Professional Documents
Culture Documents
Food Policy
journal homepage: www.elsevier.com/locate/foodpol
a r t i c l e i n f o a b s t r a c t
Article history: Drought is one of the major constraints to sustainable improvement of agricultural productivity in the
Received 12 December 2014 rainfed farming systems. In this paper, we have examined the frequency, severity and spread of droughts
Received in revised form 13 May 2015 in India, and assessed their impact on rice production. Our results show that one-third of the rice area in
Accepted 10 July 2015
India is affected by droughts, mostly moderate droughts, the frequency of which has increased in recent
Available online 31 July 2015
years. The severity of droughts, however, has reduced. An assessment of their impact reveals a decline in
drought-induced losses in rice yield, in absolute as well as relative terms. The weakening effect of
Keywords:
droughts is attributed to improvements in farmers’ adaptive capacity due to expansion of irrigation facil-
Drought
Rice
ities and increased availability of improved varieties for the rainfed production systems, along with other
Irrigation coping strategies.
Varieties Ó 2015 Elsevier Ltd. All rights reserved.
India
Introduction (GoI, 2009). India has experienced 13 major droughts since 1966,
the year of introduction of Green Revolution in the country. Four
Climate change has become a major challenge to sustainable of these have occurred quite recently, i.e., between 2001 and 2012.
development of agriculture, an activity more sensitive and linked Severe and prolonged droughts can have significant adverse
directly to climate change as compared to many other economic effects on agriculture and food security (Kumar et al., 2014), and
activities (Stern, 2006; Mendelsohn and Dinar, 2009; De Salvo may cause depletion of productive assets, exacerbate rural poverty,
et al., 2013). The challenge becomes pronounced in the case of force outmigration, reduce demand for non-agricultural goods, and
unusual or extreme changes in climate, typified by droughts, lead to over-exploitation of natural resources (Sivakumar and
floods, heat waves, etc., the frequencies of which are predicted to Kerbart, 2004; Pandey et al., 2007; Lal et al., 2012). From a study
increase in the future (World Bank, 2013). Developing countries, in eastern India, Pandey et al. (2007) have estimated that during
like India, are more vulnerable to such shocks because of their a drought year household income falls by 25–60% and
heavy dependence on agriculture and lack of technical and finan- head-count poverty ratio rises by 12–33%.1 Similar results have
cial resources to cope up with them (Lal et al., 2012). In India, been reported from the southern state of Tamil Nadu (Selvaraj and
the agricultural sector, despite its low and declining share in gross Ramasamy, 2006). These studies have also reported that despite
domestic product (15% in 2013–2014), still engages half of the the use of coping mechanisms to regain their pre-drought normal
country’s population, and is dominated by small landholders; the livelihood levels, farm households are rarely able to recover fully
holdings measuring less than or equal to one hectare (ha) comprise the loss of productive assets in the subsequent period.
two-thirds of the total landholdings. Frequent occurrence of droughts has raised concerns about the
Drought is a recurrent phenomenon in India. More than food and livelihood security. In fact, these concerns have been
two-thirds of the country’s geographical area is vulnerable to reshaping India’s drought management policy for quite some time.
droughts. The probability of occurrence of a drought is about During the past two decades, India’s strategy to cope up with
35%, i.e., once in three years (Pandey et al., 2007), ranging from droughts has undergone a paradigm shift, i.e., from crisis manage-
20% in the dry-humid regions to 40% or more in the arid regions ment to risk management. The strategy emphasizes on (i)
1
This study covered drought-prone states of Chhattisgarh, Jharkhand and Odisha.
⇑ Corresponding author. Tel.: +91 11 25842665; fax: +91 11 25842684. Approximately 60% of the rice area in Chhattisgarh and Jharkhand and 20% in Odisha
E-mail address: psbirthal@ncap.res.in (P.S. Birthal). are prone to droughts (Pathak et al., 2011).
http://dx.doi.org/10.1016/j.foodpol.2015.07.005
0306-9192/Ó 2015 Elsevier Ltd. All rights reserved.
