RESPONSIVENESS OF CROP YIELD AND

ACREAGE TO PRICES AND CLIMATE

RUIQING MIAO, MADHU KHANNA, AND HAIXIAO HUANG

We investigate the effect of crop price and climate variables on rainfed corn and soybean yields
and acreage in the United States using a large panel dataset for the 1977–2007 period. Instrumen-
tal variables are used to control for endogeneity of prices in yield and acreage regressions, while
allowing for spatially auto-correlated errors. We find that an increase in corn price has a statisti-
cally significant positive impact on corn yield, but the effect of soybean price on soybean yields
is not statistically significant. The estimated price elasticities of corn yield and acreage are 0.23
and 0.45, respectively. Of the increase in corn supply caused by an increase in corn price, we find
that 33.8% is due to price-induced yield enhancement and 66.2% is due to price-induced acreage
expansion. We also find that the impact of climate change on corn production ranges from −7%
to −41% and on soybean ranges from −8% to −45%, depending on the climate change scenarios,
time horizon, and global climate models used to predict climate change. We show that the aggre-
gate net impact of omitting price variables is an overestimation of the effect of climate change on
corn yield by up to 9% and on soybean yield by up to 15%.

Key words: Climate change, crop acreage, crop yield, prices, spatial correlation.
JEL codes: Q11, Q15, Q54.

Growing concern about the impact of cli-
mate change on agriculture has led to
numerous studies to estimate the effect of
changed climatic conditions on yields, land
values, and profits. Many of these studies
have used field experiments or simulation
approaches to assess the impact of climate
variables on yields (see a review in Long
et al. 2006). These studies omit consider-
ation of the potential for farmers to adapt
to climatic changes by changing input use,
crop rotations, and agronomic practices,
and, therefore, may be over-estimating the
Ruiqing Miao is a postdoctoral research associate, Energy Bio-
science Institute, University of Illinois at Urbana-Champaign,
Urbana, Illinois. Madhu Khanna is a professor, Department of
Agricultural and Consumer Economics and Energy Biosciences
Institute, University of Illinois at Urbana-Champaign, Urbana,
Illinois. Haixiao Huang is a senior analyst, NV Energy, Las
Vegas, Nevada. He contributed to this research while he was
a post-doctoral fellow at the Energy Biosciences Institute.
All correspondence should be addressed to Madhu Khanna
at khanna1@illinois.edu. The authors gratefully acknowledge
funding from the Energy Biosciences Institute, University
of California, Berkeley. They thank Kathy Baylis for com-
ments and suggestions on the spatial econometric techniques
applied in this article; Wolfram Schlenker, Atul Jain, and
David LeBauer for providing climate data; Michael Roberts,
Peter Berck, and two anonymous referees for their insightful
and helpful comments. Any remaining errors are their own.
damages due to climate change (Mendelsohn,
Nordhaus, and Shaw 1994). Other stud-
ies have used the Ricardian approach to
estimate the impact of climate on land
rent or value of farmland while allow-
ing for the possibility of adapting to it
(Mendelsohn, Nordhaus, and Shaw 1994;
Schlenker, Hanemann, and Fisher 2005, 2006;
Deschênes and Greenstone 2007). The Ricar-
dian approach also accounts for the impact
of climate on the agricultural sector as a
whole by including both the direct impact of
climate on yields and indirect impact due to
changes in inputs, crops, and other practices.
It relies on cross-sectional data and inher-
ently assumes that input and output prices
remain constant over time even with climate
change (Darwin 1999; Schlenker, Hanemann,
and Fisher 2005; Schlenker and Roberts
2009). However, as Deschênes and Green-
stone (2007) and Schlenker and Roberts
(2009) note, climate change may affect crop
prices by affecting world crop production.
This raises questions about the extent to
which forecasts of the long-term effects on
land value or crop yield from these models
can be attributed to climate change.
Recent studies have used panel data for
county-specific yield and weather to examine

Amer. J. Agr. Econ. 98(1): 191–211; doi: 10.1093/ajae/aav025

Published online May 29, 2015

c The Author 2015. Published by Oxford University Press on behalf of the Agricultural and Applied Economics
Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 192 January 2016
nonlinearity in the effect of temperature
and the interactions between temperature
and precipitation on crop yields. Schlenker
and Roberts (2009) project that warmer tem-
peratures can severely reduce crop yields
in the rainfed United States by 30%–82%
depending on the speed of global warming
by the end of the century. Deschênes and
Greenstone (2007) and Kucharik and Serbin
(2008) find that an increase in precipitation
is beneficial for corn and soybean yields
whereas McCarl, Villavicencio, and Wu
(2008) suggest that increased precipitation
can be harmful for wheat yield, beneficial for
cotton yield, but have no significant impact
on corn, sorghum, and soybean yield.
These studies treat the crop production
function to be a technical relationship
between yield per acre and climate vari-
ables while controlling for time invariant
factors such as soil quality using fixed effects.
Although this is appropriate at a field scale,
observed yield per acre of a crop at a more
aggregate level (such as a county) represents
the direct effect of input application and crop
variety decisions (determined at least in part
by input and output prices) and the indi-
rect effect of the land quality in the county
used for the crop. Expected crop prices can
influence the adoption of improved crop
practices, seed varieties, and other technolo-
gies which can affect crop yields (Kaufmann
and Snell 1997). Additionally, land quality
in the county used for the crop is likely to
be affected by input and output prices too
because they affect incentives for substi-
tution of acres among crops as well as for
making changes in crop acreage at the exten-
sive margin. This determines the extent to
which low-quality marginal land is brought
into crop production and affects the average
yield in the region. However, aforemen-
tioned studies examining the effect of climate
variables on crop yields do not control for
farmers’ responses to expected input and
output prices.
A few earlier econometric studies focus
on the positive responsiveness of crop yields
to prices while controlling for the effect of
climate variables (Houck and Gallagher
1976; Menz and Pardey 1983; Choi and
Helmberger 1993; Kaufmann and Snell
1997). But these studies do not account
for the endogeneity of crop prices. Fur-
thermore, data used in these studies are
aggregated at the national level and are
quite outdated (prior to 1989). Recent public
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
debate on indirect land-use impact of corn-
based ethanol production indicates that price
elasticities of crop yield and acreage play a
key role in quantifying this impact (Khanna
and Cargo 2012; Gohin 2014). It is therefore
important to examine the responsiveness
of crop yields and acreage to prices using
recently developed econometric methods and
updated data.
Climate can affect crop supply through its
effects not only on crop yields but also on
crop acreage. For example, precipitation can
influence the date when crops are planted
and hence can affect crop choice and acreage
planted. Excessive precipitation in spring
has been observed to delay planting of corn
and lead to an increase in soybean acreage
in the Midwest several times in the last
three decades (Sacks et al. 2010; Weinraub
2013). Studies analyzing the determinants
of crop acreage have focused primarily on
the effect of economic variables without
controlling for the effect of climate variables
(see table 1 for a list of relevant studies). An
exception is an early study by Orazem and
Miranowski (1986) that controls for the effect
of spring precipitation when studying how
crop acreage allocation responds to different
price expectation regimes.1
The purpose of this article is threefold.
First, we seek to examine the effect of cli-
mate variables on yield and acreage of corn
and soybeans and therefore on the supply of
these crops while controlling for the effect
of crop input and output prices. Second, we
project the impact of climate change on crop
production under alternative climate scenar-
ios and temporal frameworks by using the
most updated climate change forecasts as of
2014 from the Intergovernmental Panel on
Climate Change (IPCC)’s fifth Assessment
Report (AR5). Third, based on the estimates
for the price elasticities of crop yield and
acreage, we examine the extent to which
changes in crop prices lead to changes at the
intensive versus extensive margin by affecting
yield per acre and crop acreage. Our analysis
focuses on the rainfed area of the United
States (east of the 100th meridian) for the
period 1977–2007.
The inclusion of crop prices in the reduced
form regressions of crop yield and acreage
1
However, because their focus is on comparing the effect of
different expectation regimes, Orazem and Miranowski (1986) do
not explicitly discuss the effect of precipitation on crop acreage.
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate 193

Table 1. Estimates of Crop Yield and Acreage Elasticities in Different Studies

Crop Yield Elasticities
Study Price Used Crop Type Elasticity
Choi and Helmberger (1993) Futures Corn 0.27
Soybeans 0.13
Houck and Gallagher (1976) Received Corn 0.24–0.76
Lyons and Thompson (1981) Received Corn 0.22
Menz and Pardey (1983) Received Corn 0.61
This article Received Corn 0.23
Soybeans Not statistically significant
Futures Corn 0.26
Soyabeans Not statistically significant
Crop Acreage Elasticities
Cross-price
Own-price Elasticity between
Study Price Used Crop Type Elasticity Corn and Soybeans
Barr et al. (2011)a Futures Total acreage 0.007–0.029
Chavas and Holt (1990)a Received Corn 0.15 −0.15
Soybeans 0.45 −0.30
Chembezi and Womack (1992)a Received Corn 0.10 −0.05
Lee and Helmberger (1985)a Received Corn 0.05 −0.15
Soybeans 0.25 −0.15
Lin and Dismukes (2007)a Futures Corn 0.17–0.35 N/A
Soybeans 0.30 N/A
Miller and Plantinga (1999)a Received Corn 0.95 −0.45
Soybeans 0.95 −0.40
Orazem and Miranowski (1994)a Futures Corn 0.10 0.02
Soybeans 0.33 0.01
Tegene, Huffman, and Received Corn 0.20 N/A
Miranowski (1988)b
This articlea,c Received Corn 0.45 N/A
Soybeans 0.63 N/A

Note: a Acreage regression models do not control for lagged acreage. b Acreage regression models control for lagged acreage. c See footnote 12 for a
discussion regarding results using futures prices.

raises several econometric issues. First, we
need to consider how farmers form their
expectations of crop prices when making
decisions about crop acreage and cropping
practices at the time of planting and the
appropriate proxies for these expected prices.
Existing studies have used various proxies for
expected prices, including lagged-year real-
ized prices and current-year futures prices,
but there is no strong evidence to suggest
that one outperforms the other in describing
farmers’ expectations. We examine the
robustness of our results to both proxies
for expected prices. Second, the inclusion
of crop prices as explanatory variables has
the potential to lead to endogeneity bias
because yield and acreage in a given year
can affect crop prices in that year (Roberts
and Schlenker 2013). Moreover, since crop
prices are serially correlated and supply
shocks in one year can persist over time,
even lagged prices are likely to be correlated
with unobservable factors that affect current
yields and acreage decisions.2 We therefore
use an instrumental variable (IV) approach
to address this endogeneity problem. Third,
since county specific weather data and unob-
servable factors that affect crop yield and
land allocation can be quite similar across
a wide region (say, a state), county-level
crop yields and acreage are expected to be
spatially correlated. Therefore, we allow for
spatial autocorrelation of the error terms in
the yield and acreage regressions.
2
High correlation between futures price and received prices
(as found to be the case by Chavas, Pope, and Kao 1983) implies
that futures prices are endogenously determined as well.

