Q UANTIFYING THE Y IELD S ENSITIVITY

OF M ODERN R ICE V ARIETIES TO W ARMING

T EMPERATURES : E VIDENCE FROM
THE P HILIPPINES

R UIXUE W ANG , R ODERICK M. R EJESUS , J ESSE B. T ACK ,

J OSEPH V. B ALAGTAS , AND A NDY D. N ELSON

This study examines the relationship between yields of modern rice varieties and warming temperatures.
Data from a long-running farm-level survey in the Philippines, with rich information on planted rice vari-
eties, allow us to estimate fixed effect econometric models of rice yields. We find that increases in temper-
ature, especially minimum temperatures, have statistically significant negative impacts on rice yields.
Point estimates of the marginal effect of higher temperatures on rice yields indicate that early modern
varieties bred primarily for higher yields, pest resistance, and/or grain quality traits (i.e., not necessarily
abiotic stress tolerance) tend to be more resilient to heat events than traditional rice varieties. Moreover,
the marginal effect point estimates also suggest that more recent rice varieties bred for better tolerance to
abiotic stresses are likely more resilient to warming than both traditional varieties and early modern vari-
eties. Notwithstanding the heat resilience pattern suggested by these point estimates, we are unable to find
statistically significant differences in the marginal yield response to warming across these three rice vari-
etal groups. These results provide suggestive evidence that rice breeding efforts have improved resilience
to warming temperatures and point to several interesting future research directions.

Key words: Central Luzon, climate change, rice yield, rice varieties.

JEL codes: Q12, Q16, Q18.

Rice is the most important food crop in the Asian countries, and it is also becoming an
world, with nearly half of the world’s popula- increasingly important food crop in Africa
tion relying on it for sustenance every day. It and in Latin America (Nigatu et al. 2017;
is the main staple food across a number of USDA-ERS 2020). Over 144 million farms
cultivate rice across an area of about 167 mil-
lion hectares (ha) in more than 100 countries
Ruixue Wang is a former PhD student, Department of Agricul- (FAOSTAT 2019). Rice-based farming sys-
tural and Resource Economics, North Carolina State University.
Roderick M. Rejesus is a professor and extension specialist, tems have also been the main source of income
Department of Agricultural and Resource Economics, North Car- for a large proportion of rural farmers located
olina State University. Jesse B. Tack is an associate professor,
Department of Agricultural Economics, Kansas State University. in a number of developing countries (Fan
Joseph V. Balagtas is an associate professor, Department of Agri- et al. 2005).
cultural Economics, Purdue University. Andy D. Nelson is a pro-
fessor, Faculty of Geo-Information Science and Earth
Given the importance of rice as a major
Observation, University of Twente. food staple and a source of income for farmers
We would like to thank the editor, Terry Hurley, and two anony- worldwide, a key challenge is to find strategies
mous referees for their helpful comments and suggestions. We
would also like to acknowledge Fe Gascon, Kei Kajisa, Alice that would maintain or improve rice produc-
Laborte, Jose Yorobe Jr., and Jauhar Ali for their assistance on tivity in the presence of climate change. Based
the data and interpretation of results. The work of Rejesus was sup-
ported in part by the USDA NIFA Hatch Project No. NC02696. on the recent climate assessment reports of the
Correspondence may be sent to: [email protected] Intergovernmental Panel on Climate Change

Amer. J. Agr. Econ. 00(00): 1–22; doi:10.1111/ajae.12210

© 2021 The Authors. American Journal of Agricultural Economics published by Wiley Periodicals LLC. on behalf of
Agricultural & Applied Economics Association.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License,
which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial
and no modifications or adaptations are made.
2 Month 2021 Amer. J. Agr. Econ.

(IPCC), global warming has intensified over the aimed at developing climate-tolerant traits,
last fifty years, and this warming trend is pre- have led to modern varieties that are more
dicted to persist in the future (see figure S1 in resilient to warming in farmer fields. To
the online supplementary appendix). A warm- achieve this objective, we utilize farm-level
ing climate has the potential to adversely affect survey data collected every four to five years
rice yields and rice quality (Peng et al. 2004; from 1966 to 2016 in the Central Luzon region
Iizumi et al. 2006; Lyman et al. 2013; Kawasaki of the Philippines (Moya et al. 2015; Laborte
and Uchida 2016)1. For example, extremely et al. 2015). Examining the Philippine case is
high temperatures can lead to spikelet sterility especially relevant because it is one of the
and reduce rice yields (Wassmann et al. 2009; top ten rice-producing countries in the world
Nguyen et al. 2014; Bheemanahalli et al. 2016). (FAOSTAT 2019), and the evolution of major
These adverse warming effects then have the varietal group releases in this country is repre-
potential to compromise food security in coun- sentative of other major rice-producing coun-
tries that rely on it as a food staple or a source tries like India, Indonesia, Bangladesh, and
of income. Vietnam (Brennan and Malabayabas 2011;
One strategy that may help address the cli- Pandey et al. 2012). Because farmers are
mate change challenge in rice production is tracked over time in the data set utilized, we
the development and use of rice varieties that are able to develop fixed effects econometric
are better able to adapt to a progressively models, which then allow us to identify “varie-
warming climate. Over the years, develop- tal-group-specific” yield response to several
ment and adoption of new rice varieties have weather variables (e.g., minimum tempera-
been utilized to overcome a variety of produc- ture, maximum temperature, and precipita-
tion challenges that have historically arisen in tion).2 Therefore, the study results provide
this sector. Since the Green Revolution in the insights on the effectiveness of prior breeding
1960s, there have been development and con- investments and rice varietal development
sequent adoption of several generations of efforts, specifically in terms of mitigating
modern rice varieties (MVs) aimed at addres- adverse impacts of climate change.
sing various production challenges such as Due to concerns about the effect of climate
lodging, low fertilizer responsiveness, pest change on agriculture, there is now a large lit-
problems, and adverse weather conditions erature that uses econometric methods to
(see next section for more details). The release examine how weather variables influence crop
and subsequent adoption of these MVs have yield outcomes (see, for example, Auffham-
led to remarkable increases in rice yields over mer, Ramanathan, and Vincent 2006; Welch
time (Barker, Herdt, and Rose 1985; Hayami et al. 2010; Sarker, Alam, and Gow 2012;
and Otsuka 1994; Otsuka, Gascon, and Lyman et al. 2013; and Kawasaki and
Asano 1994; Estudillo and Otsuka 2006), espe- Uchida 2016 for rice; Schlenker and Rob-
cially as compared to the traditional rice varie- erts 2009 for corn; Tack, Barkley, and Nal-
ties (TVs) available prior to the Green ley 2015 for wheat). There is also another
Revolution. strand of literature that explores the determi-
With this history of rice varietal develop- nants and economic impacts of particular cli-
ment over time, it is important to examine mate change adaptation practices for
whether there is heterogeneity in each vari- different crops (see Chen, Wang, and
ety’s (or varietal group’s) yield response to Huang 2014; Wang et al. 2010; Deressa
weather variables. The objective of this study et al. 2009; Di Falco, Veronesi, and Yesuf 2011;
is to determine the yield response of different Butler and Huybers 2013; Huang, Wang, and
rice varietal groups to warming temperatures. Wang 2015). Despite this rich literature on cli-
Findings from this study have important impli- mate change adaptation and climate change
cations with regards to whether past rice effects on yields, to the best of our knowledge,
breeding investments, especially recent efforts there are a limited number of studies that
investigate how the yield impact of weather
1
The effects of high temperatures on rice quality include
increased broken (or cracked) grain percentage and chalkiness.
2
The quality effects of warming can substantially influence the As noted in Launio et al. (2008) and Laborte et al. (2015) there
eventual revenues received by rice farmers (as mentioned in the are numerous specifically-named MVs that have been released in
studies above). However, quality information is not available for the Philippines since 1966, and it would have been impossible to
the farm level data utilized in this study. As such, estimating the estimate yield response for each of these specifically-named rice
effects of high temperatures on rice quality (based on farm-level varieties. Hence, in this study, we focus on the yield response of
data) is left for future research. varietal groups (as further defined below) to weather variables.
Wang et al. Modern Rice Varieties and Warming Temperatures 3

variables may vary depending on the rice vari- farm-level data set we have also has rich infor-
ety, or the rice varietal group, used by farmers. mation on the rice varieties used, as well as the
Tack et al. (2016), using a long time-series of other inputs utilized by the grower
field trial data in the U.S., examined variety- (e.g., fertilizer, insecticide). Much of the indi-
specific yield response to higher temperatures vidual data sets used for climate-change stud-
for wheat but not for rice. Hasan, Sarker, and ies in the past do not have rich varietal
Gow (2016) examined how the yield response information that would allow one to estimate
of TVs differ from high yielding rice varieties variety-specific (or varietal-group-specific)
(HYVs), using more aggregate region-specific yield response to weather variables. Disre-
data from Bangladesh. We have not found any garding heterogeneity in the yield response
study that utilizes individual farm-level data to of specific rice varieties may lead to inaccurate
econometrically examine the relationship inferences regarding the yield effects of warm-
between rice varietal use and yield response ing. Hence, having this unique and novel data
to weather variables. set gives us the rare opportunity to study the
Our main contribution is to disentangle the interactions of rice varietal traits and the envi-
impact of warming on rice yields by allowing ronment it grows in, over a long period of time.
for and econometrically identifying varietal- The rest of the paper is organized as follows.
group-specific effects. This is important Section 2 discusses the empirical setting, evo-
because it will allow us to know which rice lution of rice varieties in the Philippines, and
varietal group is most effective in attenuating data sources. The modeling framework is
the adverse effects of warming temperatures described in Section 3. Section 4 presents the
and whether the older MVs had some climate estimation results. Section 5 provides various
change adaptation features (Wassmann robustness checks, and Section 6 discusses
et al. (2009)). Although not all previously the conclusions.
released rice MVs are widely used anymore
(Laborte et al. 2015), it is still important to
determine whether these older varietal groups
have historically been effective climate change Empirical Setting and Data Sources
adaptation tools, especially because they were
not specifically bred for this purpose (see more The empirical setting for this study covers six
discussion on this issue below). If climate major rice-producing provinces from two
change attenuation effects are present for administrative regions in the Philippines: (a)
these earlier MVs, then these are important La Union and Pangasinan provinces in Region
“spillover” rice breeding effects that need to I (called the Ilocos region), and (b) Nueva
be recognized. But more importantly, given Ecija, Pampanga, Bulacan, and Tarlac prov-
that newer rice varieties were developed to inces in Region III (usually called the Central
be more tolerant to adverse climatic condi- Luzon region). For the purpose of this study
tions, providing empirical evidence to show (and consistent with Laborte et al. 2015), the
the climate change attenuation effects of these six provinces in the study area are collectively
newer varieties on farmers’ fields allows one to referred to here as Central Luzon. In 2013, the
see whether there has been “on-the-ground” total harvested area in the six provinces was
progress from breeding efforts to produce 0.9 million ha, with the majority (82%) of the
climate-change-tolerant varieties. area under irrigation (i.e., which is slightly
The second contribution is that we exploit higher than the ~70% of rice area irrigated
actual farm-level panel data in our analysis nationally). The study area is considered as
rather than using more aggregate rice produc- one of the major rice producing regions in
tion data (e.g., district level, province level) or the Philippines, where average rice yield was
experimental field trial data, which are the two 4.7 tons per ha, per cropping season in 2013,
most commonly used data types in previous lit- which is slightly higher than the national aver-
erature. The novel data set used in this study age. The average farm size in the study area is
allows one to better examine rice yield around 1 ha (Moya et al. 2015) and is consis-
response under actual farmer-managed field tent with the national average. The sociode-
conditions. The data set used is also unique mographic profile of farmers in the study
in terms of the decades-long time period it area is also roughly in line with the national
spans, which is relatively rare in terms of the average (i.e., age in the mid-50s, with about
few climate-change studies that utilize individ- nine years of education). Rice is planted twice
ual farm-level data sets. Furthermore, the a year in the study area: (a) the wet season
4 Month 2021 Amer. J. Agr. Econ.

