livestock, and shadow flickers (Khatari,
2004; Groothuis, Groothuis, and Whitehead, 2008). Potential positive effects include more beneficial
weather (Li et al., 2018) and even an increase in crop yields (Chen, 2019; Kaffine, 2019).
Additionally, there is the potential for a positive network effect if lease payments or the ability to
“opt in” to a wind energy lease are affected by having existing turbines in place nearby. Indeed, the
wind turbine data exhibits some clustering in the location of turbines (Figures 1 and 2), suggesting
there may be incentives to agglomerate (e.g., Moreno-Cruz and Taylor, 2017). The various potential
negative and positive impacts of having turbines nearby mean that the net impact could be offsetting.
To account for proximity effects of a parcel to a turbine, we use two distance measures that
have been used in previous hedonic analyses of wind turbines in residential settings. The first is the
inverse of the linear distance to the nearest turbine. The reasoning behind using the inverse distance
measure is that parcels closer to a turbine are most likely to experience adverse environmental
effects or, alternatively, positive agglomeration effects (see, e.g., Heintzelman and Tuttle, 2012).
To compute the inverse distance, we calculate for each year the linear distance between the location
of a parcel that sold and the nearest turbine that is active in that year. The second distance measure
is a set of dummies representing whether a parcel belongs to a set of spatial rings around a turbine.
We define the rings at 0–2 km, 2–4 km, and 4–6 km around the turbine.10 Using spatial bands
set at 2 km permits more observations to be exploited in the treatment groups (i.e., compared to
narrower, 1 km bands). Additionally, the cutoff of 6 km was chosen based on visual extents used
in previous hedonic studies (e.g., Vyn and McCullough, 2014). Together, these dummy variables
capture potential nonlinear effects in proximity to a turbine.

Descriptive Statistics

Table 3 presents summary statistics for the wind turbine distance measures. The number of post-
turbine installation sales are relatively few, largely because wind capacity grew slowly between
2001 and 2009 and turnover of agricultural land is generally low. We observe 44 land transactions
with an active turbine located on the parcel. This represents about 0.3% of our total observations.
Sales of parcels that are located 0–2 km (178 total), 2–4 km (193 total), or 4–6 km (314 total) away
9 Consider an example: Suppose there are two identical parcels for sale: one with turbines and one without. Suppose the

parcel without turbines retains the option to contract with a wind company at any future time. In this case, potential buyers
might not bid more for the parcel with turbines because they could readily trigger that option on the alternative parcel when/if
they desire.
10 Analyses using 0–3 km and 3–6 km rings produce similar results (Table S1 in the Online Supplement

[www.jareonline.org]).
418 September 2020 Journal of Agricultural and Resource Economics

Table 4. Farm Characteristics for Parcels That Install Turbines and Those That Do Not
Turbine on Turbine Not on
Variable Parcel Parcel
Price per acre ($) 2, 475.8 2, 602.7
(198.40) (23.00)
Commute time to 10,000 population (hrs) 0.7 0.8∗∗
(0.02) (0.00)
Commute time to 40,000 population (hrs) 1.7 1.6
(0.06) (0.01)
Proportion of parcel irrigated 5.5 8.3∗
(1.40) (0.20)
Root zone available water storage (mm) 236.6∗∗ 228.1
(4.40) (0.50)
Soil organic carbon (kg/m2 ) 9.6 9.6
(0.17) (0.03)
Acidic soils (proportion of land) 0.0 0.8∗
(0.03) (0.06)
Basic soils (proportion of land) 58.3∗∗∗ 50.2
(3.20) (0.40)
Slope (%) 3.5 3.5
(0.17) (0.02)
Elevation (ft) 584.8 591.7
(13.40) (1.93)
Growing season precipitation (inches) 17.7 17.9
(0.27) (0.03)
Evapotranspiration (inches) 34.6 34.4
(0.11) (0.01)
Degree days between 10◦ and 32◦ Celsius (degrees × days) 2, 054.6 2, 051.7
(6.04) (0.92)
Degree days over 32◦ Celsius (degrees × days) 42.1 42.9
(0.56) (0.09)
Notes: Mean values are shown and standard errors appear in parentheses. Parcels that choose to install turbines are compared with those that
do not install turbines using a t-test. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%, 5%, and 1% level.

from an active turbine each represent about 1%–2% of our total observations. A potential limitation
of this study is therefore the low number of post-turbine observations in the data.
Before discussing the econometric analysis, it is useful to contextualize the setting by
summarizing the key farm characteristics and land values from the data. Table 4 presents summary
statistics across parcels that have turbines and those that do not. Mean values are compared using
a t-test. We find no statistically significant difference in the real market price per acre for parcels
with turbines and parcels without turbines. We do find that parcels with turbines are statistically
more likely to be located closer to towns with populations of 10,000 or more, to have less irrigation,
to have a higher proportion of basic soils, and to have greater root zone water storage. There is
no statistically significant difference in climate characteristics for parcels with and without wind
turbines. However, a more detailed econometric analysis is required to determine whether these
differences are truly the result of wind turbines (rather than coincidental correlation).

