Effects of Temperature On Agricultural Input Decisions in Moderate Climates

In the Weeds: Effects of Temperature on Agricultural Input

Decisions in Moderate Climates

C. Barrett¹; M. Jagnani²; Y. Liu³; L. You³

1: Cornell University, Charles H Dyson School of Applied Economics & Management, United
States of America, 2: Cornell University, Dyson School of Applied Economics and Management,
United States of America, 3: International Food Policy Research Institu
Corresponding author email: [email protected]
Abstract:
Does heat affect agriculture in regions where temperatures are not high enough to directly, adversely affect
crop growth? Combining daily weather data with a qualitatively rich, longitudinal survey of Kenyan
agricultural households in rural maize-growing areas where daily average temperatures are well below
30C, we find that higher temperatures early in the growing season increase the use of pesticides, while
reducing fertilizer use, with comparatively modest effects of temperature later in the growing season.
Suggestive evidence indicates that greater heat increases the incidence of pests, crop diseases and weeds,
compelling farmers to divert investment from productivity-enhancing technologies like fertilizer to
adaptive, loss-reducing inputs like pesticides.
Acknowledegment: Tim Johnson and Yating Ru provided excellent research assistance. We thank Nicholas
Flores, Teevrat Garg, Ariel Ortiz Bobea, Cynthia Lin Lawell and Vis Taraz for helpful comments. We are
grateful for feedback from seminar participants at Cornell University and the International Food Policy
Research Institute, as well as conference participants at the 2017 CU Boulder -- Environmental and
Resource Economics Workshop, and the 2017 Association of Environmental and Resource Economists
Annual Meetings. Funding support came from the African Development Bank through the Structural
Transformation of African Agriculture and Rural Spaces (STAARS) project. All errors are our own
JEL Codes: Q54, Q16
#605

In the Weeds: Effects of Temperature on Agricultural Input
Decisions in Moderate Climates
This version: January 7, 2018

Abstract
Does heat affect agriculture in regions where temperatures are not high enough to di-
rectly, adversely affect crop growth? Combining daily weather data with a qualitatively
rich, longitudinal survey of Kenyan agricultural households in rural maize-growing ar-
eas where daily average temperatures are well below 30C, we find that higher tem-
peratures early in the growing season increase the use of pesticides, while reducing
fertilizer use, with comparatively modest effects of temperature later in the growing
season. Suggestive evidence indicates that greater heat increases the incidence of pests,
crop diseases and weeds, compelling farmers to divert investment from productivity-
enhancing technologies like fertilizer to adaptive, loss-reducing inputs like pesticides.
JEL Codes: O13, Q15, Q16, Q54

Keywords: climate change, technology adoption, agriculture
1 Introduction
Agricultural livelihoods in developing countries are especially fragile in the face of climate
change (e.g. Rosenzweig and Parry, 1994; Mendelsohn, 2008). The effects of climate change
on agriculture, however, may be spatially heterogeneous (e.g., The World Bank, 2010; Zhao
et al., 2017). An interesting issue in this context concerns whether global warming adds
to the weather risk faced by agricultural households in moderate climates, where the like-
lihood of crop physiological stress from higher temperature is low–indeed modestly higher
temperature might boost yields–especially in regions that have traditionally cultivated heat-
resistant crops, like maize.1 If, however, heat affects farmers’ input use, there might exist
indirect non-physiological pathways by which temperature increases associated with climate
change will affect farmers. This paper tries to make progress on this issue by exploring a
household-level panel of rural maize-growers in Kenya (Figure 1).
Almost 50% of households in this sample noticed a change in temperature in the last 10
years, and over 80% of those households indicated that they were affected by said change
(Table 1). Such responses are significant since average daily temperatures in maize growing
agro-ecological zones of Kenya range from 12-29C (Figure 2), and because maize yields
only decline due to physiological stress above 29-30C (Lobell et al., 2011; Schlenker and
Roberts, 2009). Large swaths of maize farms in Africa are in similar agro-ecological zones
(Figure 3). Since these households are unlikely to experience physiological heat stress in
maize itself, qualitative data suggests implications for agricultural livelihoods beyond the
direct, physiological relationship between temperature and maize output, and illustrates the
empirical questions that motivate this paper: might climate change affect smallholder farmer
productivity via pathways other than crop physiology?
Qualitative data and significant ecological literature (e.g., Coakley, Scherm and Chakraborty,
1999; Garrett et al., 2006; Chakraborty, 2008; Patterson et al., 1999; Rosenzweig et al., 2001)
suggest that higher temperatures increase the incidence of pests, weeds and crop diseases.
Perhaps not coincidentally, we find that higher temperatures early in the growing season, in
the pre-planting and initial vegetative growth stages, increase pesticide use. Moreover, we
1 Heat-resistant or C4 crops have a greater ability for photosynthetic acclimation to heat stress than C3 crops, like rice, in
which photosynthesis is inhibited even at moderately high temperatures. Other examples of C4 crops include sorghum, pearl
millet, sugar cane, finger millet and pasture grasses.
1
show that higher temperatures early in the growing season reduce the uptake of fertilizer.
Presumably this occurs due to binding financial constraints or changing risk profile of fu-
ture harvest. High temperatures confront farmers with a trade-off, inducing them to divert
spending from yield-increasing fertilizer to loss-reducing pesticides and weeding.2
Our analysis combines daily temperature data at the village-level with a household level
panel survey of Kenya’s maize growing regions. We find that a 10% increase in degree days
over 21C during the initial vegetative growth stage increases pesticide use on the extensive
margin by around 10 percent, and reduces uptake of fertilizer by approximately 2 percent,
compared to the baseline. On the intensive margin, a 10% increase in degree days over 21C
increases the intensity of pesticide use by 15 percent, while reducing quantity of fertilizer
used by over 5 percent.
In this paper, we connect two distinct literatures. The first is an environmental and
agricultural economics literature that examines the relationship between temperature and
agriculture. We make three contributions to this literature. One, existing studies have almost
exclusively focused on estimating the reduced form relationship between temperature and
agricultural output or yield (e.g., Deschênes and Greenstone, 2007; Guiteras, 2008; Feng,
Krueger and Oppenheimer, 2010; Welch et al., 2010). Yields are the joint product of crop
physiological response to higher temperatures holding input use constant, and the effects of
induced change in input application patterns on crop yields. The first effect is well established
in environmental science (e.g., Lobell and Burke, 2008; Schlenker and Lobell, 2010; Lobell
et al., 2011). We establish that high temperatures can also affect agricultural input use,
both loss-reducing and productivity-enhancing.
Two, we document an ecological channel, beyond the physiological relationship be-
tween heat and crop growth. The dependence of plant diseases and pests on weather has
been well-known amongst plant pathologists and entomologists (e.g., Coakley, Scherm and
Chakraborty, 1999; Garrett et al., 2006; Chakraborty, 2008). In fact, the broader ecological
literature concludes that climate change will increase challenges to agriculture from pests,
weeds and diseases, in part due to higher than normal temperatures (e.g., Patterson et al.,
1999; Rosenzweig et al., 2001). However, as far as we know, we are first to provide eco-
2 Fungicides, herbicides and insecticides are all pesticides.
2
nomic evidence for such an ecological channel while examining the temperature-agriculture
relationship.
Three, we estimate the impact of heat during each stage of the crop growth cycle sep-
arately, and highlight that farmers are extremely quick to adopt adaptation strategies in
response to changing agro-ecology driven by higher temperatures. The existing literature
has often inferred agricultural adaptation by using cross-sectional variation to compare out-
comes in hot versus cold areas (e.g., Mendelsohn, Nordhaus and Shaw, 1994; Schlenker,
Hanemann and Fisher, 2006), or by comparing estimates from annual temperature fluctua-
tions for a given area under hotter versus cooler conditions (e.g., Deschênes and Greenstone,
2007; Deschênes and Greenstone, 2011 Schlenker and Roberts, 2009; Dell, Jones and Olken,
2012; Taraz, 2017), or by differentiating estimates from annual temperature fluctuations with
long-run impacts of higher than normal temperatures (e.g., Burke and Emerick, 2016). Such
approaches rely on aggregate data at the county- or district-year level, and fail to explicitly
examine adaptive behavior. In this paper, using household-level data, and by disaggregating
temperatures in the growing season by different stages of the crop growth cycle, we combine
the intent of agroeconomic models with panel data analyses to observe adaptation strate-
gies in the short-run. We demonstrate that farmers adjust purchased inputs after observing
temperatures pre-planting or early in the growing season.3
The second literature to which we contribute is in development economics that stud-
ies the determinants of agricultural technology use.4 This literature provides a number
of explanations for low adoption of modern agricultural inputs: learning (e.g., Foster and
Rosenzweig, 1995; Conley and Udry, 2010), insurance or credit constraints (e.g., Moser and
Barrett, 2006; Dercon and Christiaensen, 2011; Cole et al., 2013), heterogeneity in returns
(e.g., Marenya and Barrett, 2009; Suri, 2011) and behavioral anomalies (e.g., Duflo, Kremer
and Robinson, 2011). Our findings suggest that increase in temperatures associated with
global warming will also reduce adoption of productivity-enhancing technologies like fertil-
3 Few papers have investigated how farmers adjust their inputs to adapt to higher temperatures. However, these studies
either examined effects on irrigation (Kurukulasuriya and Mendelsohn, 2007; Kurukulasuriya, Kala and Mendelsohn, 2011;
Oehninger, Lawell and Springborn, 2017; Seo and Mendelsohn, 2008a), or crop mix (Kurukulasuriya and Mendelsohn, 2008;
Seo and Mendelsohn, 2008b). Furthermore, a small number of these use household or farm level data, but none of these papers
disaggregate growing season temperatures by different stages of the crop growth cycle. Therefore, these studies only observe
aggregate response to growing season, annual, or longer-run temperature.
4 Feder, Just and Zilberman (1985) and Foster and Rosenzweig (2010) provide a comprehensive overview of this literature.
3
izer, in favor of loss-reducing adaptive inputs like pesticides, presumably due to the financial
constraints faced by poor farmers in developing countries.
2 Data
We use a qualitatively rich, household level panel data set, representative of maize farmers in
Kenya’s main crop cultivating provinces. We augment these with detailed village level data
on daily weather variables including temperature, rainfall, humidity and soil moisture.
2.1 Household Data
The household panel survey data are representative of the main maize-growing areas in
Kenya. The survey was designed and implemented under the Tegemeo Agricultural Moni-
toring and Policy Analysis Project (TAMPA), a collaboration among Tegemeo Institute of
Egerton University, Michigan State University, and the Kenya Agricultural Research Insti-
tute. Figure 1 maps the survey villages across Kenya. These villages were selected randomly
from each of eight predetermined agro-economic zones and then households were sampled
randomly from each selected village. We use data from a balanced panel of 1242 households
collected over five rounds: 1996-97, 1999-00, 2003-04, 2006-07, and 2009-10. The survey in-
cludes detailed agricultural input and output data, demographics, credit and infrastructure
information. The 2009-10 round collected rich subjective data on farmers perceptions of the
impacts of ‘climate change’, as well as reasons for non-adoption of fertilizer. Villages were
geo-referenced, allowing us to merge the household data with daily temperature, precipita-
tion, relative humidity and soil moisture data at the village level as well as agro-ecological
zone crop calendars.
Figure 4 shows both uptake and intensity of pesticide use; these detailed data were only
collected in 2003-04, 2006-07, and 2009-10. While answering questions on inputs, respon-
dents often used pesticides and specific pests, weeds and disease repellents (e.g., herbicide,
insecticide, fungicide) interchangeably. Therefore, our measure of pesticide use: takes the
binary value of 1 if a farmer uses any chemical or biological agent that protects crops from
pests, weeds or crop diseases, 0 otherwise. Almost 30% of households in our balanced panel
4
adopted some variety of pesticides in 2003, use then increased to 65% in 2006, before some-
what dropping to 50% in 2009. The average maize farmer used 0.25 kg/acre of pesticides in
2003, increasing to over 0.5 kg/acre by 2009. Figure 5 shows the average number of labor
days spent in weeding activities. Figure 6 depicts fertilizer use in the main growing season,
1997-2010. Fertilizer use is high amongst maize farmers in rural Kenya. In 1997, almost 65%
of households used fertilizer, while the corresponding figure is 75% for 2010. The average
maize farmer used around 45 kgs/acre in 1997, average quantity use then increased to over
55 kgs/acre in 1999, before dropping to 50 kgs/acre in 2009.
Figure 7 shows average expenditure (cash or credit) on fertilizer, pesticides and hired
labor for manual weeding, conditional on use, by adopters of each ‘technology’, respectively.
The mean expenditure on pesticides and hired-labor for weeding by adopters is over 30% of
the mean expenditure on fertilizer by fertilizer adopters.
Figure 8 shows average maize yields over time, while Figures 9 and 10 show the distri-
bution of maize yields, pooled for all rounds, by technology. Ignoring possible selection bias,
these suggest that maize yields are higher, and variance lower for fertilizer and pesticide
users, indicating first order stochastic dominance of use of these modern purchased inputs.
Table A.1 presents summary statistics for our balanced sample from 1997-2010. Finally, Ta-
bles 2 and 3 show household-level transitions of pesticide and fertilizer use in the data, with
30% (60%) of households switching in or out of fertilizer (pesticide) use across survey rounds.
So there is clearly considerable across round variation in input use patterns by Kenyan maize
farmers around the broader trend of expanding purchased input use over time.
2.2 Kenyan Maize Calendar
To uncover the underlying mechanisms that influence farmer climate adaptation strategies,
and related spillover effects on productivity-enhancing inputs, we need to disaggregate the
main growing season. So as to parse the information set available to farmers as they make
season-specific input use choices. We use maize crop calendars specific to each agro-ecological
zone (AEZ) in Kenya, broken into three distinct stages of the agricultural cycle.5 This
calendar gives the usual start and end dates of the planting period and harvest period for
5 The maize calendar was downloaded from http://www.fao.org/agriculture/seed/cropcalendar/welcome.do.
5
each AEZ and for long and short rainy seasons. We use the calendar for long rainy season
(main growing season). We define as the ‘pre-planting’ period the two months right before
the start of the planting or basal fertilizer application during which land preparation occurs.6
We define the four to six weeks right after the start of the planting period as the post-planting
period; this is also the recommended period for top dressing application of fertilizer. Thus,
the three phases of the main agricultural cycle are: 1) ‘PP’: land preparation period (from
onset of pre-planting to onset of planting) 2) ‘GS1’: planting and basal fertilizer application
period (from onset of planting to onset of top dressing fertilizer application), and 3) ‘GS2’:
post-planting and top dressing fertilizer application period (after top dressing fertilizer to
onset of harvest).
2.3 Weather Data
Because of incomplete coverage of ground weather stations in Kenya, we use daily tem-
perature, precipitation, relative humidity and soil moisture data from various gridded and
satellite data sets. Daily temperature data are the land surface temperature from the Noah
2.7.1 model in Global Land Data Assimilation System (GLDAS).7 The data are in 0.25 de-
gree resolution, from 1990. The temporal resolution is three-hour (Rodell et al., 2004). A
point shapefile for each village in the TAMPA sample was used to generate the value of each
point for each daily temperature pixel it intersects with. We generated a table of every date
and the temperature values of the points for each village coordinate point for every day in
a year. Similarly, we generated daily precipitation data from the Climate Hazards Group
InfraRed Precipitation Station (CHIRPS) data set of daily 0.05 degree resolution grided data
for all of Africa.8 Daily relative humidity data came from NASA.9 These satellite and model
derived solar and meteorological data cover the global surface at 1 by 1 degree resolution.
Lastly, daily soil moisture data are sourced from the European Space Agency. This global
soil moisture data set has been generated using active and passive microwave spaceborne
instruments and covers the 37 year period from 1978 to 2015. It provides daily surface soil
6 Please see http://nafis.go.ke/agriculture/maize/establishment-of-maize/ for recommendation on land preparation and
http://www.nafis.go.ke/agriculture/maize/field-management-practices/ for recommendation on fertilizer application.

