Impact of Climate Change On The Indian Economy: Evidence From Foodgrain Yields

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/256042157
Impact of Climate Change on the Indian Economy: Evidence from Foodgrain

Yields
Article in Climate Change Economics · September 2012

DOI: 10.2139/ssrn.2191010
CITATIONS READS
30 6,066
3 authors:
Shreekant Gupta Partha Sen

University of Delhi (Delhi School of Economics) delhi school of economics
47 PUBLICATIONS 785 CITATIONS 68 PUBLICATIONS 599 CITATIONS
SEE PROFILE SEE PROFILE
Suchita Srinivasan
Graduate Institute of International and Development Studies
1 PUBLICATION 30 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Indian Economic Review View project
Capital flows to India View project
All content following this page was uploaded by Partha Sen on 14 May 2018.
The user has requested enhancement of the downloaded file.

CDE
Accepted in September 2012
IMPACT OF CLIMATE CHANGE ON THE INDIAN

ECONOMY: EVIDENCE FROM FOODGRAIN YIELDS
SHREEKANT GUPTA
Email: sgupta@econdse.org
Delhi School of Economics
University of Delhi
&
LKY School of Public Policy (NUS)
PARTHA SEN
Email: partha@econdse.org
South Asian University
SUCHITA SRINIVASAN
Email: suchita.srinivasan18@gmail.com
American Express
Working Paper No. 218
Centre for Development Economics

Department of Economics, Delhi School of Economics
Impact of Climate Change on the Indian Economy:
Evidence from Foodgrain Yields
Shreekant Gupta, Partha Sen and Suchita Srinivasan*#
Abstract
This paper estimates the impact of climate change on foodgrain yields in India, namely rice
and millets. We estimate a crop-specific agricultural production function with exogenous
climate variables, namely, precipitation and temperature and control for key inputs such as
irrigation, fertilizer and labour. Our analysis is at the district level using a panel dataset for
physical yield (output per hectare – gross cropped area) for the period 1966-99. Thus, we
eschew crop simulation approaches that rely on experimental data. We do not also estimate
reduced form relationships between economic variables such as profits or the monetary value
of yield and weather measures. Consistent with other studies at the district and state level we
find significant impacts of climate change (temperature and precipitation) on Indian
agriculture. The implication of our results for inter-state disparities and corrective measures is
elaborated.
Key words: Climate Change; Agricultural Impacts; Developing Countries
JEL Classification: O13; Q54; R11
*
Authors are in alphabetic order. They are, respectively, with Delhi School of Economics and LKY School of
Public Policy, South Asian University and American Express. Corresponding author: Gupta
(sgupta@econdse.org, sgupta@nus.edu.sg)
#
We thank Centre for Development Economics, Delhi School of Economics for financial support through UK
Department of International Development (DfID) Purchase Order No. 40048622. We alone are responsible for
the findings and conclusions.
1. Introduction
Meteorological data compiled over the past century suggest the earth is warming. In keeping
with this, for India as a whole mean annual temperature shows a significant warming trend of
0.51 degrees Celsius per 100 years during the period 1901-2007 (Kothawale et al., 2010).
Similarly, global projections of temperature and for precipitation augur a warmer and wetter
world, on average. Simulations with regional climate models project similar trends for both
variables for India—by 2030s annual mean temperatures and summer monsoon rainfall are
both expected to increase on average (GoI 2010). But for a country of sub-continental
proportions there are, as expected, significant regional variations in past and future trends.
Both these facts have significant implications for the Indian economy in terms of the impact
of climate change and also its regional incidence.
This is particularly true for the agricultural sector where climate is a direct input into
production. While the salience of this sector in India has declined over the years, it still
remains important in the socio-economic fabric of the country. Though agriculture alone
(other than forestry and fisheries) accounts for only about 15% of GDP, it employs 55% of
the workforce [ref]. Moreover, according to the 2011 census 69% of the population remains
rural and intimately connected to this sector. There are also important forward and backward
linkages to agriculture from the rest of the economy.
Our primary interest in this sector, however, is from the perspective of poverty which remains
widespread despite significant progress in reducing it—the country is still estimated to have a
third of the world's poor. According to some estimates 32.7% of the population is below the
international poverty line of US$ 1.25 per day (PPP) while 68.7% live on less than US$ 2 per
day (World Bank 2011)1. A key aspect of poverty is its incidence which is rural and
concentrated among agricultural labour. Thus, according to the Indian Planning Commission,
in the year 2009-10 more than 1 in 3 of the rural population was poor (33.8%) whereas the
figure for urban areas was about 1 in 5 (20.9%). Equally important for our analysis, nearly
50% of agricultural labourers were below the poverty line in rural areas (GoI 2012).
1
The definition and measurement of poverty in India is controversial and politically charged and somewhat of
an obsession with Indian economists. We do not enter into this area other than providing illustrative numbers
for purposes of our argument.
1
Within agriculture we focus on two key food crops, namely, rice and millets. Using district
level panel data on annual yields (output per hectare), inputs, and climate variables for 1966-
99, we find significant impacts of rainfall and temperature on yields of both crops.
Extrapolating forward, we find climate change is likely to reduce yields of these crops
significantly2. We also focus on regional variations in our results to highlight the uneven
impact of climate change across India. The latter finding has policy implications in terms of
regional imbalances and also for the literature on convergence and divergence across Indian
states3.
Our methodology is based on estimating an agricultural production function with exogenous

climate variables, namely, precipitation and temperature. Thus, we eschew crop simulation
approaches that rely on experimental data4. We do not also estimate reduced form
relationships between economic variables such as profits or the monetary value of yield and
various forms of weather measures.
Our focus on rice and millets is deliberate. The former is the most important cereal food crop
in India accounting for 23.3% of gross cropped area and about 43% of total food grain
production as well as 46% of cereal production (Singh 2009). In economic terms, rice
(including paddy) ranked highest by value among all agricultural products in India with a
total output of about $38.4 billion in 2010 (FAO 2011a). India is considered to be one of the
original centers of rice cultivation covering 44 million hectares. Its rice harvesting area is the
largest in the world. Around 65% of the total population in India eats rice and it provides the
main source of income and employment for more than 50 million households (IRRI).
2
This is consistent with Guiteras (2009).
3
See for instance Kalra and Sodsriwiboon (2010) and Purfield (2006).
4
This approach is interestingly referred to as ‘production function’ approach a la Deschênes and Greenstone
(2007) and Guiteras (2009). A comprehensive discussion of the pros and cons of this approach is found in
Schlenker and Roberts (2009). For examples of the hedonic approach (also known as the Ricardian approach)
which focuses on variations in land prices or profits see the seminal paper by Mendelsohn et al. (1994). This
paper has spawned an extensive literature across several countries. Early applications of the Ricardian
approach to India are Dinar et al. (1998) and Kumar and Parikh (2001). A recent application to Africa
(Ethiopia) is Deressa and Hassan (2009).
2
From our perspective of rural poverty in particular, we note rice (actually paddy)5 as
cultivated in most parts of India, is a highly labour intensive crop. Also, much of this labour
especially during sowing and transplanting is provided by women. Thus, there is an
important gender dimension as well (FAO 2004 and FAO 2011b)6.
Millets on the other hand are traditional ‘coarse cereals’ whose importance is more in terms
of their role as a staple crop consumed by the poor7. In terms of food grain production
millets ranked fourth in India behind rice, wheat and maize (FAO 2011a)8. Within millets
we concentrate on the two key varieties, namely, pearl millet (bajra) and sorghum (jowar).
Post Green Revolution, millets have lost ground to other food crops, especially wheat and
rice--the production of millets has more or less remained constant between 1966-2006
whereas that of rice and wheat has increased by 125% and 285%, respectively (MNI et al.,
2009).
Table A1 provides normal (averaged over 2005-06 to 2009-10) area, production and yield of
foodgrains in India. As the table shows, rice and wheat now account for more than three
fourths of foodgrain output in India with ‘coarse cereals’ contributing only about one sixth
(16 percent). But as late as 1970 the picture was quite different--the contribution to foodgrain
output of rice, wheat and ‘coarse cereals’ in 1970-71 was respectively, 39%, 22% and 28.2%.
As we discuss later and anticipating our empirical results, this has important policy
implications vis-à-vis climate change—we find millets in particular to be much less sensitive
to temperature and rainfall than rice and by corollary more climate resilient.
5
Paddy (Oryza sativa) is the rice grain with husk (which is also known as rice hull). Paddy becomes rice after
the removal of husk by threshing. In 2010 India produced approx. 143 million tons of paddy and about 95
million tonnes of rice (FAO 2012). Rice husk is mainly used as fuel in boilers in rice mills.
6
Conversely, in Asia (and in sub-Saharan Africa), women who are employed are more likely to be employed in
agriculture than in other sectors--almost 70 percent of employed women in Southern Asia work in agriculture
(FAO 2011b).
7
In India ‘coarse cereals’ (a loaded and highly normative term) comprise millets, barley and maize (Table A1).
Millets in turn primarily comprise pearl millet (bajra) and sorghum (jowar) which dominate in area and
production. This is followed by finger millet (ragi) and small millets. Bajra, jowar and maize together account
for over 90% each of area and production of ‘coarse cereals’ in India (Table A1).
8
This was a distant fourth rank, however—production of these four crops in 2010 was, respectively, 95, 81,
14.1 and 13.3 million tons (FAO 2011a). By value too, millet output was about $2.3 billion for the same year
(compared to $38.4 billion for rice).
3
Again, from a poverty perspective both rice and millets are largely grown by small and
marginal farmers (i.e., those with holdings less than 2 hectares) and with much less irrigation
as compared to wheat which is almost entirely irrigated and cultivated by relatively bigger
farmers. Table A2 shows 57% of area under rice is cultivated by marginal and small farmers
compared to 44% for wheat. At the other end of the spectrum large wheat farmers (holdings
of 10 hectares or more) account for more than 10% of the area under their crop compared to
similar large rice farmers who account for only about 5.5% of the area under rice. Finally,
90% of area under wheat is irrigated compared to 59% for rice and only 29% for other cereals
(coarse cereals).
The plan of the paper is as follows. The following section provides the context in terms of
projected trends and regional variation in climate variables (rainfall and temperature) for
India. It also provides further detail on how, when and where our two crops are grown in the
country and the potential role of climate variables. Section 3 presents our conceptual
framework and situates it in the literature on the impact of climate on agriculture, particularly
with regard to India. Section 4 describes the data and econometric methodology. Section 5
presents and interprets the results of our analysis including the likely effect of climate
change. Section 6 concludes the paper.
2. Climate change and agriculture in India
2.1 Trends and regional variation in climate variables