Empirical Approach

We use the hedonic approach to measure the effects of wind turbines on agricultural land values.
Following Rosen (1974), any parcel i can be expressed as a bundle of observable attributes, with the
market price for parcel i being determined by these attributes. In competitive markets, agricultural
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 419

land parcels having desirable attributes will be bid up by potential buyers. The amount by which
parcel i is bid up depends on how much potential buyers value the attribute. The estimated price
premium associated with a particular attribute of agricultural land thus provides evidence of how the
market values that attribute.
Formally, the models we estimate can be described by two regression equations. The first uses
spatial dummies of varying resolution:
price
(1) ln = wi,t β + xi α + τt + ηm + λl + µi,t ,
acre i,t
where ln price
acre i,t is the log of the real price per acre for parcel i in year t, wi,t represents the treatment
variables, which are defined as the different measures of turbine proximity at the time of the sale
(Table 3), xi is a set of other property attributes (e.g., irrigation, soil organic carbon), τt captures
unobserved temporal heterogeneity using year dummies, ηm is a set of month-specific dummies, λl
is a set of spatial dummy variables ranging from no controls to controls for the 461 townships (6-
mile-square) in our study region, and µi,t is the error term, which is clustered at the township. Thus,
when township dummies are specified, the effects of wind turbines on land values in equation (1)
are identified through cross-sectional and time-series variation within townships that is not common
to all Kansas parcels.
Two main concerns in the identification of the effects of wind turbines on land values are omitted
variable bias and endogeneity. If unobserved factors impact land values and are also correlated with
the treatment variables, then estimates of the regression parameters will be biased. Similarly, if the
location of wind turbines is driven in part by the value of land in that location, then this relationship
would result in endogeneity bias. For example, if wind turbines are strategically located in areas
that have lower land values, then one might erroneously conclude that wind turbines negatively
affect land values. These issues are addressed by the spatial dummy variables, which absorb any
time-invariant heterogeneity that affects land values (i.e., land endowment effects). In this way,
any heterogeneity that is clustered (e.g., at the township level) will no longer be omitted in the
regression. Likewise, any time-variant heterogeneity that affects all land values will be controlled
for using the year and month dummies. Controlling for endogeneity bias in this setting therefore
relies upon the assumption that wind turbines are located at random within the unit of spatial control
(e.g., township).
A final concern is spatial autocorrelation in land values. That is, it may be the case that the
market price for parcel i is driven to some degree by the value of nearby parcel j via neighborhood
effects. Additionally, factors that are unobserved for parcel i may correlate with the unobserved
factors for parcel j. One way to address spatial autocorrelation is to develop an empirical model that
explicitly accounts for the particular structure of spatial dependence (e.g., Elhorst, 2003; Schnier and
Felthoven, 2011; Sampson, 2018).11 However, the particular specification of the spatial weighting
matrix is often arbitrary, making it difficult to determine which model of spatial dependence best
represents the true data-generating process (Gibbons and Overman, 2012). In this paper, we correct
for spatial correlation in land values through the use of spatial controls. To control for correlation
in remaining unobserved factors, we cluster the model errors at the township level. Thus, if certain
geographies have higher or lower land values on average, then this will to a large extent be controlled
for by the spatial dummies. Additionally, we allow for arbitrary correlation in the error term within
each township.
Our second model exploits data from repeated sales of parcels. We have data on 1,530 parcels
that sold at least twice from 2001 to 2017, for a total of 3,263 repeat transactions. The regression
model in equation (1) can be adapted as
price
(2) ln = wi,t β + τt + ηm + λi + µi,t ,
acre i,t
11 In this context, the township dummy variables can be viewed as analogous to a spatial weighting matrix, where 1 is

assigned to parcel pairs within a township and 0 otherwise.

420 September 2020 Journal of Agricultural and Resource Economics

where λi is a parcel fixed effect and all other variables are as described previously. In equation (2),
parcel characteristics will drop out because they do not vary over time. Identification of the effect
of wind turbine proximity is thus achieved from parcels that transact more than once during the
period of analysis and vary in their situational exposure to turbines over time. While repeat sales
data are statistically powerful, it is worth noting that the reduction in observations is likely to
reduce the precision of the estimates, while inclusion of parcel-level fixed effects may exacerbate
any attenuation bias resulting from measurement error.

Results

We first present results for the analysis using equation (1) as the regression model. Columns 1–3
of Table 5 show coefficient estimates for the model specifying the log of the inverse distance to
the nearest turbine while columns 4–6 show estimates for the model specifying a set of dummies
indicating proximity to the nearest turbine. In columns 1–3, the coefficient is positive and statistically
significant at 0.10 or better and similar in magnitude across specifications. Because the main variable
of interest is specified using a log-log, the coefficient on the log of the inverse distance can be
interpreted as an elasticity of land price with respect to the inverse of the distance to the nearest
turbine. In particular, the coefficient value indicates that a 10% increase in the inverse distance results
in a 0.26%-0.65% increase in land value. Thus, this implies that proximity to a turbine increases land
values.
Looking at columns 4–6 of Table 5, we find that the coefficient for having a turbine on the
parcel switches between negative and positive, though the effect is not statistically significant in
any of the specifications. For the 0–2 km and 2–4 km treatment groups, we find positive but
statistically insignificant effect on land values. For the 4–6 km treatment group, the coefficient
is negative in columns 4 and 5 and marginally statistically significant in column 5. In column 6,
which includes spatial controls at the township level, the effect is very small in magnitude and
not statistically significant, implying that the use of township controls absorbs some of the effect
previously attributed to the 4–6 km treatment group. As mentioned previously, it is possible that
the null effect for turbines near the parcel stems from offsetting positive and negative impacts. In
said cases, a more detailed set of data and empirical approach would be necessary to decompose the
total effect. Taken together, columns 1–6 of Table 5 suggest that the impacts of turbines on farmland
values range from positive to statistically insignificant.
With respect to the other model covariates, we find that irrigation, soil organic carbon, and the
number of degree days favorable to growing conditions positively impact agricultural land values (as
expected). In particular, a 10% increase in the percentage of the total parcel that is irrigated increases
land value by about 7%–9%, and this effect is significant at 0.01, which is consistent with previous
studies of irrigation premiums (Sampson, Hendricks, and Taylor, 2019). An additional kg/m2 of soil
organic carbon increases land values by about 1%–3%, and the effect is statistically significant at
0.10 or better. An additional degree day between 10◦ and 32◦ Celsius increases land values by about
0.1%–0.3%, in certain specifications (i.e., columns 1, 3, and 4).