7 The data are located on the OPENDAP NASA web server as GRIB and netcdf files.
8 CHIRPS was downloaded from http://chg.geog.ucsb.edu/data/chirps/
9 The relative humidity data are from https://power.larc.nasa.gov/cgi-bin/agro.cgi?na
6
moisture with a spatial resolution of 0.25 degrees.10
From daily data, we generate aggregate weather indicators for each stage of the crop
growth cycle, across five rounds of the TAMPA data. For our primary variable of interest,
temperature, we use the concept of cumulative growing degree days (GDD), which measures
the intensity of daily exposure to temperatures above a lower bound and below an upper
bound at which heat stress might begin, to estimate the effects of temperature on fertilizer
and pesticide use, as well as weeding labor days.11 We use daily average temperatures to
calculate the number of days each village is exposed to temperatures above a lower bound
and below an upper bound, and then sum these daily exposures for each of the three phases
during the main growing season for those bounds.12 Figure 4 shows the distribution of daily
average temperatures in each phase of the agricultural cycle during the main growing season
for all villages in TAMPA data. Table 4 present summary statistics for GDD above 21C in
each phase of the agricultural cycle for all five rounds of the household survey. For each phase
of the agricultural cycle, there exists substantial variation in degree days across households
in each round, as well as significant round-on-round variation in GDD for all households in
the TAMPA data.
3 Background
3.1 Temperature, Pests, Weeds and Pesticides
In the TAMPA survey, almost 40% of maize-farmers affected by changes in temperature

pointed to an increase in incidence of pests, weeds and crop diseases as one of the primary
consequences of such changes (Table 5). Furthermore, another almost 45% of survey re-
spondents reported decline in yields, which is counterintuitive given that in these temperate
AEZs, additional GDDs should increase not decrease maize yields physiologically. This sug-
gests that some other mechanisms–such as disease, insects or weed pressure–was experienced,
resulting in increased crop losses, that is, lower net yields. Unfortunately, our data do not
10 The soil moisture data are downloaded from http://www.esa-soilmoisture-cci.org/node/145
11 Past literature has demonstrated the relationship between temperature and agricultural outcomes using GDDs (e.g., Lobell
et al., 2011; Schlenker and Lobell, 2010; Schlenker and Roberts, 2009; Schlenker, Hanemann and Fisher, 2006).
12 We have a pre-determined upper bound, since average daily temperatures in the data are less than 30C.
7
contain information on incidence of specific crop diseases or pests. But the ecological lit-
erature broadly concurs that climate change will increase disease, pest and weed pressure.
We look to existing studies that examine plant pathology and crop management to better
understand the effects of heat stress on diseases and pests specific to maize.
Grey leaf spot is a major maize disease in Kenya. It was first reported in Kenya during
1995. Small-scale farmers have continued to experience considerable yield losses from grey
leaf spot in Kenya (Simons, 2006). Infection and growth of grey leaf spot are most likely
to occur following a humid and warm period. Specifically, at 100% relative humidity, the
optimum temperature for sporulation is between 25-30C. Similarly, the highest rates of lesion
expansion were observed at 25C and 30C (Paul and Munkvold, 2005). Moreover, experiments
indicate that fungicide treatment should be initiated after the disease was observed but before
high levels were present (Ward, Laing and Rijkenberg, 1997). So higher temperatures from
the historical range in our sites should increase gray leaf spot incidence and induce early
season adaptive responses by farmers.
Insect behavior, distribution, development and survival are strongly coupled with envi-
ronmental conditions, especially temperature, since insects do not use their metabolism, but
rather depend on ambient temperature to control their body temperature. Global warm-
ing will favor insect proliferation and increase the incidence and severity of insect-related
damages in maize (Cairns et al., 2012).
The most common insect maize pest in Kenya is the stem borer. Damage caused by
stem borers is one of the main causes of low maize yields (Songa, Guofa and Overholt, 2001).
Female stem borer moths lay eggs on maize leaves. The newly emerged larvae enter into
the whorls of young maize plants and feed actively on the tender leaves. Later, the larvae
bore into the stem and start tunneling. Stem borers can be controlled by applications of
insecticides to the leaf whorl early in crop growth cycle to kill early larval instars; this method
has limited effectiveness once the larvae bore into the stem (Gianessi, 2014). So as with gray
leaf spot disease, the stem borer pest pressure on maize in Kenya should increase with higher
temperatures inducing early season response through pesticide application.
Weeds compete with crops for nutrients, moisture, light and space, adversely affect-
ing crop yields. Weed control during the first eight weeks after planting is crucial because
8
weeds compete vigorously with the maize crop for nutrients and water during this period
(du Plessiss, 2003). So weeding continues beyond the period of fertilizer top dressing. Specif-
ically, maize fields should be kept weed-free for the first 56 days after planting to achieve
maximum yields (I. O Akobundu, 1987). One week’s delay in first weeding may reduce maize
yields by as much as one-third (Orr, Mwale and Saiti, 2002). Herbicides can be used before
planting and during the growing season (Gianessi, 2010). Weed growth is also influenced
by abiotic conditions such as temperature and humidity (Dukes et al., 2009; Peters, Breit-
sameter and Gerowitt, 2014; Singer, Travis and Johst, 2013). For instance, milder winters
are likely to increase the survival of some winter annual weeds, whereas warmer summers
may allow other type of weeds to grow in regions previously inhospitable (Bloomfield et al.,
2006; Hanzlik and Gerowitt, 2012; Walck et al., 2011). As with maize disease and pests,
higher temperatures are thus expected to induce greater weed competition with crops, forc-
ing farmers to devote more labor and pesticides to combating weeds. That would continue
later in the season through disease and pest reduction efforts. These predictions from the
agro-ecological literature mirror what we find in the data.
3.2 Fertilizer Use Under Liquidity Constraints
Almost 60% of all non-adopters of fertilizer pointed to financial liquidity constraints as the
reason for non-adoption, while a little less than 40% claimed they didn’t need to use fertilizer,
which could reflect either highly fertile or poor, non-responsive soils (Table 6). Higher than
normal temperatures increase the prevalence of pests and diseases, plausibly forcing farmers
to divert resources from productivity-enhancing technologies like fertilizer, towards expendi-
ture on adaptive loss-reducing inputs like pesticides (e.g., herbicides, insecticides, fungicides)
or labor for manual weeding. Such effects on fertilizer uptake might be driven by ex ante
credit constraints faced by poor farmers, or alternatively, farmers might anticipate increased
risk of crop losses due to an increase in pest pressure, and opt for less risky technologies
to avoid lasting damage. For instance, Binswanger and Rosenzweig (1993), show that poor
farmers facing increased rainfall variability tend to hold a portfolio that is less influenced by
rainfall, although wealthier farmers facing varying exposure to risk do not exhibit changing
portfolios of investments. More recently and nearby, Dercon and Christiaensen, 2011, find
9
that both ex ante credit constraints and the possibility of low consumption outcomes when
harvests fail discourage the application of fertilizer in Ethiopia. Marenya and Barrett (2007)
find that enhanced financial liquidity due to non-farm employment boosts fertilizer uptake
among Kenyan maize farmers.
Farmers usually apply fertilizer twice on maize. Basal fertilizer applications occur at
planting. Top dressing fertilizer seldom occurs without basal fertilizer application. But if
fertilizer is used at planting, top dressing often occurs post-germination. In high rainfall ar-
eas, topdressing fertilizer is applied in two splits. The first split is done 6 weeks after planting
and the second split is done 10-15 days later or just before tasseling. In low rainfall areas
fertilizer is applied only once, typically six or so weeks after planting (NAFIS, 2011).
Thus, higher than normal temperatures in moderate climates might be expected to
increase use of pesticides and weeding labor, and reduce fertilizer use. Existing evidence in-
dicates that farmers in Africa make their decisions sequentially, adapting to new information
as it emerges (Dillon, 2014; Fafchamps, 1993). Duflo, Kremer and Robinson (2011) find that
in Western Kenya, 96-98% of farmers who used fertilizer had bought it just before applying
it. Moreover, fertilizer and pesticide adoption rates are quite high amongst maize farmers in
Kenya, suggesting that agricultural input markets are relatively well-developed compared to
other countries in Sub-Saharan Africa (Sheahan and Barrett, 2017). Thus, it is likely that
the effects of increased temperatures on input decisions will be driven by changes during
the pre-planting or early vegetative growth phase, only extending deeper into the growing
season for manual weeding as farmers can adjust labor inputs later in the season.
4 Temperature and Agricultural Input Use Response