India’s climate system has unique features that are often not well captured by global climate
models, the most widely used being coupled atmosphere-ocean-sea-ice-land-surface models
(AOGCMs) with a resolution of 250-300 km grids. India’s climate is dominated by the
summer or southwest monsoon (and to a lesser extent the winter or northeast monsoon) and
by the country’s physiological features such as the western and eastern ghats, the central
plateau and the Himalayas. The summer monsoon and the rains that it brings are a major
weather phenomenon in the Indian subcontinent and deeply influences the lives of its
inhabitants. It is a four-month period (June-September) when massive convective
thunderstorms dominate India's weather, and is the Earth's most productive wet season
(Collier and Webb 2002). This season provides over 80% of India’s rainfall (Bagla 2006,
2012). Thus the quantity, temporal and spatial distribution of the precipitation (rainfall)
accompanying the monsoon is its most monitored component and is particularly important for
4
agriculture. The summer monsoon is also the most economically important weather pattern
to the extent a recent Indian Finance Minister (Pranab Mukherjee) called it “the real finance
minister”9. Yet it is “only partially understood and notoriously difficult to predict”
(Wikipedia 2012).
Though the dependence of agriculture on the summer monsoon has reduced somewhat lately
due to increased irrigation and better drought management, Figure 1 shows clearly how the
two have marched in lockstep. This is unsurprising since, out of the total net sown area of
141.0 million hectares (Mha) in India, rainfed area accounts for 85.0 Mha spread over 177
districts. This constitutes approximately 60 percent of the total farming area in the country.
Rainfed agriculture contributes 44% of the total food grain production of the country and
produces 75% of pulses and more than 90% of sorghum, millet and groundnut from arid and
semiarid regions. Even after half a century of lopsided policies that have focused on pockets
of the country and specific crops, rainfed regions provide livelihood to nearly 50% of the
total rural workforce and sustain 60% of cattle population of the country (MNI et al., 2009).
Keeping in view the peculiarities of India’s climate and of the summer monsoon in particular,
in a recent exercise a regional climate model (RCM) with 50 km resolution, namely, PRECIS
was deployed to dynamically downscale global model simulations and superimpose regional
details from India (GoI 2010). This exercise coupled with long instrumental records allows
us to capture past trends and also make projections for key climate variables such as
temperature and rainfall within the country at a disaggregated level.
As mentioned earlier, for India as a whole mean annual temperature shows a significant
warming trend of 0.51 degrees Celsius per 100 years during the period 1901-2007
(Kothawale et al., 2010). More important, accelerated warming has been observed in the last
approximately 40 years (1971-2007), mainly due to intense warming in the recent decade
1998-200710. Increases in the mean have been accompanied by a rise in both maximum and
minimum temperatures at the all India level—by 0.71 and 0.27 degrees Celsius, respectively,
9
“India cheers as monsoon arrives; hope of better farm output raised” Hindustan Times May 31, 2010.
10
All four major Indian seasons, namely, summer/pre-monsoon, monsoon, post-monsoon and winter contribute
to this trend. But the increase in winter and post-monsoon temperatures is most marked—by 0.80 and 0.82
degrees Celsius, respectively over the last hundred years (GoI 2010). This has significant implications for rice
yields as noted below.
5
per 100 years during the period 1901-2007. Also, as with mean temperature, there has been
an acceleration in trends of both maximum and minimum temperatures during 1971-2007. At
the regional level, the homogenous regions11 of east coast, west coast and the peninsula show
an increasing trend in the frequency12 of hot days but northern India (north of 22 degree N)
does not. With respect to the frequency of cold days, however, all seven homogenous regions
show a decreasing trend (in the frequency of cold days).
With regard to precipitation, Indian monsoon rainfall from 1871-2009 shows only a slight
negative trend. But there is significant spatial variation in these trends over this period.
Also, there is an increase in extreme rainfall events and their intensities (GoI 2010, Sen Roy
and Balling 2004).
Projections for climate in the medium-run for India seem to indicate it will be warmer and
wetter but with significant regional variation. Overall there will be (i) an increase in average
surface temperature by 2-4 degrees C, (ii) changes in the distribution of rainfall (inter-temporal
and spatial) during both monsoon and non-monsoon months, (iii) decrease in the number of rainy
days by more than 15 days, (iv) an increase in the intensity of rainfall by 1-4mm/day, and (v) an
increase in the frequency and intensity of cyclonic storms (Ranuzzi and Srivsatava 2012)
2.2 Rice and millets in India: key issues

Rice is the most important food crop in India13. It accounts for 23.3% of gross cropped area
and 43% of total food grain production as well as 46% of cereal production (Singh 2009). In
economic terms, rice (including paddy) ranked highest by value among all agricultural
products in India with a total output of about $38.4 billion in 2010 (FAO 2011a). The
production of rice alone has more than quadrupled from around 21 million tons in 1950 to 95
million tons in 2010.
There are three seasons for growing rice in India—autumn (pre-kharif), winter (kharif) and
summer (rabi)—named according to the season of harvest (though all crops are not grown in
11
A uniform or homogenous region is an area in which everyone shares in one or more distinctive
characteristics, in this case climate.
12
Defined as days per decade.
13
This discussion is based on Singh (2009).
6
all regions). Winter or kharif rice (sown during Jun-Jul and harvested in Nov-Dec) is the
main growing season and accounts for 84% of the country’s rice crop14. This is followed by
summer rice (sown during Nov-Feb and harvested in Mar-Jun) at 9% and autumn rice (sown
during May-Aug and harvested in Sept-Oct) which accounts for 7% of the rice crop.
Among millets, pearl millet (bajra) is the most widely grown type of millet followed by
sorghum (jowar)15. Because of its tolerance to difficult growing conditions such as drought,
low soil fertility and high temperature, it can be grown in areas where other cereal crops, such
as maize or wheat would not survive (Basavaraj et al. 2010). Pearl millet in India is grown as
a single season crop. Cultivation predominantly takes place on marginal lands and unirrigated
lands. It is also grown in a small area as summer crop under irrigation particularly in the
northwestern states of India mainly as a fodder crop. Area trends of pearl millet in India are
constantly declining. Between 1972–73 and 2004–05, nearly 3 million ha has been diverted
from pearl millet cultivation to other crops. Pearl millet production is concentrated in
Gujarat, Maharashtra and Rajasthan which account for 70% of production in India. These
states also have the highest concentration of pearl millet consumers since bulk of the
consumption for food use takes place in the growing areas.
3. Framework and relevant literature
3.1 Conceptual framework

Our methodology is based on estimating an agricultural production function with exogenous
climate variables, namely, precipitation and temperature. Our analysis is at the district level
using a panel dataset for physical yield (output per hectare – gross cropped area) for rice and
pearl millet. Thus, we eschew crop simulation approaches that rely on experimental data.
We do not also estimate reduced form relationships between economic variables such as
profits or the monetary value of yield and arbitrary forms of weather measures.
3.2 Relevant literature

Several studies have looked at the impact of climate related variables on crop yields
specifically for India. Lahiri and Roy (1985) (L-R) look at the supply response of rice yields
14
As can be noted sowing and transplanting of winter rice is during the summer monsoon (Jun-Sept).

15
This discussion is based on Basavaraj et al. (2010).
7
at the all India level and also include monthly rainfall (either in the sowing season or in the
growing season). The paper is in the agricultural economics tradition of acreage and yield
response to price (and also to “supply shifters” such as rainfall) and models this response in a
Nerlovian partial adjustment framework (for which they actually do not find evidence). L-R
postulate a gamma distribution for the effect of rainfall on yield (right skewed and bounded at
zero), i.e., less rainfall (droughts) is worse than too much (floods). For yield they find the
optimal monthly average rainfall is about 293mm for the months of July and August. They
also argue that with the spread of HYVs post-mid 1960s (1965 onwards) Indian agriculture
has become more rainfall dependent esp. since water requirement has gone up and the spread
of irrigation has not kept pace with it.
Kanwar (2006) extends this type of work to several food grains. He looks at supply response
using a state level panel dataset and again finds rainfall matters considerably for supply
response (it’s a ‘supply shifter’)16. Auffhammer et al. (2012) extend Auffhammer et al
(2006) and explicitly look at the impact of too little/too much rainfall (akin to gamma
rainfall) on rice yields using a state level panel dataset. They too find significant impacts.
A problem in doing state or national level analysis is the need to aggregate rainfall and other
weather data (there are several observation stations in a state) to one value at the state or
national level. This is problematic since several Indian states are large, often bigger than
countries in Europe and elsewhere17. Given the variation in rainfall and other weather
variables for a state the resulting measurement error may bias the coefficients on weather
variables downward Auffhammer et al. (2012).
As in this paper a district level panel dataset for India is used by Guiteras (2009) and also by
Fishman (2011) and Krishnamurthy (2012). Guiteras (2009) examines the impact of
temperature and rainfall on combined yield (in money terms) for five major food and one
cash crop, namely, rice, wheat, jowar, bajra, maize and sugarcane. He finds that climate
16
“In other words, rainfall is the single most important factor determining supply response even today. Despite
decades of massive irrigation schemes, the food crops continue to be rainfall-dependent.” (op. cit., p. 80)
17
For instance, the five biggest states of India, namely, Rajasthan, Madhya Pradesh (MP), Maharashtra, Andhra
Pradesh (AP) and Uttar Pradesh (UP) range from 342,000 to 241,000 sq kms. The biggest state, Rajasthan is
almost as big as Germany, whereas the next two (MP and Maharashtra) are almost the size of Poland and bigger
than Italy and the Philippines each. AP and UP, respectively are bigger than or the same size as United
Kingdom.
8
change could reduce yields by 4.5 to 9% in the medium-run (2010-39) and by as much as
25% in the long-run (2070-2099) in the absence of long-run adaptation. The main drawback
of Guiteras as highlighted by Sarker et al. (2012) and by Krishnamurthy (2012) is combining
different crops which are impacted differently by climate change. The dependent variable is
akin to district income (from 6 crops) 18 normalised by area to arrive at gross revenue per
hectare, and is hard to interpret.
Fishman (2011) also uses a district level panel and shows the impact of intra-seasonal
variability of rainfall on yields. He shows that this effect can be moderated by spread of
irrigation (so called adaptation) but the effect varies with groundwater depletion. Further, no
evidence is found that irrigation mitigates increased heat exposure 19.
4. Data and methodology
4.1 Data sources