Heterogeneity in Turbine Size

Regression estimates thus far have grouped all turbine sizes together. If turbine disamenity
value is driven by factors related to visibility, then it stands to reason that taller turbines would
have a relatively greater effect. Turbine hub heights in the data (distance from platform to
center of rotor) range from about 65 m to about 95 m, with an average of 73 m. Table S2 in the
Online Supplement presents results for a subset of the data based on turbines that have hub heights of
Table 5. Regression Results
Variable 1 2 3 4 5 6
ln(Inverse distance to nearest turbine) 0.0452∗∗∗ 0.0264∗ 0.0651∗∗∗
(0.014) (0.014) (0.015)
Turbine on parcel −0.0452 −0.0772 0.1130
(0.1400) (0.1460) (0.1580)
Nearest turbine 0–2 km 0.0536 0.0411 0.0911
(0.0906) (0.0849) (0.0689)
Nearest turbine 2–4 km 0.0667 0.0036 0.0732
(0.0837) (0.0779) (0.0823)
Sampson, Perry, and Taylor

Nearest turbine 4–6 km −0.0935 −0.126∗ 0.0012

(0.0710) (0.0737) (0.0668)
Proportion of parcel irrigated 0.0091∗∗∗ 0.0078∗∗∗ 0.0074∗∗∗ 0.0091∗∗∗ 0.0078∗∗∗ 0.0074∗∗∗
(0.0007) (0.0007) (0.0006) (0.0006) (0.0007) (0.0006)
Root zone available water storage (mm) −0.0005 −0.0007∗ −0.0004 −0.0005 −0.0007∗ −0.0003
(0.0004) (0.0004) (0.0004) (0.0004) (0.0004) (0.0004)
Soil organic carbon (kg/m2 ) 0.0246∗∗∗ 0.0279∗∗∗ 0.0110∗ 0.0242∗∗∗ 0.0278∗∗∗ 0.0112∗
(0.0064) (0.0059) (0.0059) (0.0064) (0.0058) (0.0059)
Acidic soils (proportion of land) −0.0031∗ 0.0001 0.0004 −0.0030∗ 0.0001 0.0004
(0.0018) (0.0017) (0.0015) (0.0018) (0.0016) (0.0015)
Basic soils (proportion of land) −0.0008 −0.0002 −0.0003 −0.0007 −0.0003 −0.0003
(0.0005) (0.0005) (0.0005) (0.0005) (0.0005) (0.0005)
Slope (%) 0.0264∗∗∗ 0.0189∗∗∗ −0.0120 0.0274∗∗∗ 0.0190∗∗∗ −0.0114
(0.0074) (0.0073) (0.0076) (0.0073) (0.0072) (0.0075)
Elevation (ft) −0.0001 −0.0002 −0.0002∗ −0.0002 −0.0002 −0.0003∗
(0.0002) (0.0001) (0.0001) (0.0002) (0.0001) (0.0001)
Growing season precipitation (inches) −0.0151 0.0475∗ 0.0011 −0.0250 0.0429∗ −0.0024
(0.0164) (0.0250) (0.0409) (0.0162) (0.0248) (0.0409)
Evapotranspiration (inches) −0.0548 0.0638 0.0955 −0.0369 0.0723 0.0874
(0.0503) (0.0790) (0.1480) (0.0510) (0.0788) (0.1470)
Degree days between 10◦ and 32◦ Celsius (degrees × days) 0.0011∗∗∗ 0.0001 0.0034∗ 0.0013∗∗∗ 0.0001 0.0033
(0.0004) (0.0009) (0.0020) (0.0004) (0.0009) (0.0020)
Degree days over 32◦ Celsius (degrees × days) −0.0069 0.0008 −0.0202 −0.0111∗∗ −0.0008 −0.0198
(0.0056) (0.0106) (0.0234) (0.0054) (0.0106) (0.0234)
Commute time to 10,000 population (hrs) 0.1660∗∗∗ −0.0049 0.1210 0.1540∗∗∗ −0.0127 0.1160
(0.0545) (0.0877) (0.1840) (0.0542) (0.0868) (0.1850)
Commute time to 40,000 population (hrs) 0.1540∗∗∗ 0.0721 0.0419 0.1380∗∗∗ 0.0669 0.0368
(0.0504) (0.0748) (0.1590) (0.0507) (0.0744) (0.1600)
Spatial controls None County Township None County Township
Adjusted R2 0.23 0.26 0.35 0.22 0.26 0.35
No. of obs. 12,157 12,157 12,153 12,189 12,189 12,185
Effects of Wind Turbines on Land Values 421

Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%, 5%, and 1% level. In all specifications, the dependent variable is the log of
real price per acre. Models also include year and month dummies.
422 September 2020 Journal of Agricultural and Resource Economics

at least 73 m, which is approximately the cutoff between small- and large-scale wind turbines.12
That is, the regressions in Table S2 use only turbines having hub heights of 73 m or larger. Note
that analysis of heterogeneity in turbine size in this approach may exacerbate attenuation bias from
measurement error due to the reduction in observations. Looking across columns 1–4 of Table S2,
the results provide no evidence that taller turbines have greater effects on agricultural land values
than smaller turbines.