If higher temperatures increase the incidence of pests, weeds and diseases, then farmers may
incur greater adaptive expenditures on pesticides and weeding labor, and simultaneously
reduce use of productivity enhancing fertilizer, as just explained. In this section, we show
that higher than normal temperatures indeed affect adoption of agricultural inputs in ways
consistent with the just described ecological channel. We subsequently rule out alternative
mechanisms.
10
4.1 Research Design
To examine the effect of temperature on agricultural input use, we estimate the following
model:
Yijdqt = β1 (30C > GDDP P > 21C)jdqt + β2 (30C > GDDGS1 > 21C)jdqt +
β3 (30C > GDDGS2 > 21C)jdqt + f (Rainjdqt ) + αj + µqt + ijdqt (1)
Yijdqt is fertilizer or pesticide use (a binary variable equal to one if used) for household i
in village j, in district d in province q in round t. We control for cumulative rainfall using
upper/lower terciles indicators calculated for each period in the agricultural cycle using
daily data, and include village fixed effects (αj ). We also include province-by-round fixed
effects (µqt ) to control for unobservables that vary by province over time, such as input
prices or seasonal climate forecasts. (30C > GDD > 21C)jdqt is the sum of degree days
over 21C and below 30C during each stage of the main growing season in Kenya.13 14
Thus, β1 is the marginal effect of an extra growing degree days during the pre-planting
phase, and so on for other coefficients. We cluster standard errors at the village level.
The identifying assumption is that changes in the number of degree days experienced by a
household during each phase of the agricultural cycle is exogenous to household or village
level unobservable characteristics that vary over time. The assumption is plausible given
the randomness of weather fluctuations and the inability of rural households to predict such
fluctuations beyond common spatial features such as season climate forecasts for which we
control with province-by-round fixed effects (µqt ). As robustness checks, we also control
for time invariant household level characteristics (e.g., farming skill, access to groundwater,
education, relationship with input suppliers), as well as district level attributes that vary
over time (e.g., local elections), and provide plausibly causal estimates for the effects of
temperature on agricultural input use.
13 21C is the 75th percentile of the distribution of average daily temperatures for villages in our sample, between 1990-2010.
In Section 4.2, we demonstrate that our results are robust to the choice of lower bound used to calculate growing degree days.
14 We have a pre-determined upper bound of 30C, since didn’t observe a single day with an average temperature over 30C in
the data.
11
4.2 Results
4.2.1 The Response of Pesticides Use and Weeding Days to Temperature
We estimate equation (1) and find that an extra 1 degree day above 21C in the initial growth
period (GS1) increases pesticide use by over 0.6 percentage points (Table 7: Column 1). In
2003, almost 30% of maize-farmers in TAMPA adopted pesticides. Thus our point estimates
imply that an extra 3 DDs in GS1 leads to an approximately 6 percent increase in pesticide
use. Similarly, on the intensive margin, an increase in 3 DDs in GS1 leads to a 15 percent
increase in quantity of pesticide used. Note that since pesticide application typically occurs
soon after pests are found on germinated crop, the effect should be most pronounced in GS1,
not PP or GS2. This is precisely what we find.
If greater heat increases the incidence of weeds, we should also observe an increase in
manual weeding. Indeed, we find that an extra degree day in the post-planting period (PP)
is associated with a 1% increase in weeding labor (Table 7: Column 5). Since the most
intense crop-weed competition typically occurs after top dressing, these results are precisely
what one would expect.
Combined with the qualitative evidence presented in Table 5, these results strongly
suggest an ecological mechanism beyond physiological heat stress, which should not be an
issue in these data (which we confirm below). Early growing season temperatures, in the
pre-planting and initial vegetative growth stages, increase the incidence of pests and dis-
eases, driving use of adaptive inputs like pesticides in the early crop growth stages, with no
significant impact of heat during latter stages of the growing season, by which time farmer
response is likely unproductive. Effects on weeding labor is pronounced deeper into the grow-
ing season as the ability to reverse the adverse effects of weed competition persists longer as
well. Household or hired labor can clear weeds manually if they survive initial application of
herbicides, or to tackle encroachment of weeds that arises later in the growing season, due
to higher than normal temperatures.
12
4.2.2 The Effects of Temperature on Fertilizer Use
Next, we examine effects on productivity-enhancing inputs like fertilizer. Descriptive data

indicate that pesticide adopters spend on pesticides the equivalent of 10-20% of the total
amount spent on fertilizer by fertilizer adopters. Households hiring labor for manual weeding
spend on hired labor around 25-45% of the total amount spent on fertilizer by fertilizer
adopters (Figure 7). Thus, if credit constraints bind for farmers, expenditure on loss-reducing
adaptive inputs may necessitate reduced fertilizer use. Alternatively, increased ex ante maize
yield risk due to an increase in the prevalence of pests and weeds could also adversely affect
fertilizer uptake.
We indeed find that an extra 3 DD above 21C in the initial planting or basal fertilizer
application period (GS1) decreases fertilizer use by 0.6 percentage point (Table 7: Column 3).
These effects coincide temporally with the pesticide effects observed, entirely consistent with
a liquidity constraint mechanism, moreso than a risk response. Almost 65% of households
in our balanced panel applied fertilizer in 1997, so a 0.6 percentage point decrease translates
into a 1 percent decrease from adoption levels in Round 1. Similarly, an extra 3 DD over 21C
in GS1 reduce fertilizer use by around 6 percent on the intensive margin. These effects are
driven by early growing season temperatures, by which time financial constraints typically
begin to bind. Higher temperatures before planting or late in the growing season, after
fertilizer top dressing is usually finished, do not have a significant impact on fertilizer use.
The associated point estimates are of both low magnitude and low confidence.
4.2.3 Robustness Checks
We exploit plausibly random round-by-round variation in temperature at the village level