(WS) production that ranges from May/June released after 1995. In this period, public rice
to September/October, and (b) the dry season breeding programs started to focus on the
(DS) production that ranges from November/ research and development of varieties specifi-
December to March/April (Moya et al. cally for adverse rice production environ-
2015).3 Like many other countries of the ments, such as those subject to salinity,
world, the Philippines (and the study area floods, and drought (Laborte et al. 2015).4
under consideration) have experienced signif- The main data source utilized for this study
icant warming trends over the years. Estimates is from the so-called “ Central Luzon Loop
from the Philippine Atmospheric, Geophysi- Survey” or simply the “ Loop Survey.” It is
cal and Astronomical Services Administration called the Loop Survey because of the sam-
(PAGASA) suggest that, between 1951 to pling strategy used, where the farm house-
2010, average maximum and minimum tem- holds included in the sample are located
peratures in the Philippines have increased along the loop of the main highway that passes
by 0.36
C and 1.0
C, respectively. through the six provinces (figure 1). Face-to-
As previously mentioned, the evolution of face interviews were conducted to collect vari-
Philippine rice varietal group development ous socio-demographic, input use, and rice
roughly follows the pattern for other major production information from the sample
rice-producing countries in Asia (Brennan respondents (See Moya et al. 2015 for more
and Malabayabas 2011; Pandey et al. 2012). details on how the survey was conducted over
The first-generation MVs (called MV1) were the years and the different sets of information
released from the mid-1960s to the mid- collected). The loop survey data included WS
1970s, which included the IR5 to IR34 varie- information for the following cropping years:
ties developed by the International Rice 1966, 1970, 1974, 1979, 1982, 1986, 1990,
Research Institute (IRRI) and the C4 series 1994, 1999, 2003, 2008, 2011, and 2015;
developed by the University of the Philippines whereas DS information was available for
(UP). Specifically, the release of IRRI’s IR8 1967, 1971, 1975, 1980, 1987, 1991, 1995,
variety in the Philippines and India is widely 1998, 2004, 2007, 2012, and 20165.
considered as the event that ignited the Green Note that the Loop Survey collected pro-
Revolution for rice production. Compared to duction and input use data for each parcel
taller TVs, the semi-dwarf MV1s achieved (or field) the farmer uses (i.e., there could be
higher yields primarily due to their resistance three rice parcels for a particular farm house-
to lodging, their ability to make more efficient hold, and input use information, say on fertil-
use of solar energy, and their responsiveness izer, was collected for each of the three
to fertilizer (Launio et al. 2008). Although parcels, where the input applied for each par-
MV1 are typically higher yielding (relative to cel may vary). However, there was no unique
TVs), they were more susceptible to pests identifier used to consistently track parcels
and diseases. The second-generation MVs over time. Hence, only a farm-level panel data
(called MV2) were released in the mid-1970s set can be constructed with the loop survey
to mid-1980s and included such IRRI- because only the farm households can be
developed varieties like IR36 to IR62. These uniquely tracked over time (and not the par-
MV2 varieties incorporated multiple pest and cels for each farm household). Nevertheless,
disease resistance traits (relative to MV1). we still “carry-over” the parcel level data rows
The third-generation MVs (called MV3) were (for each farm household) and run our empir-
developed and released between the mid- ical models using parcel-level observations.
1980s to the late-1990s and incorporated
better grain quality and stronger host plant
resistance (Launio et al. 2008). Last, the 4
As noted in Laborte et al. (2015), there was an additional vari-
fourth-generation MVs (called MV4) were etal group called MV5 that refers to modern rice varieties released
after 2005. However, these varieties do not have substantially dif-
ferent characteristics relative to MV4. Hence, MV4 and MV5 are
considered as a same varietal group—we call them “Recent
3
The seasonal production ranges coincide with the climate MVs” in this study. Further, note that hybrid rice varieties are
regime in the study area—one with a distinct wet monsoon season excluded from the analysis given that only a small proportion of
and a distinct dry season. Note that the Philippines is a spatially this variety is adopted in the study area, especially in the wet sea-
heterogeneous country with four major climate regimes: (a) dis- son (Moya et al. 2015).
5
tinct wet monsoon season and dry season; (b) no distinct dry sea- Not all households have a complete set of data for all years
son but a strong wet monsoon season; (c) intermediate between (i.e., attrition). This is common for studies based on repeated sur-
type 1 and 2, where there is a short wet monsoon and short dry sea- veys. However, further analysis on this issue suggest that attrition
son; and (d) an even distribution of rainfall throughout the year is likely random in our case, though one cannot completely rule
(Stuecker, Tigchelaar, and Kantar (2018)). it out.
Wang et al. Modern Rice Varieties and Warming Temperatures 5

Figure 1. The study area: Central Luzon loop survey.

Source: “Changes in rice farming in the Philippines: Insights from five decades of a household-level survey” (http://irri.org/resources/publications/books/changes-
in-rice-farming-in-the-philippines-insights-from-five-decades-of-a-household-level-survey)

But, as discussed further in the next section, pooled data are inappropriate. With the con-
we can only account for farm-level fixed struction and operation of large scale irriga-
effects (and not parcel-level fixed effects) tion systems and wide use of small pumps
given the data structure described here. used for irrigation, the population of farmers
As noted above, the loop survey includes having access to irrigated water was growing
data for two growing seasons (DS and WS). rapidly for the period considered. In the data
It is likely that the rice yield effect of weather set, we used for empirical analysis, 79% of
variables varies by season. From 1966 to observations are irrigated operations. For this
1975, only around 20% of farmers in the Cen- reason, in this study, the sample of interest was
tral Luzon region can plant a DS rice because limited to irrigated rice production planted in
of lack of irrigation. For this reason, our DS the WS.6
sample has a relatively small number of obser-
vations. Given the limited size of the dry sea-
6
son data, we focus on the analysis of the WS Limiting the sample to irrigated WS rice production produces a
more homogeneous sample that allows us to better tease out the
data. Another major concern is that yield effect of warming on yields for different varietal groups. In addi-
response to weather variables and input use tion, focusing on irrigated production in the WS makes it possible
are likely to vary depending on whether the to have a more parsimonious empirical specification. Including DS
and non-irrigated observations would require at least a doubling of
farm is irrigated or not. Thus, pooling them the already sizable number of parameters to be estimated (see
together and fitting the model for this kind of equations 1 and 2 below). The number of parameters need to at
6 Month 2021 Amer. J. Agr. Econ.

Aside from the loop survey data, we also Modeling Framework

utilized data on the monthly average of daily
values for minimum temperature (in
C) and We use multivariate regression methods to
maximum temperature (in
C), and monthly estimate econometric models of the following
total precipitation (in mm/month) in our anal- general form:
ysis. The two main sources of raw climate
data used are: (a) the University of East

ð1Þ ln yijmt
Anglia’s Climatic Research Unit (CRU)  
time-series (TS) data (version 4.01), and (b) = αj + f tminkmt , tmaxkmt , precmt ,Vijmt ; δ,β,ψ
the WorldClim data (version 2).7 The + γXijmt + ηt + εijmt
CRU-TS data, which is a gridded, historic,
monthly climate data set at 0.5 degrees reso- where ln(yijmt) is the natural log of rice yield y
lution (Harris et al. 2018, were downscaled (in kg/ha) for parcel i and farm j, located in
to a 30 arc-second (approx 1 km) resolution municipality m, for year t. The other terms in
by using the delta method and the high- Equation (1) are defined as follows. The
spatial resolution reference climatology data parameter αj accounts for unobservable time-
from WorldClim (Fick and Hijmans 2018). invariant, farm-level fixed effects such as soil
In particular, the GlobalClimateData.org quality and farmer management ability. The
Downscaling Package (used with the function f() is what we call the climate func-
MATLAB software) is used to produce the tion that includes the following explanatory
30 arc-second monthly downscaled climate variables: (a) a vector of weather variables:
data (Mosier, Hill, and Sharp 2014, 2018).8 municipality-level maximum and minimum
For each municipality covered by the loop temperature for a particular kth growing
survey, we then overlay the municipal bound- phase, as well as cumulative growing season
aries from the Global Administrative Areas precipitation (i.e., see the subsection below
(GADM) database (version 3.6) on the describing the climate function specification
downscaled climate data to calculate the for more details); and (b) a vector of parcel-
mean monthly climate data values used in level rice varietal group dummy variables
the study.9 Therefore, the climate data in this Vijmt.
study are at the municipality level and For parsimony and ease of interpretation,
reported at a monthly time scale for the years we classify the hundreds of varieties in the
covered in the loop survey. This climate data Loop Survey data set into three main varietal
were then merged to the loop survey data in groups: the “TV” group, the “Early MVs”
order to have one unified data set to run our group, and the “Recent MVs” group.10 The
empirical models. TV group is the omitted category in the regres-
sions, which includes the varieties prior to the
Green Revolution. Rice varieties commonly
considered as MV1, MV2, and MV3 are
included in the “Early MVs” group, where
least double because of the additional interaction terms “Early MV” is a dummy variable equal to
(e.g., double and triple interactions) needed to differentiate the
warming effects for wet versus dry season and for irrigated versus
one if the rice variety planted is either consid-
non-irrigated environments (for proper interpretation). Parsi- ered as MV1, MV2, or MV3, zero otherwise.
mony of the specification is compromised in this case, with likely In addition, rice varieties commonly classified
smaller gains in the accuracy of the estimates, given the sample size
of the survey. as MV4 and MV5 are included in the “Recent
7
See http://wwww.worldclim.org for the WorldClim data and MVs” group, where it is represented as a
https://crudata.uea.ac.uk/cru/data/hrg/ for the CRU data. For dummy variable equal to one if the rice variety
more information on how these two data sets were constructed
see Harris et al. (2018) and Hijmans et al. (2005), respectively. planted is commonly considered as “Recent
8
The resulting downscaled, monthly climate data were com- MVs,” zero otherwise.
pared against the Global Historical Climatology Network
(GHCN) station data (see Lawrimore et al. 2011) to assess the reli-
The term Xijmt is a vector of control variables
ability and accuracy of the downscaled data. We find that the mean that includes parcel-level input applications
absolute errors (MAE) and mean weighted absolute percentage
errors (MWAPE) between downscaled and observed station data
are reasonable for the purposes of this study (given the variation in
the observed data).
9
The Global Administrative Areas (GADM) database is
10
located at http://www.gadm.org (see GADM 2018 for more infor- This means that, for the purpose of parsimony, we did not use
mation). Calculation of the climate data per municipality was done the more common MV1 to MV5 varietal group classification as
in the software R version 3.5.1 using the raster package v2.6–7 described in the previous section (and as utilized in previous stud-
(Hijmans 2015). ies like Launio et al. 2008 and Laborte et al. 2015).
Wang et al. Modern Rice Varieties and Warming Temperatures 7