2 P.S. Birthal et al. / Food Policy 56 (2015) 1–12
To begin, we plot in Fig. 1 the standardized deviations in the or historical trend (Xiao-jun et al., 2012). The ‘value at risk’
seasonal rainfall and mean monthly temperature from their approach has also been used to estimate probable loss from
respective historical mean values. There is a negative relationship extreme climatic events, where the loss is measured as the product
between the two, evoking that simultaneous occurrence of abnor- of severity and frequency of the event, value of assets exposed to
mally high temperature and abnormally low rainfall (quadrant IV the event and their vulnerability to the event (see, ECA, 2009).
of Fig. 1) could be more damaging to crops than the Some studies have used linear and non-linear mathematical pro-
rainfall-deficit alone (quadrant I of Fig. 1). We are interested in gramming models to simulate economic impacts of droughts
quadrant IV that contains a large number of dry and hot weather (Dono and Mazzapicchio, 2010; Peck and Adams, 2010; Jenkins
events. et al., 2003; Booker et al., 2005). Others have used econometric
Mathematically, the drought index can be written as: models to assess the damage due to droughts at macro-level or
at disaggregated regional or crop levels (e.g., Quiroga and
DIit ¼ f minð0; TRDsd sd
it Þ maxð0; MTDit Þg ð1Þ Iglesias, 2009). Computable general equilibrium and input–output
models have also been used to evaluate welfare impacts of
where subscripts i and t denote district and time, respectively. DI is droughts (see, Pérez y Pérez and Barreiro-Hurlé, 2009;
drought index. TRDsd normal
it ¼ ðTRit TRi Þ=sdðTRÞi is the standardized Martin-Ortega and Berbel, 2010).
deviation in the seasonal rainfall from its normal. Here, TR is the We follow the econometric approach, viz., fixed effects panel
total rainfall, and sd(TR) is the standard deviation in rainfall. regression, where crop yield is regressed on drought index, its
MTDsd normal
it ¼ ðMT it MT i Þ=sdðMTÞi is the standardized deviation in squared term, and their interactions with time trend. If the distribu-
mean monthly temperature. In this expression, MT is the mean tion of drought events remains constant over time, then the change
monthly temperature and sd(MT) is its standard deviation. The in tolerance level of a crop to drought translates directly into the
index has been standardized so as to make it comparable across dis- change in yield risk (Yu and Babcock, 2010). An increasing (decreas-
tricts and time period. ing) drought-tolerance results in decreasing (increasing) yield risk.
DI can also be computed as the sum of the standardized devia- Farmers follow a number of strategies, such as irrigation and
tions of temperature and rainfall. However, the preferred specifica- water management, adoption of drought-tolerant varieties,
tion is the one based on the product of their standardized increased application of fertilizers and shifting of planting dates
deviations that lays relatively greater emphasis on larger devia- to cope up with droughts. We try to capture the effects of such
tions in both rainfall and temperature. Higher the temperature strategies on drought-induced losses. To do so, we include irriga-
and lower the rainfall, larger is the value of DI. tion as a control in our model. Irrigation is considered to be one
DI so constructed has three main advantages (for details, see of the best bets against drought shocks, more so for the
Yu and Babcock, 2010). One, it is based on district-level data on water-intensive crops, such as rice, which are also more vulnerable
temperature and rainfall; and being location-specific it is more to droughts (O’Toole, 2004). Further, our choice of irrigation is
relevant to the agricultural activities for that location. Two, it guided by the availability of information on different adaptation
comprises of two important causes of crop damage, viz., lack of measures that the farmers follow. Crop-wise irrigation statistics
moisture and excess heat; hence, a single index rather than multi- are available at district-level, but no such information is available
ple weather variables, offers an easy means of assessing impacts of on other strategic variables. In our model, we allow rice area irri-
extreme events on agriculture. Third, it provides severity level of a gated (in %) to interact with drought index, squared drought index
drought, making it convenient to quantify crop loss at different and time trend so as to explicitly bring out the role of irrigation in
severity levels. managing drought risks. Implicitly, we also assess the contribution
of adaptation strategies other than irrigation. For this purpose, we
simulate yield loss at varying levels of irrigation against the coun-
Measuring impact of drought
terfactual of no irrigation.