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 194 January 2016
Accounting for endogeneity of prices by
using IV methods and controlling for spa-
tial autocorrelation of the error terms in the
yield and acreage regressions by applying
recently developed techniques in spatial
panel econometrics are two contributions
of this article compared to earlier studies
regarding the effects of prices on yields and
acreage (Houck and Gallagher 1976; Menz
and Pardey 1983; Choi and Helmberger 1993;
Kaufmann and Snell 1997). Moreover, by
including both prices and climate variables
to explain crop yield and acreage, we avoid
possible estimation biases caused by omitted
variables in previous studies (e.g., Schlenker
and Roberts 2009).3
Like Schlenker and Roberts (2009), we
find that climate change will decrease crop
yields. We also find that spring precipitation
has a statistically significant effect on crop
acreages. The net combined impact of cli-
mate change on corn production ranges from
−7% to −41% and on soybean ranges from
−8% to −45%, depending on the climate
change scenarios, time horizon considered,
and global climate models used to predict
climate change. More importantly, we show
that omitting price variables from the corn
yield regression will underestimate the effect
of growing degree days (GDD) by 10.3%
and overestimate the effect of overheat
degree days, temperature deviation, and
precipitation by up to 41.3%, 18.1%, and
9.6%, respectively. The net impact of omit-
ting price variables is an overestimation of
the effect of climate change on corn yield
by up to 9% and on soybean yield by up to
15%, depending on climate change scenarios
and time horizon considered. These results
indicate that the negative impact of climate
change on crop supply may be partly miti-
gated by changes in inputs and production
practices due to the increase in crop prices
that is likely to be induced by reduction in
production caused by climate change.
We find a statistically significant own price
elasticity of corn yield of 0.23 but not of
3
Schlenker and Roberts (2009) used year fixed effects to
control for technology and time effects in their sensitivity analysis.
One may argue that, based on the law of one price, the effect
of prices is controlled for by year fixed effects. Crop prices,
however, differ across states and over time and their effects are
therefore difficult to control while also controlling for county-
specific unobservable using county fixed effects. Moreover, we are
interested in estimating the effect of prices on yield and acreage,
which cannot be achieved using year fixed effects models while
assuming that the law of one price holds.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
soybean yield. For corn and soybean acreage,
the own price elasticities are 0.45 and 0.63,
respectively. Based on the own price elastic-
ities of corn yield and acreage, we calculate
that 33.8% of an increase in corn production
caused by an increase in corn price is due
to yield increase, whereas 66.2% is due to
acreage increase. Failing to account for the
price elasticity of yield will therefore overes-
timate the land-use change effect of biofuel
production because it ignores the mitigating
effect of changes at the intensive margin
that can increase supply. Our findings have
implications for the recent public debate
on indirect land-use impact of corn-based
ethanol production. That is, models assuming
a zero price elasticity of corn yields to prices,
such as Searchinger et al. (2008), are likely
to be overestimating the impact of biofuel
production on indirect land use change.
The rest of the article is organized as
follows. We first present a conceptual frame-
work to motivate the effect of prices, climate,
and agricultural policy on crop yield and
acreage. Then we illustrate econometric mod-
els and data used in the analysis, followed by
estimation results and the implications of
these results. The last section concludes.
Conceptual Framework: Determinants of
Crop Yield and Acreage
A crop production function at a field level is
typically modeled as a technical relationship
between yield per acre and factors affect-
ing yield such as input use, soil fertility,
farm management practices, and climate.
Although climate and soil fertility are exoge-
nously determined to farming, input use and
management practices can be controlled
by farmers. Moreover, observed yield per
acre of a crop at the county-level repre-
sents the effect of farming decisions at both
intensive margin (e.g., input application,
management practice, and crop choices, etc.)
and extensive margin (e.g., acreage expan-
sion). These decisions at the intensive and
extensive margins are likely to be affected
by input and expected output prices as well
as by agricultural policies that affect crop
prices received by farmers and the flexibility
of making crop acreage changes. A reduced
form relationship examining the determi-
nants of county-level average crop yield
and crop acreage is therefore likely to be a
function not only of soil and climate variables
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate
but also of input and expected output prices
as well as agricultural policies.
Effects of Output and Input Prices
An increase in the output price of a crop can
be expected to increase the acreage devoted
to the crop by creating incentives to switch
land from other uses to this crop. However,
the effect of expected output price on crop
yield per acre at the county-level is concep-
tually ambiguous. On the one hand, ceteris
paribus, an increase in expected price of a
crop could lead farmers to increase input-use
and adopt more productive management
practices and crop varieties (Feng and
Babcock 2010), which will increase the aver-
age county level yield per acre of this crop.
On the other hand, the increase in expected
price could also lead to a change in rotation
practice (e.g., an increase in corn price can
be expected to lead land converting from a
corn-soybean rotation to a continuous-corn
rotation) and expanding acreage under the
crop to marginal, low quality acres (Feng and
Babcock 2010), which could decrease county-
average crop yield per acre. The net effect of
expected crop price on county-average crop
yield depends on the relative magnitude of
these effects and is an empirical question.
Conceptually, the effect of input prices
on county-level average yield and acreage
is ambiguous as well. For instance, the
direct effect of the reduction in fertilizer use
induced by an increase in fertilizer price
is likely to be negative for corn yield on a
specific field. However, with an increase in
fertilizer price, land with low corn productiv-
ity may be converted to other crops so that
only land with high corn productivity is left
for corn production, which may increase the
average corn yield in a county. The effect
of increasing fertilizer price on crop acreage
could be negative if it leads farmers to switch
to crops that require less fertilizer to save
input costs (e.g., from corn to soybeans). It
could also lead farmers to substitute land for
fertilizer and expand crop acreage.
Due to biological lags in agricultural pro-
duction, planting decision and post-planting
crop management practices have to be based
on farmers’ expectations of output prices.4
Gardner (1976) provides empirical evidence
4
Irwin and Thraen (1994) and Gouel (2010) provide
comprehensive surveys on this issue.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
195
to support the use of futures price as being
consistent with rational expectations. Chavas,
Pope, and Kao (1983) examine the per-
formance of futures price and one-year
lagged received price in determine farm-
ers’ acreage decision. They show that since
futures price and one-year lagged received
price are highly correlated, it makes little
difference whether futures price or one-
year lagged received price is used, and the
issue of multi-collinearity will likely arise
if both prices are included in a regression.
Like the results of Chavas, Pope, and Kao
(1983), Shideed and White (1989) also find
that using futures prices or lagged received
price generates similar price elasticities of
acreage. In sum, the existing literature sug-
gests that futures prices and lagged received
price can be used interchangeably and there
is no strong evidence to suggest that one
outperforms the other in describing farmers’
expectations.
In empirical studies, both lagged-year
received prices and current-year futures
prices are widely used as proxies for expected
prices (see table 1). For example, Houck
and Gallagher (1976) and Menz and Pardey
(1983) use one-year lagged received prices
as proxy for the expected prices to explain
crop yield whereas Choi and Helmberger
(1993) use futures prices. Among studies
analyzing the response of crop acreage to
prices, Chavas and Holt (1990), Tegene,
Huffman, and Miranowski (1988), Chembezi
and Womack (1992), Lee and Helmberger
(1985), and Miller and Plantinga (1999)
use lagged received prices as an explana-
tory variable. Studies that use futures prices
include Lin and Dismukes (2007), Barr et al.
(2011), and Orazem and Miranowski (1994).
In this study, we examine the effect of both
one-year lagged received price and, sepa-
rately, current-year futures price as proxies
of expected crop prices.
Effects of Climate Variables on Yield and
Acreage
It is well documented that GDD have a
nonlinear effect on yield and that tem-
perature higher than a threshold may be
harmful for the growth of crops (Schenkler
and Roberts 2009). In addition to GDD,
precipitation is another key factor that
affects crop yields. Yields are also expected
to be affected by temperature variability
(McCarl, Villavicencio, and Wu 2008).
 196 January 2016
Therefore, in our crop yield models we
include variables for GDD, precipitation,
and temperature deviation. Specifically, we
include GDD and its squared terms to cap-
ture the non-linear effect of temperature.
We also include overheat degree days in
order to capture temperature’s threshold
effect. Since the effect of precipitation on
yield will depend on the timing and intensity
with potentially nonlinear effects, unlike
previous studies we include monthly mean
precipitation during the growing season and
their squared terms to capture the non-linear
effects of precipitation and the potential
impact of timing and seasonal variation on
yield. We also include monthly tempera-
ture deviations over growing season in yield
models.
Within a growing season, information
about post-planting climate conditions is not
available when acreage allocation decisions
are made. Therefore, post-planting climate
conditions should not be included in acreage
models although they affect crop yields.
However, we believe that it is reasonable to
include spring precipitation in acreage mod-
els because a wet spring will make it difficult
for corn being planted on time, and hence
corn acreage may be switched to soybean
acreage (Sacks et al. 2010; Weinraub 2013).
Therefore, we include monthly precipitation
in March to May to control for the effect of
the pre-planting climate on crop acreage.5
Moreover, since farmland faces competition
from population growth and urban devel-
opment, we include population density in
the model to control for its effect on crop
acreage in a county.
Effects of Farm Policy
Crop yields and acreage are also likely to
be affected by agricultural policies that can
either directly affect expected returns from
cropping or the flexibility to make farm level
(and therefore aggregate) acreage decisions.
A significant change in US farm policy took
place in 1996 with the passage of the Fed-
eral Agriculture Improvement and Reform
(FAIR) Act, which decoupled government
payments from farmers’ actual production
decisions. Prior to the FAIR Act, crops were
classified as program and nonprogram crops
5
We are indebted to two anonymous referees for comments
on how climate affects crop acreage that led to this model
specification.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
and government payments that a farmer
received in a year were limited to program
crops based on the actual acreage devoted
to that program crop and the yield specified
for the farmer.6 These coupled payments that
were calculated based on actual acreage and
yield rewarded more intensive production
and higher crop yields. After the FAIR Act
was enacted, farmers received Production
Flexibility Contract (PFC) payments that
were calculated based on historical acreage
and yield, which indicates that the FAIR Act
allowed for planting flexibility.
Several studies have found that decoupled
payments affect farmers’ risk attitudes and
have implications for input use, technical effi-
ciency, and output, although these effects are
typically small (e.g., Serra, Zilberman, and
Gil 2008; Bhaskar and Beghin 2009; Serra,
Goodwin, and Featherstone 2011; Weber
and Key 2011). Serra et al. (2006) find that
decoupling may decrease the mean and vari-
ance of farm-level output through reducing
the use of risk-increase input. Simulations
by Ray (1996) show that the 1996 FAIR Act
could be expected to decrease the acreage
of program crops such as corn and cotton. In
this study, we control for the effects of the
FAIR Act by including a dummy variable
that equals 0 in years earlier than 1996 and 1
in 1996 and thereafter.
Empirical Framework
We model the county-specific average yield
per acre, Yijt , of crop i ∈ {c, s}, with c for corn
and s for soybeans, in county j and in year
t as
(1) Yijt = βi0 + βi1 ijt + βi2  ijt + βi3  ijt
+ Dt + uij + ijt
where βi0 is a constant, and βi1 to βi3 are
vectors of parameters to be estimated; , ,
and  are vectors of prices, climate variables,
and time trend, respectively; Dt is a dummy
6
The approach employed to specify yield for a farmer prior to
the FAIR Act evolved over time. Prior to the Food, Agriculture,
Conservation, and Trade (FACT) act of 1990, actual yields were
used to calculate program payments. The 1990 FACT froze the
yields used to calculate program payment at 1985 levels. The 1996
FAIR Act continued the 1990 FACT act in determining payment
yield. The 2002 Farm Security and Rural Investment (FSRI) Act
allowed producers to voluntarily adjust base yield to match the
real productivity.
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate
variable reflecting the enactment of the 1996
FAIR Act (Dt = 0 when t < 1996 and Dt = 1
when t ≥ 1996); uij is a county-level fixed
effect, and ijt is an error term. The price
vector  in equation (1) includes both input
price and output prices. The input price is
represented by a fertilizer price index. The
output price is one-year lag of state-level
received prices, or separately, planting-
season prices of futures contracts that mature
in harvest season. The climate vector, , in
equation (1) includes GDD and its squared
term, overheat degree days, monthly mean
precipitation and their squared terms, and
monthly deviation in temperature, which is
included to control for variability in tem-
perature. We include disaggregated climate
variables (e.g., monthly precipitation and
monthly temperature deviation) to examine
the potential impact of timing and seasonal
variation in climate variables on agricul-
tural production. The definitions of these
climate variables are given below. Soil qual-
ity variables are not included directly in the
regression since they are time invariant and
cannot be distinguished from county-specific
effects. The time trend vector  includes a
linear and a quadratic form of time trend to
reflect advances in technology and agronomic
practices.
Similarly, the reduced-form of acreage
models can be written as
(2) Aijt = γi0 + γi1 ijt + γi2  ijt + γi3  ijt
+ γi4 Pjt + γi5 Dt + vij + eijt
where the subscripts, variables, and param-
eters are defined similarly as in equation (1)
except that A is crop acreage and P is pop-
ulation density. Unlike the yield model, the
climate vector, , in the acreage equation
includes monthly spring precipitation (March
to May) only.
Climate conditions, soil quality, and
input and output prices in a large region
(say, a state) can be quite similar, and the
county-level yield and crop acreage may
be correlated with those in neighboring
counties. Therefore, we use Moran’s I test
to examine spatial autocorrelation of yield
and acreage. Results show that Moran’s I
statistics for corn and soybean yields are 0.73
and 0.85, respectively. For corn and soybean
acreage, Moran’s I statistics are 0.69 and 0.67,
respectively. All these statistics are significant
at 1% level, so we can conclude that both
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
197
yield and acreage are positively spatially
correlated. Therefore, in yield and acreage
models described in equations (1) and (2),
we allow for spatial autocorrelation of the
error terms ijt and eijt . Failure to take spatial
autocorrelation into account may generate
biased and inconsistent estimates (LeSage
1997) and lead to underestimation of the
standard errors (Schlenker, Hanemann, and
Fisher 2006).
We define i (t) ≡ (i1t , . . . , iNi t ) in equation
(1) and ei (t) ≡ (ei1t , . . . , eiMi t ) in equation
(2) where Ni and Mi are numbers of coun-
ties in yield models and acreage models,
respectively, and i ∈ {c, s}. Then the spa-
tial autocorrelation takes the form: i (t) =
ρi WiNi i (t) + ξi (t) and ei (t) = φi WiMi ei (t) +
ηi (t), where ρi and φi are scalar autoregres-
sive parameters; WiNi and WiMi are Ni × Ni
and Mi × Mi weight matrix of constants that
do not change over year t; and ξi (t) and ηi (t)
are Ni × 1 and Mi × 1 vectors of innova-
tions in year t, respectively. To construct the
weight matrix, we define neighbors as the
8 nearest counties to a county.7 Following
Schlenker, Hanemann, and Fisher (2006), we
first obtain generalized method of moments
(GMM) estimators of ρi and φi and then
transform the data by pre-multiplying the
yield data by I − ρ̂i WiNi (I − φ̂i WiMi for the
acreage data). The command “spgm” in R
package “splm” is applied for estimating the
GMM estimators.
To address the endogeneity of input and
output prices, we apply a panel data IV
estimator with county fixed effects. The
fixed-effect IV approach is appropriate since
explanatory variables such as lagged crop
prices and fertilizer price index may not be
strictly exogenous and the time-invariant
county characteristics such as geography
may be correlated with those explanatory
variables. Following Roberts and Schlenker
(2013), we use one-year lagged climate vari-
ables and one-year lagged crop stocks as
IVs because they are exogenous and affect
current yields and acreage only through
affecting prices. Specifically, across all corn
and soybean yield models as well as corn
7
For soybean acreage models, we define neighbors as the 5
nearest counties. Baylis, Paulson, and Piras (2011) define neighbors
as the ten nearest counties. We define a smaller number of nearest
counties as neighbors because in our dataset the selected counties
are more sparsely distributed than those in Baylis, Paulson, and
Piras (2011) due to the fact that some counties do not have corn
or soybean production.
 198 January 2016
acreage models, the IVs are: one-year lagged
corn stocks, one-year lagged soybean stocks,
and one-year lagged GDD. For soybean
acreage models, since the aforementioned
three IVs do not pass the over-identification
test, we use one-year lagged corn stocks, one-
year lagged soybean stocks, and one-year
lagged annual precipitation as IVs.
We specify a linear functional form for
the yield models and a log-linear functional
form for the acreage models. Crop prices are
specified in logarithmic form in the acreage
model while other independent variables
are included in levels. This model specifica-
tion is supported by results from Box-Cox
regressions (see Greene 2003, 500). Based
on the original datasets (i.e., non-spatially
transformed), for the yield models, the Box-
Cox test suggests a linear form because the
maximum likelihood estimates (MLE) of
Box-Cox transformation parameters for corn
and soybean yield models are 0.9 and 0.8,
respectively. The MLE of the Box-Cox trans-
formation parameters for corn and soybean
acreage models are 0.26 and 0.31, respec-
tively, which indicate that the log-linear
models are preferred.8
Data and Variables
We obtain county-level yield and planted
acreage data for corn and soybean over
1977–2007 from National Agricultural Statis-
tics Service (NASS) of the US Department
of Agriculture (USDA). We focus only on
rainfed counties, that is, counties on the east
of the 100th meridian. We construct bal-
anced panels because we are estimating panel
data models with spatial correlation. Spatial
panel econometrics with unbalanced panel
datasets may have problematic asymptotic
properties when the reason for missing data
is unknown (Elhorst 2010). For the corn yield
and acreage regressions, data from 1,144
counties are included, whereas for the soy-
bean yield and acreage equations, data from
926 counties are included.
The state-level received prices for each
crop are obtained from NASS as well.
Although futures prices are at national-
level, crop planting dates in each state differ.
8
When running the Box-Cox regressions for both acreage
and yield models, we treat the panel datasets as cross-sectional
datasets. We also estimate Box-Cox regressions based on spatially
transformed datasets and found similar results.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
By using this difference we generate state-
level futures prices because what matters
to farmers’ price expectation is the planting
season price of futures that mature in the
harvest season. Specifically, for a state, we
first identify its most active planting dates
based upon “Field Crops Usual Planting and
Harvesting Dates” reported by NASS (2010).
Then we calculate the average prices during
these planting dates of October corn futures
prices and November soybean futures prices
to obtain the corn and soybean futures prices
to be used in the regressions. The national-
level fertilizer price index is obtained from
data published by Economic Research Ser-
vice of USDA.9 All prices are converted to
2000 dollars using US GDP implicit price
deflator. State-level crop stocks on Decem-
ber 1st in each year over 1977–2007 are also
downloaded from NASS.
Monthly mean, minimum, and maxi-
mum temperature, as well as monthly mean
precipitation for each county from Jan-
uary 1977 to December 2007 are derived
from the Parameter-Elevation Regressions
on Independent Slopes Model (PRISM)
that was developed by the Spatial Climate
Analysis service at Oregon State University
(PRISM Climate Group 2009). Here, the
monthly mean temperature is the mean of
daily average temperature within a month.
Monthly minimum (respectively, maximum)
temperature is the mean of daily minimum
(respectively, maximum) temperature within
a month. Monthly deviation in temperature
is calculated as the monthly maximum minus
the monthly minimum temperature. In this
study, we include both a GDD variable,
a squared term of GDD, and an overheat
degree days variable in yield models. Follow-
ing Schlenker, Hanemann, and Anthony
(2006) and Deschênes and Greenstone
(2007), GDDs in the range 8–32◦ C within
growing season are included in the regres-
sions. GDDs above 32◦ C are defined to be
overheat degree days. GDDs in the range of
8–32◦ C and overheat degree days are calcu-
lated based on the PRISM data following the
standard mean temperature cut-off approach
illustrated by Chandrasekaran, Annadurai,
and Somasundaram (2010, 512). Following
Schlenker, Hanemann, and Fisher (2006), we
employ the square root of overheat degree
9
Website: http://www.ers.usda.gov/data-products/fertilizer-use-
and-price.aspx (accessed on Dec. 2nd , 2014).
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate
days in the models. In this study, we define
the growing season as March to August as in
Schlenker and Roberts (2009) because plant-
ing and harvesting of corn and soybeans in
the majority of growing states starts in March
and September, respectively (NASS 2010).
The county-level data on population den-
sity from 1977 to 2007 are obtained from
Population Division at the US Census
Bureau (2009). Table 2 contains the sum-
mary statistics for corn yield and acreage
dataset. Summary statistics for soybean yield
and acreage dataset are provided in table
S1 in the online Supporting Information (SI
hereafter).
Estimation Results
In this section, we discuss the estimation
results of the yield and acreage regressions.
We estimate three alternative specifications
for each of the yield models of corn and soy-
beans. For acreage models, we estimate two
alternative specifications for each crop.
Determinants of Crop Yields
The estimation results for crop yields are
presented in tables 3 and 4 for corn and
soybeans, respectively.10 Due to space limi-
tation, we present results based on received
prices in the main text and present results
based on futures prices in tables S2 and S3
in SI. The effects of crop prices on yield
are qualitatively the same whether received
prices or futures prices are used. As we see
in tables 3 and 4, yield model I includes only
climate variables. Yield model II includes
climate variables and the FAIR dummy.
Models III adds to yield model II by also
including fertilizer price index and the crop’s
own price. Kleibergen and Paap (2006) rk
statistic is used to test for underidentifica-
tion of model III. The test result shows that
model III is identified. We use two tests for
weak identification (i.e., excluded instru-
ments are just weakly correlated with the
endogenous variables): Cragg-Donald F
Wald statistic and Kleibergen-Paap Wald
rk F statistic. Our results show that yield
model III for both corn and soybean passes
the weak-identification test (tables 3 and 4).
10
To keep tables 3 and 4 in a reasonable size, we only present
estimations of key variables in these two tables. For a complete
presentation of estimations, see tables S2 and S3 in SI.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
199
That is, the instrument variables are strongly
correlated with the endogenous variables.
The test for overidentification using the
Hansen J statistic shows that we cannot reject
the hypothesis that the IVs are valid (i.e.,
uncorrelated with the error term) at any rea-
sonable significance levels. Therefore, we can
conclude that the IVs used in the regressions
are strongly correlated with endogenous
variables and uncorrelated with error term.
The same conclusion holds when futures
prices are used in the yield models. Further-
more, we account for serial autocorrelation
and heteroskedasticity by using the robust
estimator.
From table 3 we see that received corn
price has a statistically significant and posi-
tive effect on corn yield, with the coefficient
being about 7.4 in model III. When evaluated
at sample means, this implies that the elastic-
ity of corn yield with respect to received corn
price is 0.23. This indicates that the positive
effect of corn price on corn yield dominates
its negative effect through induced changes
in land use. We compare our elasticity esti-
mates with those in the literature (table 1).
Our estimates are close to those estimated
by Choi and Helmberger (1993) and Lyons
and Thompson (1981). We also estimate the
price elasticity of corn yields using futures
prices and the estimated coefficient of 7.9 of
corn futures prices (see model IV in table S2
of SI). When evaluated at sample means, the
elasticity of corn yield with respect to corn
futures price is 0.26.
As expected, corn yield responds neg-
atively to fertilizer index. Table 4 shows
that soybean yield is not sensitive to its own
price. The reason could be that soybean
yield increase may be mainly caused by the
adoption of genetic technology and by the
advances in agronomic practices over the
last few decades (Specht et al. 2014). It could
also be indicative of the intensive and exten-
sive margin effects offsetting each other.
The result that soybean yields are not sensi-
tive to fertilizer price index is not surprising
because not very much fertilizer is applied
to soybeans. Table 3 also shows that corn
yield has a positive and statistically significant
time trend; corn yield increases at an increas-
ing rate over time. Table 4 shows that the
coefficient of the linear term of time trend
in the soybean yield model is positive and
statistically significant in models I and II and
weakly significant (at 25% significance level)
in model III, respectively.
 200 January 2016 Amer. J. Agr. Econ.