beyond time invariant village level characteristics and time varying spatial or administrative
features, for which we control with province-by-round fixed effects, to provide plausibly
causal estimates for the effects of temperature on agricultural input use.15 After removing
village and province-by-round fixed effects, remaining temperature variation pertains only to
within-province-round deviations from village means, as for example, the amount by which
15 Including household fixed effects doesn’t affect our estimates since the treatment (temperature) is at the village level (Table
A.2).
13
western parts of Nyanza province, are warmer than normal in a given survey round compared
to how much eastern Nyanza is warmer than normal in the same round.
If provinces are large enough, it is plausible that we can control for time varying admin-
istrative features at a much smaller spatial unit like district, and still have enough variation
to precisely estimate our coefficients of interest. However, generally whenever eastern Migori,
a district in Nyanza province, is warmer than normal, so is western Migori, because tem-
peratures vary smoothly in space due to thermodynamics. Thus, it is unlikely that we have
sufficient identifying variation in temperature after removing household and district-by-year
fixed effects. We report within-province and within-district temperature variation in Table
A.3. The entries report the percentage of households by round observations with deviations
at least as large as 5 or 10 degrees, averaged over the five survey rounds. For example,
the “Removed Province*Round Effects” degree-days column indicate that 29% and 10% of
households by round observations had deviations larger than 5 and 10 degree-days in the
planting period (GS1), respectively. The corresponding percentages for the “Removed Dis-
trict*Round Effects” degree-days column are only 11% and 3%, respectively. Indeed, when
we estimate equation (1) with district-by-round fixed effects instead of province-by-round
effects we lose precision for ‘Pesticides’, although the point estimate remains relatively un-
changed (Table A.4). In fact, our point estimates for fertilizer use increase for all three peri-
ods, although the effects are still largest for GS2. Lastly, for ‘Log Weeding Days’, we observe
a substantial decrease in the GS2 DD point estimate, and a corresponding increase in GS1
DD point estimate. We have low confidence in our GS2 point estimates for fertilizer as well
as weeding labor because there isn’t sufficient variation in GS2 DD after removing district-
by-round fixed effects. The “Removed Province*Round Effects” (“Removed District*Round
Effects”) degree days column indicates that 13% (6%) and 4% (3%) of households by round
observations had deviations larger than 5 and 10 degree days in the post-planting period,
respectively.
A sizable proportion of households, across rounds, did not use fertilizer, pesticides, or
use weeding labor. Thus, limited dependent variable models might be appropriate for esti-
mating the effect of temperature on intensity of input use. However, fixed effects in tobit
models based on the normal distribution yield inconsistent estimates, as fixed effects cannot
14
be treated as incidental parameters without biasing the other model coefficients (as long as
N>T) (Hsiao, 1986). Thus, for consistent estimation, first, we provide regression estimates
using Honoré semi-parametric fixed effect tobit estimator (Honoré, 1992).16 As before, the
effects on pesticide and fertilizer use are driven by early growing season temperatures. More-
over, early growing season estimates are statistically significant as well. Similarly, we provide
log-level regression estimates for weeding labor as well. Table A.6 shows the effects of tem-
perature on intensity of pesticide and fertilizer use based on Honoré household fixed effects
tobit. For comparison, the standard tobit is also presented in Table A.5. Qualitative con-
clusions drawn from our main results presented in Table 7 remain unchanged.
In Table A.7, we cluster standard errors at the level of district-round. Our main results
allow for heteroskedasticity and autocorrelation of households within each village and across
rounds, but assume that households are identically distributed across villages. Here, we
allow for spatial correlation of villages (households) within a district in a round. The core
story line remains, although the significance of the estimated effects on fertilizer use weakens
appreciably. Lastly, in Table A.8 we demonstrate that effects of temperature on agricultural
input decisions are robust to the choice of lower bound used to calculate cumulative growing
degree days (GDDs). The estimation results are consistent with predictions from the agro-
nomic literature and with farmers qualitative comments, and stand up to various robustness
tests.
4.2.4 Hetereogenous Effects - by Wealth
Disentangling the effects of credit constraints and ex ante risk falls outside the scope of this
paper. However, apart from the qualitative evidence presented earlier, we provide suggestive
empirical evidence for a broader wealth channel driven by the relationship between rising
temperatures and the increased incidence of pests and diseases. To examine such a mecha-
nism, the key thought experiment involves the question of whether, ceteris paribus, changes
in ex ante income or income risk affect input use. We exploit plausibly exogenous changes in
temperature over time across relatively ‘poor’ and ‘wealthy’ households. We show suggestive
evidence that household wealth differences may confer different abilities to accommodate ex
16 We use Honoré’s Pantob program, accessible here: http://www.princeton.edu/ honore/stata/
15
ante risk or to absorb income shocks that are differentially influenced by an increase in pest
and disease pressure arising due to higher temperatures.
We use average land ownership across all five rounds as a proxy for wealth. We separate
the balanced sample by terciles, and denote households in the bottom tercile as relatively
‘poor’. We then estimate the relationships between heat and agricultural input use, now
adding interaction terms between degree days in each phase of the crop cycle (PP, GS1 and
GS2), and a 0-1 binary wealth variable which takes value 1 if wealth for household i is in the
bottom tercile, that is if the 1995-2010 average land holding is less than 2.5 acres, 0 otherwise.
We find that poorer households are less likely to adapt to higher temperatures via pesticide
use. These effects are consistent with the binding liquidity constraints hypothesis, but less
so with a risk aversion story if pesticide purchases reduce risk and farmers exhibit constant
or decreasing absolute risk aversion. We also examine the relationship between GDDs and
fertilizer use by household wealth. We find that poorer households use less fertilizer due to
higher temperatures (Tables A.9, A.10 and A.11).17
These results suggest that not only do wealthier farmers adapt more than their poorer
neighbors to an increase in incidence of pests, weeds and diseases, through increased pes-
ticide use, they also seem to reduce their expenditure on fertilizer less in face of higher
temperatures. These associations suggest that higher than normal temperatures may lead
to regressive distributional yield effects within communities in these moderate agro-ecological
zones. Limited liquidity constrains uptake of loss-reducing inputs and aggravates the reduc-
tion in fertilizer application as temperature increases.
4.2.5 Alternative Explanations
In this section, we rule out some alternative channels that could potentially explain the ob-
served relationship between temperature and agricultural input use. Specifically, we consider
three alternative explanations: (1) the direct impact of temperature on agricultural yields,
(2) influence of humidity on the incidence of pests and crop disease, and (3) higher than
normal temperatures can affect soil moisture, in turn reducing fertilizer uptake.
17 In Tables A.12, A.13 and A.14, we use baseline (Round 1) land ownership as a proxy for wealth for all rounds. The point
estimates remain largely unchanged.
16
Temperature and Maize Yields: Temperature could affect agricultural input use
by directly affecting agricultural production: heat stress during the growing season in the
previous year could reduce agricultural yields through physiological mechanisms, affecting
input use in the current growing season. In fact, 45% of maize-growers in our sample said that
changes in temperature reduced crop yields (Table 5). However, it is likely that these answers
indicate an indirect relationship between temperature and yields due to the aforementioned
ecological channel rather than direct heat stress on maize plants.
We expect the link from higher temperatures to reduced maize yields is indirect for two
reasons. First, optimum maize growth occurs at temperatures of 24-30C (Pingali, 2001).
Schlenker and Roberts (2009) find that maize yields in the United States increase with
temperature up to 29C. Average daily temperatures for villages in our sample are well below
30C, in fact, the 99th percentile of the distribution of maximum daily temperatures for
villages in our sample is 32C (Figure 12).
Second, we directly examine the relationship between growing degree days over 21C
and agricultural yields amongst maize farmers in the data. We observe a reduced form
relationship between temperatures in the growing season and maize yields; that is, we observe
the net effect of at least the following channels of impact: an increase in incidence of pests,
weeds and crop diseases, consequent increase in pesticide use and manual weeding, decrease
in fertilizer use, and a possible direct effect of higher temperatures on maize yields. We find
no statistically significant relationship between GDDs and maize yields when we control for
key inputs, use of which responds to temperature change (Table A.15). Indeed, the point
estimate is positive, consistent with a positive physiological effect of temperature on maize
growth in this region, given its relatively moderate temperatures, offset by greater pest and
weed pressures to which farmers respond adaptively, as we have already established, and
reduced fertilizer inputs.18
Humidity: Grey leaf spot is a major maize disease in Kenya. Empirical results suggest
that moderate to high temperatures and prolonged periods of high relative humidity are both
18 By employing a sinusoidal interpolation between the daily minimum and maximum temperature (Snyder, 1985), we follow
Roberts, Schlenker and Eyer (2013), and also generate growing degree days accounting for within-day temperature variations,
not just the daily mean temperature, and estimate the effects on maize yields. Again, we fail to find evidence for a negative
effect of temperature on yields (Table A.16).
17
favorable for the development of gray leaf spot (Latterell and Rossi, 1983; Rupe, Siegel and
Hartman, 1982). Similarly, relative humidity is also a main factor affecting the distribution
of stem borers, the main insect pest affecting maize in Kenya (Mwalusepo et al., 2015).
Thus, given the correlation between heat and humidity, it is possible that our estimates
actually capture the influence of relative humidity on pests and crop diseases. To rule
out this explanation, we control for relative humidity at the village level, and find that
our estimates are relatively unchanged (Table A.17, Columns 1-3). Even holding humidity
constant, temperature exerts an independent effect on agricultural input use.
Soil Moisture: Higher than normal temperatures could reduce the stock of water in
the soil, and thereby reduce fertilizer effectiveness, inducing lower farmer uptake. Water and
soil nutrients (such as nitrogen and phosphorus) are essential for crop growth. Fertilizer use
adds to soil nutrients. In rain-fed agriculture, where soil moisture depends on rainfall, tem-
perature, and soil quality, the effectiveness of fertilizer can be seriously affected inadequate
soil moisture (cite agronomy paper). When moisture deficiency is the primary factor limit-
ing crop growth, yield is less responsive to fertilizer use, in line with von Liebig’s law of the
minimum which states that yield is determined by the amount of the most limiting nutrient
(Marenya and Barrett, 2009; Paris, 1992). In addition, soil nutrients are taken up by plant
roots in a water solution, so water availability affects how efficiently applied fertilizer can be
used by crops. Farmers are less likely to adopt fertilizer in zones where soil moisture supply
is deficient (at least partially) due to low yield response to fertilizer (Jha and Hojjati, 1995;
Lele and Stone, 1989; Matlon, 1990; Thompson P. and Baanante, 1989).
Moreover, both air temperature and soil temperature affect soil moisture through the
evapotranspiration process, the predominant water cycle in the absence of precipitation (Lon-
gobardi and Khaertdinova, 2015). Temperature plays a critical role in evapotranspiration.
Higher temperature increases transpiration of water in the surface soil, just like in the plants.
Komuscu, Erkan and Oz (1998) assess the implications of climate change for soil moisture
availability in southeast Turkey, finding substantial reductions in availability during summer.
Also, local effects of heat stress on soil moisture will vary with soil characteristics. Boix-
Fayos et al. (1998), for example, show that infiltration and the water-holding capacity of soils
18
on limestone are greater with increased frost activity and infer that increased temperatures
could lead to increased surface or shallow runoff.
Since we include household fixed effects in our model, we control for time invariant
qualities of the soil. We also control already for time varying attributes of soil at province
level. However, if changes in heat across years are correlated with changes in soil moisture
within a province, the estimated relationship between temperature and fertilizer use may be
susceptible to the soil moisture channel. To rule out this explanation, we control for daily
soil moisture at the village level. Our findings remain unchanged when we hold soil moisture
constant (Table A.17, Columns 4-5).19
5 Conclusion
The use of modern agricultural technologies is key to agricultural productivity growth, and
a pathway out of poverty for agrarian households in poor economies. The dramatic growth
in agricultural yields in Asia, the stagnation of yields in Africa, and subsequent effects on
economic growth, may be partly explained by increased use of modern agricultural inputs
in Asia and continued low use in Africa (e.g., McArthur and McCord, 2017; Morris et al.,
2007; Sánchez, 2010).
In this paper, we find that higher than normal temperatures can affect agriculture even
in regions where temperatures are not high enough to directly, adversely affect crop growth
through physiological heat stress. Agricultural input decisions are sensitive to higher tem-
peratures even in moderate climates due to an ecological relationship between temperature
and the prevalence of pests, weeds and crop diseases. As far as we are aware, this is the first
economic study to provide evidence for such indirect effects of temperature on agriculture,
apparently due to a ecological mechanism underlying the relationship between climate and
agriculture.
Moreover, because different crops and different stages in the agricultural cycle are af-
fected by temperature in different ways, the underlying mechanisms of impact as well as
19 Unfortunately, we do not have daily soil moisture data for the entire sample. Columns 1 and 3 presents results using our
baseline specification for the subset of observations for which we could find matching soil moisture estimates, while Columns 2
and 4 control for soil moisture.
19
the suitable adaptive responses, may differ too. This matters because farmers make their
decisions sequentially, adapting to new information as it emerges. By separately estimating
the impact of temperatures during each stage of the agricultural cycle, we are not only able
to identify an ecological mechanism, but also demonstrate that farmers promptly adapt to
temperature variation in the short-run.
20
References
Binswanger, Hans P., and Mark R . Rosenzweig. 1993. “Wealth , Weather Risk
and the Composition and Profitability of Agricultural Investments.” Economic Journal,
103(416): 56–78.
Bloomfield, J. P., R. J. Williams, D. C. Gooddy, J. N. Cape, and P. Guha.

2006. “Impacts of climate change on the fate and behaviour of pesticides in surface and
groundwater-a UK perspective.” Science of the Total Environment, 369(1-3): 163–177.
Boix-Fayos, C., A. Calvo-Cases, A. C. Imeson, M. D. Soriano-Soto, and I. R.

Tiemessen. 1998. “Spatial and short-term temporal variations in runoff, soil aggrega-
tion and other soil properties along a mediterranean climatological gradient.” Catena,
33(2): 123–138.
Burke, Marshall, and Kyle Emerick. 2016. “Adaptation to Climate Change: Evidence
from US Agriculture.” American Economic Journal: Economic Policy, 8(3): 106–140.
Cairns, J.E., K. Sonder, P.H. Zaidi, N. Verhulst, G. Mahuku, R. Babu, S.K.

Nair, B. Das, B. Govaerts, M.T. Vinayan, Z. Rashid, J.J. Noor, P. Devi,
F. San Vicente, and B. M. Prasanna. 2012. “Maize production in a changing climate:
Impacts, adaptation and mitigation strategies.” Advances in Agronomy, 114: 1–65.
Chakraborty, Sukumar. 2008. “Impacts of global change on diseases of agricultural crops

and forest trees.” CAB Reviews: Perspectives in Agriculture, Veterinary Science, Nutrition
and Natural Resources, 3(54): 1–5.
Coakley, Stella Melugin, Harald Scherm, and Sukuma Chakraborty. 1999. “Climate
Change and Plant Disease Management.” Annual Review of Phytopathology, 37: 399–426.
Cole, Shawn, Xavier Giné, Jeremy Tobacman, Petia Topalova, Robert Townsend,
and James Vickery. 2013. “Barriers to Household Risk Management: Evidence from
India.” American Economic Journal: Applied Economics, 5(1): 104–135.
Conley, Timothy G, and Christopher R Udry. 2010. “Learning about a New Technol-
ogy: Pineapple in Ghana.” American Economic Review, 100(1): 35–69.
Dell, Melissa, Benjamin F Jones, and Benjamin a Olken. 2012. “Climate Shocks and
Economic Growth: Evidence from the Last Half Century.” American Economic Journal:
Macroeconomics, 4(3): 66–95.
Dercon, Stefan, and Luc Christiaensen. 2011. “Consumption risk, technology adop-
tion and poverty traps: Evidence from Ethiopia.” Journal of Development Economics,
96(2): 159–173.
Deschênes, Olivier, and Michael Greenstone. 2007. “The economic impacts of climate
change: Evidence from agricultural output and random fuctuations in weather.” American
Economic Review, 97(1): 354–385.
21
Deschênes, Olivier, and Michael Greenstone. 2011. “Climate change, mortality, and
adaptation: Evidence from annual fluctuations in weather in the US.” American Economic
Journal: Applied Economics, 3(4): 152–185.
Dillon, Brian. 2014. “Risk and Resilience among Tanzanian Farmers.” Mimeo.
Duflo, Esther, Michael Kremer, and Jonathan Robinson. 2011. “Nudging farmers
to use fertilizers: theory and experimental evidence from Kenya.” American Economic
Review, 101(6): 2350–2390.
Dukes, Jeffrey S., Jennifer Pontius, David Orwig, Jeffrey R. Garnas, Vikki L.
Rodgers, Nicholas Brazee, Barry Cooke, Kathleen A. Theoharides, Erik E.
Stange, Robin Harrington, Joan Ehrenfeld, Jessica Gurevitch, Manuel Lerdau,
Kristina Stinson, Robert Wick, and Matthew Ayres. 2009. “Responses of insect
pests, pathogens, and invasive plant species to climate change in the forests of northeastern
North America: What can we predict?” Canadian Journal of Forest Research, 39(2): 231–
248.
du Plessiss, Jean. 2003. “Maize Production.” Directorate of Agricultural Information Ser-

vices, Department of Agriculture South Africa.
Fafchamps, Marcel. 1993. “Sequential Labor Decisions Under Uncertainty : An Estimable

Household Model of West- African Farmers.” Econometrica, 61(5): 1173–1197.
Feder, Gershon, Richard E. Just, and David Zilberman. 1985. “Adoption of Agricul-
tural Innovations in Developing Countries.” Economic Development and Cultural Change,
33(2): 255–298.
Feng, S., A. B. Krueger, and M. Oppenheimer. 2010. “Linkages among climate change,
crop yields and Mexico-US cross-border migration.” Proceedings of the National Academy
of Sciences, 107(32): 14257–14262.
Foster, Andrew D., and Mark R. Rosenzweig. 1995. “Learning by Doing and Learning
from Others : Human Capital and Technical Change in Agriculture.” The Journal of
Political Economy, 103(6): 1176–1209.
Foster, Andrew D., and Mark R. Rosenzweig. 2010. “Microeconomics of Technology

Adoption.” Annual Review of Economics, 2(1): 395–424.
Garrett, K. A., S. P. Dendy, E. E. Frank, M. N. Rouse, and S. E. Travers. 2006.