The fixed effects regression equation that we estimate is:
There are several methods for assessing the impact of drought
depending on its nature (direct or indirect) and the level of aggre-
gation (farm, household, regional or economy-wide). In its simplest X
N
Y it ¼ DCT i þ /i ½DCT i T þ b1 DIit þ b2 ðDIit DIit Þ þ b3 ðDIit TÞ
form, the impact of a drought can be measured as the negative
i¼1
deviation in crop yield in a drought year from its previous normal
þ b4 ðDIit DIit TÞ þ b5 IRRit þ b6 ðDIit IRRit Þ
þ b7 ðDIit DIit IRRit Þ þ b8 ðDIit IRRit TÞ
þ b9 ðDIit DIit IRRit TÞ þ it ð2Þ
Eq. (2) decomposes rice yield into three components: (i) deter-
ministic trend yield with district- specific mean,
P
ðDCT i þ Ni¼1 /i ½DCT i TÞ, that includes improvements in yield
due to technological advances, increased use of inputs, etc.; (ii)
variation in trend yield due to droughts, irrigation and their inter-
actions, and (iii) a residual term, it , representing the effect of other
random factors. The change in yield due to droughts can be esti-
mated taking partial derivation of Y with respect to DI.
@Y it
Loss ¼ ¼ b1 þ 2b2 DIit þ b3 T þ 2b4 DIit T þ b6 IRRit
@DIit
þ 2b7 ðDIit IRRit Þ þ b8 ðIRRit TÞ þ 2b9 ðDIit IRRit TÞ ð3Þ
If the crop is susceptible to drought, then marginal effect of
drought in Eq. (3) should be negative. From Eq. (3) it is possible
to test a number of hypotheses related to the change in
drought-induced losses, and the effects and efficacy of irrigation
as a coping strategy against droughts. The change in yield loss Fig. 2. Frequency distribution of drought index.
due to drought over time can be estimated taking partial derivative
of Eq. (3) with respect to T. Likewise, the change in yield loss due to
change in irrigation level can be obtained taking partial derivative
of Eq. (3) with respect to IRR. We have looked for the changes in frequency and severity of
droughts over time, taking 1987 as the cut-off year. We took
@Loss 1987 as cut-off because that year experienced a severe drought
DLossT ¼ ¼ b3 þ 2b4 DIit þ b8 IRRit þ 2b9 ðDIit IRRit Þ ð4Þ
@T affecting more than 60% of the cropped area (GoI, 2009), and after-
wards there was a perceptible shift in India’s drought management
@Loss policy, from crisis management to risk management. A comparison
DLossIRR ¼ ¼ b6 þ 2b7 DIit þ b8 T þ 2b9 ðDIit TÞ ð5Þ
@IRR of the frequency of drought events before and after 1987 shows (i)
The role of irrigation in mitigating the harmful effects of a decline in the incidence of severe drought events, from 21% dur-
droughts on crop yield can be assessed from Eqs. (4) and (5). The ing 1969–1987 to 9% during 1988–2005, and (ii) an increase in the
partial differentiation of Eq. (4) with respect to IRR, and of Eq. (5) incidence of moderate drought events, from 77% to 88%. The sever-
with respect to T results in the same expression, i.e., Eq. (6) below. ity level of droughts too has declined; the mean value of DI fell
from 0.97 to 0.57.
@ DLossT @ DLossIRR To see if there is a long-term trend in the drought events, we
¼ ¼ b8 þ 2b9 ðDIit Þ ð6Þ
@IRR @T regressed occurrence of a drought (drought = 1, otherwise zero)
Eq. (6) estimates the change in drought-induced yield loss due on linear time trend using Probit model.5 The results are presented
to improvements in irrigation or in other words the change in in Table 1(b) and point toward a tendency of droughts becoming
effectiveness of irrigation in mitigating the adverse effects of more frequent. In case of moderate droughts, the coefficient of time
droughts on rice yield over time. Eqs. (4)–(6) can be posed as sta- trend is positive and statistically significant, implying an increase in
tistical test of the following hypothesis. Using Eq. (4), one can test the probability of their occurrence. The coefficient of severe
the hypothesis that yield loss due to droughts has remained con- droughts, however, is negative but not significant.
stant over time. Similarly, from Eq. (5) it is possible to test the null Table 1(a) also presents the information on rice area affected by
hypothesis that improvements in irrigation have not made any sig- droughts of different intensities. On an average, about one-third of
nificant difference in the drought-induced yield loss; and from Eq. the rice area is vulnerable to droughts, an estimate close to that
(6) one can test the hypothesis that the efficacy of irrigation in mit- reported by Pandey et al. (2007) and Pathak et al. (2011). Of the
igating the drought impacts has not changed over time. total drought-affected rice area, about 82% is prone to moderate
droughts and 15% to severe droughts. Further, we find a marginal
decline in the drought-affected area during 1988–2005, mainly
Frequency, severity and spread of droughts
because of fewer incidences of severe droughts during this period.