Impacts over Time

Previous work highlights the possibility that over long enough periods of time, the hedonic price
coefficients may vary (Kuminoff, Parmeter, and Pope, 2010). Given the relatively long period of
analysis (2001–2017) there is potential for this issue here. One main concern is the real estate market
crash and economic downturn of 2007–2009. To test the impact of the real estate market crash, we
run the models after excluding observations from 2007–2009. Estimates are reported in columns 1
and 3 of Table S3 in the Online Supplement. In short, the results are consistent with Table 5.
A parsimonious way to test temporal variance over the full sample period is to specify a linear
time trend and interact the various treatment groups with the trend. We report these estimates
in columns 2 and 4 of Table S3 for specifications including township dummies. For the model
using the log of the inverse distance to the nearest turbine, the baseline coefficient is 0.116 and
is statistically significant at p < 0.01 (consistent with column 3 of Table 5). Additionally, the linear
trend interaction indicates the magnitude of the effect declined by about 0.6 percentage points per
year on average. For the model with the 0–2 km, 2–4 km, and 4–6 km treatment groups, we detect
that the baseline impacts of on-farm and near-farm wind turbine proximity are not statistically
significant (consistent with column 6 of Table 5). Additionally, the linear trend interacted with the
treatment groups is not statistically significant.

Repeat Sales

Table 6 presents coefficient estimates when the sample is restricted to only sales occurring more than
once for the same parcel. In the repeat sales data, there are 8 transactions with a turbine on-parcel, 44
transactions in the 0–2 km group, 40 transactions in the 2–4 km group, and 72 transactions in the 4–6
km group. The smaller sample size and inclusion of parcel fixed effects may reduce precision and/or
exacerbate measurement error. Column 1 of Table 6 presents estimates for the model specifying the
log of the inverse distance to the nearest turbine with parcel-level controls. Column 2 of Table 6
presents estimates for the model specifying the set of dummies indicating proximity to the nearest
turbine with parcel-level controls.
In column 1, we find a negative and statistically insignificant coefficient on the log of the inverse
distance to the nearest turbine. In column 2, we also find that the coefficients on all treatment groups
are negative and statistically insignificant. The point estimates on some of the treatment groups in
column 2 go up in absolute magnitude relative to the pooled model estimates in Table 5, but the
uncertainty also increases because less variation is exploited in the smaller sample. Consistent with
the earlier estimates from columns 4–6 of the pooled cross section in Table 5, the model estimates
in Table 6 provide no evidence of statistically significant proximity effects of wind turbines on
agricultural land values. Thus, our results do not confirm that parcels that opt to have turbines
installed capitalize a premium compared to similar parcels. Additionally, parcels that have turbines
installed nearby do not suffer any systematic disamenity value.
We also investigate different definitions of the on-parcel effects of turbines in the repeat sales
data. Table S4 in the Online Supplement presents estimates using the number of total turbines

12 For example, Vaisala Energy defines a large-scale turbine (> 1 MW) as typically having hub heights of 80 m or larger

(https://www.3tier.com/en/support/wind-online-tools/what-prospecting-hub-height/).
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 423

Table 6. Effects of Turbine Proximity to a Parcel Using Repeat Sales

Variable 1 2
ln(inverse distance to nearest turbine) −0.0091
(0.0280)
Turbine on parcel −0.0983
(0.0760)
Nearest turbine 0–2 km −0.1490
(0.1060)
Nearest turbine 2–4 km −0.1940
(0.1450)
Nearest turbine 4–6 km −0.0957
(0.1490)
Spatial dummies Parcel Parcel
Within R2 0.26 0.26
No. of obs. 3, 250 3, 263
Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. In all specifications, the dependent variable is the log of real price per acre. Models also include year and month dummies.

installed on the parcel (column 1) and the total installed capacity (in MW) on the parcel (column 2).
In both cases, the effect of turbines is not statistically significant.

Sensitivity to Existence Date

As mentioned previously, we assume the relevant existence date for the turbines began at the time
the turbines became operational. However, impacts to property values could precede the turbine
operation date if awareness of the turbine locations is known by buyers and sellers. To investigate
this possibility, we define an alternative existence date as the full calendar year prior to the turbine’s
operation date. Estimates on turbine construction times range from 2 to 6 months, so a full calendar
year setback should capture pre-construction-phase impacts (European Wind Energy Association,
2016). We report estimates using the pooled sample with township dummies in columns 1 and 3 of
Table S5 in the Online Supplement. Columns 2 and 4 of Table S5 present estimates from the repeat
sales subsample with parcel-level controls. Generally, the point estimates are robust to the alternative
existence date. One exception is the 2–4 km treatment group in the repeat sales subsample, which
changed from −0.19 and was not statistically significant using the operation date to −0.31 and
marginally statistically significant (p < 0.10) using the alternative existence date.

Residential versus. Nonresidential Farmland

We investigate whether parcels having a residence are affected differentially from parcels without
a residence. To do this, we factor in the value of residence in the real price per acre and create an
indicator variable if the parcel has a positive residential value associated with it. We then interact
the indicator with the treatment variables. Note that we could not estimate differential effects for on-
parcel turbines because there is only one transaction having both a turbine and a residence. Results
are presented in Table S6 in the Online Supplement. In short, post-estimation F-tests do not reject
the null hypothesis of no differential impacts between parcels with and without residential value.

Land Value Impacts of Electricity Production

The on-farm hedonic price impact of having wind turbines can be converted to a value per unit
of electricity production. To this end, we conduct a back-of-envelope analysis of the largest dollar
value of a MWh of electricity that cannot be rejected given the econometric estimates. The following
424 September 2020 Journal of Agricultural and Resource Economics

calculations are meant as first-order approximations only. Focusing on the on-farm impact of wind
turbines, our preferred coefficient estimate is 0.113, with an upper 95% confidence interval of 0.424
using the pooled data (Table 5). Using the pooled sample average of $2,605/acre and 172 acres per
transaction (Table 2) implies an average parcel value of $448,060. Assuming 1.3 turbines per parcel
(following footnote 8) and an average size of 1.8 MW with average capacity factor of 0.35 implies
annual electricity production of 7,174 MWh. The upper 95% confidence interval estimate equates
to a 53% increase in parcel value. Dividing this premium by 7,174 MWh gives $33/MWh, which is
in the range of recent power purchase agreements in the Midwest (Berkeley Lab, 2019). Thus, the
largest on-farm impact that we cannot reject from the econometric analysis is about $33/MWh.