“Climate Change Effects on Plant Disease: Genomes to Ecosystems.” Annual Review of
Phytopathology, 44(1): 489–509.
Gianessi, L. 2014. “Importance of Pesticides for Growing Maize in Sub-Saharan Africa.”

Crop Foundation International Pesticide Benefit Case Study, 104.
Gianessi, Leonard. 2010. “Solving Africa’s Weed Problem : Increasing Crop Production
& Improving the Lives of Women.” Aspects of Applied Biology.
22
Guiteras, R. 2008. “The Impact of Climate Change on Indian Agriculture.” Mimeo.
Hanzlik, Kristin, and Barbel Gerowitt. 2012. “Occurrence and distribution of impor-
tant weed species in German winter oilseed rape fields.” Journal of Plant Diseases and
Protection, 119(3): 107–120.
Honoré, Bo E . 1992. “Trimmed Lad and Least Squares Estimation of Truncated and
Censored Regression Models with Fixed Effects.” Econometrica, 60(3): 533–565.
Hsiao, Cheng. 1986. “Analysis of panel data.” Cambridy University Press.
I. O Akobundu. 1987. “Weed Science in the Tropics: Principles and Practices.” Wiley-
lnterscience.
Jha, Dayantha, and Behjat Hojjati. 1995. “Fertilizer use on smallholder farms in Eastern
Province, Zambia.” Field Crops Research, 42(2-3): 147–148.
Komuscu, A L I Umran, Ayhan Erkan, and Sukriye Oz. 1998. “Possible Impacts
of Climate Change on Soil Moisture Availability in the Southeast Anatolia Development
Project Region (GAP): An Analysis from an Agricultural Drought Perspective.” Climate
Change, 40: 519–545.
Kurukulasuriya, Pradeep, and Robert Mendelsohn. 2007. “Endogenous Irrigation:

The Impact of Climate Change on Farmers in Africa.” World Bank Policy Research Work-
ing Paper 4278.
Kurukulasuriya, Pradeep, and Robert Mendelsohn. 2008. “Crop switching as a strat-

egy for adapting to climate change.” African Journal of Agricultural and Resource Eco-
nomics, 2(1): 105–126.
Kurukulasuriya, Pradeep, Namrata Kala, and Robert Mendelsohn. 2011. “Adap-

tation and Climate Change Impacts: A Structural Ricardian Model of Irrigation and Farm
Income in Africa.” Climate Change Economics, 2(2): 149–174.
Latterell, Frances M., and Albert E. Rossi. 1983. “Gray Leaf Spot of Corn: A Disease
on the Move.” Plant Disease, 67(8): 842–847.
Lele, Uma, and Steven W Stone. 1989. “Population Pressure, the Environment and
Agricultural Intensification; Variations on the Boresup Hypothesis.” Managing Agricul-
tural Development in Africa (MADIA) Discussion Paper, 1(4): 1–132.
Lobell, David B., and Marshall B. Burke. 2008. “Why are agricultural impacts of
climate change so uncertain? The importance of temperature relative to precipitation.”
Environmental Research Letters, 3(3): 34007.
Lobell, David B., Marianne Bänziger, Cosmos Magorokosho, and Bindiganavile

Vivek. 2011. “Nonlinear heat effects on African maize as evidenced by historical yield
trials.” Nature Climate Change, 1(1): 42–45.
23
Longobardi, Antonia, and Elina Khaertdinova. 2015. “Relating soil moisture and air
temperature to evapotranspiration fluxes during inter-storm periods at a Mediterranean
experimental site.” Journal of Arid Lands, 7(1): 27–36.
Marenya, Paswel P., and Christopher B. Barrett. 2007. “Household-level determi-

nants of adoption of improved natural resources management practices among smallholder
farmers in western Kenya.” Food Policy, 32(4): 515–536.
Marenya, Paswel P., and Christopher B. Barrett. 2009. “State-conditional fertilizer

yield response on Western Kenyan Farms.” American Journal of Agricultural Economics,
91(4): 991–1006.
Matlon, P J. 1990. “Improving productivity in sorghum and pearl millet in semi-arid

Africa.” Food Research Institute Studies, 22(1): 1–43.
McArthur, John W., and Gordon C. McCord. 2017. “Fertilizing growth: Agricultural
inputs and their effects in economic development.” Journal of Development Economics,
127(February): 133–152.
Mendelsohn, By Robert, William D Nordhaus, and Daigee Shaw. 1994. “The

Impact of Global Warming on Agriculture : A Ricardian Analysis.” American Economic
Review, 84(4): 753–771.
Mendelsohn, Robert. 2008. “The impact of climate change on agriculture in developing

countries.” Journal of Natural Resources Policy Research, 1(1): 5–19.
Morris, Michael, Valerie a Kelly, Ron J Kopicki, and Derek Byerlee. 2007. “Fer-
tilizer Use in African Agriculture.” Experimental Agriculture, 44(1): 162.
Moser, Christine M., and Christopher B. Barrett. 2006. “The complex dynamics
of smallholder technology adoption: The case of SRI in Madagascar.” Agricultural Eco-
nomics, 35(3): 373–388.
Mwalusepo, Sizah, Henri E.Z. Tonnang, Estomih S. Massawe, Gerphas O.

Okuku, Nancy Khadioli, Tino Johansson, Paul André Calatayud, and
Bruno Pierre Le Ru. 2015. “Predicting the impact of temperature change on the future
distribution of maize stem borers and their natural enemies along East African mountain
gradients using phenology models.” PLoS ONE, 10(6): 1–23.
NAFIS. 2011. “Field Management.” National Famers Information Service - Kenya.
Oehninger, Ernst Bertone, Cynthia Lin Lawell, and Michael R Springborn. 2017.
“The effects of climate change on groundwater extraction for agriculture and land-use
change.” Mimeo.
Orr, A., B. Mwale, and D. Saiti. 2002. “Modelling agricultural ’performance’: Small-
holder weed management in Southern Malawi.” International Journal of Pest Manage-
ment, 48(4): 265–278.
24
Paris, Quirino. 1992. “The von Liebig Hypothesis.” American Journal of Agricultural Eco-
nomics, 74(4): 1019–1028.
Patterson, D.T., J.K. Westbrook, R.J.V. Joyce, P.D. Lingren, and J. Rogasik.
1999. “Weeds, Insects, and Diseases.” Climate Change, 43: 711–727.
Paul, P. A., and G. P. Munkvold. 2005. “Influence of Temperature and Relative Humid-
ity on Sporulation of Cercospora zeae-maydis and Expansion of Gray Leaf Spot Lesions
on Maize Leaves.” Plant Disease, 89(6): 624–630.
Peters, Kristian, Laura Breitsameter, and Bärbel Gerowitt. 2014. “Impact of cli-
mate change on weeds in agriculture: A review.” Agronomy for Sustainable Development,
34(4): 707–721.
Pingali, Prabhu L. 2001. “CIMMYT 19992000 World Maize Facts and Trends. Meeting
World Maize Needs: Technological Opportunities and Priorities for the Public Sector.”
D.F.: CIMMYT.
Roberts, Michael J., Wolfram Schlenker, and Jonathan Eyer. 2013. “Agronomic
weather measures in econometric models of crop yield with implications for climate
change.” American Journal of Agricultural Economics, 95(2): 236–243.
Rodell, M., P. R. Houser, U. Jambor, J. Gottschalck, K. Mitchell, C-J. Meng,
K. Arsenault, B. Cosgrove, J. Radakovich, M. Bosilovich, J. K. Entin, J. P.
Walker, D. Lohmann, and D. Toll. 2004. “The Global Land Data Assimilation Sys-
tem.” Bulletin of the American Meteorological Society, 85(3): 381–394.
Rosenzweig, Cynthia, Ana Iglesias, X B Yang, Paul R Epstein, and Eric Chivian.
2001. “Climate change and extreme weather events.” Global Change and Human Health,
2(2): 90–104.
Rosenzweig, Cynthia, and Martin L. Parry. 1994. “Potential Impact of Climate Change
on World Food Supply.” Nature, 367.
Rupe, J.C., M.R. Siegel, and J.R. Hartman. 1982. “Influence of Environment and
Plant Maturity on Gray Leaf Spot of Corn Caused by Cercospora zeae-maydis.” Ecology
and Epidemiology, 82: 11–51.
Sánchez, P A. 2010. “Tripling crop yields in tropical Africa.” Nature Geoscience, 3(5): 1–2.
Schlenker, Wolfram, and David B Lobell. 2010. “Robust negative impacts of climate
change on African agriculture.” Environmental Research Letters, 5(1): 8 pp.
Schlenker, Wolfram, and Michael J. Roberts. 2009. “Nonlinear temperature effects
indicate severe damages to U.S. crop yields under climate change.” Proceedings of the
National Academy of Sciences, 106(37): 15594–15598.
Schlenker, Wolfram, W Michael Hanemann, and Anthony C Fisher. 2006. “The
Impact of Global Warming on U. S. Agriculture : an Econometric Analysis of Optimal
Growing Conditions.” Review of Economics and Statistics, 88(1): 113–125.
25
Seo, S N, and R Mendelsohn. 2008a. “Measuring impacts and adaptations to climate
change: a structural Ricardian analysis of climate change impacts and adaptations in
African agriculture.” Agricultural Economics, 38: 151–165.
Seo, S. Niggol, and Robert Mendelsohn. 2008b. “An analysis of crop choice: Adapting
to climate change in South American farms.” Ecological Economics, 67(1): 109–116.
Sheahan, Megan, and Christopher B. Barrett. 2017. “Ten striking facts about agri-
cultural input use in Sub-Saharan Africa.” Food Policy, 67: 12–25.
Simons, Sarah. 2006. “Management strategies for maize grey leaf spot (Cercospora zeae-
maydis) in Kenya and Zimbabwe.” DFID Technical Report, R7566: 1–67.
Singer, Alexander, Justin M J Travis, and Karin Johst. 2013. “Interspecific interac-
tions affect species and community responses to climate shifts.” Oikos, 122(3): 358–366.
Snyder, R. L. 1985. “Hand calculating degree days.” Agricultural and Forest Meteorology,
35(1-4): 353–358.
Songa, Josephine Moraa, Zhou Guofa, and William Allan Overholt. 2001. “Rela-
tionships of Stemborer Damage and Plant Physical Conditions to Maize Yield in a Semi-
Arid Zone of Eastern Kenya.” Insect Science and its Application, 21(3): 243–249.
Suri, Tavneet. 2011. “Selection and Comparative Advantage in Technology Adoption.”