The rice area affected by moderate droughts, however, has
Fig. 2 depicts the distribution of drought index, i.e. DI. The index
increased.
ranges from zero to eight; zero means seasonal rainfall being
To understand better the temporal changes in the frequency
greater or equal to normal rainfall, and seasonal monthly temper-
and severity of droughts, we plot the distribution of
ature being less or equal to normal temperature. Abnormally low
drought-affected rice area for each year in Fig. 3. A few important
values of DI can be considered to be representative of the normal
observations that have emerged from it are: (i) the country has
weather. DI is skewed toward its lower bound, indicating that most
experienced a drought most of the times in one or the other region,
drought events during 1969–2005 were of low or moderate inten-
(ii) the droughts of 1972, 1974, 1979, 1982, 1987, 1995, 1998, and
sity. The incidence of severe drought events, say of DI P 3 was rare.
2002 were widespread, affecting over 50% of the rice area, and (iii)
Based on the dispersion of DI, drought events have been classi-
the severe droughts have become less frequent, while the moder-
fied into low, moderate and severe intensity droughts. A drought is
ates have increased.
of low intensity if DI is one standard deviation less than its own
mean. It is of moderate intensity if DI lies within ±1 standard devi-
ation around the mean; and of high intensity if DI is equal to or
more than one standard deviation above the mean. Accordingly, 5
Likelihood-ratio test shows that the panel-level variance component is unimpor-
only 15% of the total drought events in the country during the tant, and the panel estimator is not different from the pooled estimator; therefore, we
period 1969–2005 were of severe nature, and 82% represented a have presented results of the pooled estimator (probit) with district-clustered
moderate drought (Table 1(a)). standard errors.
P.S. Birthal et al. / Food Policy 56 (2015) 1–12 5
Table 1
Panel (a) Distribution of drought events, their severity levels, and rice area affected by droughts.
Period Severity level Mean drought index % of total events % rice area affected
1969–2005 Low 0.05 (0.07) 2.7 1.1
Moderate 0.47 (0.45) 82.7 25.7
Severe 2.59 (1.26) 14.6 4.6
Mean 0.77 (0.99) 100.0 31.4
1969–1987 Low 0.04 (0.04) 2.3 0.7
Moderate 0.49 (0.47) 76.8 24.5
Severe 2.84 (2.55) 20.9 6.5
Mean 0.97 (1.21) 100.0 31.8
1988–2005 Low 0.06 (0.08) 3.1 1.5
Moderate 0.45 (0.42) 88.4 26.8
Severe 2.00 (0.78) 8.5 2.7
Mean 0.57 (0.64) 100.0 31.0
Figures in parentheses are standard deviations.
Panel (b) Trends in the frequency of drought events by their severity level.
Variable (1) (2) (3) (4)
Low intensity drought = 1, Moderate drought = 1, Severe drought = 1, Drought event = 1,
zero otherwise zero otherwise zero otherwise zero otherwise
Time trend 0.0038 0.0107*** 0.0025 0.0094***
(0.0041) (0.0015) (0.0026) (0.0015)
Number of events 56 1712 302 2070
100.0
Low
90.0
Moderate
80.0
Severe
70.0
60.0
per cent
50.0
40.0
30.0
20.0
10.0
0.0
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
6
The deviations were estimated using Hodrick–Prescott (HP) filter. Fig. 4. Relationship between drought index and deviation in rice yield, 1969–2005.
6 P.S. Birthal et al. / Food Policy 56 (2015) 1–12
Table 2
Deviation in rice yield from its historical trend at different severity levels of drought.