(e.g., fertilizer use, pesticide applications, and the entire season (as previously defined). This
labor), as well as other farmer/farm sociode- alternative specification is also used in Welch
mographic characteristics (e.g., age, education, et al. (2010).
land tenure). The term ηt is a linear time trend In our main empirical specification, we use
that is common to all farms in the sample, and, tmin and tmax by k growing phase, instead of
in previous studies, it typically represents tech- by month. We decided to do this in order to
nological evolution. However, note that use of have a parsimonious specification, to facilitate
rice varietal group dummies in the specification estimation, and for ease of interpretation.
allows us to separate at least the “varietal Because our focus is on the WS, it is important
development” part of the technological change to note that this growing season spans
from this time trend. The term εijmt is the 3–6 months and the lengths of the growing
parcel-level idiosyncratic error term, and δ, β, season vary across provinces. One can then
ψ, and γ are parameter vectors to be estimated. designate the main growing phases in each
Note that the farm-level fixed effects (αj) season as k = 1,2,3, where 1 = vegetative phase,
allow one to control for potential endogeneity 2 = reproductive phase, and 3 = ripening
caused by farm-level, time-invariant unobserv- phase. For example, tmax3mt would represent
ables that do not vary across parcels within a the maximum temperature for the ripening
farm (i.e., like unobserved farmer management phase (k = 3).
ability). Given that farm size in our data only However, the climate data set only contains
averages around 1 to 2 hectares, it is reasonable the monthly average of daily minimum tem-
to expect that these farm-level fixed effects ade- peratures and maximum temperatures, as well
quately control for potential endogeneity as the monthly cumulative precipitation
caused by time-invariant unobservables. Fur- (i.e., the sum of daily observations within a
thermore, we cluster standard errors at the vil- month). To construct weather variables by
lage level to account for potential correlations growing phase, we need to assign the monthly
among the parcels within a farm and the spatial weather values to each growing phase for each
correlations among farms within a village. year and across all provinces in the survey
data. Therefore, data on the “rice growing
windows” (i.e., the dates from planting to har-
Climate Function Specification vesting) for each growing season in the data
are required. For this purpose, we utilized
To estimate Equation (1), the function f
the RiceAtlas (Laborte et al. 2017), which con-
(tminkmt, tmaxkmt, precmt, Vijmt; δ, β, ψ) needs
tains the planting and harvesting dates for all
to be specified. The weather variables used
of the provinces covered by the Central Luzon
are minimum temperature (tmin), maximum
Loop Survey.12 However, the RiceAtlas
temperature (tmax), and precipitation (prec),
mainly focused on the “growing windows”
which are the same weather variables typically
from 1979 onwards, whereas the Loop Survey
used in previous studies (Welch et al. 2010;
data covers a longer period of time (i.e. from
Hasan, Sarker, and Gow 2016).11 Note how-
1966 to 2016). Information about “growing
ever that these weather variables were only
windows” for the earlier years of the Loop
available at the municipality level (m) and
Survey is not available. Thus, we needed to
not at the farm or parcel level. As discussed
make reasonable assumptions about the
further below, we also run an alternative spec-
months to include in each phase for earlier
ification with the following weather variables:
years of the Loop Survey data. Before 1979,
tavg, dtr, and prec. In this case, the variable
when TVs and MV1 are the major varieties
tavg is mean temperature (in ∘C), dtr repre-
adopted, growing seasons typically lasted
sents the diurnal temperature range (which is
around five to six months, and the wet season
equal to the difference between tmax and
starts around June and ends in November.
tmin), and prec is cumulative precipitation fo
The vegetative phase usually lasts seventy
five–ninety five days (i.e., three months), with
11
Minimum temperature is normally associated with nighttime the duration of both the reproductive and rip-
temperatures and maximum temperature is associated with day- ening phases around one month (see http://
time temperatures. Welch et al. (2010) have shown that these
two variables may have differing effects on rice yield and can enter
linearly in the specification. For the temperature range of our data,
the linear relationship between rice yields and temperature is sup-
12
ported by previous agronomic studies (Peng et al. 2004, Nagarajan See Table S20 in the online supplementary appendix for infor-
et al. 2010) and other past rice yield and temperature studies mation on the average maturity lengths and growing phase lengths
(Sarker, Alam, and Gow 2012, Pattanayak and Kumar 2014). for each province.
8 Month 2021 Amer. J. Agr. Econ.

Wet Season

200
Area(ha)
100
0

1966 1970 1974 1979 1982 1986 1990 1994 1999 2003 2008 2011 2015

Dry Season
150
100
Area(ha)
50
0

1967 1971 1975 1980 1987 1991 1995 1998 2004 2007 2012 2016

TV Early MVs Recent MVs

Figure 2. Adoption area of rice varietal group by survey year

www.knowledgebank.irri.org/step-by-step- monthly weather values to appropriate grow-

production/pre-planting/crop-calendar). ing season phases for these years.
Based on the information above, for the years Another major component of the climate
prior to 1979, we take the average weather function f() is the rice varietal group dummies
values from June to September as the vegeta- (Vijmt). In this study, we designate TV as the
tive phase value, the average of September base group (e.g., the omitted category) and
and October as the reproductive phase value, then use the notation Vr to represent the two
and the average of October and November as other varietal groups we defined in the previ-
ripening phase value. With the adoption of ous section (i.e., r = 1, 2 corresponds to
MV2, the average growth period declined 1 = “Early MVs” and 2 = “Recent MVs”,
from about 150 days in the 1960s and 1970s respectively. The area planted to each varietal
to about 110–120 days in the 1980s and 1990s grouping (for each survey year) is presented in
(Moya et al. 2015). For growing seasons after figure 2.
1979, the RiceAtlas provides accurate planting Given the notations discussed above, the cli-
and harvesting dates, and we, therefore, use mate function f() can then be fully specified as
this information to properly assign the follows:

X
2 X
3 X
3
ð2Þ βr Vrijmt + δ1k tminkmt + δ2k tmaxkmt + δ3 precmt + δ4 ðprecmt Þ2 +
r=1 k=1 k=1
3 X
X 2
 X3 X
2

ψ r1k tminkmt × Vrijmt + ψ r2k tmaxkmt × Vrijmt +
k=1 r=1 k=1 r=1

X
2
 X2

ψ r3 precmt × Vrijmt + ψ r4 ðprecmt Þ2 × Vrijmt
r=1 r=1
Wang et al. Modern Rice Varieties and Warming Temperatures 9

Table 1. Descriptive Statistics for the Economic Variables

Variable Units/definition Obs Mean St dev Min Max
Yield kg/ha 1,151 3889.51 1555.22 306.67 11250.00
Land tenure 1 = owner; 0 = other 1,151 0.42 0.49 0.00 1.00
Farm size ha 1,151 1.32 0.97 0.03 9.00
Age of head No. of years 1,149 52.63 13.64 22.00 94.00
Educ. of head No. of years 1,151 7.25 3.34 0.00 16.00
Labor man-days/ha 1,151 70.13 28.70 0.00 257.75
Nitrogen fert. kg/ha 1,151 81.89 50.50 0.00 483.91
Potassium fert. kg/ha 1,151 11.03 13.50 0.00 127.80
Phosphorus fert. kg/ha 1,151 9.20 8.28 0.00 67.10
Insecticide kg/ha 1,151 1.50 2.64 0.00 70.27
Herbicide kg/ha 1,151 0.89 2.42 0.00 32.00

Quadratic precipitation terms is added to included in the specification are: land tenure
the climate function to allow for nonlinear pre- status, age, and education of household head
cipitation effects, which is similar to the speci- (in no. of years), and farm size (ha). Land ten-
fication used in previous research (Tack, ure status is represented by a dummy variable
Barkley, and Nalley 2015, Lobell, Schlenker, Own where this variable is equal to 1 if the
and Costa-Roberts 2011, Schlenker and land is owned, and it is zero otherwise
Lobell 2010).13 The climate–MV interaction (e.g., share tenant, fixed rent leaseholder, or
terms make it possible to examine whether other tenurial arrangements). Table 1 pro-
there is heterogeneity in each varietal groups’ vides descriptive statistics for the “economic
response to weather variables. variables” included in the empirical model,14
and table 2 presents the summary statistics
for the weather variables.
Specification of Control Variables
The next component of Equation (1) that
needs to be specified is the vector Xijmt, which
accounts for a number of control variables Marginal Effects
such as parcel-level input applications and One of the main goals of this study is to inves-
other sociodemographic farm characteristics. tigate heterogeneity in the yield response of
Including these variables in the specification different rice varietal groups to weather vari-
allows us to control for observable factors ables. The yield response is measured by the
(i.e., varying over time and space) that can marginal effect of changes in weather vari-
influence rice yields, thereby improving the ables on rice yield. Given the climate function
accuracy and efficiency of our estimations. specified in equation (2), the marginal effect of
The input application variables included in minimum and maximum temperatures can be
the specification are fertilizer (e.g., nitrogen, calculated using the following:
phosphorus, and potassium applications in
kg/ha), insecticide use (in kg/ha), herbicide ∂y

use (in kg/ha), and labor (in man-days/ha). ð3Þ = δ1k + ψ r1k × Vrijmt ,
These are considered major inputs in Philip- ∂tmink
pine rice production (Moya et al. 2015). ∂y

Sociodemographic and farm characteristics ð4Þ = δ2k + ψ r2k × Vrijmt
∂tmaxk

13
Although it would have been ideal to include quadratic tem-
perature terms in the climate function (i.e., to capture non-linear
14
temperature effects), an out-of-sample forecasting analysis in the See table S22 in the online supplementary appendix for more
spirit of Schlenker and Roberts (2009) indicate that adding these details about how the sample statistics have evolved over time
quadratic terms in equation (2) actually decrease model perfor- for the irrigated WS observations. In addition, we point the inter-
mance. This suggests that the inherent variation in the weather ested reader to Moya et al. (2015) for a description about the full
data, and perhaps the thinness of the temperature data at the tails survey sample and the evolution of the pertinent survey statistics
(as in Welch et al. 2010), precludes improvement in model perfor- over the 1966–2012 period. Note that the sample size in our study
mance even if one adds more quadratic terms. Overfitting does not exactly match the ones in Moya et al. (2015) because we
becomes an important concern. For these reasons, only linear tem- drop observations with missing data, unrealistic values, and hybrid
perature terms are used in the climate function specification. varieties (see footnote 4).
10 Month 2021 Amer. J. Agr. Econ.