4.1.1 Agricultural data- The data on the agricultural variables span the time period 1966-
1999, and have been collected from the ICRISAT VDSA (Village Dynamics in South Asia)
Apportioned Meso database. This is a district-level database, which includes information on
crop area and production for several crops, irrigated area under different crops, land use, farm
harvest prices, rainfall, livestock and agricultural implements, operational holdings, and
population census data for 16 states of the country. The boundaries of the districts included
in this database are defined as of 1966, i.e. any data on districts that were created after 1966
are given ‘back’ to the parent districts from which the newer districts were created. The final
database thus includes data on the parent districts only, which is inclusive of the data
‘apportioned’ amongst the districts created later.
The variables of interest in this database include the area under, and production of, rice, pearl
miller and sorghum (measured in hectares and tons respectively), district-wise consumption
of fertilizers (tons of nitrogen, phosphate and potash fertilizers used), total gross cropped area
in each district (measured in hectares, and accounting for multiple cropping), gross irrigated
18
As stated by Guiteras (op. cit.) “these comprise roughly 75% of total revenues.” (p. 9 footnote 6).
19
Inter alia, he uses a quadratic time trend fs(t) which is state specific—it reflects technological progress and
productivity gains, which are allowed to differ from state to state because of the large variance in agricultural
performance across India.
9
area under each of the three crops (measured in hectares, again accounting for multiple
cropping), and total agricultural labour of each district. While the data on the area and
production have been compiled into the ICRISAT database from the Statistical Abstracts of
India, State Statistical Abstracts and the State Season and Crop reports, the data on fertilizer
consumption has been obtained from the Fertilizer Statistics for India. The agricultural labour
data has been procured from the decadal population censuses. Barring data gaps for a few
variables in some districts, the area, input and output-related data is available at an annual
frequency for both crops, whereas the data on agricultural labour is only available for the
census years in the relevant range (1971, 1981, and 1991). For the other years in the sample,
this data has been interpolated (by fitting a linear trend on population growth).
The dependent variable considered is the logarithm of yield (tons of output per hectare), for
each of the three crops. The three main independent variables created from the list of
ICRISAT variables are fertiliser consumption, labour consumption, and irrigation. Since
information on fertiliser consumption is only available at the district level in overall form (i.e.
for all crops combined), this variable had to be made relevant for the two crops in question.
This was done by weighing total fertiliser consumption in a district by the ratio of area under
the crop to the gross cropped area of that district (assuming uniform distribution of fertilisers
over the cropped area). Similarly, data on labour consumption is also available at the overall-
level for each district. The same weight was applied to total labour consumption of a district,
to create the labour variables for the crops. The irrigation variable has been created by taking
the ratio of gross cropped area under rice/ millet to the gross cropped area in the entire
district. The fertilizer consumption and labour consumption variables, in turn, have been
divided by the total area under the crop.
4.1.2 Climate data- The climatic data has been procured from the India Water Portal domain
(www.indiawaterportal.org). The portal contains 102 years of climatic data for the districts of
India, which includes monthly data on precipitation, temperatures, cloud cover, humidity, and
ground frost frequency, for example. The database that is used to compile this meteorological
dataset is the publicly available Climate Research Unit (CRU) TS2.1 dataset, out of the
Tyndall Centre for Climate Change Research, School of Environmental Sciences, University
of East Anglia in Norwich, UK. This dataset consists of interpolated (on 0.5 degree latitude-
longitude grid) global monthly data on variables such as rainfall and temperature from 1901
10
to 2002. The CRU data was transformed to the district level by simple linear averaging from
the gridded data of the CRU dataset.
Two independent variables have been created using this database, i.e. rainfall and
temperature. In case of rice, the rainfall variable has been defined as the total growing-season
precipitation for a given district, i.e. total rainfall in the months of June, July, August and
September (Deschenes and Greenstone 2007 and Guiteras 2009). The rainfall in the growing
season/ during transplantation of the crop is found to contribute immensely to the growth of
the rice crop, especially the Kharif crop, which comprises about 84% of the total rice output
of the country. The temperature variable for rice has been defined as the average of the
maximum temperatures in the months of November and December of each year, normalized
by long-term (1901-1999) average of the maximum temperatures for these two months,
divided by the long-term standard deviation of the maximum temperature for these two
months. When seasonal variations are present within a set of data, it is useful to express the
data in terms of standardized anomalies. Moreover, standardized/normalised variables
generally provide more information about the magnitude of anomalies in the data, because
influences of dispersion have been removed. The average of the maximum temperatures in
the months of November and December has been used, as it is the maximum temperature
(especially at night) which during the time of harvest is found to affect the yield of rice
significantly. In the case of pearl millet and sorghum, the rainfall variable is the (normalised)
total annual rainfall for a given district, whereas the temperature variable is the (normalised)
annual average of the average temperature. Both climactic variables have also been used in
the quadratic form (defined as temp2 and rain2), to understand the incremental impact of
rainfall and temperature on the yield of all three crops. Table A3 provides summary statistics
of the main variables used in the analysis.
4.2 Methodology
In the case of rice, 102 districts were chosen across the top 9 rice producing states of the
country (West Bengal, Andhra Pradesh, Uttar Pradesh, Tamil Nadu, Orissa, Madhya Pradesh,
Karnataka, Punjab, and Bihar) 20. In case of pearl millet, 69 districts were chosen amongst
20
These nine states account for about 90% of the total production of rice in India, as of 1995. Adding additional
states to the sample does not increase this proportion significantly. The top 102 districts have been chosen in a
manner so as to account for about 80-85% of the output of rice in each state (adding extra districts to the sample
increases this proportion only marginally).
11
the top 9 pearl-millet producing states of India (Rajasthan, Gujarat, Maharashtra, Madhya
Pradesh, Karnataka, Andhra Pradesh, Tamil Nadu, Orissa and Haryana) 21, whereas for
sorghum the top 58 producing-districts were chosen amongst the top seven states
(Maharashtra, Madhya Pradesh, Karnataka, Andhra Pradesh, Tamil Nadu, Gujarat and
Rajasthan). Table A4 in the appendix lists the districts considered in each state for the three
crops. As previously mentioned, the districts included in the ICRISAT database are those
that existed as of 1966. However, the climatic dataset has been created taking into account
the district boundaries as of 2002, which are remarkably different from those of 1966. The
districts that comprise the panel-sample have been selected on the basis of the districts that
existed in the ICRISAT database, and the climatic variables for these districts have been
approximated by the district to which the largest area of the parent district was allocated
(provided that it is more than 50% of the total area of the parent district).
In the presence of AR (1) serial autocorrelation (the outcomes are correlated across years for
a given district) and cross-sectional dependence (the outcomes are correlated across districts
in a given year), along with heteroscedasticity, FGLS (feasible generalised least squares) with
fixed effects was found to an appropriate method of estimation. However, one of the
drawbacks of FGLS estimation is that it produces overly optimistic standard error estimates.
Moreover, the estimates are only feasible if N<T, i.e. the number of observations are less than
the number of time period, which is not the case for any of the three crops. To correct for this,
panel-corrected standard error (PCSE) estimates are obtained, where the parameters are
estimated using a Prais-Winsten (or OLS) regression. Equations have been estimated with
district and year fixed effects, and district fixed effects and district-by-year fixed effects.
In the case of rice, it was observed that that the errors exhibited the presence of
heteroscedasticity, serial autocorrelation, and also contemporaneous correlation. A Prais –
Winsten regression was thus estimated, under two different assumptions on correlation:
1) Within panels, there is AR (1) autocorrelation and that the coefficient of the AR (1)
process is common to all of the panels, and
2) Within panels, there is AR (1) autocorrelation and that the coefficient of the AR (1)
process is specific to each panel (i.e. panel-specific AR (1) autocorrelation)
21
The states considered for pearl millet and sorghum account for nearly the entire output of these crops, while
the districts known to be the largest producers have been chosen.
12
However, in the case of pearl millet, the errors were found to exhibit heteroscedasticity, but
no serial autocorrelation or contemporaneous correlation, whereas for sorghum, the errors
exhibited heteroscedasticity, and contemporaneous correlation, but no serial correlation. A
linear panel-data regression was then estimated, under the different assumptions on
autocorrelation than which held for rice.
The regression equations which are estimated all three crops are as follows:
Ln (YIELDit) = αi+ γt + Xit'β + ∑ ε* f (Wit) + µit,