Econometrica, 79(1): 159–209.
Taraz, Vis. 2017. “Can Farmers Adapt to Higher Temperatures? Evidence from India.”
Mimeo.
The World Bank. 2010. World Development Report 2010: Development and Climate
Change. The World Bank, 2009.
Thompson P., Thomas, and Carlos A. Baanante. 1989. “A Socioeconomic Study

of Farm-Level Constrainsts To Fertilizer Use in Western Niger.” International Fertilizer
Development, Alabama -(IFDC - P6): 1–6.
Walck, Jeffrey L., Siti N. Hidayati, Kingsley W. Dixon, Ken Thompson, and Pe-
ter Poschlod. 2011. “Climate change and plant regeneration from seed.” Global Change
Biology, 17(6): 2145–2161.
Ward, J. M. J., M. D. Laing, and F. H. J. Rijkenberg. 1997. “Frequency and Timing

of Fungicide Applications for the Control of Gray Leaf Spot in Maize.” Plant Disease,
81(1): 41–48.
Welch, Jarrod R, Jeffrey R Vincent, Maximilian Auffhammer, Piedad F Moya,

Achim Dobermann, and David Dawe. 2010. “Rice yields in tropical/subtropical Asia
exhibit large but opposing sensitivities to minimum and maximum temperatures.” Pro-
ceedings of the National Academy of Sciences, 107: 14562–14567.
26
Zhao, Chuang, Bing Liu, Shilong Piao, Xuhui Wang, David B. Lobell, Yao
Huang, Mengtian Huang, Yitong Yao, Simona Bassu, Philippe Ciais, Jean-
Louis Durand, Joshua Elliott, Frank Ewert, Ivan A. Janssens, Tao Li, Erda
Lin, Qiang Liu, Pierre Martre, Christoph Müller, Shushi Peng, Josep Peñuelas,
Alex C. Ruane, Daniel Wallach, Tao Wang, Donghai Wu, Zhuo Liu, Yan Zhu,
Zaichun Zhu, and Senthold Asseng. 2017. “Temperature increase reduces global yields
of major crops in four independent estimates.” Proceedings of the National Academy of
Sciences, 114(35): 9326–9331.
27
Tables and Figures
Figures
Figure 1: Location of Sample Villages
28
Figure 2: Daily Average Temperature in TAMPA Sample (1990-2012)
Notes: Distribution of average daily temperatures in villages in TAMPA from 1990-2012.

According to existing literature, temperature affects maize yields only after 30C (represented
by the red line). (Lobell et al., 2011; Schlenker and Roberts, 2009)
29
Figure 3: Agro-Ecological Zones and Maize Production in Africa
Source: Agro-ecological zones - IFPRI Harvest Choice (www.harvestchoice.org.);

Maize Production in Africa: Spatial Production Allocation Model (SPAM), 2005
(www.mapSPAM.info)
Agro-Ecological Zones (AEZs): Agro-ecological zones (AEZs) are geographical areas

sharing similar climate characteristics (e.g., rainfall and temperature) with respect to their
potential to support (usually rain-fed) agricultural production. Because of the general sim-
ilarity of production conditions, many agricultural technologies, practices and production
systems tend to behave or respond consistently within a specific AEZ. AEZs therefore pro-
vide a useful spatial framework for identifying the potential area extent of applicability of
given innovations and, futhermore, the likely potential for production related innovations to
“spillover” from one country (or continent) to another. AEZs provide an ecology-based di-
vision of geographic space as opposed to administrative or political boundaries within which
environmental conditions could vary significantly. The tabulation of rural population by
AEZ for Sub-Saharan Africa indicates that almost 23% of the rural population lives in more
humid highland regions.
30
Figure 4: Household Pesticide Use
Figure 5: Household Weeding Labor
31
Figure 6: Household Fertilizer Use
Figure 7: Amount Spent on Fertilizer, Pest Repellents and Hired Labor for Weeding by
Adopters
32
Figure 8: Mean Maize Yield (in Kgs/Acre)
Figure 9: Distribution of Yields by Pesticide Use
33
Figure 10: Distribution of Yields by Fertilizer Use
34
Figure 11: Daily Average Temperature by Phases in the Agricultural Cycle (1990-2012)
Notes: Distribution of average daily temperatures from 1990-2012 for three phases of the
agricultural cycle. Phase 1: pre-planting or land preparation - onset of planting; Phase 2:
planting or basal fertilizer application - onset of top dressing fertilizer; Phase 3: top dressing
fertilizer application - onset harvest. We calculate cumulative growing degree days from a
lower bound of 21C (represented by red vertical line)
35
Figure 12: Daily Maximum Temperature in TAMPA Sample (1990-2013)
Tables
Table 1: Climate Change in Kenya?

(1) (2)
Farmer Noticed Change in Temperature? Famer Affected by Changes in Temperature?
2009
No 53.14 17.70
Yes 46.86 82.30
Notes: Sample includes 1242 households, balanced over 5 survey rounds, in the 2009-10 TAMPA survey.
36
Table 2: Pesticide Use Transitions
Fraction of Households
NNN 0.22
(0.42)
NYY 0.26
(0.44)
NNY 0.08
(0.27)
NYN 0.16
(0.37)
YNY 0.02
(0.15)
YNN 0.02
(0.15)
YYN 0.07
(0.25)
YYY 0.17
(0.37)
Observations 1242
Notes: This table shows all possible three transitions in our sample of farmers and the fraction of our sample that experiences
each of these transitions. The three periods correspond to the 2003-04, 2006-07 and 2009-10 survey rounds. In the first column,
the three letters represent the transition history with respect to pesticide adoption, where “Y” represents the use of pesticides
and “N” represents non-adoption of pesticides. These are ordered by survey round. For example, the transition “YYY” stands
for farmers who used pesticides in all three periods; they make up about 17% of our sample. “YYN” represents the 7% of the
sample that use pesticides in 2003-04 and 2006-07 but not in 2009-10.
37
Table 3: Fertilizer Use Transitions
Fraction of Households
NNNNN 0.16
(0.37)
NYYYY 0.06
(0.23)
NNYYY 0.03
(0.17)
NNNYY 0.02
(0.12)
NNNNY 0.03
(0.16)
NYN/YNY 0.14
(0.35)
YNNNN 0.00
(0.07)
YYNNN 0.01
(0.07)
YYYNN 0.00
(0.04)
YYYYN 0.02
(0.12)
YYYYY 0.54
(0.50)
Observations 1242
Notes: This table shows all possible five transitions in our sample of farmers and the fraction of our sample that experiences
each of these transitions. The three periods correspond to the 1996-97, 1999-00, 2003-04, 2006-07 and 2009-10 survey rounds.
In the first column, the five letters represent the transition history with respect to fertilizer adoption, where “Y” represents the
use of pesticides and “N” represents non-adoption of fertilizer. For example, the transition “YYYYY” stands for farmers who
used fertilizer in all five periods; they make up about 54% of our sample. “NYN/YNY” stands for farmers who transitioned
both in and out of fertilizer use within these five rounds of data. All other sequences are unidirectional.
Table 4: Growing Degree Days: Mean and Standard Deviations – Rounds 1-5
All 1997 2000 2004 2007 2010
CY PP DD >21C 47.58 37.86 46.79 38.16 70.93 44.17
(74.61) (68.55) (73.81) (81.04) (81.95) (60.63)
CY GS1 DD >21C 30.03 24.60 23.04 38.17 24.08 40.25
(49.28) (44.84) (42.70) (61.47) (39.35) (52.03)
CY GS2 DD >21C 16.46 17.73 12.28 21.46 14.37 16.44
(42.88) (40.89) (33.21) (50.92) (41.24) (45.63)
Observations 6210 1242 1242 1242 1242 1242
Notes: Sample includes 1242 households balanced over 5 survey rounds (1996-97, 1999-00, 2003-04, 2006-07 and 2009-10).
Temperature data was generated at the village level, so the table reports mean and standard deviations for degree days (DD)
over 21C for each survey round. CY: current year; PP: pre-planting or land preparation - onset of planting; GS1: planting or
basal fertilizer application - onset of top dressing fertilizer; GS2: top dressing fertilizer application - onset harvest. Standard
deviations are in parentheses.
38
Table 5: How was farming affected by this change in temperature?
(1)
Affected by Changes in Temperature, How?
2009
Decline in Yields 44.68
Decrease in Land Quality 4.38
Difficult to Time Seasons 6.89
Increase in Yields 5.43
Other 1.88
Weeds/Pests/Diseases 36.74
Notes: Sample includes 1242 households in the 2009-10 TAMPA survey.
Table 6: Why Didn’t You Use Fertilizer?

(1)
Why No Fertilizer?
2009
Fertilizer Not Available 0.92
Lack of Advice 3.06
No Money/Too Expensive 57.80
No Need To Use Fertilizer 38.23
Notes: Sample includes 1242 households in the 2009-10 TAMPA survey.
Table 7: Temperature, Fertilizer and Pesticide Use

(1) (2) (3) (4) (5)
Pesticides Ln Pesticide/Acre Fertilizer Ln Fertilizer/Acre Ln Weeding Days/Acre
β / SE β / SE β / SE β / SE β / SE
CY PP DD >21C 0.0019 0.0084 -0.0003 -0.0054 0.0017
(0.0014) (0.0090) (0.0005) (0.0055) (0.0021)
CY GS1 DD >21C 0.0063** 0.0450*** -0.0018** -0.0180** -0.0004
(0.0026) (0.0159) (0.0008) (0.0087) (0.0048)
CY GS2 DD >21C -0.0004 -0.0108 0.0003 0.0005 0.0084***
(0.0015) (0.0079) (0.0004) (0.0044) (0.0031)
Village FE Yes Yes Yes Yes Yes
Prov-by-Year FE Yes Yes Yes Yes Yes
Rainfall Controls Yes Yes Yes Yes Yes
Observations 3726 3726 6210 6210 3726
R2 0.336 0.354 0.594 0.656 0.177
Notes: Sample includes 1242 households balanced over 5 survey rounds (1996-97, 1999-00, 2003-04, 2006-07 and 2009-10) for
fertilizer use and 3 survey rounds (2003-04, 2006-07 and 2009-10) for pesticides and weeding labor days. The table presents the
effects of temperature (captured via degree days (DD) over 21C) on agricultural input use. CY: current year; PP: pre-planting
or land preparation - onset of planting; GS1: planting or basal fertilizer application - onset of top dressing fertilizer; GS2: top
dressing fertilizer application - onset harvest. Standard errors are in parentheses, clustered by village. *Significant at 10%.
**Significant at 5%. ***Significant at 1%.
39
A Appendix
Table A.1: Summary Statistics

1997 2000 2004 2007 2010
Pesticides 0.27 0.65 0.53
(0.45) (0.48) (0.50)
Pesticide/Acre(kgs) 0.24 0.50 0.56
(1.68) (1.01) (3.96)
Total Weeding Days/Acre 9.59 4.67 4.56
(11.64) (6.94) (6.43)
Pesticide Exp/Acre(KES) 46.79 274.87 312.50
(199.58) (617.82) (894.09)
Hired Weeding Exp/Acre(KES) 340.23 236.16 397.80
(638.27) (418.45) (705.59)
Fertilizer 0.63 0.69 0.71 0.75 0.75
(0.48) (0.46) (0.45) (0.43) (0.44)
Fertilizer/Acre(kgs) 46.07 57.48 51.37 54.53 51.25
(76.02) (91.09) (70.20) (63.80) (57.05)
Fertilizer Exp/Acre(KES) 0.00 0.00 1540.48 1989.29 3019.45
(0.00) (0.00) (1926.76) (2168.13) (3442.11)
Maize Output/Acre(kgs) 292.33 355.18 406.68 489.37 394.87
(333.03) (908.16) (424.91) (445.54) (353.66)
Notes: Standard deviations are given in parentheses. Sample includes 1242 households balanced over 5 survey rounds (1996-97,
1999-00, 2003-04, 2006-07 and 2009-10). Detailed data on pesticides and weeding labor days was only collected in 2003-04,
2006-07 and 2009-10.
Table A.2: Household FE: Temperature, Fertilizer and Pesticide Use