Table 3
to 1.3% from 3.6%. In comparison to 1969–1987, the absolute yield
loss in the latter period was 23% less under severe drought condi- Panel (a) Model specification tests.
tions and 45% under moderate drought conditions. However, the Items Linear Log linear
yield is not affected adversely by the low intensity droughts. F-test for joint significance district specific 13.04*** 4.16***
These trends provide an indication of the rice production becoming intercept F(170, 5578)
resilient to droughts. Hausman specification test (Fixed effects vs 2217.59*** 708.01***
The negative deviation in the rice yield from its trend (yield Random effects) Chi2(176)
F-test for common trend vs agroclimatic zone- 37.06*** 27.72***
loss) cannot solely be attributed to droughts. Some other factors, specific trend F(3, 5578)
for example, insect pests and diseases, can also lead to reduction F-test for district specific trend vs agro climatic 12.69*** 3.75***
in yield. The decline in yield loss can also be due to droughts zone specific-trend F(167, 5578)
becoming less severe, and improvements in irrigation and water Modified Wald test for group-wise 5.7e + 10*** 2.1e + 08***
heteroskedasticity Chi2(171)
management practices, increased adoption of drought-tolerant
F-test for significance of fixed effects regression 39.73*** 21.86***
crop varieties, fertilizers, agronomic practices, etc. The trends in F(180, 5578)
yield loss presented in Table 2 have been validated by estimating
***,**, and * denote significance at the 1%, 5%, and 10% levels, respectively.
the panel fixed effects regressions.
There is a possibility of dependent variable (i.e., rice yield) being Panel (b) Estimated regression coefficients of Eq. (1) with heteroskedastic
non-stationary. Similarly, the drought index, constructed using robust standard errors.
temperature and rainfall, may also be non-stationary due to the Explanatory variables Linear model Log linear model
presence of a stochastic trend in either of these. The DI (drought index) 551.5 13.5
non-stationarity in the dependent and independent variables (6944.1) (10.9)
may lead to spurious estimates. To check for stationarity in the DI ⁄ DI 2902.4 8.3
time series of these variables, we have conducted panel unit root (2068.3) (5.2)
DI ⁄ Trend 0.161 0.007
tests, and the results are reported in the appendix Table A1. All
(3.50) (0.005)
the tests reject the null hypothesis that panel contains unit roots, DI ⁄ DI ⁄ Trend 1.481 0.004
and thus rule out the possibility of any of the series being (1.043) (0.003)
non-stationary. IRR (proportion of rice area irrigated) 284.014*** 0.132**
Table 3 presents the coefficients and standard errors of the (107.431) (0.063)
DI ⁄ IRR 236.6 4.7
explanatory variables for linear and log-linear specifications of (10268.1) (13.6)
Eq. (2). Table 3(a) presents specification tests to check appropriate- DI ⁄ DI ⁄ IRR 4534.1 12.2*
ness of the chosen fixed effects model with district-specific trend. (2997.8) (6.4)
The F-test for joint significance of district fixed effects is significant DI ⁄ Trend ⁄ IRR 0.142 0.002
(5.171) (0.007)
at 1% level, endorsing our preference for the fixed effects regression
DI ⁄ DI ⁄ Trend ⁄ IRR 2.3 0.006*
over pooled regression. And, the Hausman test favors the fixed (1.5) (0.003)
effects regression over the random effects regression for both lin- Constant 44838.7*** 20.7***
ear and non-linear specifications. Having chosen the fixed effects (819.4) (0.7)
model, we further conducted F-tests to justify the inclusion of No. of observations 5929 5929
district-specific trend. These tests suggest supremacy of the Average marginal effects
region-specific7 trend over the common trend; and the Trend 23.4*** 0.016***
district-specific trend over the region-specific trend. We used mod- (0.42) (0.0003)
ified Wald test to check for heteroskedasticity. The Chi2 statistic for DI 194.0*** 0.246***
(14.1) (0.022)
Wald test is significant at 1% level indicating the presence of
IRR 289.2*** 0.181***
heteroskedasticity. To take care of heteroskedasticity and serial cor- (106.8) (0.063)
relation we estimated district-clustered standard errors.