Conclusion

For many agricultural producers, the value of their land constitutes the largest component of their
wealth (U.S. Department of Agriculture, 2019). Fluctuations in that value can therefore have major
implications for well-being and even for solvency. In this paper, we leverage a large dataset on
agricultural land transactions to estimate the impact of wind turbines on nearby land values and on-
farm land values. Across all our analyses, the preponderance of results suggests that wind turbines
do not affect agricultural property values, either on-farm or nearby, in a statistically significant way.
This contrasts with some recent studies that find economically significant disamenity effects for
residential properties, but it supports recent related studies on agricultural land values (Vyn and
McCullough, 2014; Shultz, Hall, and Strager, 2015) and rural residential land values in Kansas
(Center for Economic Development and Business Research, 2019).
While the preponderance of our results indicates no statistically significant impacts to
agricultural land values, this does not strictly rule out any positive or negative impacts from occurring
on a case-by-case basis. Some of the treatment group estimates in our regression analysis have large
standard errors, which may suggest that some properties have experienced large positive or negative
impacts from turbine proximity. In fact, community opposition to turbines has arisen over time in
some regions of Kansas (Lefler, 2019; Shorman, 2019) and community sentiment can shape the
ways in which turbines affect property values (Vyn, 2018; Boyle et al., 2019). Thus, our study
cannot strictly refute that some properties have been impacted by proximity to turbines. Rather,
our analysis indicates that any such impacts have not occurred on a systematic basis across the 23
Kansas counties with utility-scale turbines.
Our findings have several implications for policy makers and agricultural landowners. As
previously noted, the number of wind farms are slated to increase by nearly 4 times over the next 10
years. Additionally, the size and energy generation capacity of wind turbines are increasing, which
has the potential to increase lease payments to landowners but also strengthen any external effects
to neighboring parcels in the future. Our results indicate that, at least so far, such effects have not
materialized on a systematic basis for agricultural properties in Kansas. Concerning landowners,
for those that already have turbines on their property, our findings cannot confirm that the value of
their land will increase. Similarly, landowners interested in contracting with a wind firm should not
expect a wind farm to raise (or lower) the value of their land. One interpretation of the absence of any
on-farm effect is that the bargaining process between landowners and wind energy companies has
resulted in lease payments that approximate landowners’ minimum willingness to accept on average
across Kansas. Another possibility is that properties with turbines are still not exclusive enough to
warrant a significant premium. Indeed, Myrna, Odening, and Ritter (2019) note that suitable land for
turbine installation is highly scarce in Germany, which may explain their finding of positive on-farm
impacts.
There are several potential directions for future research. Here we have focused on agricultural
properties, but it would be interesting to extend the analysis to exurban properties, as in Vyn and
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 425

McCullough (2014) and the Center for Economic Development and Business Research (2019).13
A second possibility is to compare the impacts across state borders or across different types of
agricultural producers (e.g., field crops vs. pasture). Both a limitation of the present study and a
possibility for the future concerns the breadth of the data. Wind turbines have only been in operation
for less than 20 years in Kansas, and thus data on land sales with and near turbines, particularly
repeat sales data, are still somewhat limited. Thus, it will be important to revisit the results of this
analysis, particularly the on-farm analysis, as more data are collected.

[First submitted July 2019; accepted for publication January 2020.]

References

American Wind Energy Association. “Wind Energy in Kansas.” 2019. Available online at
https://www.awea.org/Awea/media/Resources/StateFactSheets/Kansas.pdf.
Baker, S., D. Klein, B. Sherrick, and B. Haight. “Wind Energy and Farmland Values – Illinois
Society of Professional Farm Managers and Rural Appraisers.” Illinois Society of Professional
Farm Managers and Rural Appraisers (2018). Available online at https://ispfmra.org/2018/
03/22/wind-energy-and-farmland-values/.
Berkeley Lab. “Wind Power Purchase Agreement (PPA) Prices.” 2019. Available online at
https://emp.lbl.gov/wind-power-purchase-agreement-ppa-prices.
Boyle, K. J., J. Boatwright, S. Brahma, and W. Xu. “NIMBY, Not, in Siting Community Wind
Farms.” Resource and Energy Economics 57(2019):85–100. doi:
10.1016/j.reseneeco.2019.04.004.
Center for Economic Development and Business Research. Wind Project Effects on Kansas
CountiesâĂŹ Property Values. Wichita, KS: Wichita State University, W. Frank Barton School
of Business, 2019. Available online at https://www.greaterhutch.com/media/userfiles/
subsite_24/ files/Wind Power Property Value Analysis.pdf.
Chen, T. “Wind Energy and Agricultural Production – Evidence from Farm-Level Data.” Paper
presented at the annual meeting of the Agricultural & Applied Economics Association, July
21-23, Atlanta, Georgia, 2019.
Coase, R. H. “The Problem of Social Cost.” Journal of Law and Economics 3(1960):1–44.
Davis, W. “Farmer’s Data Shows Land Beneath Wind Farm Gets More Rain, Less Hail.”
FarmProgress (2018). Available online at https://www.farmprogress.com/weather/farmers-data-
shows-land-beneath-wind-farm-gets-more-rain-less-hail.
Dodge, Z. “Mixed Reaction in Neosho County, Kan as Commission Looks at Wind Farm.” KOAM
News Now (2019). Available online at https://www.koamnewsnow.com/mixed-reaction-in-
neosho-county-kan-as-commission-looks-at-wind-farm/.
Elhorst, J. P. “Specification and Estimation of Spatial Panel Data Models.” International Regional
Science Review 26(2003):244–268. doi: 10.1177/0160017603253791.
European Wind Energy Association. “Wind Energy Frequently Asked Questions (FAQ).” 2016.
Available online at http://www.ewea.org/wind-energy-basics/faq/.
Gibbons, S., and H. G. Overman. “Mostly Pointless Spatial Econometrics?” Journal of Regional
Science 52(2012):172–191. doi: 10.1111/j.1467-9787.2012.00760.x.
Groothuis, P. A., J. D. Groothuis, and J. C. Whitehead. “Green vs. Green: Measuring the
Compensation Required to Site Electrical Generation Windmills in a Viewshed.” Energy Policy
36(2008):1545–1550. doi: 10.1016/j.enpol.2008.01.018.
Heintzelman, M. D., and C. M. Tuttle. “Values in the Wind: A Hedonic Analysis of Wind Power
Facilities.” Land Economics 88(2012):571–588. doi: 10.3368/le.88.3.571.