Table 2. Summary Statistics of Dataset for Corn Yield and Acreage Analyses
(1,144 counties, 35,464 observations, 1977–2007)
Variables Mean SD Min Max
Corn yield (bushel/acre) 104.8 32.6 0.0 204.0
Corn received price ($/bushel) 3.2 1.2 1.6 6.8
Soybean received price ($/bushel) 7.9 2.9 4.0 15.0
Corn futures price ($/bushel) 3.4 1.2 2.0 6.3
Soybean futures price ($/bushel) 8.3 3.1 4.2 17.1
Corn acreage (planted acres) 49558 54539 200 397000
Fertilizer price index 111 28 72 209
Population density ( residents/squared mile) 132 282 2 5614
Precipitation in March (mm) 86.2 53.3 0.4 456.0
Precipitation in April (mm) 91.4 47.3 0.2 524.6
Precipitation in May (mm) 106.5 55.0 0.4 514.7
Precipitation in June (mm) 106.8 51.3 2.0 485.1
Precipitation in July (mm) 108.2 51.8 0.0 557.4
Precipitation in August (mm) 100.7 50.4 0.3 458.5
Growing degree days (GDD) (8–32◦ C) 2166 476 947 3589
Overheat degree days (>32◦ C) 22.9 55.2 0.0 539.3
Temperature deviation, March (◦ C) 12.1 1.8 6.2 19.1
Temperature deviation, April (◦ C) 13.2 1.6 7.2 19.4
Temperature deviation, May (◦ C) 13.0 1.5 7.2 19.7
Temperature deviation, June (◦ C) 12.5 1.4 6.9 20.7
Temperature deviation, July (◦ C) 12.1 1.3 6.8 19.5
Temperature deviation, August (◦ C) 12.3 1.2 7.2 18.1
Note: Statistics reported in this table are for the data used for the corn yield and corn acreage regressions. The data used for the soybean yield and
acreage regressions are presented in table S1 of SI. Prices are in 2000 dollars.