Ln(YIELDit) = αi+ εt*t + Xit'β + ∑ ε* f(Wit) + µit, and
where αi refer to the district-level fixed effects, which are quite useful in capturing
unobserved heterogeneity across districts and γt refer to the year-specific dummies which
control for annual differences in yield, common to all the districts (Deschenes and
Greenstone, 2007). The second equation estimates the same equation, but with district-by-
year fixed effects, rather than simple year fixed effects. The Xit refer to the district and year-
specific agricultural variables, whereas the Wit refers to the climactic variables (namely
rainfall and temperature). State and year fixed effects were not estimated, given that the
sample size would be too small to run a robust panel-data estimation.
5. Results
5.1 Rice
Tables 1 and A5 (in the appendix) give the results the regression estimation for rice, taking
the temperature variable as the average of the maximum temperatures in the months of
November and December, and defining the rainfall variable as the total growing season
(June, July, August and September) rainfall. Table 1 gives the regression results with district
and district-year fixed effects, while table A5 gives the results with district fixed effects, and
year fixed effects. The former specification accounts for both district-specific time-invariant
unobservable characteristics, and also for district-specific time-varying unobservable
characteristics, that trend smoothly over the sample period. These tables give the results,
assuming AR (1) serial autocorrelation exists. The results are robust, even if panel-specific
AR (1) serial autocorrelation is assumed. For these results, refer to tables A6 and A7 in the
appendix. The coefficients on the district and year fixed effects have been suppressed.
13
Table 1: Rice with district level and district- year fixed effects
Number of obs = 3464 R-squared 0.6422
Wald chi2 (41) = 5564.04 Prob > chi2 0.0000
Panel
lnyield Coeff. Corrected z P > |z| 95% Confidence Interval
Std. Errors
rainfall 0.000658 0.0001507 4.37 0 0.000363 0.000953
rain2 -1.60E-07 3.92E-08 -4.09 0 -2.37E-07 -8.33E-08
maxtemp -0.02046 0.0062746 -3.26 0.001 -0.03275 -0.00816
temp2 0.000308 0.0001743 1.77 0.077 -3.40E-05 0.00065
labour 0.178417 0.0907443 1.97 0.049 0.000561 0.356273
fertilizer 0.003202 0.0003903 8.2 0 0.002437 0.003967
irrigation 0.196569 0.0795174 2.47 0.013 0.040718 0.352421
Intercept -0.96013 0.1804848 -5.32 0 -1.31387 -0.60639
Table 1 reveals that with district and district-year fixed effects, both the rainfall and
maximum temperature variables are found to be significant, even at the 1% level of
significance. Higher rainfall leads to higher yield of rice, whereas higher temperatures lower
the yield. The quadratic variable for rainfall is significant, and negative, which means that
higher rainfall leads to higher yield, but at a decreasing rate. Interestingly, the coefficient on
the quadratic term for temperature is positive: higher temperatures would mean lower yield
rates, but the incremental impact of a one unit increase in the temperature variable becomes
more severe as the temperature increases. Both the irrigation and fertilizer consumption
variables are highly significant, which is expected given that rice production is highly input-
driven in large parts of the country, and the signs are intuitive (higher the proportion of land
under rice irrigated, higher is the yield, and higher the fertilizer consumption used for rice,
higher is the yield of rice). Labour too is found to have a positive impact on the yield of rice,
even though the variable is slightly less significant (still significant at the 5% level). This
particular specification is appropriate, given that it captures the effects of district-specific
time-invariant unobservable characteristics, and also district-specific time-varying
unobservable characteristics, that trend smoothly over the sample period.
Table A5 (in the appendix) captures the district and year fixed effects, without accounting for
the time-varying unobservable factors. All the variables retain the same signs as in Table 1.
The only difference is that labour now has a negative sign, which is counter-intuitive: more
the labour consumption, lower is the yield of rice. Some variables are no longer significant,
14
such as the temperature quadratic term, labour consumption, and the temperature term is
significant only at the 10% level of significance. These results also hold once panel-specific
errors are taken into account (tables A6 and A7 in the appendix).
5.2 Pearl millet (bajra)

Tables 2 and A8 (in the appendix) produce similar results for pearl millet as for rice, with the
difference being in the definition of the climactic variables. Rainfall has been defined as the
(standardized, using long term mean and standard deviation) total annual rainfall for each
district, while temperature has been defined as (standardized, using long term mean and
standard deviation) average annual temperature. Due to lack of serial correlation in the data,
the errors are assumed to be independent in this case. Table 2 gives the results for district
and district-year fixed effects, while table A8 (in the appendix) gives results with district
fixed effects and year fixed effects. The coefficients on fixed effects have been suppressed.
Table 2: Pearl Millet with district-level and district-year fixed effects

Wald chi2 (75) = 5567.92 Prob > chi2 0.0000
Panel Corrected
lnyield Coeff. z P > |z| 95% Confidence Interval
Std. Errors
rainfall 0.135329 0.011123 12.17 0 0.113529 0.157129
rain2 -0.0609 0.008664 -7.03 0 -0.07788 -0.04392
avgtemp -0.00034 0.097327 0 0.997 -0.19109 0.190422
temp2 -0.13075 0.145426 -0.9 0.369 -0.41578 0.15428
irrigation -1.11178 0.405889 -2.74 0.006 -1.90731 -0.31625
fertilizer 0.003491 0.000499 7 0 0.002513 0.004469
Labour -0.26723 0.11032 -2.42 0.015 -0.48345 -0.05101
Intercept -0.81681 0.083299 -9.81 0 -0.98007 -0.65355
Table 2 reveals a similar impact of higher rainfall on pearl millet production, as in case of
rice (greater the rainfall, higher the yield). The variable is also found to be highly significant.
While the average temperature variable is (highly) insignificant, the sign is intuitive. The
sign of the rainfall quadratic term is identical to that of rice (higher rainfall means that the
incremental impact of rainfall on pearl millet yield declines), however the sign of the
temperature quadratic is negative: higher temperatures mean that the incremental impact of
temperatures on pearl millet yield is dampened. It is also highly insignificant. The impact of
15
the other agricultural variables appears to be significantly diluted in case of pearl millet: both
labour usage and irrigation seem to have a negative impact on the yield of rice, even though
they are significant at the 5% level of significance.
Table A8 gives mostly similar results to table 2, in terms of the signs and significance of the
climactic variables. The difference in the agricultural control variables is that the fertilizer
variable is now insignficant. Labour consumption and irrigation still have a negative impact
on the yield of pearl millet, whereas fertilizer consumption has a positive impact. Another
difference from table 2 is that the quadratic term for temperature (though insignificant) has a
positive sign: as temperatures increase, yield of pearl millet declines; moreover, the negative
impact is getting exacerbated over increases in temperature. The coefficient for quadratic for
rainfall retains its positive sign.
5.3 Sorghum (jowar)

Tables 3 and A9 (in the appendix) present the results for sorghum, assuming the same
dependent variable as for the other two crops, while defining the climactic variables in a
similar manner as was done for pearl millet. Similar to pearl millet, there is a lack of serial
correlation in the data, hence the errors are assumed to be independent. Table 3 gives the
results for district and district-year fixed effects, whereas table A9 (appendix) gives results
with district fixed effects and year fixed effects. The coefficients on the fixed effects have
been suppressed.
Table 3: Sorghum with district and district-year fixed effects

Wald chi2 (34) = 9015.6366 Prob > chi2 0.0000
Panel
Std. Errors
rainfall 0.068505 0.021986 3.12 0.002 0.025414 0.111597
rain2 -0.05289 0.015469 -3.42 0.001 -0.08321 -0.02257
avgtemp -0.10864 0.220206 -0.49 0.622 -0.54023 0.322958
temp2 -0.03371 0.337748 -0.1 0.92 -0.69568 0.628264
labour -0.04893 0.181818 -0.27 0.788 -0.40529 0.307423
fertilizer 0.001117 0.000612 1.82 0.068 -8.30E-05 0.002317
irrigation -0.24422 0.3488 -0.7 0.484 -0.92785 0.439417
Intercept -1.24453 0.131083 -9.49 0 -1.50145 -0.98761
16
According to Table 3, the climactic variables seem to have similar effects on the yield of
sorghum as they did on the other two crops. Higher rainfall means higher the yield, higher
the average temperature, lower the yield. Higher the temperature, lower the rate of decrease
of yield with temperature (as was the case with pearl millet), and higher the rainfall, lower the
rate of increase of yield with rainfall. However, the temperature variables are found to be
highly insignificant. Moreover, the agricultural control variables seem to have counter-
intuitive signs, i.e. higher the labour or irrigation, lesser is the yield of sorghum; these
variables are also insignificant. Table A9 also gives similar results, with respect to the signs
of the variables, other than the fact that fertilizer consumption is found to have a negative
impact on yields, while irrigation is found to have a positive impact. The fertilizer and
irrigation variables are highly insignificant, along with the temperature variable, while even
labour is insignificant at the 5% level of significance.
5.3 Regional differences

The regressions for rice, pearl millet and sorghum were also run using region-level interactive
terms, to gauge the impact of rainfall and temperature on specific regions. These regions
have been decided on the basis of geographical proximity, and similarity of climate,
agricultural conditions, etc. The states comprising the sample for the three crops have been
divided as follows:
Rice:
West Bengal and Bihar- Eastern region
Punjab and UP- Northern region
MP and Orissa- Central region
AP, Karnataka and Tamil Nadu-Southern Region
Pearl Millet (bajra):

Haryana, Rajasthan and UP- Northern region
Gujarat, MP and Maharashtra- Central region
Sorghum (jowar):
Rajasthan- Northern region
Gujarat, MP and Maharashtra- Central region
17
Table A10 (in the appendix) gives the results for rice, where the regression has been carried
out using district and year fixed effects, along with the interactive terms between the region
indicators and both the rainfall and temperature variables, assuming AR(1) serial
autocorrelation. While the significance of the original climactic variables drops
tremendously, this exercise is useful in understanding the differential impact of climate
change across regions. The district and year fixed effect terms have been suppressed. Table
A11 in the appendix gives the results once panel-specific AR (1) serial autocorrelation has
been assumed.
Tables A12 and A13 in the appendix illustrate similar results for pearl millet and sorghum
respectively, assuming heteroscedasticity, and no AR (1) serial autocorrelation. Again, the
district and year fixed effect terms have been suppressed.
Tables A10 and A11 illustrate the differential regional impact of climate change in case of
the rice crop. For instance, for the central region, the positive effect of higher rainfall on yield
of rice has been exacerbated, i.e. the (net) coefficient on rainfall for the central region is
higher than the overall coefficient. This means that yield in this region responds more, to a
given change in rainfall. However, given that over time, rainfall has declined in this region,
so would have the yield of rice. For the northern and eastern regions, the positive effect of
more rainfall is dampened: it would lead to a lesser increase in yield, even though the final
coefficient still remains positive. However, again given that the rainfall in these regions too
has declined over time, it seems fair to say that the reduction in yield would not be as much
as it would be at the overall level. Similarly, in case of temperature, for the southern and
eastern regions, the negative impact of higher temperatures on rice yield is getting dampened:
in case of the southern region, the final effect has still remained negative, while in case of the
eastern region, the final effect has turned positive! Given that temperatures have been rising
in both eastern and southern India, this means that while the yield of rice in southern India
has fallen less than it has at an overall level (even though it has still fallen), the yield of rice
in eastern India has actually gone up due to higher temperatures. Conversely, in central India,
the negative impact of higher temperatures on yield has only been worsened: higher
temperatures would reduce yield even further in central India than they would at an overall
level.
18
Tables A12 and A13 also gives some interesting results: in the central and northern regions,
the positive impact of higher rainfall on the yield of pearl millet is getting strengthened:
however, given that rainfall has declined in both these regions over time, the impact on yield
of this crop in these regions would have been more negative than at an overall level. For the
temperature effects, in both southern and central India, the (net) effect of temperature on the
yield of pearl millet is turning positive. This bodes well for the pearl millet crop, if
temperatures have been rising in both these regions of the country. Similarly, in case of
sorghum, the net effect of rainfall on its yield remains positive in both the central and
southern states, even though the (positive) effect is mellowed down. For the temperature
variable however, it appears as if in the central states, the net effect is becoming positive, i.e.
higher the temperature, higher the yield, whereas the effect remains negative for the three
southern states (Tamil Nadu, AP and Karnataka). This is well-supported by evidence, which
suggests that the yield of sorghum is negatively affected by declines in temperature.
5.4. Interpretation of Regression Results