Table 2. Descriptive Statistics for the Weather Variables

Variable Unit Obs Mean St dev Min Max
vtmin Deg. C 1,151 22.85 0.62 19.91 24.05
vtmax Deg. C 1,151 30.50 0.83 27.56 32.00
vtavg Deg. C 1,151 26.66 0.67 24.16 28.00
vdt Deg. C 1,151 7.65 0.74 5.14 9.45
retmin Deg. C 1,151 22.63 0.74 20.15 24.31
retmax Deg. C 1,151 30.40 0.78 27.78 32.45
retavg Deg. C 1,151 26.47 0.68 24.03 28.07
redt Deg. C 1,151 7.76 0.74 5.00 9.50
ritmin Deg. C 1,151 22.48 0.81 19.83 24.34
ritmax Deg. C 1,151 30.55 0.83 27.62 32.57
ritavg Deg. C 1,151 26.43 0.72 24.02 28.13
ridt Deg. C 1,151 8.07 0.87 6.00 10.51
precip mm 1,151 1386.85 357.70 692.84 3038.72
Notes: The table above displays the descriptive statistics of weather variables used in the regressions. The first four rows are the growing season averages of the
daily minimum, maximum, and mean temperatures, as well as the diurnal temperature range for the vegetative phase. The second four rows are the weather
variables for the reproductive phase, and the third four rows show the weather variables for the ripening phase. The last row is cumulative precipitation for the
entire growing season.

where Vrijmt is the parcel-level rice varietal for ease of interpretation and to facilitate mak-
group dummy variables. For example, sup- ing inferences, we focus on estimating the mar-
pose the rice variety adopted belongs to the ginal effect of a particular “warming scenario,”
“Early MVs” group, then V1ijmt = 1. In this case, where we are interested in the cumulative
marginal effect of a 1∘C increase in both tmin
the marginal yield effect of a one-unit change
and tmax in all three rice-growing phases
in the minimum (maximum) temperature for
(or for a particular phase).15 The marginal
the kth phase is δ1k + ψ r1k (δ2k + ψ r2k ) (i.e., the
effect of this specific “warming scenario” can
coefficient associated with the weather vari-
then be calculated respectively for the TVs,
able plus the coefficient associated with the
Early MVs, and Recent MVs as follows:
interaction of the weather variables and the
varietal grouping dummy). Because TV is des-
ignated as the base varietal grouping, the mar- X
3
∂y j V = TV X
3
∂y j V = TV
ð6Þ +
ginal effects of weather variables tminkmt and
k=1
∂tmink k=1
∂tmaxk
tmaxkmt on TV rice yield are δ1k and δ2k,
respectively. On the other hand, the marginal X
3 X
3
= δ1k + δ2k
effect of growing season cumulative precipita-
k=1 k=1
tion is:
X
3
∂y j V = EarlyMVs X
3
∂y j V = EarlyMVs

 ð7Þ + =
∂y ∂tmink ∂tmaxk
ð5Þ = δ3 + ð2 × δ4 × precÞ + ψ r3 × Vrijmt
k=1 k=1

∂prec X X X X


3 3 3 3
δ1k + δ2k + ψ 1k1 + ψ 2k1
+ 2 × ψ r4 × prec × Vrijmt k=1 k=1 k=1 k=1

The simple marginal effect expressions in

Equations (3) and (4) can easily be interpreted
if there are only a few weather variables to 15
Even though the specific “warming scenario” discussed here is
consider for each growing phase and if there mainly for the purpose of facilitating interpretation, it is important
are only one or two rice varietal groups. How- to note that minimum and maximum temperatures in the Philip-
pines tend to move together and are usually positively correlated
ever, our empirical model includes six “tem- (see Welch et al. 2010; Peng et al. 2004). Our data also support this
perature-growing-phase” variables for each behavior (see figure S2 and table S3 in the online supplementary
of two MV groups. Given the number of appendix). Therefore, the base “warming scenario” examined
here is still fairly reasonable based on this positive correlation
parameters involved, drawing sensible and between tmin and tmax. Nevertheless, given that minimum and
consistent inferences using the simple mar- maximum temperatures are likely not to move together in exactly
1
C intervals in reality, we also explore marginal effects for the
ginal effect expressions in Equation (3) and case where tmin and tmax changes based on projections from cli-
(4) would be difficult and complex. As such, mate models (See Section 4 below).
Wang et al. Modern Rice Varieties and Warming Temperatures 11

X
3
∂y j V = RecentMVs X
3
∂y j V = RecentMVs parsimonious models, which then build towards
ð8Þ + =
the full specification results from Equations (1)
k=1
∂tmink k=1
∂tmaxk
and (2). The first parsimonious model (Model
X
3 X
3 X
3 X
3
1) is our baseline where we do not include any
δ1k + δ2k + ψ 1k2 + ψ 2k2
k=1 k=1 k=1 k=1 of the control variables Xijmt in the specification
and only include weather variables, varietal
From these equations, we can calculate the group dummies, and relevant interactions. The
warming yield response of Early MVs and second parsimonious model (Model 2) adds in
the Recent MVs as compared to TVs. This the sociodemographic and farm characteristics
allows us to make inferences on whether or variables: land tenure, age, education and farm
not the Early MVs and/or Recent MVs are size. The fully specified empirical model
more resilient to warming temperatures rela- (Model 3) includes everything in Model 2 plus
tive to the TVs. the input application variables. The pertinent
For calculating the impact of cumulative marginal effects for Models 1 to 3 under a vari-
precipitation (prec), we can directly derive ety of warming scenarios are presented in
the marginal effect because we utilize a single table 3.16 Marginal effects for the “baseline”
cumulative growing-season precipitation vari- model (Model 1) and the corresponding P-
able in the specification instead of precipita- values are in columns 2 and 3. Model 2 results
tion in each of the three growing phases. For are presented in columns 4 and 5. Marginal
example, the estimated marginal effect of a effects and their P-values for Model 3 are in col-
1 mm increase in the cumulative precipitation umns 6 and 7.
for the TVs, Early MVs and Recent MVs can For all model specifications, a warming sce-
be calculated as follows: nario that increases both tmin and tmax by
1∘C in all growing phases substantially reduces
∂y j V = TV rice yields, and these estimated warming
ð9Þ = δ3 + 2 × δ4 × prec effects are statistically significant at the usual
∂prec
levels of significance (i.e., see warming sce-
∂y j V = EarlyMVs nario in the top panel of table 3). The magni-
ð10Þ = δ3 + 2 × δ4 × prec
∂prec tudes of our marginal effects range from
+ ψ 31 + 2 × ψ 41 × prec −13% (for Recent MVs in Model 3) to
−27.6% (for the TVs under Model 1). Results
∂y j V = RecentMVs presented in the other two warming scenarios,
ð11Þ = δ3 + 2 × δ4
∂prec where only tmin or tmax are increased sepa-
× prec + ψ 32 + 2 × ψ 42 × prec rately by 1∘C (see middle panels of table 3),
indicate that tmin is the likely source of the
observed negative yield impact of warming
Given that a squared precipitation term
(given the strong statistically significant nega-
and its interaction with the varietal group
tive effect of tmin and the largely statistically
dummy are included in Equation (2), the
insignificant effect of tmax). This result is con-
marginal impacts of precipitation in
sistent with results from Welch et al. (2010)
Equations (9) to (11) are a function
where tmin effects were also found to be the
involving the value of prec. In this study,
stronger determinant of rice yield losses due
we calculate the marginal impact of cumu-
to warming temperatures. It is also important
lative precipitation at the mean of prec. In
to note that the estimated adverse warming
addition, we also measure and report the
effects observed in Model 1 became smaller
marginal effect of a one standard devia-
as socio-demographic and input variables
tion increase in precipitation (at the mean
were added to the specification (Models
of prec).
2 and 3). This suggests that controlling for
farm-level time-varying confounding factors
(like input use and sociodemographic

Estimation Results

The fully specified empirical model for this 16

The main warming scenario considered in table 3 is a 1
C
study is primarily based on Equations (1) and increase in tmin and/or tmax. We also provide the marginal effects
for a warming scenario that increases tmin and tmax by 1 standard
(2) above. However, in this section, we also pre- deviation in table S2 and figure S3 in the online supplementary
sent estimation results from two more appendix. The pattern of results in both cases are similar.
12 Month 2021 Amer. J. Agr. Econ.