Table 3. Determinants of Corn Yield (yield and crop price are in levels)
Variables Model I Model II Model III
Received corn price – – – – 7.367∗∗ (3.43)
Lag fertilizer price index – – – – −1.381∗∗∗ (0.28)
Growing degree days 0.0269∗∗∗ (0.00860) 0.026∗∗∗ (0.009) 0.033∗∗∗ (0.01)
Degree days squared −1.00e-05∗∗∗ (2.09e-06) −9.87e-6∗∗∗ (2.09e-6) −1.18e-5∗∗∗ (2.19e-6)
Overheat degree days −0.688∗∗∗ (0.0940) −0.695∗∗∗ (0.094) −0.492∗∗∗ (0.105)
Temp. deviation, Mar. 0.430∗∗ (0.196) 0.418∗∗ (0.196) 0.053 (0.216)
Temp. deviation, Apr. 1.160∗∗∗ (0.214) 1.160∗∗∗ (0.213) 1.018∗∗∗ (0.226)
Temp. deviation, May 0.297 (0.233) 0.311 (0.233) 0.802∗∗∗ (0.262)
Temp. deviation, Jun. −1.016∗∗∗ (0.232) −1.079 ∗∗∗ (0.233) −1.488∗∗∗ (0.264)
Temp. deviation, Jul. −1.379∗∗∗ (0.272) −1.340∗∗∗ (0.272) −0.515 (0.324)
Temp. deviation, Aug. −0.517∗∗ (0.229) −0.502∗∗ (0.229) −0.776∗∗∗ (0.243)
Time trend 0.268∗ (0.154) 0.244 (0.155) 1.332 (0.858)
Time trend squared 0.0362∗∗∗ (0.00486) 0.0417∗∗∗ (0.006) 0.134∗∗∗ (0.024)
Fair dummy – – −3.321∗∗ (1.362) −12.84∗∗∗ (2.894)
Number of observations 34,320 34,320 34,320
Kleibergen-Paap rk LM – – 387.0
statistic
Cragg-Donald Wald F – – 120.2
statistic
Kleibergen-Paap rk Wald – – 130.4
F statistic
p-value of Hansen J – – 0.226
statistic
Spatial autocorrelation 0.829 0.828 0.828
Note: Overheat degree days are in square root form. Robust standard errors in parentheses. Coefficients of monthly precipitation over the growing
season are presented in table S2 in SI.
∗∗∗ 1% level.
∗∗ 5% level.
∗ 10% level.