In the case of rice, the results reveal the strong positive effect of rainfall on its yield, across
different specifications of the regression equation. Moreover, the positive impact of rainfall
on its yield is dampened, as the level of rainfall increases, or conversely, the negative impact
of lesser rainfall on yields is exacerbated as rainfall reduces. Rainfall has deteriorated over
time in most regions, other than the southern, which has experienced higher rainfall. Hence,
in the sample considered, in 4 out of 9 states (MP, Orissa, West Bengal and Bihar), declining
rainfall may be a contributing factor towards declining yields of rice. Incidentally, these are
also some of the economically weaker states of the country.
The direction of impact of maximum temperature on the yield of rice is as expected- as

temperatures increase, the yield of rice declines. Moreover, higher temperatures have meant
that the rate of decline of yield with increase in temperature also increases. The regional
effects are also interesting: in the states of AP, Karnataka, Tamil Nadu, West Bengal and
Bihar, higher temperatures have not had as severe an impact on the yield of rice as at the
overall level. In fact, in West Bengal and Bihar, they may even have had a slightly positive
impact. However, in the states of MP and Orissa, higher temperatures have meant an even
more severe decline in the yield of rice than at the overall level.
19
The agricultural controls such as irrigation, labour and fertilizer consumption have positive
and strong impact on the yields. Labour, however, is found to affect yield negatively when
time-variant unobservable characteristics are not taken into account. This may be attributed to
the surplus labour available for rice production in the country. Moreover, labour input
intensity is higher in districts which are already poor in terms of yield, where yields may be
falling irrespective of higher labour usage (which also explains the insignificance of this
variable in these specifications).
The story of pearl millet is more subtle- the impact of rainfall on the yields remains identical
to that in case of rice: higher rainfall leads to higher yields of the crop, albeit at a declining
rate. Regionally, the declines in rainfall in both the central and northern regions over time
have contributed to lowering the yields of this crop (in states like Rajasthan, UP, MP,
Gujarat, Maharashtra and Haryana).
The effects of higher (average) temperature on the yield of pearl millet continue to remain
negative, even though the variable turns out to be (highly) insignificant, pointing to the hardy
nature of this crop. In states such as AP, Karnataka, Tamil Nadu, Gujarat, MP and
Maharashtra, the response of yield to higher average temperatures is in net, positive. Such
results seem to suggest the slightly weaker impact increases in temperature may have on the
yield of pearl millet, than in case of rice: different regions would reveal different impacts on
yield, depending on whether the average temperatures have risen or fallen.
The agricultural controls however exhibit interesting effects: the labour consumption variable
has a negative impact on the yield of pearl millet (again, this might be attributed to the
availability of surplus labour in the production of this crop). Irrigation, on the other hand,
while an important factor in explaining the yield of the crop, has a negative impact on its
yield with both district and year fixed effects, and district and district-by-year fixed effects.
This is counter-intuitive. Fertiliser consumption, on the other hand, has the expected positive
impact on the yield of pearl millet, even though it appears to be insignificant in one
specification. Such results reinforce the notion of pearl millet being a crop that doesn’t
require much input application. In addition, it appears to withstand climate change better than
crops such as rice.
20
Sorghum, also considered as a ‘hardy crop’ much like pearl millet, seems to be more
sensitive to decreases in rainfall, than to an increase in the average temperature (the
temperature variable is insignificant) even though there is still a negative relationship
between temperatures and sorghum yield. The agricultural control variables also show
counter-intuitive trends with the sorghum yield. There is no consistent trend as far as these
signs are concerned: different specifications of the model yield different signs for the
irrigation and fertiliser variables, even though labour seems to have a consistent negative
impact (again, possibly due to the availability of surplus labour). Moreover, these controls
show up as insignificant in certain specifications. This points to the uncertainty surrounding
the determinants of the yield of sorghum, much like pearl millet.
The regional effects indicate that states such as AP, Tamil Nadu, Karnataka, Gujarat,
Maharashtra and Madhya Pradesh react in a similar manner to an increase in rainfall (i.e.
when rainfall increases, the yield of sorghum will also increase). In the southern states, which
have experienced an increase in rainfall over time, this may have contributed towards
increasing the yield of sorghum. In states such as Gujarat, Maharashtra and Madhya Pradesh,
which are some of the largest producers of the crop, the effect of average temperature on
yields has been positive: however, it is difficult to say whether the crop would be able to
withstand further increases in temperature in the future. For instance, in the southern states of
Karnataka, AP and Tamil Nadu, higher temperatures have already begun showing negative
effects on the yield.
5.5 Climate Change Predictions

An attempt has been made in this paper at predicting the impact of climate change on the
yields of the three foodgrains. The climate change scenario considered for these forecasts is
the short-term (2010-2039) A1F1 (business-as-usual) scenario for south Asia of the IPCC.
This involves a 0.5 °C uniform increase in temperature, and a 4% increase in precipitation
over the values in 1961-1990. While it is understandable that these predictions would not
capture the differences in distributions of rainfall and temperature across different regions of
the country, it is a start towards understanding the probable impact of climate change on
yields.
For rice, the mildly positive impact of increasing precipitation on yield was outweighed by
the negative impact of increasing temperatures on yield, both in the wettest and driest rice-
21
producing areas (the entire population was divided into quartiles, on the basis of declining
levels of rainfall). The expected decline in the yield of rice is 1.21% in the wettest areas, and
0.92% in the driest areas. However, in case of pearl millet, the yield of the crop is expected to
increase across all quartiles of regions. The yield of sorghum is expected to decline across all
regions, the decline being steepest in the wettest regions, where it is expected to go down
1.42%. For details on the percentage change in the predicted yields of the crops, refer to the
table A14 in the appendix.
These results reinforce some of the results of the regressions; rice appears to be susceptible to
climate change, especially in the wettest and driest parts of the country. Whilst pearl millet
appears to withstand climate change well (its yield is predicted to go up across regions), the
yield of sorghum appears to decline, given the short-run expected climate change. The
relative magnitude of temperature and rainfall changes appear to play a pivotal role in
deciding the direction of change in the yield: evidence suggests that yields of crops such as
pearl millet and sorghum may decline with a drop in temperature; the expected temperature
increases may therefore be beneficial to these crops. However, that does not appear to be the
case with sorghum; its yield is declining across regions. The extent of increase in the
temperature appears to be more harmful to sorghum, than it is for pearl millet. On the other
hand, the expected increase in rainfall is beneficial for all three crops. Thus, in the short-run
scenario, more rainfall seems to be conducive to crop growth, whereas higher temperatures
have an ambiguous effect on the yields.
6. Concluding remarks
Climate change is caused primarily through carbon emissions. In this nature does not
discriminate. A ton of carbon (greenhouse gases more generally) burnt anywhere in the
world and at any point in time, broadly speaking, causes the same increase in the stock of
greenhouse gases. In other words, it is a stock externality caused by a uniformly mixed
pollutant. But the effects (i.e., damages) of global warming are very asymmetric in nature. It
is expected to affect, for instance, low-lying coastal areas more than areas inland. Indeed,
some areas could gain from global warming--very cold areas could become habitable and
cultivable.
22
As mentioned in the Introduction, in this paper we wanted to look at the possible effects of
climate change on certain crops in India. We have looked at rice and millets--the former is
the leading foodgrain in terms of output and area sown, while the latter (group) is a hardy
crop that can withstand the vicissitudes of weather.
For rice the evidence is overwhelmingly that both rainfall and temperature (the two climate
variables) matter. But so do other inputs viz. labour, fertiliser and irrigation. For millets,
rainfall is the sole determinant (that is statistically significant). If one takes the regressions
seriously, millets grow on their own--none of the other variables matter.
There are now a number of recent studies that address the issues dealt with in this paper for
India and other countries, for example (but not limited to) Deschenes and Greenstone (2007),
Guiteras (2009), Fishman (2011), Auffhammer et al. (2006, 2012) and Krishnamurthy 2012.
Our study has a monthly district-wise focus. We are able to analyse fixed effects emanating
from district and state specific characteristics. Also unlike some studies (e.g., Auffhammer
2006, Schlenker and Roberts 2009, Poudel and Kotani 2012 and Sarker et al. 2012), where
only temperature and/or rainfall appear on the right-hand side, we have a number of other
inputs (controls) on the right hand side. These can give us some idea of the trade-offs
involved in the process of climate change--if the temperatures rise, causing yields of paddy to
fall, can it be compensated for by the use of more fertiliser?22 Our focus is also deliberately
crop-based. There are studies where temperature is seen to cause changes in income (in a
multi-crop setting). The interpretation of these can be problematic (joint outputs, no input
variation, the use of prices). We set ourselves the limited task of tracking (carefully) the
changing yields over a large panel. In that the crops of interest matter to the lives some of the
world's poorest people, who would be affected by climate change, this analysis seems
worthwhile.
22
This presupposes that the past is a good guide to the future, and no tipping points or other nonlinearities have
been set in motion.
23
References
Auffhammer, Maximilian, V. Ramanathan, and Jeffrey R. Vincent. 2012. “Climate change,

the monsoon, and rice yields in India.” Climatic Change, 111:411-424.
Auffhammer, Maximilian, V. Ramanathan, and Jeffrey R. Vincent. 2006. “Integrated model

shows that atmospheric brown clouds and greenhouse gases have reduced rice harvests in
India.” Proceedings of the National Academy of Sciences (PNAS), 103(52): 19668-19672
(December 26).
Bagla, Pallava. 2012. “Drawing a bead on India’s enigmatic monsoon.” Science,