Table 3. Marginal Percentage Yield Impact of Weather Variables for Different Warming
Scenarios and Varietal Groups
Model 1 Model 2 Model 3
No economic With farm With farm char. &
variables characteristics inputs
Variables Estimates P-value Estimates P-value Estimates P-value
∘
1 C warming scenario:
tmin&tmax: TV −0.276 0.021 −0.270 0.031 −0.214 0.086
tmin&tmax: Early MVs −0.235 0.000 −0.219 0.001 −0.167 0.007
tmin&tmax: Recent MVs −0.190 0.007 −0.161 0.021 −0.134 0.070
1∘C increase in tmin:
tmin: TV −0.668 0.006 −0.717 0.003 −0.624 0.014
tmin: Early MVs −0.242 0.000 −0.218 0.001 −0.196 0.002
tmin: Recent MVs −0.318 0.019 −0.263 0.048 −0.266 0.046
1∘C increase in tmax:
tmax: TV 0.392 0.167 0.447 0.135 0.410 0.183
tmax: Early MVs 0.007 0.892 −0.001 0.978 0.029 0.593
tmax: Recent MVs 0.128 0.132 0.102 0.223 0.131 0.114
One standard deviation increase in cumulative precipitation:
prec: TV −0.200 0.191 −0.194 0.247 −0.236 0.145
prec: Early MVs −0.168 0.000 −0.154 0.000 −0.147 0.000
prec: Recent MVs −0.084 0.211 −0.081 0.234 −0.036 0.589
Notes: (a) The table displays coefficients and P-values of marginal yield effect of 1∘C warming scenarios and one standard deviation of increase in prec from three
farm fixed-effect models. Standard errors for each regression are clustered at the village level. (b) The different models are as follows. Model 1 includes tmin and
tmax variables in all the growing phases (e.g., the vegetative [vtmin and vtmax], reproductive [retmin and retmax], and the ripening phase [ritmin and ritmax]),
linear and quadratic cumulative precipitation in the growing season (prec and prec2) and their interactions with dummies for rice varietal groups. Model 2 adds
farm characteristics (age and education of household head, land tenure and farm size) to Model 1. Model 3 adds input variables (labor, fertilizer [n, p, k],
insecticide and herbicide) to Model 2. (c) The first column indicates what weather variables on which the marginal effects are based and to which varietal group it
pertains. The three rows of the first panel indicate the marginal effect of a 1∘C increase in both tmin and tmax for the TV, Early MVs, and Recent MVs varietal
groups separately. The rows of panel 2 refer to the marginal effect of a 1∘C increase in tmin for the TV, Early MVs, and Recent MVs. The rows of the third panel
refer to the marginal effect of a 1∘C increase in tmax for the TV, Early MVs and Recent MVs. Last, the rows of the fourth panel indicate the marginal effect of a 1
standard deviation of increase in prec for the TV, Early MVs, and Recent MVs.

characteristics) may be important in our point estimates. In figure 3, we graphically pre-

empirical context.17 sent the marginal percentage yield effects of
Another important result from table 3 is the the main warming scenario (e.g, a 1∘C increase
apparent heterogeneity in the warming in both tmin and tmax across the vegetative,
impacts across the three varietal groups based reproductive, and ripening phases) for the
solely on the magnitudes of the marginal effect three varietal groups. For all three model spec-
ifications, the warming impact point estimate
17
is smallest for the Recent MVs varietal group.
It should be noted here that although including farm inputs in In addition, we observe that point estimates of
the specification can help control for confounding factors, it can
also raise endogeneity concerns especially if there are parcel-level the negative warming effect on yields is smal-
unobservables correlated with input use and yield outcomes that ler for the Early MVs as compared to the
are not adequately controlled for by the farm-level-fixed effects,
time trends, and other control variables. Nonetheless, this concern
TVs (across all model specifications). These
is mitigated by the result that the magnitudes and statistical signif- point estimate results are suggestive of
icance of the estimated effects in Models 1 and 2 (without the input improved heat resilience for Early MVs and
use variables) are roughly similar to the one in Model 3 (when
input use variables are included). Results from a battery of robust- Recent MVs relative to the TVs.
ness checks (see next sub-section) also supports the main findings The point estimate result indicating
here. Last, there may also be concerns about endogeneity of the improved heat tolerance of Early MVs relative
varietal group dummies per se. However, Moya et al. (2015,
p. 59) point out that seed cost is a small proportion of total costs to TVs is particularly interesting given that
for the loop survey farmers (<5%) and real seed prices have been Early MVs were not primarily bred to have
relatively stable over time. Hence, it is likely that agronomic, soil,
and climate factors are the main drivers of varietal decisions (not
enhanced tolerance to abiotic stresses.
unobserved economic-related factors) and these factors are Though it should be noted that there are previ-
already sufficiently controlled for through farm fixed effects, ous studies that point out the agronomic and
farm/farmer characteristics, time trends, and weather variables.
Hence, endogeneity associated with varietal group dummies is physiological basis for why post-green-
not a major issue. revolution Early MVs may have better heat
Wang et al. Modern Rice Varieties and Warming Temperatures 13

0
-10
Yield Impact (%)
-20
-30
-40
-50

Model 1 Model 2 Model 3

TV Early MVs Recent MVs

Figure 3. Predicted impacts of the 1∘C warming scenario on three rice varietal groups for three
model specifications described by table 3.
Notes: Impacts are reported as the percentage change in yield. Whiskers show 90% confidence interval

tolerance than TVs (Wassmann et al. withstand heat given that drought events
(2009)). For example, Wassmann et al. are usually associated with lack of moisture
(2009) explains that the semi-dwarf plant and above-average temperatures.
architecture of these Early MVs, as com- Notwithstanding the pattern of results for the
pared to the taller TVs, makes it sturdier marginal effect point estimates in table 3 and
and more tolerant to heat events. Specifi- figure 3, the confidence bands in figure 3 indi-
cally, the semi-dwarf architecture of the cate that the marginal yield response to warm-
Early MVs allows them to have panicles that ing do not differ statistically across varietal
are surrounded by canopy, which makes it groups (i.e., confidence interval “whiskers”
possible for these Early MVs to have more across varietal group vertically overlap with
efficient transpiration cooling during the each other). Formal statistical tests of equality
heat sensitive flowering period and, conse- (e.g., F-tests) among the coefficients used for
quently, better heat resilience. In addition, calculating the marginal effect point estimate
the shorter growing season of Early MVs, for each varietal group also suggest that there
as compared to TVs, allows for reduced are no statistically significant differences among
exposure to heat events during critical these coefficients (and the marginal effects
growth stages (i.e, “heat avoidance” con- themselves). Thus, even with the apparent vari-
cept), which may then lead to smaller heat- etal group heterogeneity in the point estimates
induced yield damages. For the more Recent of the marginal yield response to warming, the
MVs, the point estimate results suggesting lack of statistical difference among the marginal
better heat tolerance of this varietal group effects across varietal groups highlights the
relative to the Early MVs and TVs, is in line need for further rice breeding, agronomic, and
with rice variety releases in the Philippines economic research on this topic (i.e., more on
with abiotic stress tolerance traits. For exam- this in the conclusions).
ple, drought-tolerant varieties have been Next, we utilize the parameter estimates
released in the Philippines since the mid- to from our fixed effect models to investigate
late-2000s, and arguably these drought toler- how projected future climate change will likely
ant varieties have some abilities to better influence potential rice yields of the three
14 Month 2021 Amer. J. Agr. Econ.

varietal groups examined in this study.18 To under CNCM3 model (as compared to the
complete this climate projection and rice yield other two GCMs).
simulation exercise, we utilize the projected The percentage change in rice yields due to
climate change values from PAGASA, the the projected temperature changes are pre-
main meteorological government agency in sented in the online supplementary appendix
the Philippines. The climate change values figures S4 and S5 for Model 2 (i.e, specification
from PAGASA are the projected change in with farm characteristics, but without input vari-
seasonal minimum temperature, maximum ables). The detailed yield effect point estimates
temperature, and precipitation from the aver- for all models are presented in supplementary
age over the period 1971–2000 to the average table S8. In general, our yield prediction point
over the period 2011–2040. These projected estimates suggest that the Recent MVs are still
changes are generated based on the statistical the ones that are more tolerant to projected
downscaling of three global climate models warming temperatures for most of the GCM-
(GCMs): (a) the BCM2, (b) the CNCM3, and emission-scenario combinations examined
(c) MPEH5; and two plausible emissions sce- (with the exception of the results from the
narios: (a) the A1B emission scenario, and CNCM3 projection model). Point estimate
(b) the A2 emission scenario.19 results from this analysis also suggest that Early
The projected changes in tmin and tmax and MVs exhibit better tolerance to projected
prec for each of the six provinces in this study warming temperatures (as compared to the
are presented in tables S5–S7 of the online sup- TVs). These climate projection results are con-
plementary appendix. In addition, the summary sistent with the marginal effect point estimates
statistics for the average across the six Loop from the earlier analysis (table 3), as well as
Survey provinces by growing phase (in the the statistically insignificant yield response dif-
WS) are provided in table S4 of the online sup- ferences across varietal groups.
plementary appendix. Note that table S4 shows So far, we have focused on the differential
that both tmin and tmax are predicted to warming impacts across different varietal groups
increase in the future. Under most of the “emis- using both the warming scenario and climate
sion-scenario-GCM-growing phase” combina- projection models. Precipitation effects have
tions examined, the magnitudes of the changes not been discussed. In figure S7 in the online
in tmin and tmax are similar (which validates supplementary appendix, we also show the mar-
the original “warming scenario” examined ginal rice yield response due to a one standard
above). However, specifically under the “A1B- deviation increase in growing season cumulative
CNCM3-Vegetative Phase” combination and precipitation prec (evaluated at the mean of
the “A2-CNCM3-Vegetative Phase” combina- prec). Increases in prec (at the mean) tend to
tion, the incremental increase in tmin is double reduce yields of all three varietal groups. Among
that of the increase in tmax, which typically the three varietal groups, the estimated reduc-
leads to relatively different climate predictions tion in the Recent MVs yield is the smallest.
These point estimates indicate that the Recent
MVs is the rice varietal group that tend to be
18
Simulating the effect of projected future climate on rice yields more tolerant to increases in cumulative precip-
also provides additional insights relative to the 1
C warming sce- itation. Although, it should be noted that the
nario examined in table 3 because this simulation exercise does Early MVs also exhibit resilience to increases
not implicitly assume that tmin and tmax change by the same
amount (i.e., dtr is not assumed to be constant in the future climate in cumulative precipitation (as compared to the
projections).
19
TVs). However, similar to the findings on yield
Note that GCMs are powerful computer programs that use
physical processes to replicate, as accurately as possible, the func-
effects of higher temperatures, we do not find
tioning of the global climate system (Comer, Fenech, and statistically significant differences in the
Gough 2007). The BCM2 model was established by the Bjerknes varietal-group-specific yield response to
Centre for Climate Research. On the other hand, the CNCM3
GCM was developed by the Météo-France (Centre National de increases in cumulative precipitation (even
Recherches Météorologiques). Last, the MPEH5 was developed though there is apparent heterogeneity in the
by the Max Planck Institute for Meteorology. These three GCMs marginal effect point estimates).
are considered the most effective at simulating climate for the Phil-
ippines (Tolentino et al. 2016).
On the other hand, the A1B and A2 are two emissions scenar-
ios used in the regional climate projections of the Intergovernmen-
tal Panel on Climate Change (IPCC) Fourth Assessment Report
(AR4) and were generated by the Geophysical Fluid Dynamics Robustness Checks
Laboratory (GFDL) model. The A1 family of scenarios assumes
a more integrated world and A1B is based on a balanced techno-
logical emphasis on all energy sources. The A2 scenarios, on the As a robustness check, we also estimate similar
other hand, assumes a more divided world. models as described in Equations (1) and (2),
Wang et al. Modern Rice Varieties and Warming Temperatures 15