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate 201

Table 4. Determinants of Soybean Yield (yield and crop price are in levels)
Variables Model I Model II Model III
Soybean price – – – – 0.186 (0.572)
Lag fertilizer price index – – – – 0.0948 (0.0879)
Growing degree days 0.0190∗∗∗ (0.00311) 0.0188∗∗∗ (0.00311) 0.0192∗∗∗ (0.0035)
Degree days squared −5.19e-06 ∗∗∗ (7.45e-07) −5.17e-06∗∗∗ (7.45e-7) −5.10e-06∗∗∗ (7.69e-7)
Overheat degree days −0.159∗∗∗ (0.0275) −0.160∗∗∗ (0.0275) −0.171∗∗∗ (0.0303)
Temp. deviation, Mar. 0.0279 (0.0622) 0.0257 (0.0622) 0.0548 (0.0695)
Temp. deviation, Apr. 0.289∗∗∗ (0.0646) 0.288∗∗∗ (0.0646) 0.302∗∗∗ (0.0660)
Temp. deviation, May 0.0502 (0.0681) 0.0522 (0.0681) 0.0237 (0.0759)
Temp. deviation, Jun. −0.0535 (0.0764) −0.0648 (0.0768) −0.0224 (0.0899)
Temp. deviation, Jul. −0.0596 (0.0787) −0.0520 (0.0790) −0.115 (0.100)
Temp. deviation, Aug. −0.552∗∗∗ (0.0746) −0.549∗∗∗ (0.0745) −0.523∗∗∗ (0.0790)
Time trend 0.316∗∗∗ (0.0506) 0.312∗∗∗ (0.0510) 0.514 (0.468)
Time trend squared 0.00159 (0.00157) 0.00263 (0.00181) −0.00818 (0.0128)
FAIR dummy – – −0.631 (0.446) −0.435 (0.697)
No. of observations 27,780 27,780 27,780
Kleibergen-Paap rk LM – – 287.0
statistic
Cragg-Donald Wald F – – 78.2
statistic
Kleibergen-Paap rk Wald – – 96.4
F statistic
p-value of Hansen J – – 0.591
statistic
Spatial autocorrelation 0.839 0.839 0.839
Note: Overheat degree days are in square root form. Robust standard errors in parentheses. Coefficients of monthly precipitation over the growing
season are presented in table S3 in SI.
∗∗∗ 1%level.
∗∗ 5% level.
∗ 10% level.

The sign and statistical significance of
the effect of climate variables on corn and
soybean yield are robust across the three
models and consistent with expectation. The
magnitude of the coefficients of the climate
variables in the corn yield model I are close
to those in model II but quite different from
those in model III. For example, the coeffi-
cients of GDD are 0.027 and 0.026 in models
I and II, respectively. In model III, however,
the coefficient of GDD is 0.033. This indi-
cates that inclusion of price variables affects
estimates of the impact of climate variables
on corn yield. A similar conclusion holds for
soybean yield models. We discuss this further
in the next section.
The coefficients of GDD and squared
GDD are statistically significant across
all model specifications for both corn and
soybean yield. The positive coefficients of
GDD and negative coefficients of squared
GDD indicate an inverse U-shaped rela-
tionship between yield and GDD. Based
upon model III, the estimated coefficients
of GDD and squared GDD suggest that,
ceteris paribus, corn and soybean yields peak
at 1,403 and 1,882 GDD, respectively. At the
GDD sample means, 2,166 for corn and 2,167
for soybean, the effect of a marginal increase
in GDD on corn and soybean yield is nega-
tive. Specifically, when evaluated at sample
means, a 1% increase in GDD will decrease
corn yield by 0.37% (about 0.4 bushel per
acre) and soybean yield by 0.19% (about 0.06
bushel per acre).
Extreme temperature that is higher than
32◦ C (i.e., overheat degree days) during the
growing season is found to be harmful for
corn yield. This result is consistent across all
the yield model specifications. For soybean
yield, the same conclusion holds except that
the magnitude of the coefficient is much
smaller than that for corn. Our results further
show that monthly temperature deviations
in March to May are beneficial for both
corn and soybean yields, while the monthly
temperature deviations in June to August
are harmful. An explanation for the above
pattern of effect could be that larger temper-
ature deviations in March to May indicate

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 202 January 2016
that weather is warmer than usual but with-
out extreme heat, which is beneficial for
crop yield. In June to August, however, a
larger temperature deviation is more likely
an indicator of extreme heat, which is harm-
ful for crop yields. McCarl, Villavicencio,
and Wu (2008) found that growing season
temperature deviation is harmful to crop
yield. Compared with their study, our study
shows that the timing of this temperature
deviation also matters. Our results show
that for both corn and soybean, the effects
of temperature deviation in June to August
dominate those in March to May. Therefore,
we expect the aggregate effects of tempera-
ture deviation in our study to be consistent
with those in McCarl, Villavicencio, and Wu
(2008). From tables S2 and S3 in SI, we see
that the monthly precipitation variables and
their squared terms generally have statisti-
cally significant impacts on corn and soybean
yields. The generally negative coefficients of
the squared precipitation variables indicate
that the relationship between crop yield and
precipitation is also inverse U-shaped.
The 1996 FAIR Act is found to have a
negative and statistically significant impact on
corn yield. This is consistent with the findings
of Serra et al. (2006) who show that decou-
pling may decrease the mean and variance
of farm-level output through reducing the
use of risk-increase input. On the other hand,
however, we do not find evidence that soy-
bean yield is sensitive to the enactment of the
FAIR Act.
Determinants of Crop Acreage
We estimate two acreage models for each
crop.11 Acreage model I does not include
spring precipitation while model II does.
Table 5 presents the regression results of
these two acreage specifications for both corn
and soybeans based on received prices.12
11
Corn acreage models with a quadratic term of time trend
rarely pass the over-identification tests and generate noisy pre-
dictions; we therefore do not include a quadratic term of time
trend in corn acreage models.
12
Table S4 in SI includes regression results for acreage models
based on futures prices. The results of the acreage models based
on futures prices are relatively noisy (e.g., the magnitude of price
coefficients is large, and particularly in soybean acreage models,
the coefficients of soybean prices are not statistically significant).
This may be because the regression on futures prices are based
on the state-level futures prices that are constructed using the
national-level futures prices. Chavas, Pope, and Kao (1983) also
find that futures prices may not be good proxies for expected
prices in the presence of government support programs because,
as their results show, “futures markets do not reflect the effects
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
Similar to yield models, test results for IVs
presented in the bottom of table 5 indi-
cate that the IVs of the acreage modes pass
the tests for under-identification, weak-
identification, and over-identification. That is,
the instrument variables used in the acreage
models are strongly correlated with endoge-
nous variables and uncorrelated with error
terms.
Table 5 shows that both corn and soybean
acreage positively respond to an increase
in their own prices. For corn acreage, this
response is statistically significant at 1% level
across all two specifications. For soybean
acreage, the coefficients of soybean price
are only significant at 5% or 10% levels.
In model I of corn acreage, the estimated
acreage elasticity with respect to corn price is
0.55. However, when monthly precipitation
variables over March to May are controlled
for, then this acreage elasticity decreases
to 0.45. A similar conclusion holds for soy-
bean acreage models. The coefficients of
fertilizer price index are all positive and sta-
tistically significant across all specifications of
both corn and soybean acreage models. This
may indicate that a higher fertilizer price
reduces the intensity of cultivation and leads
to changes at the extensive margin. Soybean
acreage could be increasing due to a switch
from continuous corn to corn soybean rota-
tions while corn acreage might be increasing
as some marginal land are converted to corn
production. We find that population density
has a negative and a statistically significant
effect on acreage across all specifications
with a coefficient at about 0.0009. For corn
acreage models, the coefficient of linear
term of time trend is negative and statisti-
cally significant, which indicates that, all else
equal, corn acreage is decreasing every year.
For soybean acreage models, the coefficient
of the linear term of time trend is positive
whereas that of quadratic term is negative,
indicating an inverse-U shaped soybean
acreage time trend.
We also find that monthly precipitation
in March to May has a negative effect on
corn acreage. However, only the coefficient
of April precipitation is statistically signifi-
cant. The coefficients of May precipitation
of government programs” (Chavas, Pope, and Kao 1983, 31).
Moreover, table 1 shows that the majority of previous studies
used lagged received prices when examining price elasticities of
crop acreage. Therefore, we prefer the set of results based on
the received price and use this set of results in the simulation in
the next section.
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate 203

Table 5. Determinants of Corn and Soybean Acreage (Dependent variable: log(planted

acres))
Corn Acreage Soybean Acreage
Variables Model I Model II Model I Model II
log(Own price) 0.550∗∗∗ 0.448∗∗∗ 0.749∗∗ 0.625∗
(0.182) (0.147) (0.364) (0.330)
Lag fertilizer price index 0.045∗∗∗ 0.040∗∗∗ 0.0373∗∗ 0.0355∗∗∗
(0.011) (0.008) (0.0145) (0.0127)
Population density −0.000914∗∗∗ −0.000924∗∗∗ −0.000947∗∗∗ −0.000873∗∗∗
(0.0001) (0.0001) (0.000284) (0.000283)
Time trend −0.070∗∗∗ −0.066∗∗∗ 0.107∗∗∗ 0.0971∗∗∗
(0.020) (0.016) (0.0300) (0.0272)
Time trend squared – – −0.00538∗∗∗ −0.00508∗∗∗
(0.00134) (0.00117)
Fair dummy −0.265∗∗ −0.211∗∗∗ 0.0179 0.0118
(0.082) (0.066) (0.0801) (0.0714)
Precipitation in March −5.64e-5 −0.000442
(0.0002) (0.000273)
Precipitation in April −0.000451∗∗∗ −0.000378
(0.0001) (0.000273)
Precipitation in May −4.27e-5 0.000628∗∗
(0.0001) (0.000267)
Observations 34,320 34,320 27,780 27,780
Kleibergen-Paap rk LM statistic 92.45 127.65 209.12 234.83
Cragg-Donald Wald F statistic 36.43 47.6 62.92 71.90
Kleibergen-Paap rk Wald F statistic 30.82 42.7 74.04 83.00
p-value of Hansen J statistic 0.57 0.25 0.20 0.13
Spatial autocorrelation 0.85 0.82 0.62 0.59
Note: Robust standard errors in parentheses.
∗∗∗ 1% level.
∗∗ 5% level.
∗ 10% level.

in soybean models are positive and statisti-
cally significant, indicating that an increase in
May precipitation discourages corn planting
and encourages soybean planting. The coef-
ficients of FAIR dummy are negative and
statistically significant across all models and
price specifications for corn acreage. Based
on results of corn acreage model II in table 5,
we find that, all else equal, the passage of the
FAIR Act decreases corn acreage by about
21%. For soybean acreage models, the coeffi-
cients of FAIR dummy are positive but none
of them are statistically significant.
Implications of the Estimation Results
In this section, we quantify the effects of
omitting price variables when estimating
the impact of climate change on crop yields.
We also decompose the effect of a corn price
increase on corn production into that through
a change in yield and that through a change
in acreage. Then we discuss the potential
implications of the estimation results for
the land use effects of price changes. We
focus on corn in the main text and leave the
calculation results for soybeans in SI (see
tables S5–S8 in SI).
Impact of Climate Change on Corn Yield
We use parameter estimates from corn yield
models II and III in table 3 for estimating
the impact of climate change on corn yield.
The only difference between these two mod-
els is that model III includes prices variables
whereas model II does not. We first obtain
changes in climate variables under different
future climate scenarios from two climate
models (to be described below) and then use
those changes with the parameter estimates
in yield models II and III in table 3, as well as
acreage Model II in table 5, to quantify the
effect of omitting price variables on the effect
of climate on corn yield and production.
Due to different model specifications, data,
methodology, and research purposes, results
in our study are not directly comparable with
those in Schlenker and Roberts (2009).