335(6071):910, 24 February.
Bagla, Pallava. 2006. “Controversial rivers project aims to turn India's fierce monsoon into a
friend.” Science, 313(5790):1036–1037, 25 August.
Barrios, Salavador, Bertinelli, Luisito, Strobl, Eric. 2006. “Climatic change and rural-urban
migration: the case of sub-Saharan Africa.” Journal of Urban Economics, 60(3): 357-371.
Basavaraj, G., P. Parthasarathy Rao, Shraavya Bhagavatula and Wasim Ahmed. 2010.
“Availability and utilization of pearl millet in India.” SAT eJournal (ejournal.icrisat.org):
Vol. 8 (December):1-6.
Cagliarini, Adam and Anthony Rush. 2011. “Economic development and agriculture in
India.” Reserve Bank of Australia Bulletin, June Quarter:15-22.
Cameron, Colin A. and Pravin K Trivedi. 2009. Microeconometrics Using Stata. College
Station, Texas: StataCorp LP.
Collier, Michael and Robert H. Webb. 2002. Floods, Droughts, and Climate Change,
Tucson, AZ: University of Arizona Press (published 1 November 2002)
Dell, Melissa, Benjamin F. Jones and Benjamin A. Olken. 2009. “Temperature and income:
reconciling new cross-sectional and panel estimates.” American Economic Review: Papers &
Proceedings, 99(2):198-204.
Deressa, Temesgen Tadesse and Rashid M. Hassan. 2009. “Economic impact of climate
change on crop production in Ethiopia: evidence from cross-section measures.” Journal of
African Economies, 18(4):529-554.
Deschênes, Olivier and Greenstone, Michael. 2007. “The economic impacts of climate
change: evidence from agricultural output and random fluctuations in weather.” American
Economic Review, 97(1):354-385.
Dinar, Ariel, Robert Mendelsohn, Robert Evenson, Jyoti Parikh, Apurva Sanghi, Kavi
Kumar, James McKinsey and Stephen Lonergan. 1998. “Measuring the imapct of climate
change on Indian agriculture.” Technical Paper 402, World Bank, Washington, D.C.
Fishman, Ram Mukul. 2011. “Climate change, rainfall variability, and adaptation through
irrigation: evidence from Indian agriculture.” Columbia University. November.
24
Food and Agriculture Organization (FAO). 2012. Rice Market Monitor, XV(1), January.
Trade and Markets Division, FAO.
Food and Agriculture Organization (FAO). 2011a. FAOSTAT (FAO Statistical Database):
http://faostat.fao.org/DesktopDefault.aspx?PageID=339&lang=en&country=100 and
(accessed 29.09.2012).
Food and Agriculture Organization (FAO). 2011b. The state of food and agriculture (2010-
11): Women in agriculture: closing the gender gap for development. Rome.
Food and Agriculture Organization (FAO). 2004. Report of “Regional consultation on

gender dimensions in Asian rice livelihood systems in the changing milieu of technologies
and economy.” held on March 9-12. Regional Office for Asia and the Pacific, Bangkok.
RAP Publication 2004/13.
Government of India (GoI). 2012. “Press note on poverty estimates, 2009-10.” Planning
Commission. March.
Government of India (GoI). 2010. “Climate change and India: a 4 x 4 assessment—a

sectoral and regional analysis for 2030s.” Indian Network for Climate Change Assessment
(INCCA), Ministry of Environment and Forests. November.
Guiteras, Raymond. 2009. “The impact of climate change on Indian agriculture.” Mimeo.
University of Maryland, College Park. September.
Hornbeck, Richard. 2012. “The enduring impact of the American dust bowl: short- and
long-run adjustments to environmental catastrophe.” American Economic Review, 102(4):
1477–1507.
International Rice Research Institute (IRRI). “Rice in India.”

http://irri.org/index.php?option=com_k2&view=item&id=8744:rice-in-india&lang=en
(accessed September 30, 2012).
Kalra, Sanjay and Piyaporn Sodsriwiboon. 2010. “Growth convergence and spillovers
among Indian states: what matters? What does not?” IMF Working Paper, WP/10/96 (April).
Kanwar, Sunil. 2006. “Relative profitability, supply shifters and dynamic output response,
in a developing economy.” Journal of Policy Modeling, 28:67–88.
Kothawale, D.R., A.A. Munot and K. Krishna Kumar. 2010. “Surface air temperature
variability over India during 1901–2007, and its association with ENSO.” Climate Research,
42:89-104.
Krishnamurthy, Chandra Kiran B. 2012. “The distributional impacts of climate change on

Indian agriculture: a quantile regression approach.” Madras School of Economics Working
Paper 69/2012 (May).
Kumar, Hemanshu and Rohini Somanathan. 2009. “Mapping Indian districts across census
years, 1971-2001.” Centre for Development Economics (Department of Economics, Delhi
School of Economics) Working Paper No. 176 (April).
25
Kumar, K.S. Kavi. 2009. “Climate sensitivity of Indian agriculture.” Madras School of
Economics Working Paper 43/2009 (April).
Kumar, K.S. Kavi and Jyoti Parikh. 2001. “Indian agriculture and climate sensitivity.”
Global Environmental Change, 11(2):147-154.
Kumar, Krishna K., Rupa Kumar Kolli, R.G. Ashrit, N.R. Deshpande, and J.W. Hansen.
2004. “Climate impacts on Indian agriculture.” International Journal of Climatology,
24:1375-1393.
Lahiri, Ashok and Prannoy L. Roy. 1985. “Rainfall and supply response: a study of rice in
India.” Journal of Development Economics, 18(2-3):315-334.
Lal, Murari, T. Nozawa, S. Emori, H. Harasawa, K. Takahashi, M. Kimoto, A. Abe-Ouchi, T.

Nakajima, T. Takemura and A. Numaguti. 2001. “Future climate change: implications for
Indian summer monsoon and its variability.” Current Science, 81(9):1196-1207 (November).
Mall, R.K., Ranjeet Singh, Akhilesh Gupta, G. Srinivasan and L. S. Rathore. 2006. “Impact
of climate change on Indian agriculture: a review.” Climatic Change, 78: 445-478.
Mendelsohn, Robert, William Nordhaus and Daigee Shaw. 1994. “The Impact of Global
Warming on Agriculture: A Ricardian Analysis.” American Economic Review, 84(4):753-
771.
MNI (Millet Network of India), Deccan Development Society and FIAN (FoodFirst
Information and Action Network), India. 2009. Millets: future of food and farming.
Peng, Shaobing, Jianliang Huang, John E. Sheehy, Rebecca C. Laza, Romeo M. Visperas,
Xuhua Zhong, Grace S. Centeno, Gurdev S. Khush, and Kenneth G. Cassman. 2004. “Rice
yields decline with higher night temperature from global warming.” Proceedings of the
National Academy of Sciences (PNAS), 101(27): 9971-9975 (July 6).
Poudel, Santosh and Koji Kotani. 2012. “Climatic impacts on crop yield and its variability
in Nepal: do they vary across seasons and altitudes?” Climatic Change, published online
May 23 (DOI 10.1007/s10584-012-0491-8).
Purfield, Catriona. 2006. “Mind the Gap—Is economic growth in India leaving some states
behind?” IMF Working Paper, WP/06/103 (April).
Ranuzzi, Anna and Richa Srivastava. 2012. “Impact of climate change on agriculture and
food security in India.” ICRIER Policy Series, No. 16 (May). Indian Council for Research
on International Economic Relations (ICRIER), New Delhi.
Sarker, Md. Abdur Rashid, Khorshed Alam and Jeff Gow. 2012. “A comparison of the
effects of climate change on Aus, Aman and Boro rice yields in Bangladesh: evidence from
panel data.” Paper presented at the 41st Australian Conference of Economists (ACE),
Victoria University, Melbourne, July 8-12.
26
Schlenker, Wolfram and Michael J. Roberts. “Nonlinear temperature effects indicate severe
damages to U.S. crop yields under climate change.” Proceedings of the National Academy of
Sciences (PNAS), 106(37): 15594-15598 (September 15).
Sen Roy, Shouraseni and Robert C. Balling Jr. 2004. “Trends in extreme daily precipitation
indices in India.” International Journal of Climatology, 24(4):457–466, (March 30).
Singh, Madan Pal. 2009. “Rice productivity in India under variable climates.” Paper
presented at MARCO (Monsoon Asia Agro-Environmental Research Consortium)
Symposium, October 6-9, Tsukuba, Japan. Workshop 2 (October 6).
Wikipedia. “Monsoon of Indian subcontinent.”

http://en.wikipedia.org/wiki/Monsoon_of_Indian_subcontinent (accessed September 30,
2012).
World Bank. “Poverty & equity databank and PovcalNet.”

http://povertydata.worldbank.org/poverty/country/IND (accessed September 30, 2012) and
http://povertydata.worldbank.org/poverty/country/IND (accessed September 30, 2012).
27
Appendix
Figure 1. India-rainfall and agricultural production
** Percentage deviation from long-run average.

Source: Cagliarini and Rush (2011)
28
Table A1. Normal (average of 2005-06 to 2009-10) area, production and yield
of food crops in India
Area - Million Hectares
Production - Million Tonnes
Yield - Kg./ Hectare
(% of (% of (% of total as
Season Area Production Yield
total) total) of 1970-71)
Kharif 39.36 80.38
Rice Rabi 4.41 13.64
Total 43.77 35.7 94.02 42.4 (39) 2148
Wheat Rabi 27.75 22.6 77.04 34.7 (22) 2777
Kharif 3.43 3.54
Jowar Rabi 4.62 3.79
Total 8.05 6.6 7.33 3.3 911
Bajra Kharif 9.26 7.5 8.29 3.7 895
Kharif 6.96 13.04
Maize Rabi 1.05 4.00
Total 8.01 6.5 17.04 7.7 2128
Kharif 21.97 27.32
Coarse
Rabi 6.31 9.14
Cereals*
Total 28.28 23.0 36.46 16.4 (28.2) 1290
Kharif 10.65 4.99
Pulses Rabi 12.27 9.32
Total 22.92 18.7 14.31 6.5 625
Kharif 71.97 112.70
Foodgrains Rabi 50.74 109.15
Total 122.71 100.0 221.85 100.0 1808
*(Jowar, Bajra, Ragi, Small Millets, Barley and Maize)
Note: Figures in parentheses are shares in total foodgrain production in 1970-71
Source: Agricultural Statistics at a Glance, Directorate of Economics and Statistics,