Tests of hypotheses (at mean DI = 0.77, IRR = 0.50 and T = 1987)
The coefficients of DI and DI-squared are negative in the linear
H0: DLoss(T) = 0 (Eq. (4)) 0.75 0.0072***
as well as log-linear specifications of Eq. (2), but these are not sta-
(1.22) (0.0014)
tistically significant (Table 3(b)). Though the constructed drought H0: DLoss(IRR) = 0 (Eq. (5)) 22.46 0.2102***
(30.92) (0.0401)
H0: DLoss(T,IRR) = 0 (Eq. (6)) 3.38 0.0117***
7
The regions here refer to the agroclimatic regions, viz., humid, semi-arid (3.09) (0.0035)
temperate and arid-semi-arid tropics. The districts were classified into regions
following the criteria (temperature and length of growing period) outlined in TAC Figures in parentheses are district-clustered standard errors.
(1994). ***,**, and * denote significance at the 1%, 5%, and 10% levels, respectively.
P.S. Birthal et al. / Food Policy 56 (2015) 1–12 7
index is capable of capturing the negative impact of weather Three, the efficacy of irrigation in mitigating the harmful effects
shocks on rice yield, the standard errors are large. The log-linear of droughts has remained unchanged over time, i.e.,
specification provides a better precision in the estimates. The lack DLoss(T,IRR) = 0. These hypotheses have been evaluated at the
of precision in the estimates associated with DI-squared are per- respective means of DI, IRR and T. The coefficient of DLoss(T) is pos-
haps due to the large variation in yield and the fewer events of a itive in both the specifications. We reject the hypothesis that pro-
severe drought. The coefficient of DI ⁄ T is positive, suggesting that portion of yield loss due to droughts has remained constant over
rice production has become less vulnerable to droughts over time. time in favor of the alternative hypothesis, i.e., drought-induced
It has also become less vulnerable to severe droughts as is indi- loss has declined over time. From the log-linear specification we
cated by a positive coefficient of DI ⁄ DI ⁄ T. could also reject the null hypothesis of yield loss (in %) due to
The coefficient of IRR is positive and significant indicating the droughts remaining unchanged with irrigation levels. The positive
importance of irrigation in enhancing rice yield. The interaction coefficient of DLoss(IRR) indicates that irrigation is an effective
of IRR with DI-squared is also positive and significant in the means of reducing harmful effects of droughts. However, the neg-
log-linear specification, which indicates that irrigation moderates ative and significant coefficient of DLoss(T,IRR) indicates that the effi-
harmful effects of severe droughts. The effectiveness of irrigation, cacy of irrigation in reducing adverse effects of droughts has
however, seems to have declined as is implied in the negative diminished.
and significant coefficient of DI ⁄ DI ⁄ T ⁄ IRR. This is likely; as with Fig. 6 shows the predicted relationship between rice yield (in
rise in the frequency of droughts, water becomes a more acute con- kg/ha) and drought index (at 95% confidence interval) at mean
straint rendering irrigation less effective in mitigating the harmful level of rice area irrigated (50%) during 1969–2005 against the
effects of droughts. counterfactual of no irrigation. The response curve is negatively
Table 3(b) also reports the marginal effects associated with time sloped and convex in both the scenarios, implying that the drought
trend, drought and irrigation evaluated at their mean values. The reduces yield, but the incremental loss becomes smaller with
marginal effect of time indicates an increase in rice yield at a rate increase in its severity level. The yield response curve with no irri-
of 1.6% or 23.4 kg/ha per annum during the period 1969–2005, but gation lies below the response curve with irrigation at all levels of
the trend varies considerably across districts (Fig. 5). In the linear drought severity.
model, the trend coefficient of yield varies from 12 kg/ha to From the descriptive statistics (in Table 2) we had noticed a
+76 kg/ha per annum, and in the log-linear model the rate of decline in the severity level of droughts, which possibly could have
growth in rice yield ranges from 1.6% to +4.7% per annum. been one of the reasons for reduction in yield loss. Keeping the
Nevertheless, in majority of the districts rice yield has increased, level of irrigation constant (at its mean level during 1969–2005)
and the negative trend is confined to a fewer districts. This varia- in Eq. (4), we estimated the marginal effect of droughts assuming
tion in yield trends is an indication of the heterogeneity in the pat- no change in their severity level over time, i.e., DI = 0.77. The
tern and rate of adoption of improved technologies and practices results are presented in Table 4, and confirm that
across districts. The positive marginal effect of irrigation reinforces drought-induced losses, absolute as well as relative, have declined.