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 204 January 2016
Projections of future climate variables used
in our calculation are obtained from World-
Clim (http://www.worldclim.org; accessed 9
May 2015), which provides monthly climate
predictions in the medium term (average for
2041–2060) and in the long term (average
for 2061–2080) based on the most updated
climate projections as of 2014 that are used
in IPCC’s AR5. WorldClim covers climate
predictions of monthly average minimum
temperature, monthly average maximum
temperature, and monthly total precipita-
tion of 19 global climate models under four
greenhouse gas representative concentra-
tion pathways (RCP). These four pathways
are RCP2.6, RCP4.5, RCP6.0, and RCP8.5.
We utilize data from global climate mod-
els HadGEM2-ES and NorESM1-M with
spatial grids at ten minutes of a degree of
longitude and latitude (about 11.6 miles
at the equator). These two climate mod-
els are selected because they provide good
contrasts on global temperature changes
and have complete datasets under the four
RCPs (Warszawski et al. 2014).13 To obtain
the county-level climate information in the
medium term, and long term we take the
average of data for four closest grid points
of each county centroid. The changes in
climate variables are calculated by using
future climate data from WorldClim minus
sample means of climate data used in our
regressions.
Table 6 presents summary statistics of
predicted changes in climate variables under
RCP2.6 and RCP8.5 by models HadGEM2-
ES and NorESM1-M for counties included
in this study. HadGEM2-ES predicts that in
the medium term the average temperature
increases under RCP2.6 and RCP8.5 are
2.6◦ C and 4.1◦ C, respectively. In the long
term, HadGEM2-ES predicts that the aver-
age temperature increases under RCP2.6
and RCP8.5 are 2.5◦ C and 6.1◦ C, respec-
tively. Climate model NorESM1-M has lower
predictions on temperature increases than
HadGEM2-ES.
By using the climate change data described
above, we calculate the predicted yield
changes for each county based on corn
yield models II and III in table 3, respec-
tively. The acreage-weighted average yield
effects are summarized in table 7 (based
13
HadGEM2-ES predicts larger temperature increase than
does NorESM1-M.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
on HadGEM2-ES) and table S9 (based
on NorESM1-M). We find that if we use
climate predictions from HadGEM2-ES,
then corn yield model III (price variables
are included) predicts that climate change
decreases corn yield on average by 11.3%–
40.1%. Table 7 also shows that omitting
price variables in corn yield estimation
will overestimate the impact of climate
change on average corn yield by 3%–7%,
depending on climate change scenarios
and temporal term considered. If climate
predictions from NorESM1-M are used,
then omitting price variables in the yield
models overestimates the impact of climate
change on average yield by up to 9% (see
table S9).
Although the extent of over-estimation of
the net impact of climate variables on corn
yield is relatively moderate, this moderate
over-estimation masks two dimensions of
heterogeneity in estimation biases: geograph-
ical heterogeneity and climatic heterogeneity.
Maps in figure 1 depict county-level overes-
timation of the effect of climate change on
corn yield and shows that the magnitude of
this overestimation varies across regions. For
some northern counties, the overestimation
can be as high as 20% or above. In some
Southeastern states, omitting price variables
may even underestimate the effect of climate
change on yield. Geographical patterns of
the overestimation differ across results that
are based on different climate models. Using
data from climate model HadGEM2-ES
we find that the effect of climate change on
yield will be overestimated in the Midwest
when price variables are omitted (see the
four maps in the upper panel of figure 1).
However, using data from climate model
NorESM1-M leads to the opposite result
for some Midwestern counties (see the four
maps in the lower panel of figure 1). This is
because climate models differ in their predic-
tions of the magnitude for temperature and
of precipitation changes in a region. Similar
maps for soybeans are presented in figure S1
in SI.
We decompose the impact of climate vari-
ables on corn yield in table 8 using parameter
estimates from yield models II and III in
table 3 with various climate change scenarios
of HadGEM2-ES.14 We find that changes
14
Corresponding results under climate model NorESM1-M
are presented in table S10 of SI.
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate 205

Table 6. Mean Values of Predicted Changes in Climatic Variables by HadGEM2-ES and

NorESM1-M Models under Different Emission Scenarios for Rain-fed Counties Included in
this Study
2041–2060 Average 2061–2080 Average
HadGEM2-ES NorESM1-M HadGEM2-ES NorESM1-M
RCP2.6

Growing degree days (8–32◦ C) 476.1 (51.8) 319.2 (42.0) 448.0 (48.4) 290.9 (46.3)
Overheat degree days (>32◦ C) 101.0 (95.7) 36.9 (50.1) 87.7 (90.2) 30.7 (37.2)
Average temperature (◦ C) 2.6 (0.3) 1.8 (0.2) 2.5 (0.3) 1.6 (0.3)
Average precipitation in March (mm) 11.4 (11.4) 20.8 (14.4) 11.5 (11.0) 18.9 (11.7)
Average precipitation in April (mm) 10.4 (10.4) 0.6 (8.2) 13.6 (10.7) 9.4 (10.9)
Average precipitation in May (mm) 0.9 (7.7) 11.3 (13.6) 4.9 (9.8) 0.7 (16.4)
Average precipitation in June (mm) −5.6 (14.4) 2.7 (9.7) 2.0 (9.4) 3.9 (11.6)
Average precipitation in July (mm) −2.6 (12.4) 3.2 (14.9) −3.5 (11.4) 9.2 (13.4)
Average precipitation in August (mm) −6.8 (17.7) 3.9 (15.9) −5.8 (16.1) −1.8 (12.6)
Temperature deviation in March (◦ C) −0.16 (0.38) 0.00 (0.42) −0.09 (0.43) −0.21 (0.40)
Temperature deviation in April (◦ C) 0.06 (0.42) 0.52 (0.43) −0.21 (0.41) 0.08 (0.41)
Temperature deviation in May (◦ C) 0.18 (0.44) 0.42 (0.49) 0.09 (0.50) 0.61 (0.48)
Temperature deviation in June (◦ C) 0.75 (0.57) 0.58 (0.44) 0.54 (0.53) 0.65 (0.57)
Temperature deviation in July (◦ C) 0.83 (0.61) 0.86 (0.49) 0.84 (0.64) 0.59 (0.52)
Temperature deviation in August (◦ C) 0.85 (0.62) 1.00 (0.69) 0.89 (0.69) 1.14 (0.60)

RCP8.5

Growing degree days (8–32◦ C) 736.2 (77.8) 555.5 (91.4) 1077.7 (109.7) 772.1 (84.0)
Overheat degree days (>32◦ C) 265.9 (166.9) 129.2 (94.3) 511.8 (228.1) 247.2 (129.6)
Average temperature (◦ C) 4.1 (0.4) 3.1 (0.5) 6.1 (0.6) 4.3 (0.5)
Average precipitation in March (mm) 14.0 (12.2) 31.1 (17.6) 20.7 (13.2) 19.9 (14.1)
Average precipitation in April (mm) 12.4 (10.9) 7.9 (6.7) 19.6 (13.4) 9.1 (9.4)
Average precipitation in May (mm) −1.2 (10.6) 7.0 (15.1) 1.2 (10.9) 17.0 (12.3)
Average precipitation in June (mm) −9.6 (11.6) −2.0 (13.0) −15.1 (11.9) −0.6 (12.1)
Average precipitation in July (mm) −20.1 (11.1) 5.9 (15.0) −27.4 (12.9) 1.1 (12.0)
Average precipitation in August (mm) −19.1 (17.3) −0.9 (20.1) −27.1 (15.6) −2.1 (22.2)
Temperature deviation in March (◦ C) −0.20 (0.41) −0.13 (0.53) −0.20 (0.42) 0.47 (0.54)
Temperature deviation in April (◦ C) −0.22 (0.39) 0.46 (0.45) −0.43 (0.41) 0.56 (0.46)
Temperature deviation in May (◦ C) 0.11 (0.42) 0.64 (0.48) 0.24 (0.52) 0.66 (0.51)
Temperature deviation in June (◦ C) 0.97 (0.62) 1.09 (0.50) 1.37 (0.69) 1.29 (0.51)
Temperature deviation in July (◦ C) 1.52 (0.80) 1.13 (0.63) 1.80 (0.88) 1.40 (0.57)
Temperature deviation in August (◦ C) 1.38 (0.77) 1.46 (0.94) 1.53 (0.84) 1.96 (0.96)

Note: Numbers in parentheses are standard deviations.

in GDD causes the largest yield reduction.
Specifically, increase in GDD (ranging from
476 to 1078 under various RCP scenarios
and temporal framework) caused by climate
change will decrease corn yield by 7% to
24% when compared with sample mean of
corn yield, depending on the climate change
scenarios and temporal framework. Change
in overheat degree days causes the second
largest reduction in corn yield. Change in
precipitation tends to cause the least yield
reduction among the four climatic factors
included in the yield models. Comparing the
results obtained using parameter estimates
from model II and III we find that omitting
price variables will underestimate the effect
of GDD by about 10% while overestimat-
ing the effects of overheat degree days on
yield by 41.3%, of precipitation by up to
9.6%, and of temperature deviation by up
to 18.1%.
The last two columns of table 7 show the
effect of climate change on corn production
under alternative yield models and climate
scenarios. The effects on corn production
are obtained by simply summing up the
effects of climate change on yield and on
acreage. Since the negative effect of climate
change on corn acreage is small (about 1%),
the major channel through which climate