10
0
Yield Impact (%)

-10
-20
-30
-40

Model 1 Model 2 Model 3

TV Early MVs Recent MVs

Figure 4. Predicted impacts of a +1∘C increase in tavg on three rice varietal groups for three
model specifications described by table 4.
Notes: Impacts are reported as the percentage change in yield. Whiskers show 90% confidence interval

but instead of tmin and tmax, as the two main increasing more than tmax) is generally negative
temperature variables considered, we instead (as expected), though these estimates are largely
utilize average temperature (tavg) and diurnal statistically insignificant (see table 5 [middle
temperature range (dtr). Cumulative precipita- panel] and figure S8 in the online supplementary
tion prec is still included in this robustness check appendix).
specification (with both linear and quadratic Under all three model specifications, the
terms). We still follow the approach from the point estimate for the percentage negative
previous section where we examine three model yield impact of tavg is highest for TVs and low-
specifications (Models 1–3). est for the Recent MVs. This is consistent with
The estimated marginal yield effects of tavg the point estimate patterns observed in the
and dtr for various warming scenarios and model previous section. However, as seen in figure 4,
specifications are presented in table 4 (and the confidence intervals for each varietal
regression results for the specifications are in group still largely overlaps, suggesting that
table S9 in the online supplementary appendix). there are no statistically significant differences
In addition, the marginal effects of a 1∘C increase in the marginal yield response to tavg across
in tavg are graphically shown in figure 4. Our varietal groups. In addition, figure S9 in the
results indicate that increases in tavg negatively online supplementary appendix shows the
impact rice yields. However, the magnitudes of marginal yield impacts of prec at the mean
the marginal effects for tavg is smaller than the for the model using tavg and dtr, and this figure
ones in the previous section for tmin and tmax, supports the robustness of the precipitation
with the TV marginal effects being largely statis- effects from the regression runs in the previous
tically insignificant (consistent with previous section.
studies like Welch et al. 2010). This may be Another robustness check is running sepa-
because tmin and tmax have opposing rice yield rate regressions by varietal groups. The data-
impacts for most varietal groups in nearly all set was divided into three subsamples by
specifications in table 4. Thus, the opposing tem- varietal groups. We constructed a model spec-
perature impacts may partly cancel each other ification including linear terms for tmin and
out when using tavg. On the other hand, the mar- tmax, linear and quadratic terms for prec, and
ginal effect of a decrease in dtr (i.e., tmin applied this specification to each varietal
16 Month 2021 Amer. J. Agr. Econ.

Table 4. Marginal Percentage Yield Impact of Weather Variables: Alternative Specification

Using Mean Temperatures & DTR
Model 1 Model 2 Model 3
No economic variables With farm characteristics With farm char. & inputs
Variables Estimates P-value Estimates P-value Estimates P-value
1
C warming scenario:
tavg: TV −0.165 0.157 −0.153 0.177 −0.110 0.337
tavg: Early MVs −0.164 0.001 −0.152 0.002 −0.111 0.021
tavg: Recent MVs −0.117 0.024 −0.101 0.040 −0.094 0.053
1
C decrease in diurnal temperature variation:
dtr: TV −0.447 0.064 −0.512 0.030 −0.465 0.063
dtr: Early MVs −0.086 0.128 −0.070 0.202 −0.078 0.192
dtr: Recent MVs −0.145 0.116 −0.118 0.199 −0.148 0.108
one standard deviation increase in cumulative precipitation:
prec: TV −0.236 0.021 −0.210 0.050 −0.241 0.031
prec: Early MVs −0.156 0.000 −0.141 0.000 −0.135 0.000
prec: Recent MVs −0.039 0.512 −0.039 0.527 0.000 0.994
Notes: (a) The table displays coefficients and P-values of the marginal yield effect of 1
C increase in tavg and 1
C decrease dtr for all phases in the growing season
and 1 standard deviation increase in prec, based on three farm fixed-effect models estimated. Standard errors for each regression are clustered at the village level.
(b) The different models are as follows. Model 1 includes tavg and dtr variables in all the growing phases (e.g., the vegetative [vtavg and vdtr], reproductive [retavg
and redtr], and the ripening phase [ritavg and ridtr]), linear and quadratic cumulative precipitation in the growing season (prec and prec2) and their interactions
with dummies for rice varietal groups. Model 2 adds farm characteristics (age and education of household head, land tenure, and farm size) to Model 1. Model 3
adds input variables (labor, fertilizer [n, p, k], insecticide and herbicide) to Model 2. (c) The first column indicates what weather variables on which the marginal
effects are based and to which varietal group it pertains. The three rows of the first panel indicate the marginal effect of a 1
C increase in tavg for the TV, Early
MVs, and Recent MVs varietal groups separately. The rows of panel 2 refer to the marginal effect of a 1
C decrease in dtr for the TV, Early MVs, and Recent
MVs. Last, the rows of the third panel indicate the marginal effect of a one standard deviation of increase in prec for the TV, Early MVs, and Recent MVs.

group subsample (i.e., due to the smaller sam- differences among all the marginal yield
ple size for each varietal group, we did not response estimates to warming.
include the control variables for farmer/farm Since the rollout and use of the different vari-
characteristics and input use). The estimated eties occurred sequentially through time
impacts of a +1∘C warming scenario and a (i.e., TVs in earlier years, followed by the
one standard deviation increase in prec for release of Early MVs, and then Recent MVs
each varietal group subsample are seen in in more recent years), one other approach to
table S10 and the parameter estimates are check the robustness of results is by running a
reported in table S11 (see the online supple- specification with no varietal group dummy
mentary appendix). In addition, we graphi- interactions with weather but instead interact-
cally show the impact of a +1∘C warming ing the weather variables (by growing phase)
scenario based on the separate regression runs with the time trend. Parameter estimates from
in figure S10 in the online supplementary this alternative specification are reported in
appendix, whereas the impact of a one stan- table S12 in the online supplementary appendix.
dard deviation increase in prec is provided In this specification, varietal development is
graphically in figure S11. Note that in embedded in the time trend (along with other
figure S10, we only plot the confidence interval rice technologies evolving over time). Hence,
for Early MVs and Recent MVs because of the if varietal development is the main driver of rice
large confidence interval for the TV group technological change, then we would expect a
(which is likely due to the small sample size), point estimate pattern where the adverse effect
and this does not easily fit the scale of the fig- of warming would be larger in earlier years
ure. Overall, results from this robustness (where TV is predominant), and it would then
check is still consistent with the previous slowly decrease over time as more MVs are
analysis—point estimates of the marginal released. More recent years will have smaller
effect of our 1∘C warming scenario follow the point estimates of the negative warming effects
pattern where the highest negative warming than earlier years given the release of recent
impact is observed for TVs and the lowest is MVs. This pattern is indeed verified and shown
obseved for Recent MVs. Confidence bands in figure S12, which supports the robustness of
also indicate that there are no statistical our earlier point estimate results.
Wang et al. Modern Rice Varieties and Warming Temperatures 17

Another robustness check we conducted is Hence, it seems like MV4 is the main varietal
to examine a specification with both: (a) varie- group driving the resilience of the Recent
tal group interactions with the weather, and MVs varietal group in our main analysis. But
(b) time trend interactions with the weather. note that this may also be due to MV5 being
Compared to the specification in the previous adopted only for a shorter period in the data.
paragraph, this last specification separates In addition, consistent with results from the
out the warming effect of varietal groups from main model (table 3), there are no statistically
the warming effect due to other technologies. significant different marginal effects across the
Parameter estimates from this specification four varietal groups (i.e, confidence intervals
are reported in table S13 in the online supple- overlap) in the robustness check separating
mentary appendix, and the pertinent marginal out MV5.
effects are presented in figure S13. Marginal Last, we conduct three other robustness
effect point estimates from this last robustness checks: (a) a specification that includes three
check are still consistent with the main pattern growing phase precipitation variables and
of results from the previous analysis, where their respective quadratic terms (i.e., rather
the adverse warming effect is smaller for the than using a cumulative season-long precipita-
recent MVs relative to the earlier MVs and tion variable and its quadratic term;
the TVs. table S15), (b) a specification with precipita-
The number of observations for the TV tion only for the reproductive phase (i.e,
varietal group is relatively small and available because precipitation has been shown in previ-
only at the beginning of the study period (see ous agronomic studies to be critical for this
figure 2). For this reason, estimates related to phase; table S16), and (c) a specification with
TV generally have large standard errors, fixed two-month growing phase windows,
though in our main warming scenario in table 4 rather than growing phases primarily based
the marginal effects for TVs are still statisti- on the Philippine RiceAtlas estimates
cally different from zero. Due to the difficulty (table S17). Results from these alternative
of getting efficient estimators for TV, we con- models also generally support the findings
duct another robustness check where we run from our main analysis in the previous section,
our models using data that do not include especially the magnitude pattern of the mar-
observations for TV (i.e., only the Early MV ginal effect point estimates.
and Recent MV observations are included in
the data). Appendix Table S18 shows the
warming impacts on early MVs and recent
MVs when TV observations are dropped from Conclusions
the data. For our main 1∘C warming scenario,
the point estimates of the marginal effects The main objective of this study is to investi-
indicate a larger reduction in yields for Early gate whether modern rice varieties (MVs)
MVs as compared to Recent MVs. reduce the adverse yield impacts of warming
Even though the classification of MV5 is temperatures, especially the more recent vari-
mainly based on the year of release rather than eties bred to be more tolerant to abiotic stres-
the difference in its characteristics relative to ses (i.e., those in the MV4 and MV5 varietal
the previous generation of modern varieties groups). We provide unique empirical evi-
(Laborte et al. 2015), it is still interesting to dence on whether investments in breeding
determine whether resistance to warming is programs have led to farmer-planted rice vari-
different between MV4 and MV5. For this rea- eties that are more resilient to warming. By
son, we also conduct a robustness check where merging Philippine farm-level survey data
we separate recent MVs into MV4 and MV5 (from 1966 to 2016) with monthly,
and estimate the coefficients for them sepa- municipality-level climate data, we are able
rately. The marginal impact of warming esti- to estimate fixed effect econometric models
mated from these models are provided in of rice yields with “weather-varietal group”
table S14 in the online supplementary appen- interactions and assess whether there is het-
dix. The marginal effect point estimates indi- erogeneity in the warming effects across dif-
cate that MV4 tend to be more resilient to ferent rice varietal groups. Our regression
heat relative to early MVs and TVs, and early models suggest that increases in temperature,
MVs tend to be more resilient relative to the especially minimum temperatures, have sub-
TVs. However, we find that MV5 tend to be stantial negative impacts on rice yields,
affected more by heat as compared to MV4. regardless of varietal group. Point estimates
18 Month 2021 Amer. J. Agr. Econ.