13 A key limitation to the study by the Center for Economic Development and Business Research (2019) is that it used

county-level assessed values rather than parcel-level transaction data.

426 September 2020 Journal of Agricultural and Resource Economics

Hoen, B., J. Brown, T. Jackson, R. Wiser, M. Thayer, and P. Cappers. “A Spatial Hedonic Analysis
of the Effects of Wind Energy Facilities on Surrounding Property Values in the United States.”
LBNL 6362E, Lawrence Berkeley National Laboratory, Berkeley, CA, 2013.
Hoen, B., R. Wiser, P. Cappers, M. Thayer, and G. Sethi. “Wind Energy Facilities and Residential
Properties: The Effect of Proximity and View on Sales Prices.” Journal of Real Estate Research
33(2011):279–316.
Jensen, C. U., T. E. Panduro, T. H. Lundhede, A. S. E. Nielsen, M. Dalsgaard, and B. J. Thorsen.
“The Impact of On-Shore and Off-Shore Wind Turbine Farms on Property Prices.” Energy
Policy 116(2018):50–59. doi: 10.1016/j.enpol.2018.01.046.
Kaffine, D. T. “Microclimate Effects of Wind Farms on Local Crop Yields.” Journal of
Environmental Economics and Management 96(2019):159–173. doi:
10.1016/j.jeem.2019.06.001.
Khatari, M. Impact of Wind Farms on the Value of Residential Property and Agricultural Land.
London, UK: Royal Institute of Chartered Surveyors, 2004.
Koundouri, P., and P. Pashardes. “Hedonic Price Analysis and Selectivity Bias.” Environmental
and Resource Economics 26(2003):45–56. doi: 10.1023/A:1025633604579.
Krueger, A. D., G. R. Parsons, and J. Firestone. “Valuing the Visual Disamenity of Offshore Wind
Power Projects at Varying Distances from the Shore: An Application on the Delaware
Shoreline.” Land Economics 87(2011):268–283.
Kuminoff, N. V., C. F. Parmeter, and J. C. Pope. “Which Hedonic Models Can We Trust to Recover
the Marginal Willingness to Pay for Environmental Amenities?” Journal of Environmental
Economics and Management 60(2010):145–160. doi: 10.1016/j.jeem.2010.06.001.
Ladenburg, J., and A. Dubgaard. “Willingness to Pay for Reduced Visual Disamenities from
Offshore Wind Farms in Denmark.” Energy Policy 35(2007):4059–4071. doi:
10.1016/j.enpol.2007.01.023.
Lang, C., J. J. Opaluch, and G. Sfinarolakis. “The Windy City: Property Value Impacts of Wind
Turbines in an Urban Setting.” Energy Economics 44(2014):413–421. doi:
10.1016/j.eneco.2014.05.010.
Laposa, S., and A. Mueller. “Wind Farm Announcements and Rural Home Prices: Maxwell Ranch
and Rural Northern Colorado.” Journal of Sustainable Real Estate 2(2010):383–402. doi:
10.1080/10835547.2010.12091798.
Lefler, D. “Sedgwick County Bans Wind Farms, Restricts Commercial Solar.” Wichita Eagle
(2019). Available online at https://www.kansas.com/news/politics-government/
article234217332.html.
Li, Y., E. Kalnay, S. Motesharrei, J. Rivas, F. Kucharski, D. Kirk-Davidoff, E. Bach, and N. Zeng.
“Climate Model Shows Large-Scale Wind and Solar Farms in the Sahara Increase Rain and
Vegetation.” Science 361(2018):1019–1022. doi: 10.1126/science.aar5629.
Libecap, G. D. Contracting for Property Rights. New York, NY: Cambridge University Press,
1993.
Ma, S., and S. M. Swinton. “Hedonic Valuation of Farmland Using Sale Prices versus Appraised
Values.” Land Economics 88(2012):1–15. doi: 10.3368/le.88.1.1.
Moreno-Cruz, J., and M. S. Taylor. “An Energy-Centric Theory of Agglomeration.” Journal of
Environmental Economics and Management 84(2017):153–172. doi:
10.1016/j.jeem.2017.02.006.
Myrna, O., M. Odening, and M. Ritter. “The Influence of Wind Energy and Biogas on Farmland
Prices.” Land 8(2019):19.
Rosen, S. “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition.”
Journal of Political Economy 82(1974):34–55. doi: 10.1086/260169.
Sampson, G. S. “Economic Spillovers in Spatial Harvest Behavior.” Ecological Economics
145(2018):57–74. doi: 10.1016/j.ecolecon.2017.08.016.
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values 427

Sampson, G. S., N. P. Hendricks, and M. R. Taylor. “Land Market Valuation of Groundwater.”