(1) (2) (3) (4) (5)
CY PP DD >21C 0.0019 0.0084 -0.0003 -0.0054 0.0017
(0.0017) (0.0109) (0.0006) (0.0062) (0.0025)
CY GS1 DD >21C 0.0063** 0.0450** -0.0018* -0.0180* -0.0004
(0.0031) (0.0192) (0.0009) (0.0096) (0.0058)
CY GS2 DD >21C -0.0004 -0.0108 0.0003 0.0005 0.0084**
(0.0018) (0.0095) (0.0005) (0.0049) (0.0038)
Household FE Yes Yes Yes Yes Yes
Prov-by-Year FE Yes Yes Yes Yes Yes
Observations 3726 3726 6210 6210 3726
R2 0.587 0.587 0.739 0.788 0.498
1
Table A.3: Observed temperature variation: proportion of households with degree-days
below/above average (degrees) after removing province*year effects and district*year effects
Removed Prov*Round FE Removed Dist*Round FE

% HHs % HHs
CY PP DD >21C: DD below/above 5 degrees 0.37
CY GS1 DD >21C: DD below/above 5 degrees 0.29
Notes: Sample include 1242 balanced households over 5 survey rounds (1996-97, 1999-00, 2003-04, 2006-07 and 2009-10). The
table presents the leftover variation in growing degree days (DD) after removing province-by-round and district-by-round fixed
effects. CY: current year; PP: pre-planting or land preparation - onset of planting; GS1: planting or basal fertilizer application
- onset of top dressing fertilizer; GS2: top dressing fertilizer application - onset harvest.
Table A.4: District*Year FE: Temperature, Fertilizer and Pesticide Use

(1) (2) (3) (4) (5)
CY PP DD >21C 0.0009 0.0063 -0.0016** -0.0194** 0.0019
(0.0015) (0.0088) (0.0008) (0.0082) (0.0030)
CY GS1 DD >21C 0.0051 0.0480** -0.0031** -0.0323** 0.0108
(0.0038) (0.0229) (0.0014) (0.0136) (0.0085)
CY GS2 DD >21C 0.0013 0.0037 -0.0011** -0.0148*** -0.0006
(0.0016) (0.0096) (0.0005) (0.0055) (0.0041)
District-by-Year FE Yes Yes Yes Yes Yes
Observations 3726 3726 6210 6210 3726
R2 0.371 0.388 0.607 0.667 0.189
2
Table A.5: Standard Tobit Estimates: Temperature, Fertilizer and Pesticide Use
(1) (2) (3)
Ln Pesticide/Acre Ln Fertilizer/Acre Ln Weeding Days/Acre
β / SE β / SE β / SE
CY PP DD >21C 0.0349* -0.0038 0.0011
(0.0183) (0.0074) (0.0021)
CY GS1 DD >21C 0.1081*** -0.0317** -0.0032
(0.0365) (0.0129) (0.0055)
CY GS2 DD >21C -0.0064 0.0001 0.0079***
(0.0190) (0.0054) (0.0031)
Village FE Yes Yes Yes
Prov-by-Year FE Yes Yes Yes
Rainfall Controls Yes Yes Yes
Observations 3726 6210 3726
R2
effects of temperature (captured via degree days (DD) over 21C) on weeding labor. CY: current year; PP: pre-planting or land
preparation - onset of planting; GS1: planting or basal fertilizer application - onset of top dressing fertilizer; GS2: top dressing
fertilizer application - onset harvest. Standard errors are in parentheses. Standard errors are clustered by village for columns
1-2. *Significant at 10%. **Significant at 5%. ***Significant at 1%.
Table A.6: Honoré Fixed Effects Tobit: Temperature, Fertilizer and Pesticide Use
(1) (2) (3)
Ln Pesticide/Acre Ln Fertilizer/Acre Ln Weeding Days/Acre
CY PP DD >21C 0.0259* -0.0041 0.0007
(0.0144) (0.0075) (0.0027)
CY GS1 DD >21C 0.0899*** -0.0375*** -0.0012
(0.0265) (0.0140) (0.0064)
CY GS2 DD >21C -0.0054 0.0014 0.0108***
(0.0126) (0.0058) (0.0040)
R2
dressing fertilizer application - onset harvest. Standard errors are in parentheses. *Significant at 10%. **Significant at 5%.
***Significant at 1%.
3
Table A.7: Clustering Standard Errors by District-Round: Temperature, Fertilizer and Pes-
ticide Use
(1) (2) (3) (4) (5)
CY PP DD >21C 0.0019 0.0084 -0.0003 -0.0054 0.0017
(0.0015) (0.0094) (0.0005) (0.0059) (0.0015)
CY GS1 DD >21C 0.0063** 0.0450** -0.0018 -0.0180* -0.0004
(0.0029) (0.0182) (0.0011) (0.0104) (0.0040)
CY GS2 DD >21C -0.0004 -0.0108 0.0003 0.0005 0.0084**
(0.0025) (0.0156) (0.0008) (0.0078) (0.0033)
District-by-Year FE Yes Yes Yes Yes Yes
Observations 3726 3726 6210 6210 3726
R2 0.336 0.354 0.594 0.656 0.177
effects of temperature on agricultural input use. CY: current year; PP: pre-planting or land preparation - onset of planting; GS1:
planting or basal fertilizer application - onset of top dressing fertilizer; GS2: top dressing fertilizer application - onset harvest.
Standard errors are in parentheses, clustered by district-round. *Significant at 10%. **Significant at 5%. ***Significant at 1%.
Table A.8: Alternative GDD Lower Bounds: Temperature, Fertilizer and Pesticide Use
(1) (2) (3) (4) (5) (6)
Pesticides Pesticides Pesticides Fertilizer Fertilizer Fertilizer
β / SE β / SE β / SE β / SE β / SE β / SE
CY PP DD >19C 0.0015 -0.0004
(0.0010) (0.0004)
CY GS1 DD >19C 0.0029** -0.0014**
(0.0011) (0.0006)
CY GS2 DD >19C -0.0006 -0.0000
(0.0006) (0.0003)
CY PP DD >20C 0.0017 -0.0004
(0.0012) (0.0005)
CY GS1 DD >20C 0.0034** -0.0016**
(0.0016) (0.0007)
CY GS2 DD >20C -0.0007 0.0001
(0.0008) (0.0004)
CY PP DD >22C 0.0018 -0.0008
(0.0016) (0.0007)
CY GS1 DD >22C 0.0078*** -0.0021**
(0.0028) (0.0009)
CY GS2 DD >22C -0.0027 0.0006
(0.0024) (0.0006)
Village FE Yes Yes Yes Yes Yes Yes
Prov-by-Year FE Yes Yes Yes Yes Yes Yes
Rainfall Controls Yes Yes Yes Yes Yes Yes
Observations 3726 3726 3726 6210 6210 6210
R2 0.336 0.336 0.336 0.594 0.594 0.594
fertilizer use and 3 survey rounds (2003-04, 2006-07 and 2009-10) for pesticides. The table presents the effects of temperature
(captured via degree days (DD) over 19C, 20C, and 22C) on agricultural input use. CY: current year; PP: pre-planting or land
preparation - onset of planting; GS1: planting or basal fertilizer application - onset of top dressing fertilizer; GS2: top dressing
fertilizer application - onset harvest. Standard errors are in parentheses, clustered by village. *Significant at 10%. **Significant
at 5%. ***Significant at 1%.
4
Table A.9: Temperature, Pesticides and Fertilizer Use, by Wealth (Round 1-5)
(1) (2) (3) (4)
Pesticides Ln Pesticide/Acre Fertilizer Ln Fertilizer/Acre
β / SE β / SE β / SE β / SE
CY PP DD >21C 0.0026 0.0126 -0.0001 -0.0035
(0.0017) (0.0111) (0.0006) (0.0062)
CY GS1 DD >21C 0.0071** 0.0496** -0.0014 -0.0155*
(0.0031) (0.0190) (0.0009) (0.0092)
CY GS2 DD >21C -0.0011 -0.0142 0.0003 0.0013
(0.0020) (0.0114) (0.0004) (0.0044)
CY PP DD >21C*Bottom Wealth Tercile -0.0015** -0.0088** -0.0006 -0.0049
(0.0006) (0.0038) (0.0004) (0.0039)
CY GS1 DD >21C*Bottom Wealth Tercile -0.0014 -0.0073 -0.0009 -0.0058
(0.0011) (0.0077) (0.0006) (0.0057)
CY GS2 DD >21C*Bottom Wealth Tercile 0.0025 0.0138 -0.0001 -0.0032
(0.0018) (0.0163) (0.0008) (0.0076)
Household FE Yes Yes Yes Yes
Prov-by-Year FE Yes Yes Yes Yes
Rainfall Controls Yes Yes Yes Yes
Observations 3726 3726 6210 6210
R2 0.588 0.589 0.740 0.788
fertilizer use and 3 survey rounds (2003-04, 2006-07 and 2009-10) for pesticides. The table presents the heterogeneous effects of
temperature (captured via degree days (DD) over 21C) on agricultural input use, by wealth. CY: current year; PP: pre-planting
dressing fertilizer application - onset harvest. Average landholding of the 33rd percentile is 2.5 acres. Standard errors are in
parentheses, clustered by village. *Significant at 10%. **Significant at 5%. ***Significant at 1%.
Table A.10: Standard Tobit Estimates: Temperature, Pesticides and Fertilizer Use, by
Wealth (Round 1-5)
(1) (2)
Ln Pesticide/Acre Ln Fertilizer/Acre
β / SE β / SE
CY PP DD >21C 0.0382** 0.0002
(0.0189) (0.0079)
CY GS1 DD >21C 0.1177*** -0.0260*
(0.0364) (0.0138)
CY GS2 DD >21C -0.0025 0.0045
(0.0196) (0.0057)
CY PP DD >21C*Bottom Wealth Tercile -0.0123 -0.0118*
(0.0086) (0.0068)
CY GS1 DD >21C*Bottom Wealth Tercile -0.0105 -0.0192*
(0.0163) (0.0108)
CY GS2 DD >21C*Bottom Wealth Tercile -0.0045 -0.0110
(0.0200) (0.0110)
Household FE Yes Yes
Prov-by-Year FE Yes Yes
Rainfall Controls Yes Yes
Observations 3726 6210
R2
5
Table A.11: Honoré Fixed Effects Tobit: Temperature, Pesticides and Fertilizer Use, by
Wealth (Round 1-5)
(1) (2)
β / SE β / SE
CY PP DD >21C 0.0282*** -0.0012
(0.0086) (0.0050)
CY GS1 DD >21C 0.0921*** -0.0293***
(0.0174) (0.0111)
CY GS2 DD >21C -0.0049 0.0067
(0.0104) (0.0106)
CY PP DD >21C*Bottom Wealth Tercile -0.0078 -0.0099*
(0.0082) (0.0056)
(0.0131) (0.0109)
(0.0196) (0.0135)
R2
Table A.12: Temperature, Pesticides and Fertilizer Use, by Wealth (Round 1)