of the marginal yield effects of warming sug- that develops holistic management packages
gest that early MVs bred primarily for higher appropriate for abiotic stress tolerant varie-
yields, pest resistance, and/or grain quality ties, and (b) outreach and extension efforts
traits, as well as recent modern rice varieties that more effectively disseminate information
bred for better abiotic stress tolerance, are and improve uptake of the new varieties
more resilient to heat events than traditional together with complementary inputs. Further
rice varieties. However, our estimates of mar- economic research identifying potential bot-
ginal yield responses to higher temperatures tlenecks to adoption and to improved perfor-
are not statistically different across the three mance of climate change resilient varieties is
modern varietal groups. Taken together, these an important future research direction.
results do not provide strong statistical evi- Second, our main empirical results could
dence that recent MVs are more resilient to also imply that public rice breeding efforts
heat than TVs and Early MVs, even though have not yet reached their full potential such
the marginal effect point estimates provide that the released varieties with “high-
suggestive evidence of this. This is especially temperature tolerance traits” can still be
important given that yields of abiotic stress tol- improved.20 For example, molecular breeding
erant varieties (e.g., drought- and/or heat- efforts that can more precisely improve heat
tolerant varieties) have been shown to outper- tolerance in the sensitive stages of rice produc-
form non-tolerant “check” varieties in stress tion is seen as a potential area for future breed-
conditions, specifically when rice breeders ing advancement, as well as incorporating
evaluated these tolerant varieties in the labo- abiotic stress tolerance traits to several other
ratory and/or field trials prior to release widely grown “mega-varieties” that already
(Dixit, Singh, and Kumar 2014, Basu, Jonger- have favorable characteristics demanded by
den, and Ruivenkamp 2017, Dar et al. 2020). farmers: high-yielding with good grain quality
Our findings are consistent with multiple (Wassmann et al. (2009), Dixit, Singh, and
alternative explanations (or implications) and Kumar 2014). In this case, continued public
lay the groundwork for future research direc- investments in rice breeding programs at inter-
tions. First, the limited statistical evidence on national centers (i.e., like IRRI) and national
the resilience of modern rice varieties to breeding institutions (i.e., such as PhilRice in
warming (relative to traditional varieties) the Philippines and BRRI in Bangladesh)
may indicate that there could be other factors could increase the likelihood of developing
constraining the effectiveness of these varie- and releasing new rice MVs that can further
ties with regard to reducing the adverse yield decrease the adverse yield effects of high tem-
effects of warming. It is possible that farmers peratures. In studying and developing these
adopting the abiotic stress tolerant varieties new abiotic stress tolerant varieties, it is impor-
did not properly utilize complementary agro- tant to also assess how the warming response
nomic management practices appropriate for of these varieties in experimental trials differ
these new varieties. For example, it could be from the warming response observed in farmer
that farmers did not adopt the fertilizer appli- fields. An in-depth evaluation of so-called
cation rates, planting dates, and irrigation “yield gaps” can help identify new bottlenecks
schedules appropriate for the new stress- that may preclude realizing the full benefit of
tolerant varieties (Dixit, Singh, and these new MVs in farmer fields (as alluded to
Kumar 2014; Haefele, Kato, and Singh 2016; in the previous paragraph) and can further
Yamano et al. 2018). Farmers may have simply guide rice breeders and agronomists on how
used past “rule of thumb” agronomic practices to enhance the impact of abiotic stress tolerant
that are more consistent with non-tolerant varieties. If the identified bottlenecks in farmer
varieties. It is also possible that extension and fields cannot be overcome, then pivoting fund-
outreach efforts were insufficient such that ing to developing climate-change-resilient
information about proper management prac-
tices that accompany the recent MVs were
not properly disseminated. There may have
20
been shortcomings in the seed sector and Note that breeding institutions like IRRI already have existing
programs that specifically aims to develop rice varieties with toler-
national extension systems in terms of making ance to high temperatures (See Jagadish, Craufurd, and
these abiotic stress tolerant varieties more Wheeler 2007; Wassmann et al. 2009; Hirabayashi et al. 2015;
accessible to farmers in areas vulnerable to IRRI 2020). For example, IRRI breeding programs screen
improved and traditional varieties (as well as wild rice species)
warming. The discussion above points to the with heat tolerance traits to see if they can serve as donors in cross-
potential importance of funding: (a) research ing programs.
Wang et al. Modern Rice Varieties and Warming Temperatures 19

agronomic management practices for non-heat Reduced Rice Harvests in India. Proceed-
tolerant varieties may be warranted. ings of the National Academy of Sciences
Last, our inability to find statistically signifi- of the United States of America 103:
cant differences in heat tolerance across MVs 19668–72.
may also be the result of limitations in our data Barker, Randolph, Randolph W Herdt, and
and study design, which eventually did not Beth Rose. 1985. The Rice Economy of
allow us to more precisely distinguish the Asia, Vol 2. Los Baños, The Philippines:
effects. For instance, the sample size of our International Rice Research Institute.
survey data is modest and may have not Basu, Soutrik, Joost Jongerden, and Guido
allowed us to find strong statistical differences Ruivenkamp. 2017. Development of the
across varietal groups. The modest sample size Drought Tolerant Variety Sahbagi Dhan:
also constrained us to only focus on warming Exploring the Concepts Commons and
effects for irrigated rice farmers in the wet sea- Community Building. International Jour-
son. Hence, it should be noted that extrapolat- nal of the Commons 11: 144–70.
ing our inferences to rainfed rice farmers Bheemanahalli, Raju, Rajendran Sathishraj,
planting in the dry season may not be appro- Jesse Tack, Lanier L Nalley, Raveendran
priate. Nevertheless, because climate change Muthurajan, and Krishna SV Jagadish.
is likely to cause more damage to rice grown 2016. Temperature Thresholds for Spike-
in the dry season, we interpret our estimated let Sterility and Associated Warming
results as a lower bound of the warming Impacts for Sub-Tropical Rice. Agricul-
impacts across rice varietal groups. In addi- tural and Forest Meteorology 21: 122–30.
tion, the relatively modest survey sample Brennan, John P., and Arelene Malabayabas.
restricted our analysis to more parsimonious 2011. International Rice Research Insti-
models rather than developing more flexible tute’s Contribution to Rice Varietal Yield
models. Future work may consider these kinds Improvement in Southeast Asia. ACIAR
of extensions. Impact Assessment Series Report No. 74.
Furthermore, the weather data used in the Canberra: Australian Center for Interna-
study were only at the municipal level (rather tional Agricultural Research.
than at the farmer level or lower levels of Butler, Ethan E, and Peter Huybers. 2013.
aggregation). Future studies may consider col- Adaptation of US Maize to Temperature
lecting individual farm-level weather data to Variations. Nature Climate Change 3:
improve inferences going forward. Collecting 68–72.
individual information on other weather vari- Chen, Huang, Jinxia Wang, and Jikun Huang.
ables like radiation and vapor pressure deficit 2014. Policy Support, Social Capital, and
(VPD) may also be important in better under- Farmers’ Adaptation to Drought in
standing rice yield effects under climate change China. Global Environmental Change 24:
in the future (Krishnan and Rao 2005, Welch 193–202.
et al. 2010, Gourdji et al. 2013). Moreover, con- Comer, Neil, Adam Fenech, and Bill Gough.
ducting the analysis in this study for other 2007. Selecting a Global Climate Model
countries with more variability in weather pat- for Understanding Future Scenarios of
terns may also be a potential fruitful avenue for Climate Change. In Linking Climate
future research. Models to Policy and Decision-Making,
ed. Adam Fenech and James MacLellan,
133–45. Toronto, ON: Environment
Canada.
Supplementary Material Dar, Manzoor H, Showkat A Waza, Sarvesh
Shukla, Najam W Zaidi, Swati Nayak,
Supplementary material are available at Amer- Mosharaf Hossain, Arvind Kumar,
ican Journal of Agricultural Economics online. Abdelbagi M Ismail, and Uma S Singh.
2020. Drought Tolerant Rice for Ensuring
Food Security in Eastern India. Sustain-
References ability 12: 1–17.
Deressa, Temesgen Tadesse, Rashid M
Auffhammer, Maximilian, V Ramanathan, Hassan, Claudia Ringler, Tekie Alemu,
and Jeffrey R Vincent. 2006. Integrated and Mahmud Yesuf. 2009. Determinants
Model Shows that Atmospheric Brown of Farmers’ Choice of Adaptation
Clouds and Greenhouse Gases Have Methods to Climate Change in the Nile
20 Month 2021 Amer. J. Agr. Econ.