Resource and Energy Economics 58(2019):101,120. doi: 10.1016/j.reseneeco.2019.101120.
Schlenker, W., W. M. Hanemann, and A. C. Fisher. “The Impact of Global Warming on U.S.
Agriculture: An Econometric Analysis of Optimal Growing Conditions.” Review of Economics
and Statistics 88(2006):113–125. doi: 10.1162/rest.2006.88.1.113.
Schnier, K. E., and R. G. Felthoven. “Accounting for Spatial Heterogeneity and Autocorrelation in
Spatial Discrete Choice Models: Implications for Behavioral Predictions.” Land Economics
87(2011):382–402. doi: 10.3368/le.87.3.382.
Shorman, J. “‘A Landmark Case’: After Kansas Residents Stop Wind Farm, Is More Resistance
Ahead?” Wichita Eagle (2019). Available online at https://www.kansas.com/news/
politics-government/article231824228.html.
Shultz, C., J. C. Hall, and M. P. Strager. “Production of Wind Energy and Agricultural Land
Values: Evidence from Pennsylvania.” Working Paper 15-11, West Virginia University
Department of Economics, Morgantown, WV, 2015.
Sunak, Y., and R. Madlener. “The Impact of Wind Farms on Property Values: A Locally Weighted
Hedonic Pricing Model.” Papers in Regional Science 96(2017):423–444. doi:
10.1111/pirs.12197.
Sutherland, L.-A., and K. L. Holstead. “Future-Proofing the Farm: On-Farm Wind Turbine
Development in Farm Business Decision-Making.” Land Use Policy 36(2014):102–112. doi:
10.1016/j.landusepol.2013.07.004.
Tsoodle, L. J., B. B. Golden, and A. M. Featherstone. “Factors Influencing Kansas Agricultural
Farm Land Values.” Land Economics 82(2006):124–139.
U.S. Department of Agriculture. U.S. and State Farm Income and Wealth Statistics. Washington,
DC: U.S. Department of Agriculture, Economic Research Service, 2019. Available online at
https://www.ers.usda.gov/data-products/farm-income-and-wealth-statistics.
U.S. Department of Energy. “Wind Energy for Rural Economic Development.” DOE/GO
102004-1826, U.S. Department of Energy, National Renewable Energy Lab, Golden, CO, 2004.
U.S. Energy Information Administration. “Kansas State Energy Profile Overview.” 2018.
Available online at https://www.eia.gov/state/?sid=KS.
Vyn, R. J. “Property Value Impacts of Wind Turbines and the Influence of Attitudes toward Wind
Energy.” Land Economics 94(2018):496–516. doi: 10.3368/le.94.4.496.
Vyn, R. J., and R. M. McCullough. “The Effects of Wind Turbines on Property Values in Ontario:
Does Public Perception Match Empirical Evidence?” Canadian Journal of Agricultural
Economics 62(2014):365–392. doi: 10.1111/cjag.12030.
White, E. M., A. T. Morzillo, and R. J. Alig. “Past and Projected Rural Land Conversion in the US
at State, Regional, and National Levels.” Landscape and Urban Planning 89(2009):37–48. doi:
10.1016/j.landurbplan.2008.09.004.
Windustry. Wind Energy Easements and Leases: Compensation Packages. Minneapolis, MN:
Windustry, 2009.
Journal of Agricultural and Resource Economics 45(3):S1–S4 ISSN 1068-5502
Copyright 2020 the authors doi: 10.22004/ag.econ.302463

Online Supplement: The On-Farm and Near-Farm Effects of

Wind Turbines on Agricultural Land Values

Gabriel S. Sampson, Edward D. Perry, and Mykel R. Taylor

Table S1. Regression Results Using 0–3 km and 3–6 km Treatment

Variable 1 2 3
Turbine on parcel −0.0463 −0.0787 0.1170
(0.1400) (0.1460) (0.1570)
Nearest turbine 0–3 km 0.0756 0.0578 0.1270
(0.0776) (0.0752) (0.0776)
Nearest turbine 3–6 km −0.0644 −0.112∗ −0.0035
(0.0662) (0.0660) (0.0605)
Proportion of parcel irrigated 0.0091∗∗∗ 0.0078∗∗∗ 0.0074∗∗∗
(0.0006) (0.0007) (0.0006)
Root zone available water storage (mm) −0.0005 −0.0007∗ −0.0003
(0.0004) (0.0004) (0.0004)
Soil organic carbon (kg/m2 ) 0.0241∗∗∗ 0.0277∗∗∗ 0.0111∗
(0.0064) (0.0058) (0.0059)
Acidic soils (proportion of land) −0.0030∗ 0.0001 0.0004
(0.0018) (0.0016) (0.0015)
Basic soils (proportion of land) −0.0007 −0.0003 −0.0003
(0.0005) (0.0005) (0.0005)
Slope (%) 0.0277∗∗∗ 0.0192∗∗∗ −0.0113
(0.0073) (0.0072) (0.0075)
Elevation (ft) −0.0002 −0.0002 −0.0003∗
(0.0002) (0.0001) (0.0001)
Growing season precipitation (inches) −0.0251 0.0427∗ −0.0016
(0.0162) (0.0248) (0.0410)
Evapotranspiration (inches) −0.0378 0.0719 0.0874
(0.0510) (0.0787) (0.1470)
Degree days between 10◦ and 32◦ Celsius 0.0013∗∗∗ 0.0001 0.0033∗
(degrees × days) (0.0004) (0.0009) (0.0020)
Degree days over 32◦ Celsius (degrees × days) −0.0110∗∗ −0.0007 −0.0200
(0.0054) (0.0106) (0.0234)
Commute time to 10,000 population (hrs) 0.1530∗∗∗ −0.0144 0.1120
(0.0542) (0.0865) (0.1840)
Commute time to 40,000 population (hrs) 0.1390∗∗∗ 0.0679 0.0401
(0.0508) (0.0744) (0.1600)
Spatial controls None County Township
Adjusted R2 0.22 0.26 0.35
No. of obs. 12,194 12,194 12,190
Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. In all specifications, the dependent variable is the log of real price per acre. Models also include year and month dummies.
S2 September 2020 Journal of Agricultural and Resource Economics