Department of Agriculture and Cooperation, Government of India (Adapted from table 4.3)
29
Table A2. Irrigated and unirrigated area by size classes under different food crops, 2005-06
( '000 hectares)
Irrigated Unirrigated
Marginal Small Semi- Medium Large Total Marginal Small Semi- Medium Large Total
medium medium
Paddy 8,325.5 5,707.5 5,117.2 3,980.8 1,522.0 24,653.0 5,160.7 4,458.6 4,071.9 2,655.7 820.8 17,168
Wheat 5,362.5 4,071.2 4,696.2 4,985.1 2,269.8 21,384.8 444.0 528.0 620.8 585.0 205.0 2,382.9
Other
cereals 40.7 28.7 28.9 18.7 6.6 123.6 46.3 63.2 73.2 75.2 44.4 302.3
Total
Cereal 14,346.2 10,502.8 10,757.5 9,966.4 4,233.9 49,806.8 8,632.7 8,790.9 9,350.1 8,106.6 4,041.7 38,922
Marginal less than 1 hectare

Small 1 - less than 2 hectares
Semi-medium 2 - less than 4 hectares
Medium 4 - less than 10 hectares
Large 10 hectares or more
Crop % area marginal and small % area large % area irrigated

Paddy 56.6% 5.6% 58.9%
Wheat 43.8% 10.4% 90.0%
Source: Department of Agriculture and Cooperation, Agricultural Census Division
30
Table A3. Summary Statistics
Variable/Crop Rice Pearl Millet (Bajra) Sorghum( Jowar)
Unit Mean St. dev. Mean St.dev. Mean St.dev.
Rainfall mm 850.58 398.83 804.92 358.69 884.84 411.57
Normalised
rainfall - Not applicable 0.03 0.96 0.00 0.93
Temperature °C 27.51 1.40 26.36 1.67 26.25 1.24
Normalised
temperature - 0.51 1.90 0.07 0.15 0.09 0.14
Area 000 hectares 202.15 163.93 126.58 161.66 174.16 148.71
Production 000 tons 324.47 328.28 60.28 69.87 116.59 106.84
Fertiliser
consumption tons/hectare 58.08 55.31 43.38 44.45 39.20 44.92
Labour labour in 000's
consumption / area 0.49 0.34 0.35 0.31 0.37 0.28
Irrigation - 0.37 0.30 0.03 0.05 0.05 0.10
Yield
(Production/area) tons/hectare 1.58 0.88 0.64 0.41 0.68 0.38
Table A4. States and districts considered for each crop

Rice
West Bengal Burdwan, 24 Parganas, Midnapore, Birbhum, Murshidabad, Bankura, Hooghly, Nadia, Malda, Purulia
Allahabad, Azamgarh, Bahraich, Ballia, Barabanki, Bareilly, Basti, Budaon, Deoria, Faizabad,
Uttar Pradesh Ghazipur, Gonda, Gorakhpur, Hardoi, Jaunpur, Kheri, Moradabad, Pilibhit, Pratapgarh, Rampur, Rae
Bareilly, Shahjahanpur, Sitapur, Sultanpur, Unnao
Anantapur, Cuddapah, East Godavari, Guntur, Karimnagar, Khammam, Krishna, Kurnool, Nalgonda,
Andhra Pradesh
Nellore, Nizamabad, Warangal, West Godavari
Bihar Bhagalpur, Champaran, Gaya, Muzzafarpur, Patna, Purnea, Saharsa, Shanabad
Tamil Nadu Madurai, North Arcot (Vellore), Tanjore, Tiruchirapalli, Tirunelveli, South Arcot (Cuddalore)
Orissa Balangir, Balasore, Cuttack, Dhenkanal, Ganjam, Keonjhar, Kalahandi, Mayurbhanj, Sundargarh
Balaghat, Chattarpur, Chhindwata, Damoh, Jabalpur, Mandla, Narsimhapur, Panna, Rewa, Sagar,
Madhya Pradesh
Satna, Seoni, Sidhi, Tikamgarh
Karnataka Bellary, Chikmaglur, Gulbarga, Hassan, Mandya, Mysore, Raichur, Shimoga, Tumkur
Punjab Bhatinda, Firozpur, Jalandhar, Kapurthala, Patiala, Ludhiana, Rupnagar, Sangrur
Pearl Millet (Bajra)
Rajasthan Alwar, Barmer, Bikaner, Churu, Jaipur, Jalore, Jhunjhunun, Jodhpur, Nagaur, Sikar
Uttar Pradesh Aligarh, Mainpuri, Farukkhabad, Etawah, Varanasi, Deoria, Pratapgarh
Andhra Pradesh Anantapur, Chittoor,Guntur, Kurnool, Nalgonda, Nellore, Nizamabad, Medak
Gujarat Ahmadabad, Banas Kantha, Bhavnagar, Kheda, Kachch, Mehsana, Rajkot, Sabar Kantha
Madurai, North Arcot (Vellore), Salem, Tiruchirapalli, Tirunelveli, South Arcot (Cuddalore),
Tamil Nadu
Ramanathapuram
Nasik, Dhule, Jalgaon, Ahmadnagar, Pune, Satara, Sangli, Solapur, Aurangabad, Parbhani, Beed,
Maharashtra
Osamanabad, Buldhana, Akola, Amravati
Madhya Pradesh Gwalior, Guna, Morena, Indore, Jhabua, Rajgarh
Karnataka Bellary, Belgaum, Gulbarga, Bijapur, Raichur, Shimoga,
Haryana Hisar, Mahendragarh, Rohtak
31
Sorghum (Jowar)
Rajasthan Ajmer, Bundi, Jaipur, Jhalawar, Kota, Pali, Tonk, Jodhpur, Nagaur
Anantapur, Adilabad, Hyderabad, Khammam, Kurnool, Mahbubnagar, Medak, Nalgonda, Nizamabad,
Andhra Pradesh Warangal
Gujarat Bharuch, Kachch, Surat, Surendranagar, Vadodara
Tamil Nadu Coimbatore, Madurai, Tiruchirapalli
Nagpur, Nanded, Jalgaon, Ahmadnagar, Yavatmal,Sangli, Solapur, Aurangabad, Beed, Osamanabad,
Maharashtra Buldhana, Akola, Amravati
Chindwata, Dewas, Dhar, East Nimar, Guna, Mandsaur, Rajgarh, Sehore, Shajapur, Shivpuri, West
Madhya Pradesh Nimar
Karnataka Bellary, Belgaum, Gulbarga, Bijapur, Raichur, Mysore, Chitradurga
32
Table A5. Rice with district and year fixed effects
Number of obs = 3464

R-squared = 0.6422
Wald chi2 (41) = 5560.04
Prob > chi2 = 0.0000
Panel Corrected
lnyield Coeff. Z P > |z| 95% Confidence Interval
Std. Errors
rainfall 0.000373 0.000124 3.02 0.003 0.000131 0.000616
2
rain -1.02E-07 3.38E-08 -3.02 0.003 -1.68E-07 -3.58E-08
maxtemp -0.01139 0.006877 -1.66 0.098 -0.02487 0.002088
temp2 0.000145 0.000184 0.79 0.432 -0.00022 0.000506
labour -0.06606 0.089033 -0.74 0.458 -0.24056 0.108441
fertiliser 0.001309 0.000625 2.1 0.036 8.44E-05 0.002533
irrigation 0.374491 0.069726 5.37 0 0.237832 0.511151
Intercept -0.63442 0.127209 -4.99 0 -0.88375 -0.3851
Table A6. Rice with district and year fixed effects, and panel-specific errors

R-squared = 0.7910
Wald chi2 (103) =972931.02
Prob > chi2 = 0.0000
Panel Corrected
lnyield Coeff. Z P > |z| 95% Confidence Interval
Std. Errors
rainfall 0.000334 0.000111 3.01 0.003 0.000116 0.000552
rain2 -8.80E-08 2.99E-08 -2.95 0.003 -1.47E-07 -2.94E-08
maxtemp -0.00812 0.006408 -1.27 0.205 -0.02068 0.004439
temp2 7.22E-05 0.000179 0.4 0.687 -0.00028 0.000423
labour -0.07259 0.083447 -0.87 0.384 -0.23614 0.090964
fertilizer 0.001347 0.000581 2.32 0.02 0.000209 0.002485
irrigation 0.350668 0.069585 5.04 0 0.214284 0.487052
Intercept -0.57375 0.113553 -5.05 0 -0.79631 -0.35119
33
Table A7. Rice with district fixed effects and district-year fixed effects
with panel-specific errors

-squared = 0.6898
Wald chi2 (72) = 5.94e+11
Prob > chi2 = 0.0000
Panel
Corrected
Std.
Errors
rainfall 0.000565 0.00013 4.35 0 0.00031 0.000819
rain2 -1.38E-07 3.49E-08 -3.95 0 -2.07E-07 -6.97E-08
maxtemp -0.01933 0.005388 -3.59 0 -0.02989 -0.00877
temp2 0.00028 0.00016 1.75 0.08 -3.35E-05 0.000593
labour 0.161255 0.079873 2.02 0.043 0.004706 0.317804
fertilizer 0.003364 0.000351 9.58 0 0.002676 0.004053
irrigation 0.135486 0.063285 2.14 0.032 0.011449 0.259522
Intercept -0.86551 0.157207 -5.51 0 -1.17363 -0.55739
Table A8. Pearl millet with district and year fixed effects

R-squared = 0.6884
Wald chi2 (108) = 5840.23
Prob > chi2 = 0.0000
Panel
Corrected
Std.
Errors
rainfall 0.082627 0.012097 6.83 0 0.058917 0.106337
rain2 -0.0467 0.008493 -5.5 0 -0.06335 -0.03006
avgtemp -0.01335 0.121461 -0.11 0.912 -0.25141 0.224707
temp2 0.061548 0.160674 0.38 0.702 -0.25337 0.376464
irrigation -0.94739 0.373304 -2.54 0.011 -1.67905 -0.21573
fertilizer 0.000506 0.000633 0.8 0.424 -0.00073 0.001747
labour -0.23855 0.117362 -2.03 0.042 -0.46858 -0.00852
Intercept -0.37694 0.076133 -4.95 0 -0.52615 -0.22772
34
Table A9. Sorghum with district and year fixed effects

R-squared = 0.7086
Wald chi2(44) = 91495.10
Prob > chi2 = 0.0000
Panel
Std. Errors
rainfall 0.048642 0.022121 2.2 0.028 0.005286 0.091999
rain2 -0.04208 0.014794 -2.84 0.004 -0.07107 -0.01308
avgtemp -0.16316 0.275169 -0.59 0.553 -0.70248 0.376165
temp2 0.426281 0.412435 1.03 0.301 -0.38208 1.234638
labour -0.36072 0.197358 -1.83 0.068 -0.74753 0.026095
fertilizer -4.20E-05 0.000605 -0.07 0.945 -0.00123 0.001145
irrigation 0.417307 0.324852 1.28 0.199 -0.21939 1.054005
Intercept -0.7501 0.075956 -9.88 0 -0.89897 -0.60123
Table A10. Rice with district and year fixed effects, and regional interactive terms