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 206
January 2016
Table 7. Effect of Climate Change on Corn Yield, Acreage, and Production under Climate Model HadGEM2-ES
Decrease in yield caused by Overestimation of Decrease in corn production caused
climate change under Percentage Change under Decrease in corn acreage by climate change under
alternative yield models and Model II compared with under a climate change alternative yield models and
climate scenarios (%) Models III (%) scenario (%) climate scenarios (%)
Yield Model II Yield Model III
and Acreage and Acreage
Climate scenarios Yield Model II Yield Model III Acreage Model II Model II Model II
RCP2.6 Medium term 14.4 13.8 5.6 0.6 15.0 14.3
(7.8) (8.0) (5.3) (0.5) (7.9) (8.1)
Long term 12.6 12.0 5.9 0.7 13.3 12.7
(7.9) (8.2) (6.5) (0.5) (8.0) (8.3)
RCP4.5 Medium term 20.6 19.5 6.3 1.1 21.6 20.6
(11.3) (11.6) (4.6) (0.6) (11.3) (11.6)
Long term 24.8 23.4 6.6 0.6 25.4 24.0
(11.0) (11.6) (4.7) (0.5) (11.0) (11.6)
RCP6.0 Medium term 11.6 11.3 3.0 0.6 12.1 11.9
(9.8) (10.1) (5.3) (0.5) (10.0) (10.3)
Long term 23.6 22.5 6.0 0.6 24.3 23.1
(13.0) (13.5) (4.3) (0.5) (13.1) (13.6)
RCP8.5 Medium term 26.7 25.1 7.2 0.8 27.5 26.0
(13.4) (13.9) (4.6) (0.5) (13.3) (13.9)
Long term 41.7 40.1 4.7 1.3 43.0 41.4
(17.5) (18.4) (3.6) (0.6) (17.3) (18.2)

Amer. J. Agr. Econ.

Notes: Here corn yield models II and III are as specified in table 3. Acreage model II is as specified in table 5. Medium term and long term refer to average climate conditions over 2041–2060 and in 2061–2080, respectively.
Numbers in parentheses are standard deviations.

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate 207

< –10
–10 - –5
–5 - 0
0-5
5 - 10
10 - 15
15 - 20
> 20

Figure 1. Estimation biases of the effect of climate change on corn yield across eastern US
counties under climate models HadGEM2-ES and NorESM1-M when price variables are
omitted (%)
Note: Positive values indicate overestimation whereas negative values indicate underestimation.

change reduces corn production is by nega- III (see table 3) to the corn acreage elasticity
tively impacting corn yield. Everything else 0.45 from corn acreage model II based on
equal, climate change can reduce corn pro- received prices (table 5). This implies that
duction by 7%–41%, depending on RCPs 33.8% of the increase in corn production
and temporal term considered (see the last caused by a corn price increase will be due
columns in table 7 and S9). Omitting price to the increase in corn yield per acre and the
variables also results in overestimation of the remaining 66.2% of the increase will be due
effect of climate change on corn production to the increase in corn acreage. Therefore,
by 2%–9%. Figures S2 and S3 in SI depicts any assessment of the land-use effect of corn
the impact of climate change on corn and ethanol should consider the responsiveness of
soybean production at the county-level across corn yield to prices. Previous studies, such as
the rainfed United States, from which we Searchinger et al. (2008) and Fargione et al.
can see that larger impacts (measured in per- (2008), which ignored the yield elasticity with
centage of mean yield in a county) occur in respect to corn prices when estimating the
Southeastern states. land use effects of corn ethanol production,
have significantly overestimated the indi-
Land-Use Effects of Corn Price Increase rect land use change caused by corn ethanol
production. As is pointed out in Dumortier
Our estimation results indicate that corn et al. (2011), inclusion of the response of
supply elasticity with respect to corn price yield to prices can significantly mitigate or
is 0.68; this is obtained by adding the corn even reverse the findings in Searchinger et al.
yield elasticity 0.23 from corn yield model (2008).

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 208 January 2016 Amer. J. Agr. Econ.

Table 8. Decomposition of the Effect of Climate Variables on Corn Yield under Various
Climate Change Scenarios and Model Specifications under Climate Model HadGEM2-ES
Change in corn yield caused
by climate change under Overestimation of
alternative yield models Change Percentage
and climate scenarios (%) under Model I
Climate Temporal compared
Scenario Range Effects of Model I Model II with Models II (%)
RCP2.6 Medium term growing degree days −7.3 −8.0 −8.1
overheat degree days −3.1 −2.2 41.3
temperature deviation −2.3 −2.1 13.6
precipitation −1.7 −1.5 8.3
RCP2.6 Long term growing degree days −6.6 −7.2 −8.0
overheat degree days −2.3 −1.7 41.3
temperature deviation −2.5 −2.1 18.1
precipitation −1.1 −1.1 8.4
RCP4.5 Medium term growing degree days −9.1 −10.0 −8.7
overheat degree days −4.8 −3.4 41.3
temperature deviation −3.6 −3.3 10.3
precipitation −3.0 −2.8 6.8
RCP4.5 Long term growing degree days −12.3 −13.6 −9.2
overheat degree days −6.5 −4.6 41.3
temperature deviation −3.4 −2.9 17.7
precipitation −2.5 −2.3 8.6
RCP6.0 Medium term growing degree days −6.5 −7.1 −8.1
overheat degree days −2.1 −1.5 41.3
temperature deviation −2.2 −2.0 10.8
precipitation −0.7 −0.7 −1.1
RCP6.0 Long term growing degree days −12.2 −13.4 −9.2
overheat degree days −6.3 −4.5 41.3
temperature deviation −3.0 −2.6 13.0
precipitation −2.2 −2.0 8.8
RCP8.5 Medium term growing degree days −12.6 −13.9 −9.3
overheat degree days −6.9 −4.9 41.3
temperature deviation −3.7 −3.2 16.5
precipitation −3.5 −3.2 9.6
RCP8.5 Long term growing degree days −21.7 −24.2 −10.3
overheat degree days −10.8 −7.7 41.3
temperature deviation −4.6 −4.0 14.4
precipitation −4.6 −4.2 8.5

Conclusions but also by crop management practices

This article investigates the determinants
of average crop yield and crop acreage
at a county-level scale for corn and soy-
beans in the rainfed United States over
the period 1977–2007. It also analyzes the
effects of climate and prices on corn and
soybean supply. The reduced form empiri-
cal framework developed here is based on
the premise that county average crop yield
is influenced not only by climate factors
and land allocation decisions of farmers in
response to input prices and expected out-
put prices. It examines the responsiveness
of crop yields to both received prices and
futures prices. Additionally, as compared to
previous studies, this study includes a more
comprehensive and temporally disaggregated
set of climate variables and a more sophisti-
cated IV panel data approach that controls
for both endogeneity of prices and spatial
autocorrelations.

Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412

by guest
on 30 August 2018
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate
We find statistically significant elastic-
ity of corn yield with respect to corn price
with a value of 0.23. In the case of soybeans,
however, the yield price elasticity is not sta-
tistically significant. Our analysis implies that
the supply response to an increase in corn
price occurs at least partly due to changes
at the intensive margin that increase yield
per acre. Like previous studies, we also find
that climate change has significant adverse
impacts on the yields of both corn and soy-
beans. More importantly, however, we find
that the timing of the variability in temper-
ature matters. For instance, the effect of
temperature deviations on crop yields differs
across the months in which it occurs.
Using regionally disaggregated pro-
jections of climate change over medium
term (average for 2041–2060) and long
term (average for 2061–2080) based on
the most updated climate projections, as
of 2014, that are used in IPCC’s AR5, we
find that climate change can be expected
to decrease corn yield on average by up
to 40.1%, depending on the climate sce-
narios, time horizon, and global climate
models used. The omission of price vari-
ables while estimating the effect of climate
variables on corn yield models leads to an
assessment of the adverse impact of climate
change that is up to 9% higher depending
on climate and temporal scenarios consid-
ered, and on global climate models used to
predict climate change. The extent of this
estimated bias differs across geographical
regions, global climate models, and cli-
mate variables, indicating the importance of
controlling for price variables when quanti-
fying the impact of climate change on crop
production.
In the article, we focus on the aggregate
yield and acreage effects of price and climate.
We leave it to future research to decompose
these aggregate yield and acreage effects and
identify the causal effects of specific compo-
nents to determine the extent to which the
effects of price and climate variables on yield
and acreage were due to input use changes,
land allocation among crops, and extensive
margin changes. This would enable quan-
tification of the mechanisms by which price
and climate affect crop yield and acreage.
Nevertheless, our analysis does show the
importance of including the potential for
behavioral response to market conditions in
estimating the impact of climate change on
crop production.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
209
Supplementary Material
Supplementary material is available at
http://oxfordjournals.org/our_journals/ajae/
online.
References
Barr, K. J., B. A. Babcock, M. A. Carriquiry,
A. M. Nassar, and L. Harfuch. 2011.
Agricultural Land Elasticities in the
United States and Brazil. Applied Eco-
nomic Perspectives and Policy 33 (3):
449–62.
Baylis, K., N. Paulson, and G. Piras. 2011.
Spatial Approaches to Panel Data in
Agricultural Economics. Journal of Agri-
culture and Applied Economics 43 (3):
325–38.
Bhaskar, A., and J. Beghin. 2009. How Cou-
pled are Decoupled Farm Payments?
a Review of the Evidence. Journal of
Agricultural and Resource Economics 34
(1): 130–53.
Chandrasekaran, B., K. Annadurai, and
E. Somasundaram. 2010. A Textbook
of Agronomy. New Delhi: New Age
International.
Chavas, J. P., and M. Holt. 1990. Acreage
Decisions Under Risk: The Case of Corn
and Soybeans. American Journal of
Agricultural Economics 72 (3): 27–33.
Chavas, J. P., R. D. Pope, and R. S. Kao.
1983. An Analysis of the Role of Futures
Prices, Cash Prices and Government
Programs in Acreage Response. Western
Journal of Agricultural Economics 8 (1):
27–33.
Chembezi, D. M., and A. W. Womack. 1992.
Regional Acreage Response for US
Corn and Wheat: The Effects of Gov-
ernment Programs. Southern Journal of
Agricultural Economics 24: 187–98.
Choi, J.-S., and P. G. Helmberger. 1993. How
Sensitive are Crop Yields to Price Chan-
ges and Farm Programs? Journal of
Agricultural and Applied Economics 25:
237–44.
Darwin, R. 1999. A Farmer’s View of the
Ricardian Approach to Measuring Agri-
cultural Effects of Climate Change.
Climatic Change 41: 371–411.
Deschênes, O., and M. Greenstone. 2007.
The Economic Impacts of Climate
Change: Evidence from Agricultural
Output and Random Fluctuations in
 210 January 2016
Weather. American Economic Review 97
(1): 354–85.
Dumortier, J., D. J. Hayes, M. Carriquiry, F.
Dong, X. Du, A. Elobeid, J. F. Fabiosa,
and S. Tokgoz. 2011. Sensitivity of Car-
bon Emission Estimates from Indirect
Land-Use Change. Applied Economic
Perspectives and Policy 33 (3): 428–48.
Elhorst, P. J. 2010. “Spatial Panel Data Mod-
els.” In Handbook of Applied Spatial
Analysis: Software Tools, Methods and
Applications, ed. M. M. Fischer and A.
Getis, 377–407. Berlin: Springer.
Fargione, J., J. Hill, D. Tilman, S. Polasky,
and P. Hawthorne. 2008. Land Clearing
and the Biofuel Carbon Debt. Science
319 (5867): 1235–38.
Feng, H., and B. A. Babcock. 2010. Impacts
of Ethanol on Planted Acreage in Mar-
ket Equilibrium. American Journal of
Agricultural Economics 92 (3): 789–802.
Gardner, B. L. 1976. Futures Prices in
Supply Analysis. American Journal of
Agricultural Economics 58 (1): 81–84.
Gohin, A. 2014. Assessing the Land Use
Changes and Greenhouse Gas Emissions
of Biofuels: Elucidating the Crop Yield
Effects. Land Economics 90 (4): 575–86.
Gouel, C. 2010. Agricultural Price Instability:
A Survey of Competing Explanations
and Remedies. Journal of Economic
Surveys 26 (1): 129–56.
Greene, W. H. 2003. Econometric Analysis.
5th ed. New York: Pearson.
Houck, J. P., and P. W. Gallagher. 1976.
The Price Responsiveness of U.S. Corn
Yields. American Journal of Agricultural
Economics 58 (4): 731–34.
Irwin, S. H., and C. S. Thraen. 1994. Ratio-
nal Expectations in Agriculture? a
Review of the Issues and the Evidence.
Review of Agricultural Economics 16 (1):
133–58.
Kaufmann, R. K., and S. E. Snell. 1997.
A Biophysical Model of Corn Yield:
Integrating Climatic and Social Determi-
nants. American Journal of Agricultural
Economics 79 (1): 178–90.
Khanna, M., and C. L. Crago. 2012. Mea-
suring Indirect Land Use Change with
Biofuels: Implications for Policy. Annual
Review of Resource Economics 4: 161–26.
Kleibergen, F., and R. Paap. 2006. Gener-
alized Reduced Rank Tests Using the
Singular Value Decomposition. Journal
of Econometrics 133 (1): 97–126.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
Amer. J. Agr. Econ.
Kucharik, C. J., and S. P. Serbin. 2008.
Impacts of Recent Climate Change
on Wisconsin Corn and Soybean Yield
Trends. Environmental Research Letters
3 (3): 034003.
Lee, D. R., and P. G. Helmberger. 1985.
Estimating Supply Response in the
Presence of Farm Programs. American
Journal of Agricultural Economics 67:
193–203.
LeSage, J. P. 1997. Regression Analysis of
Spatial Data. Journal of Regional Anal-
ysis and Policy 27 (2): 83–94.
Lin, W., and R. Dismukes. 2007. Supply
Response Under Risk: Implications for
Counter-Cyclical Payments’ Production
Impact. Review of Agricultural Econo-
mics 29 (1): 64–86.
Long, S. P., E. A. Ainsworth, A. D. B.
Leakey, J. Nosberger, and D. R. Ort.
2006. Food for Thought: Lower-than-
Expected Crop Yield Simulation with
Rising CO2 Concentrations. Science 312:
1918–21.
Lyons, D. C., and R. L. Thompson. 1981. The
Effect of Distortions in Relative Prices
on Corn Productivity and Exports: A
Cross-Country Study. Journal of Rural
Development 4: 83–102.
McCarl, B. A., X. Villavicencio, and X. Wu.
2008. Climate Change and Future Anal-
ysis: Is Stationary Dying? American
Journal of Agricultural Economics 90 (5):
1241–47.
Mendelsohn, R., W. D. Nordhaus, and
D. Shaw. 1994. The Impact of Global
Warming on Agriculture: A Ricardian
Analysis. American Economic Review 84
(4): 753–71.
Menz, K.M., and P. Pardey. 1983. Technology
and U.S. Corn Yields: Plateaus and Price
Responsiveness. American Journal of
Agricultural Economics 65: 558–62.
Miller, D. J., and A. J. Plantinga. 1999. Mod-
eling Land Use Decisions with Aggregate
Data. American Journal of Agricultural
Economics 81 (1): 180–94.
National Agricultural Statistics Service
(NASS). 2010. Field Crops Usual
Planting and Harvesting Dates. Agri-
cultural Handbook Number 628. U.S.
Department of Agriculture, October.
Orazem, P., and J. Miranowski. 1986. An
Indirect Test for the Specification
of Expectation Regimes. Review of
Economics and Statistics 68 (4): 603–9.
 Miao, Khanna, and Huang Responsiveness of Crop Yield and Acreage to Prices and Climate
———. 1994. A Dynamic Model of Acreage
Allocation with General and Crop-
Specific Soil Capital. American Journal
of Agricultural Economics 76: 385–95.
PRISM Climate Group. 2009. Monthly
Data, http://www.prism.oregonstate.edu/
(accessed 9 May 2015).
Ray, D. E. 1996. “The FAIR Act: What Does
It All Mean?” In Increasing Understand-
ing of Public Problems and Policies, ed.
S. A. Halbrook and K. W. Ward, 75–94.
Oak Brook, IL: Farm Foundation.
Roberts, M., and W. Schlenker. 2013. Identi-
fying Supply and Demand Elasticities of
Agricultural Commodities: Implications
for the US Ethanol Mandate. American
Economic Review 103 (6): 2265–95.
Sacks, W. J., D. Deryng, J. A. Foley, and
N. Ramankutty. 2010. Crop Planting
Dates: An Analysis of Global Patterns.
Global Ecology and Biogeography 19:
607–20.
Schlenker, W., and M. J. Roberts. 2009.
Nonlinear Temperature Effects Indicate
Severe Damages to U.S. Crop Yields
Under Climate Change. Proceedings of
the National Academy of Sciences 106
(37): 15594–98.
Schlenker, W., W. M. Hanemann, and A. C.
Fisher. 2005. Will U.S. Agriculture
Really Benefit from Global Warming?
Accounting for Irrigation in the Hedonic
Approach. American Economic Review
95 (1): 395–406.
———. 2006. The Impact of Global Warming
on U.S. Agriculture: An Econometric
Analysis of Optimal Growing Condi-
tions. Review of Economics and Statistics
88 (1): 113–25.
Searchinger, T., R. Heimlich, R. A.
Houghton, F. Dong, A. Elobeid, J.
Fabiosa, S. Tokgoz, D. Hayes, and T.-
H. Yu. 2008. Use of U.S. Croplands for
Biofuels Increases Greenhouse Gases
Through Emissions from Land-Use
Change. Science 319 (5867): 1238–40.
Serra, T., B. K. Goodwin, and A. Feath-
erstone. 2011. Risk Behavior in the
Presence of Government Programs.
Journal of Econometrics 162 (1): 18–24.
Downloaded from https://academic.oup.com/ajae/article-abstract/98/1/191/2195412
by guest
on 30 August 2018
211
Serra, T., D. Zilberman, and J. M. Gil. 2008.
Farms’ Technical Inefficiencies in the
Presence of Government Programs.
Australian Journal of Agricultural and
Resource Economics 52 (1): 57–76.
Serra, T., D. Zilberman, B. K. Goodwin, and
A. Featherstone. 2006. Effects of Decou-
pling on the Mean and Variability of
Output. European Review of Agricultural
Economics 33 (3): 269–88.
Shideed, K. H., and F. C. White. 1989. Alter-
native Forms of Price Expectations
in Supply Analysis for U.S. Corn and
Soybean Acreages. Western Journal of
Agricultural Economics 14 (2): 281–92.
Specht, J. E., B. W. Diers, R. L. Nelson, J.
Francisco, F. de Toledo, J. A. Torrion,
and P. Grassini. 2014. “Soybean.” In
Yield Gains in Major U.S. Field Crops,
ed. S. Smith, B. Diers, J. Specht, and B.
Carver, 311–56. Madison, WI: American
Society of Agronomy.
Tegene, A., W. E. Huffman, and J. Mira-
nowski. 1988. Dynamic Corn Supply
Functions: A Model with Explicit
Optimization. American Journal of
Agricultural Economics 70 (1): 103–11.
US Census Bureau. 2009. Population
Estimates, http://www.census.gov/popest/
archives/ (accessed 9 May 2015).
Warszawski, L., K. Frieler, V. Huber, F.
Piontek, O. Serdeczny, and J. Schewe.
2014. The Inter-Sectoral Impact Model
Intercomparison Project (ISI-MIP):
Project Framework. Proceedings of
the National Academy of Sciences of
the United States of America 111 (9):
3228–32.
Weber, J. G., and N. Key. 2012. How much
Do Decoupled Payments Affect Pro-
duction? an Instrumental Variable
Approach with Panel Data. American
Journal of Agricultural Economics 94 (1):
52–66.
Weinraub, M. 2013. “U.S. Soybean Acreage
to Rise Due to Planting Delays.”
Reuters May 31. http://www.reuters.com/
article/2013/05/31/us-usa-crops-planting-
idUSBRE94U11S20130531 (accessed 9
May 2015).