Table S2. Heterogeneity of Impact by Turbine Hub Height

Hub Height > 73 m
Variable 1 2 3 4
ln(inverse distance to nearest −0.00428 0.0239
turbine)
(0.0180) (0.0190)
Turbine on parcel −0.0247 0.0644
(0.1480) (0.1820)
Nearest turbine 0–2 km −0.0472 0.0507
(0.0770) (0.0818)
Nearest turbine 2–4 km 0.0237 0.0453
(0.0830) (0.0790)
Nearest turbine 4–6 km −0.0676 0.00557
(0.0670) (0.0680)
Spatial dummies County Township County Township
Adjusted R2 0.22 0.32 0.22 0.32
No. of obs. 5, 808 5, 789 5, 840 5, 821
Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. In all specifications, the dependent variable is the log of real price per acre. Models also include year and month dummies.

Table S3. Regression Results for Time-Variant Models

Variable 1 2 3 4
ln(inverse distance to nearest turbine) 0.0738∗∗∗ 0.116∗∗∗
(0.0176) (0.0260)
Turbine on parcel 0.1880 −0.8530
(0.1500) (1.1800)
Nearest turbine 0–2 km 0.0780 0.1620
(0.6880) (0.2940)
Nearest turbine 2–4 km 0.0410 0.1420
(0.0788) (0.3980)
Nearest turbine 4–6 km −0.0070 −0.0080
(0.0680) (0.1770)
Linear trend 0.0100 0.0410∗∗∗
(0.0104) (0.0050)
ln(inverse distance to nearest turbine) −0.0060∗∗
(linear trend)
(0.0024)
Turbine on parcel (linear trend) 0.0670
(0.0770)
Nearest turbine 0–2 km (linear trend) −0.0060
(0.0212)
Nearest turbine 2–4 km (linear trend) −0.0060
(0.0291)
Nearest turbine 4–6 km (linear trend) 0.0005
(0.0138)
Excludes Excludes
Sample 2007–2009 Full 2007–2009 Full
Adjusted R2 0.37 0.38 0.37 0.38
No. of obs. 9,818 12,162 9,848 12,194
Notes: : Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. In all specifications, the dependent variable is the log of real price per acre. Models include township, year, and month
dummies.
Sampson, Perry, and Taylor Effects of Wind Turbines on Land Values S3

Table S4. Effects of Total Number of Turbines and Installed Capacity on a Parcel
Variable 1 2
Number of turbines installed on parcel −0.0214
(0.0210)
Installed capacity on parcel (MW) −0.0144
(0.0130)
Spatial dummies Parcel Parcel
Within R2 0.26 0.26
No. of obs. 3,263 3,263
Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. Also includes year and month dummies.

Table S5. Regression Results If Turbine Existence Date Is Pushed Back One Calendar Year
Variable 1 2 3 4
ln(Inverse distance to nearest turbine) 0.0547∗∗∗ −0.0140
(0.0145) (0.0293)
Turbine on parcel 0.0693 −0.0961
(0.1530) (0.0749)
Nearest turbine 0–2 km 0.0675 −0.1680
(0.0746) (0.1330)
Nearest turbine 2–4 km 0.0942 −0.3130∗
(0.0815) (0.1720)
Nearest turbine 4–6 km −0.0290 0.0895
(0.0655) (0.1790)
Spatial dummies Township Parcel Township Parcel
R2 0.35 0.26 0.35 0.26
No. of obs. 12,155 3,250 12,190 3,263
Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. In all specifications, the dependent variable is the log of real price per acre. Models also include year and month dummies.
Columns 1 and 3 report adjusted R2 . Columns 2 and 4 report within R2 .
S4 September 2020 Journal of Agricultural and Resource Economics

Table S6. Regression Investigating Differential Effects by Residence Status

Variable 1 2
Residence 0.1680∗∗ 0.1740∗∗∗
(0.0535) (0.0270)
ln(inverse distance to nearest turbine) (no residence) 0.0591∗∗∗
(0.0150)
ln(inverse distance to nearest turbine) (residence) 0.0572∗∗
(0.0226)
Turbine on parcel 0.1090
(0.1600)
Nearest turbine 0–2 km (no residence) 0.0642
(0.0700)
Nearest turbine 2–4 km (no residence) 0.0494
(0.0810)
Nearest turbine 4–6 km (no residence) −0.0172
(0.0670)
Nearest turbine 0–2 km (residence) 0.2670
(0.2310)
Nearest turbine 2–4 km (residence) −0.0989
(0.2650)
Nearest turbine 4–6 km (residence) −0.0396
(0.1790)
Spatial dummies Township Township
Adjusted R2 0.36 0.35
No. of obs. 12,159 12,191
Notes: Standard errors clustered at township in parentheses. Single, double, and triple asterisks (*, **, ***) indicate significance at the 10%,
5%, and 1% level. In all specifications, the dependent variable is the log of real price per acre. Models also include year and month dummies.

[Received July 2019; final revision received January 2020.]