R-squared = 0.7724
Wald chi2 (79) = 8.98e+08
Prob > chi2 = 0.0000
Panel
Std. Errors
rainfall 0.000263 0.000152 1.73 0.083 -3.50E-05 0.000562
rain2 -1.01E-07 3.70E-08 -2.73 0.006 -1.74E-07 -2.86E-08
maxtemp -0.04796 0.08652 -0.55 0.579 -0.21753 0.121619
temp2 0.00014 0.000184 0.76 0.448 -0.00022 0.0005
labour -0.09154 0.088802 -1.03 0.303 -0.26559 0.082511
fertilizer 0.001232 0.000549 2.24 0.025 0.000156 0.002308
irrigation 0.385609 0.071287 5.41 0 0.245889 0.525329
southtemp 0.036116 0.088321 0.41 0.683 -0.13699 0.209222
northtemp (dropped)
eastemp 0.05291 0.100936 0.52 0.6 -0.14492 0.25074
centraltemp -0.04998 0.083762 -0.6 0.551 -0.21415 0.114185
southrain (dropped)
northrain -0.00004 0.000141 -0.28 0.776 -0.00032 0.000236
eastrain -1.74E-06 0.000134 -0.01 0.99 -0.00026 0.000261
centralrain 0.000522 0.000187 2.79 0.005 0.000155 0.000888
Intercept -0.4641 0.141979 -3.27 0.001 -0.74237 -0.18583
35
Table A11. Rice with district and year fixed effects, regional interactive terms, panel-specific errors

R-squared = 0.7935
Wald chi2 (109) = 5.88e+06
Prob > chi2 = 0.0000
Panel
lnyield Coeff. Corrected Z P > |z| 95% Confidence Interval
Std. Errors
rainfall 0.000239 0.000146 1.64 0.102 -4.70E-05 0.000525
rain2 -8.35E-08 3.46E-08 -2.41 0.016 -1.51E-07 -1.56E-08
maxtemp 0.017899 0.05927 0.3 0.763 -0.09827 0.134066
temp2 6.88E-05 1.82E-04 0.38 0.705 -2.87E-04 4.25E-04
labour -0.11647 0.084019 -1.39 0.166 -0.28114 0.048206
fertilizer 0.001273 0.000503 2.53 0.011 0.000286 0.002259
irrigation 0.395948 0.072279 5.48 0 0.254284 0.537612
southtemp -0.02661 0.058402 -0.46 0.649 -0.14108 0.087856
northtemp -0.0651 0.090262 -0.72 0.471 -0.24201 0.111811
eastemp (dropped)
centraltemp -0.09878 0.075487 -1.31 0.191 -0.24674 0.049168
southrain (dropped)
northrain -2.66E-05 0.000127 -0.21 0.835 -0.00028 0.000223
eastrain -3.00E-05 0.000126 -0.24 0.811 -0.00028 0.000218
centralrain 0.000453 0.000176 2.57 0.01 0.000108 0.000798
Intercept -0.39766 0.131838 -3.02 0.003 -0.65605 -0.13926
36
Table A12. Pearl millet with district and year fixed effects and regional interactive terms
Number of obs 2282 R-squared 0.6949

Wald chi2 (112) 6100.94 Prob > chi2 0.0000
Panel
Std. Errors
rainfall 0.032135 0.01551 2.07 0.038 0.001736 0.062534
rain2 -0.04807 0.008497 -5.66 0 -0.06472 -0.03141
avgtemp -1.55039 0.465034 -3.33 0.001 -2.46184 -0.63894
temp2 -0.13495 0.160873 -0.84 0.402 -0.45025 0.180359
irrigation -0.54575 0.381318 -1.43 0.152 -1.29312 0.201625
fertilizer 0.000678 0.000627 1.08 0.279 -0.00055 0.001907
labour -0.3416 0.118153 -2.89 0.004 -0.57318 -0.11002
northrain 0.089207 0.032144 2.78 0.006 0.026206 0.152208
centralrain 0.030679 0.021514 1.43 0.154 -0.01149 0.072845
southrain (dropped)
northtemp (dropped)
southtemp 1.702453 0.464222 3.67 0 0.792595 2.61231
centraltemp 1.617029 0.463409 3.49 0 0.708764 2.525295
Intercept -0.3484 0.076683 -4.54 0 -0.49869 -0.1981
Table A13. Sorghum with district and year fixed effects and regional interactive terms
Number of obs 1956 R-squared 0.7195

Wald chi2(48) 93250.33 Prob > chi2 0.0000
Panel
Std. Errors
rainfall 0.240363 0.087184 2.76 0.006 0.069487 0.41124
rain2 -0.04789 0.014226 -3.37 0.001 -0.07578 -0.02001
avgtemp -0.74816 0.977292 -0.77 0.444 -2.66361 1.167302
temp2 0.1707 0.390867 0.44 0.662 -0.59539 0.936785
labour -0.41868 0.193439 -2.16 0.03 -0.79782 -0.03955
fertilizer -8.20E-05 0.000603 -0.14 0.892 -0.00126 0.0011
irrigation 0.474989 0.329623 1.44 0.15 -0.17106 1.121037
northrain (dropped)
centralrain -0.21331 0.087294 -2.44 0.015 -0.38441 -0.04222
southrain -0.2374 0.089689 -2.65 0.008 -0.41319 -0.06161
northtemp (dropped)
southtemp 0.741189 0.980375 0.76 0.45 -1.18031 2.662689
centraltemp 0.759614 0.969976 0.78 0.434 -1.1415 2.660732
Intercept -0.74737 0.075717 -9.87 0 -0.89578 -0.59897
37
Table A 14. Predicted changes in yield: short-run climate change scenario (2010-2039)
Crop Regions Percentage change in yield

Rice Wettest -1.21
Wet 0.88
Dry 0.74
Driest -0.91
Pearl Millet Wettest 1.15
Wet 1.13
Dry 0.63
Driest 0.71
Sorghum Wettest -0.55
Wet -0.93
Dry -1.02
Driest -1.42
Note:
 For each crop, all districts have been divided into four buckets, on the basis of the
amount of rainfall received over the period 1966-1999.
 The predicted changes in climate change used for the calculations are a 0.5°C rise in
temperature, and 4% increase in precipitation in the short-run (2010-2039) over the base
period (1961-1990) (IPCC Emissions Scenarios 2007)
 The model equations used to calculate the predicted change in yield are given by tables
1,2 and 3 in the Results section
 The (percentage) final change in yield is given by:
Δ Y = [(∂Y/ ∂R) * ΔR + (∂Y/ ∂T) * ΔT]* 100, where

Y = yield
R= rainfall
T= temperature
and (∂Y/ ∂R) and (∂Y/ ∂T) are determined by the model equations
38
View publication stats

Impact of Climate Change On The Indian Economy: Evidence From Foodgrain Yields

Uploaded by

Copyright:

Available Formats

Impact of Climate Change On The Indian Economy: Evidence From Foodgrain Yields

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

Impact of Climate Change On The Indian Economy: Evidence From Foodgrain Yields

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Impact of Climate Change on the Indian Economy: Evidence from Foodgrain

Article in Climate Change Economics · September 2012

Shreekant Gupta Partha Sen

SEE PROFILE SEE PROFILE

Indian Economic Review View project

Capital flows to India View project

The user has requested enhancement of the downloaded file.

IMPACT OF CLIMATE CHANGE ON THE INDIAN

Working Paper No. 218

Centre for Development Economics

Shreekant Gupta, Partha Sen and Suchita Srinivasan*#

Key words: Climate Change; Agricultural Impacts; Developing Countries

JEL Classification: O13; Q54; R11

Our methodology is based on estimating an agricultural production function with exogenous

2. Climate change and agriculture in India

2.1 Trends and regional variation in climate variables

2.2 Rice and millets in India: key issues

3. Framework and relevant literature

3.1 Conceptual framework

3.2 Relevant literature

4. Data and methodology

4.1 Data sources

Ln (YIELDit) = αi+ γt + Xit'β + ∑ ε* f (Wit) + µit,

5.2 Pearl millet (bajra)

Table 2: Pearl Millet with district-level and district-year fixed effects

5.3 Sorghum (jowar)

Table 3: Sorghum with district and district-year fixed effects

5.3 Regional differences

Pearl Millet (bajra):

5.4. Interpretation of Regression Results

The direction of impact of maximum temperature on the yield of rice is as expected- as

5.5 Climate Change Predictions

Auffhammer, Maximilian, V. Ramanathan, and Jeffrey R. Vincent. 2012. “Climate change,

Auffhammer, Maximilian, V. Ramanathan, and Jeffrey R. Vincent. 2006. “Integrated model

Bagla, Pallava. 2012. “Drawing a bead on India’s enigmatic monsoon.” Science,

Food and Agriculture Organization (FAO). 2004. Report of “Regional consultation on

Government of India (GoI). 2010. “Climate change and India: a 4 x 4 assessment—a

International Rice Research Institute (IRRI). “Rice in India.”

Krishnamurthy, Chandra Kiran B. 2012. “The distributional impacts of climate change on

Lal, Murari, T. Nozawa, S. Emori, H. Harasawa, K. Takahashi, M. Kimoto, A. Abe-Ouchi, T.

Wikipedia. “Monsoon of Indian subcontinent.”

World Bank. “Poverty & equity databank and PovcalNet.”

** Percentage deviation from long-run average.

*(Jowar, Bajra, Ragi, Small Millets, Barley and Maize)

Note: Figures in parentheses are shares in total foodgrain production in 1970-71

Source: Agricultural Statistics at a Glance, Directorate of Economics and Statistics,

Marginal less than 1 hectare

Crop % area marginal and small % area large % area irrigated

Source: Department of Agriculture and Cooperation, Agricultural Census Division

Table A4. States and districts considered for each crop

Number of obs = 3464

Number of obs = 3464

Number of obs = 3464

Number of obs = 2282

Number of obs = 1956

Number of obs = 3464

Number of obs = 3464

Number of obs 2282 R-squared 0.6949