Basin of Ethiopia. Global Environmental Hayami, Yujiro, and Keijiro Otsuka. 1994.
Change 19: 248–55. Beyond the Green Revolution: Agricultural
Di Falco, Salvatore, Marcella Veronesi, and Development Strategy into the New Century.
Mahmud Yesuf. 2011. Does Adaptation Wallingford, UK: CAB International.
to Climate Change Provide Food Secu- Hijmans, Robert J. 2015. Package: Raster.
rity? A Micro-Perspective from Ethiopia. Available at: https://cran.r-project.org/
American Journal of Agricultural Eco- web/packages/raster/raster.pdf. Accessed
nomics 93: 829–46. June 2018.
Dixit, Shalabh, Anshuman Singh, and Arvind Hijmans, Robert J, Susan E Cameron, Juan L
Kumar. 2014. Rice Breeding for High Parra, Peter G Jones, and Andy Jarvis.
Grain Yield under Drought: A Strategic 2005. Very High Resolution Interpolated
Solution to a Complex Problem. Interna- Climate Surfaces for Global Land Areas.
tional Journal of Agronomy 2014: 1–15. International Journal of Climatology 25:
Estudillo, Jonna P, and Keijiro Otsuka. 2006. 1965–78.
Lessons from Three Decades of Green Hirabayashi, Hideyuki, Kazuhiro Sasaki,
Revolution in the Philippines. Developing Takashi Kambe, Ritchel B Gannaban,
Economies 44: 123–48. Monaliza A Miras, Merlyn S Mendioro,
Fan, Shenggen, Connie Chan-Kang, Keming Eliza V Simon, et al. 2015. qEMF3, A
Qian, and K Krishnaiah. 2005. National Novel QTL for the Early-Morning Flow-
and Internationa Agricultural Research ering Trait from Wild Rice, Oryza offici-
and Rural Poverty: The Case of Rice nalis, to Mitigate Heat Stress Damage at
Research in India and China. Agricultural Flowering in Rice, O. sativa. Journal of
Economics, 33: 369–79. Experimental Botany 66: 1227–36.
FAOSTAT. 2019. FAOSTAT data. Available Huang, Jikun, Yangjie Wang, and Jinxia
at: http://www.fao.org/faostat/en. Wang. 2015. Farmers’ Adaptation to
Accessed February 12, 2019. Extreme Weather Events through Farm
Fick, Stephen E, and Robert J Hijmans. 2018. Management and Its Impacts on the Mean
WorldClim 2: New 1-Km Spatial Resoul- and Risk of Rice Yield in China. American
tion Climate Surfaces for Global Land Journal of Agricultural Economics 97:
Areas. International Journal of Climatol- 602–17.
ogy 37: 1965–78. Iizumi, Toshichika, Masatake E Hori,
GADM. 2018. Global Administrative Areas Masayuki Yokozawa, Hiroshi Nakagawa,
v3.6: Philippines. Available at: https:// Yousay Hayashi, and Fujio Kimura.
gadm.org/download_country_v3.html. 2006. Impact of Global Warming on Rice
Accessed June 2018. Production in Japan Based on Five
Gourdji, Sharon M, Ky L Mathews, Matthew Coupled Atmosphere-Ocean GCMs.
Reynolds, José Crossa, and David B SOLA 2: 156–9.
Lobell. 2013. An Assessment of Wheat IRRI. 2020. Climate-Change Ready Rice.
Yield Sensitivity and Breeding Gains in Available at: https://www.irri.org/climate-
Hot Environments. Proceedings of the change-ready-rice. Accessed July 2020.
Royal Society of London B: Biological Sci- Jagadish, SV Krishna, Peter Q Craufurd, and
ences 280: 20122190. Tim R Wheeler. 2007. High Temperature
Haefele, Stephan M, Yoichiro Kato, and Stress and Spikelet Fertility in Rice
Sudhanshu Singh. 2016. Climate Ready (Oryza sativa L.). Journal of Experimental
Rice: Augmenting Drought Tolerance Botany 58: 1627–35.
with Best Management Practices. Field Kawasaki, Kentaro, and Shinsuke Uchida.
Crops Research 190: 60–9. 2016. Quality Matters More than Quan-
Harris, Ian, Timothy J Osborn, Phil Jones, and tity: Asymmetric Temperature Effects on
David Lister. 2018. Version 4 of the CRU Crop Yield and Quality Grade. American
TS Monthly High-Resolution Gridded Journal of Agricultural Economics 98:
Multivariate Climate Dataset. Scientific 1195–209.
Data 7: 1–18. Krishnan, Prameela, and AV Surya Rao. 2005.
Hasan, M Mehedi, Md Abdur Rashid Sarker, Effects of Genotype and Environment on
and Jeff Gow. 2016. Assessment of Climate Seed Yield and Quality of Rice. Journal of
Change Impacts on Aman and Boro Rice Agricultural Science 143: 283–92.
Yields in Bangladesh. Climate Change Eco- Laborte, Alice G, Mary Anne Gutierrez, Jane
nomics 7: 1650008S. Girly Balanza, Kazuki Saito, Sander J
Wang et al. Modern Rice Varieties and Warming Temperatures 21

Zwart, Mirco Boschetti, MVR Murty, and Grain Quality of Aromatic and
et al. 2017. RiceAtlas, A Spatial Database Non-Aromatic Rice in Northwestern
of Global Rice Calendars and Production. India. Agriculture, Ecosystems and Envi-
Scientific Data 4: 170074. ronment 138: 274–81.
Laborte, Alice G, Neale C Paguirigan, Piedad Nguyen, Duc-Nhuan, Kyu-Jong Lee, Da-Ik
F Moya, Andrew Nelson, Adam H Kim, Nguyen Tuan Anh, and Byun-Woo
Sparks, and Glenn B Gregorio. 2015. Lee. 2014. Modeling and Validation of
Farmers’ Preference for Rice Traits: High-Temperature Induced Spikelt Steril-
Insights from Farm Surveys in Central ity in Rice. Field Crops Research 156:
Luzon, Philippines, 1966-2012. PLoS 293–32.
One 10: e0136562. Nigatu, Getachew, James Hansen, Nathan
Launio, Cheryl C, Guadalupe O Redondo, Childs, and Ralph Seeley. 2017. Sub-
Jesusa C Beltran, and Yoshinori Saharan Africa is Projected to be the
Morooka. 2008. Adoption and Spatial Leader in Global Rice Imports. Amber
Diversity of Later Generation Modern Waves Magazine: The Economics of Food,
Rice Varieties in The Philippines. Agron- Farming, Natural Resources and Rural
omy Journal 100: 1380–9. America. Available at: https://www.ers.
Lawrimore, Jay H, Matthew J Menne, Byron usda.gov/amber–waves/2017/october/
E Gleason, Claude N Williams, David B sub–saharan–africa–is–projected–to–be–
Wuertz, Russell S Vose, and Jared the–leader–in–global–rice–imports/.
Rennie. 2011. An Overview of the Global Accessed June 2020.
Historical Climatology Network Monthly Otsuka, Keijiro, Fe Gascon, and Seki Asano.
Mean Termperature Data Set Version 3. 1994. ‘Second-Generation’ MVs and the
Journal of Geophysical Research: Atmo- Evolution of the Green Revolution: The
spheres 116: 1–18. Case of Central Luzon, 1966–1990. Agri-
Lobell, David B, Wolfram Schlenker, and cultural Economics 10: 283–95.
Justin Costa-Roberts. 2011. Climate Pandey, Sushil, Devendra Gauchan, Maria L
Trends and Global Crop Production since Malabayabas, Maria Romilee Bool-
1980. Science 333(6042): 616–20. Emerick, and Bill Hardy. 2012. Patterns
Lyman, Nathaniel B, Krishna SV Jagadish, L of Adoption of Improved Rice Varieties
Lanier Nalley, Bruce L Dixon, and Terry and Farm-Level Impacts in Stress-Prone
Siebenmorgen. 2013. Neglecting Rice Rainfed Areas in South Asia. Los Banos,
Milling Yield and Quality Underestimates Philippines: International Rice Research
Economic Losses from High- Institute.
Temperature Stress. PLoS One 88: 1–9. Pattanayak, Anubhab, and KS Kavi Kumar.
Mosier, Thomas M, David F Hill, and Kendra 2014. Weather Sensitivity of Rice Yield:
V Sharp. 2014. 30-Arcsecond Monthly Evidence from India. Climate Change
Climate Surfaces with Global Land Cov- Economics 5: 1–24.
erage. International Journal of Climatol- Peng, Shaobing, Jianliang Huang, John E
ogy 34: 2175–88. Sheehy, Rebecca C Laza, Romeo M
. 2018. Update to the Global Climate Visperas, Xuhua Zhong, Grace S
Data Package: Analysis of Empirical Bias Centeno, Gurdev S Khush, and Kenneth
Correction Methods in the Context of G Cassman. 2004. Rice Yields Decline
Producing Very High Resolution Climate with Higher Night Temperature from
Projections. International Journal of Cli- Global Warming. Proceedings of the
matology 38: 825–40. National academy of Sciences of the United
Moya, Piedad, Kei Kajisa, Randolph Barker, States of America 101: 9971–5.
Samarendu Mohanty, Fe Gascon, and Sarker, Md Abdur Rashid, Khorshed Alam,
Mary Rose San Valentin. 2015. Changes and Jeff Gow. 2012. Exploring the Rela-
in Rice Farming in the Philippines: Insights tionship Between Climate Change and
from Five Decades of a Household-Level Rice Yield in Bangladesh: An Analysis
Survey. Los Baños, The Philippines: Inter- of Time Series Data. Agricultural Systems
national Rice Research Institute. 112: 11–6.
Nagarajan, Shantha, SVK Jagadish, AS Hari Schlenker, Wolfram, and David B Lobell.
Prasad, AK Thomar, Anjali Anand, 2010. Robust Negative Impacts of Climate
Madan Pal, and PK Agarwal. 2010. Change on African Agriculture. Environ-
Local Climate Affects Growth, Yield mental Research Letters 5: 014010.
22 Month 2021 Amer. J. Agr. Econ.

Schlenker, Wolfram, and Michael J Roberts. topics/crops/rice/rice-sector-at-a-glance/.

2009. Nonlinear Temperature Effects Accessed June 2020.
Indicate Severe Damages to US Crop Wang, Jinxia, Robert Mendelsohn, Ariel
Yields Under Climate Change. Proceed- Dinar, and Jikun Huang. 2010. How Chi-
ings of the National Academy of Sciences nese Farmers Change Crop Choice to
of the United States of America 106: Adapt to Climate Change. Climate
15594–8. Change Economics 1: 167–85.
Stuecker, Malte F, Michelle Tigchelaar, and Wassmann, Reiner, SVK Jagadish, Sigrid
Michael B Kantar. 2018. Climate Variabil- Heuer, A Ismail, E Redona, Rachid
ity Impacts on Rice Production in the Phil- Serraj, RK Singh, G Howell, H Pathak,
ippines. PLoS One 13: 1–17. and K Sumfleth. 2009. Climate Change
Tack, Jesse, Andrew Barkley, and Lawton Affecting Rice Production: The Physiolog-
Lanier Nalley. 2015. Estimating Yield ical Basis for Possible Adaptation Strate-
Gaps with Limited Data: An Application gies. Advances in Agronomy 101: 59–122.
to United States Wheat. American Journal Welch, Jarrod R, Jeffrey R Vincent,
of Agricultural Economics 97: 1464–77. Maximilian Auffhammer, Piedad F
Tack, Jesse, Andrew Barkley, Trevor W Rife, Moya, Achim Dobermann, and David
Jesse A Poland, and Lawton Lanier Dawe. 2010. Rice yields in Tropical/Sub-
Nalley. 2016. Quantifying Variety-Specific tropical Asia Exhibit Large But Opposing
Heat Resistance and the Potential for Sensitivities to Minimum and Maximum
Adaptation to Climate Change. Global Temperatures. Proceedings of the
Change Biology 22: 2904–12. National Academy of Sciences of the
Tolentino, Pamela Louise M, Ate Poortinga, United States of America 107: 14562–7.
Hideki Kanamaru, Saskia Keesstra, Jerry Yamano, Takashi, Manzoor H Dar, Architesh
Maroulis, Carlos Primo C David, and Panda, Ishika Gupta, Maria L
Coen J Ritsema. 2016. Projected Impact Malabayabas, and Eric Kelly. 2018.
of Climate Change on Hydrological Impact and Adoption of Risk-Reducing
Regimes in the Philippines. PLoS One Drought-Tolerant Rice in India. Impact
11: e0163941. Evaluation Report 72. International Ini-
USDA-ERS. 2020. Rice Sector at a Glance. tiative for Impact Evaluation (3ie), New
Available at: https://www.ers.usda.gov/ Delhi, India.