its role in improving crop yield. The effect of drought on yield, Using Eq. (5) we estimate the marginal effect of irrigation
however, is negative to the extent of 194 kg/ha or 25% of the nor- assuming no change in the severity level of droughts (DI = 0.77)
mal yield. and time trend. The results are presented in Table 5. As expected,
Table 3(b) also provides tests of some important hypotheses. the simulated yield losses decline with improvements in irrigation,
One, the yield loss due to droughts has remained constant over but these cannot be completely eliminated even with entire rice
time, i.e., DLoss(T) = 0. Two, the improvements in irrigation have area being irrigated. And, this is very much likely, as irrigation
not affected the drought-induced yield loss, i.e., DLoss(IRR) = 0. water is limited and becomes scarcer during a drought. Over time,
Fig. 6. Predicted rice yield, with and without irrigation, at varying levels of drought severity.
Table 4
Equation predicted yield loss overtime at mean DI (0.772).
Table 5
Equation predicted drought-induced yield losses at different levels of irrigation. droughts, but the efficacy of incremental irrigation declines.
These are important findings that implicitly bring out the role of
Model % rice area irrigated
adaptation strategies other than irrigation in moderating the
0 20 40 60 80 100 harmful effects of droughts. The inference is that irrigation is
Linear (kg/ha) 212.6 204.9 197.3 189.7 182.1 174.5 important, but it alone is not sufficient to cope up with droughts,
(24.68) (18.64) (14.89) (15.23) (19.45) (25.70) and needs to be complemented with other risk-mitigating strate-
Log-linear (%) 37.5 32.2 27.0 21.7 16.4 11.2
gies. These strategies can include increased adoption of improved
(4.33) (3.36) (2.47) (1.78) (1.55) (1.97)
crop varieties and agronomic practices, higher fertilizer applica-
Figures in parentheses are standard errors. tion, etc. It may be noted that the inference regarding the role of
other adaptation strategies does not follow from the regression
results.
there has been a significant expansion of the groundwater-based
irrigation in India, leading to over-exploitation of the groundwater
in many parts of the country. Discussion
Does it mean that the efficacy of irrigation to cope with
droughts has reduced overtime? Using Eq. (6) we conducted some Several studies have predicted that climate change will
counterfactual exercises to understand evolution of the role of irri- adversely affect rice production in India (Kumar et al., 2011,
gation in moderating the harmful effects of droughts on rice yield. 2014; Auffhammer et al., 2011). The magnitude of the impact
Corresponding to the linear and log-linear specifications of Eq. (6), would depend on the degree of change in temperature and precip-
in Fig. 7(a) and (b) we plot the trend in yield loss under varying itation. A smaller rise in temperature (<1 °C) is unlikely to have any
irrigation situations (see Table A2 also). There emerge two impor- harmful effect on rice yield, but a significant increase would have
tant observations from this exercise. One, the loss curve without adverse effects (Aggarwal and Mall, 2002). There is considerable
irrigation lies above its irrigation counterfactuals (50% and 90% rice heterogeneity in the rice production environments in the country,
area irrigated) suggesting that the impact of drought is bigger in and the magnitude and direction of climate impacts may differ
the rainfed farming systems. Two, as compared to higher level of across the production environments. Evidence shows that in some
irrigation, the slope of the curve with lower irrigation is steeper regions rice production may benefit from climate change, and
which implies that irrigation is important to cope up with would be adversely affected in others (Kumar et al., 2011).
P.S. Birthal et al. / Food Policy 56 (2015) 1–12 9
Table 6
Trend in per cent net irrigated area in India. Source: GoI (2013). irrigated area almost doubled, from 32% to 59%. Besides, emphasis
has also been laid on the development of watersheds to harvest
Years Source of irrigation Rice area irrigated
rainwater for irrigation purposes. Joshi et al. (2005) have demon-
Canal Groundwater Others All strated positive impact of these efforts on crop yields and farm
1969–70 9.1 8.1 4.6 21.8 38.2 incomes.