(1) (2) (3) (4)
Pesticides Ln Pesticide/Acre Fertilizer Ln Fertilizer/Acre
CY PP DD >21C 0.0025 0.0130 -0.0002 -0.0046
(0.0017) (0.0112) (0.0006) (0.0062)
CY GS1 DD >21C 0.0072** 0.0493** -0.0014 -0.0165*
(0.0031) (0.0191) (0.0009) (0.0093)
CY GS2 DD >21C -0.0009 -0.0128 0.0004 0.0027
(0.0020) (0.0113) (0.0004) (0.0044)
CY PP DD >21C*Bottom Wealth Tercile -0.0013** -0.0091** -0.0003 -0.0022
(0.0006) (0.0038) (0.0004) (0.0038)
CY GS1 DD >21C*Bottom Wealth Tercile -0.0016 -0.0070 -0.0008 -0.0037
(0.0011) (0.0073) (0.0005) (0.0049)
CY GS2 DD >21C*Bottom Wealth Tercile 0.0015 0.0072 -0.0003 -0.0073
(0.0017) (0.0146) (0.0006) (0.0055)
Household FE Yes Yes Yes Yes
Observations 3726 3726 6210 6210
R2 0.588 0.588 0.740 0.788
6
Table A.13: Standard Tobit Estimates: Temperature, Pesticides and Fertilizer Use, by
Wealth (Round 1)
(1) (2)
β / SE β / SE
CY PP DD >21C 0.0381** -0.0018
(0.0191) (0.0079)
CY GS1 DD >21C 0.1193*** -0.0278**
(0.0369) (0.0140)
CY GS2 DD >21C 0.0010 0.0064
(0.0189) (0.0058)
CY PP DD >21C*Bottom Wealth Tercile -0.0113 -0.0058
(0.0090) (0.0062)
(0.0161) (0.0105)
(0.0203) (0.0083)
Household FE Yes Yes
R2
Table A.14: Honoré Fixed Effects Tobit: Temperature, Pesticides and Fertilizer Use, by
Wealth (Round 1)
(1) (2)
β / SE β / SE
CY PP DD >21C 0.0289*** -0.0025
(0.0085) (0.0050)
CY GS1 DD >21C 0.0919*** -0.0317***
(0.0174) (0.0115)
CY GS2 DD >21C -0.0029 0.0077
(0.0108) (0.0112)
CY PP DD >21C*Bottom Wealth Tercile -0.0085 -0.0054
(0.0076) (0.0053)
(0.0125) (0.0112)
(0.0179) (0.0140)
R2
7
Table A.15: Log Total Maize Output, Agricultural Input Use and Temperature
(1) (2)
Log Maize Yield/Acre (Kg.) Log Maize Yield/Acre (Kg.)
β / SE β / SE
CY PP DD >21C 0.0037 0.0038
(0.0041) (0.0035)
CY GS1 DD >21C -0.0027 0.0093
(0.0066) (0.0068)
CY GS2 DD >21C 0.0029 -0.0048
(0.0028) (0.0039)
Ln Pesticide/Acre 0.0507***
(0.0091)
Ln Fertilizer/Acre 0.0309***
(0.0088)
Ln Weeding Days/Acre 0.1067***
(0.0169)
Village FE Yes Yes
R2 0.374 0.406
Notes: Sample includes 1242 households balanced over 5 survey rounds (1996-97, 1999-00, 2003-04, 2006-07 and 2009-10). The
table presents the effects of temperature (captured via degree days (DD) over 21C) on total maize output. CY: current year;
PP: pre-planting or land preparation - onset of planting; GS1: planting or basal fertilizer application - onset of top dressing
fertilizer; GS2: top dressing fertilizer application - onset harvest. Standard errors are in parentheses, clustered by village.
*Significant at 10%. **Significant at 5%. ***Significant at 1%.
Table A.16: Accounting for Within-Day Temperature Variation: Log Total Maize Output,
Agricultural Input Use and Temperature
(1) (2)
Log Maize Yield/Acre (Kg.) Log Maize Yield/Acre (Kg.)
β / SE β / SE
CY PP DD >21C 0.0021 0.0047
(0.0043) (0.0037)
CY GS1 DD >21C -0.0083** -0.0028
(0.0033) (0.0036)
CY GS2 DD >21C 0.0002 -0.0036
(0.0029) (0.0029)
Ln Pesticide/Acre 0.0521***
(0.0090)
Ln Fertilizer/Acre 0.0306***
(0.0088)
Ln Weeding Days/Acre 0.1056***
(0.0170)
Village FE Yes Yes
R2 0.375 0.407
Notes: Sample includes 1242 households balanced over 5 survey rounds (1996-97, 1999-00, 2003-04, 2006-07 and 2009-10). The
table presents the effects of temperature (captured via growing degree days over 21C and 0-29C and over 29C) on total maize
output. CY: current year; PP: pre-planting or land preparation - onset of planting; GS1: planting or basal fertilizer application
- onset of top dressing fertilizer; GS2: top dressing fertilizer application - onset harvest. Standard errors are in parentheses,
clustered by village. *Significant at 10%. **Significant at 5%. ***Significant at 1%.
8
Table A.17: Controls for Daily Humidity: Temperature, Pesticides and Weeding Labor Days
(1) (2) (3)

Pesticides Ln Pesticide/Acre Ln Weeding Days/Acre
CY PP DD >21C 0.0023 0.0093 0.0004
(0.0014) (0.0090) (0.0024)
CY GS1 DD >21C 0.0068** 0.0470*** -0.0029
(0.0027) (0.0165) (0.0049)
CY GS2 DD >21C 0.0001 -0.0085 0.0063*
(0.0018) (0.0097) (0.0035)
Village FE Yes Yes Yes
Prov-by-Year FE Yes Yes Yes
Rainfall Controls Yes Yes Yes
Humidity Controls Yes Yes Yes
R2 0.338 0.355 0.179
Table A.18: Controls for Soil Moisture: Temperature and Fertilizer Use
(1) (2) (3) (4)
Fertilizer Fertilizer Ln Fertilizer/Acre Ln Fertilizer/Acre
CY PP DD >21C 0.0009 0.0004 0.0062 -0.0007
(0.0007) (0.0008) (0.0073) (0.0083)
CY GS1 DD >21C -0.0018 -0.0022 -0.0134 -0.0161
(0.0013) (0.0013) (0.0138) (0.0136)
CY GS2 DD >21C 0.0002 -0.0002 0.0010 -0.0041
(0.0006) (0.0004) (0.0065) (0.0047)
Village FE Yes Yes Yes Yes
Soil Moisture Controls No Yes No Yes
Observations 2352 2352 2352 2352
R2 0.587 0.588 0.644 0.645
fertilizer use. The table presents the effects of temperature (captured via degree days (DD) over 21C) on agricultural input use.
CY: current year; PP: pre-planting or land preparation - onset of planting; GS1: planting or basal fertilizer application - onset
of top dressing fertilizer; GS2: top dressing fertilizer application - onset harvest. Standard errors are in parentheses, clustered
by village. *Significant at 10%. **Significant at 5%. ***Significant at 1%.

Effects of Temperature On Agricultural Input Decisions in Moderate Climates

Uploaded by

Copyright:

Available Formats

Effects of Temperature On Agricultural Input Decisions in Moderate Climates

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Effects of Temperature On Agricultural Input Decisions in Moderate Climates

Uploaded by

Copyright:

Available Formats

In the Weeds: Effects of Temperature on Agricultural Input

C. Barrett¹; M. Jagnani²; Y. Liu³; L. You³

This version: January 7, 2018

JEL Codes: O13, Q15, Q16, Q54

2.1 Household Data

2.2 Kenyan Maize Calendar

2.3 Weather Data

http://www.nafis.go.ke/agriculture/maize/field-management-practices/ for recommendation on fertilizer application.

3.1 Temperature, Pests, Weeds and Pesticides

In the TAMPA survey, almost 40% of maize-farmers affected by changes in temperature

3.2 Fertilizer Use Under Liquidity Constraints

4 Temperature and Agricultural Input Use Response

β3 (30C > GDDGS2 > 21C)jdqt + f (Rainjdqt ) + αj + µqt + ijdqt (1)

4.2.1 The Response of Pesticides Use and Weeding Days to Temperature

Next, we examine effects on productivity-enhancing inputs like fertilizer. Descriptive data

4.2.3 Robustness Checks

We exploit plausibly random round-by-round variation in temperature at the village level

4.2.4 Hetereogenous Effects - by Wealth

4.2.5 Alternative Explanations

estimates remain largely unchanged.

Bloomfield, J. P., R. J. Williams, D. C. Gooddy, J. N. Cape, and P. Guha.

Boix-Fayos, C., A. Calvo-Cases, A. C. Imeson, M. D. Soriano-Soto, and I. R.

Cairns, J.E., K. Sonder, P.H. Zaidi, N. Verhulst, G. Mahuku, R. Babu, S.K.

Chakraborty, Sukumar. 2008. “Impacts of global change on diseases of agricultural crops

du Plessiss, Jean. 2003. “Maize Production.” Directorate of Agricultural Information Ser-

Fafchamps, Marcel. 1993. “Sequential Labor Decisions Under Uncertainty : An Estimable

Foster, Andrew D., and Mark R. Rosenzweig. 2010. “Microeconomics of Technology

Garrett, K. A., S. P. Dendy, E. E. Frank, M. N. Rouse, and S. E. Travers. 2006.

Gianessi, L. 2014. “Importance of Pesticides for Growing Maize in Sub-Saharan Africa.”

Hsiao, Cheng. 1986. “Analysis of panel data.” Cambridy University Press.

Kurukulasuriya, Pradeep, and Robert Mendelsohn. 2007. “Endogenous Irrigation:

Kurukulasuriya, Pradeep, and Robert Mendelsohn. 2008. “Crop switching as a strat-

Kurukulasuriya, Pradeep, Namrata Kala, and Robert Mendelsohn. 2011. “Adap-

Lobell, David B., Marianne Bänziger, Cosmos Magorokosho, and Bindiganavile

Marenya, Paswel P., and Christopher B. Barrett. 2007. “Household-level determi-

Marenya, Paswel P., and Christopher B. Barrett. 2009. “State-conditional fertilizer

Matlon, P J. 1990. “Improving productivity in sorghum and pearl millet in semi-arid

Mendelsohn, By Robert, William D Nordhaus, and Daigee Shaw. 1994. “The

Mendelsohn, Robert. 2008. “The impact of climate change on agriculture in developing

Mwalusepo, Sizah, Henri E.Z. Tonnang, Estomih S. Massawe, Gerphas O.

NAFIS. 2011. “Field Management.” National Famers Information Service - Kenya.

Suri, Tavneet. 2011. “Selection and Comparative Advantage in Technology Adoption.”

Thompson P., Thomas, and Carlos A. Baanante. 1989. “A Socioeconomic Study

Ward, J. M. J., M. D. Laing, and F. H. J. Rijkenberg. 1997. “Frequency and Timing

Welch, Jarrod R, Jeffrey R Vincent, Maximilian Auffhammer, Piedad F Moya,

Figure 1: Location of Sample Villages

Notes: Distribution of average daily temperatures in villages in TAMPA from 1990-2012.

Source: Agro-ecological zones - IFPRI Harvest Choice (www.harvestchoice.org.);

Agro-Ecological Zones (AEZs): Agro-ecological zones (AEZs) are geographical areas

Figure 5: Household Weeding Labor

Figure 9: Distribution of Yields by Pesticide Use

Table 1: Climate Change in Kenya?

Table 6: Why Didn’t You Use Fertilizer?

Table 7: Temperature, Fertilizer and Pesticide Use

Table A.1: Summary Statistics

Table A.2: Household FE: Temperature, Fertilizer and Pesticide Use

Removed Prov*Round FE Removed Dist*Round FE

Table A.4: District*Year FE: Temperature, Fertilizer and Pesticide Use

Table A.12: Temperature, Pesticides and Fertilizer Use, by Wealth (Round 1)

β3 (30C > GDDGS2 > 21C)jdqt + f (Rainjdqt ) + αj + µqt + ijdqt (1)

Removed ProvRound FE Removed DistRound FE