1986–87 11.8 14.9 3.8 30.5 43.6 Our results, however, have indicated a decline in the effective-
2005–06 11.8 25.6 5.7 43.1 56.0
ness of irrigation in coping with droughts, particularly severe ones,
2010–11 11.1 27.6 6.3 44.9 58.7
beyond a threshold level of irrigation. This has to be looked from
the perspective of the crop in question, and the availability of irri-
gation water. Of the total available water, more than 85% is utilized
Eastern region is predicted to be more affected by increased tem-
for irrigation. Groundwater, the main source of irrigation, has been
perature than the northern region.
over-exploited in many parts of the country, particularly in the arid
Nevertheless, farmers can adapt to climate change through
and semi-arid regions spread across north-western, western and
shifts in planting dates, use of short-duration crop varieties,
southern states. In the eastern region, groundwater is available in
increased and efficient use of irrigation, fertilizers, etc.
plenty, but farmers are unable to harness it due to lack of energy
(Wassmann and Dobermann, 2007; Kumar et al., 2011). Some
to pump it out. Rice being a water-intensive crop is more vulnera-
studies, for example ECA (2009), Kumar (2009), Aggarwal et al.
ble to droughts than many other crops. In a drought year, the
(2010) and Auffhammer et al. (2011), have shown that climate
groundwater is under stress, and its availability to the crop is less
impacts on agriculture can be moderated to a large extent using
than its requirement, reducing its ability to withstand drought
a combination of adaptation strategies. Our results have demon-
shocks. Some other studies (e.g., Pandey et al., 2007; ECA, 2009)
strated that rice production in India has gradually become resilient
also confirm this observation.
to droughts8 and irrigation has played an important role in it. Birthal
A related finding that has emerged from our analysis is that the
et al. (2014) have shown irrigation reducing the negative effects of
drought-induced loss in rice yield has declined even in the pre-
climate change on agricultural productivity, the biggest reduction
dominantly rainfed farming systems. This implicitly suggests the
being in the rainfed farming systems. Our results also indicate the
role of adaptation strategies other than irrigation in coping with
role of adaptation strategies other than irrigation in management
droughts. Here, we emphasize on the crop breeding for
of droughts, which we could not explicitly capture in our analysis
drought-tolerance that has attracted considerable research atten-
because of non-availability of data.
tion in the recent past (Lybbert and Bell, 2010). The
There has been a significant improvement in irrigation in India;
drought-tolerant traits embedded in seed act as an insurance.
the proportion of rice area irrigated has increased from 38% in
Drought-tolerant crop varieties are of shorter duration, and have
1969 to 44% in 1987 and further to 56% in 2005 (Table 6). The addi-
ability to withstand water scarcity and high temperature. In the
tional area has largely been irrigated using groundwater; between
case of rice, these have been reported to possess yield advantage
1969–1970 and 2005–2006, the share of groundwater in net
of 5–28% over the existing high-yielding varieties (Selvaraj and
Ramasamy, 2006; Virk and Witcombe, 2007; Pray et al., 2011;
8
A recent testimony to this is the small reduction in yields of major kharif crops IRRI, 2013).
during the drought of 2009 as compared to the drought of 2002. In 2009, rainfall India’s rice breeding strategy has undergone a gradual shift.
deficit was 22%, higher by 3 percentage points than in 2002. However, rice yield in
Two decades ago, rice breeding research was largely focused on
this year was 2.5% less from its inter-annual trend as compared to 12.5% in 2002. The
deviation in yields of cotton, groundnut and soybean in 2009 was almost half of that developing varieties for favorable environments (Pandey and Pal,
in 2002. 2007), but the realization about the limits on irrigation expansion
10 P.S. Birthal et al. / Food Policy 56 (2015) 1–12
Table A1
Panel unit root tests.
Variables H0: Panels contain unit roots H0: All panels contain unit roots H0: All panels contain unit roots
H1: Panels are stationary (lags chosen H1: Some panels are stationary H1: At least one panel is stationary (lags 2)
to minimize AIC) (lags chosen to minimize AIC)
LLC (adjusted statistic) p-value IPS (z-t-tilde-bar) p-value Fisher (modified inverse chi2 Pm) p-value
Rice yield (kg/ha) 10.59 0.00 17.26 0.00 36.43 0.00
Ln rice yield 14.17 0.00 32.73 0.00 55.00 0.00
Drought index 60.47 0.00 59.42 0.00 75.09 0.00
Table A2
Trend in yield loss under varying levels of irrigation at DI = 0.77.
References Guiteras, R., 2007. The Impact of Climate Change on Indian Agriculture. Working
paper, Department of Economics, Massachusetts Institute of Technology,
Cambridge, MA.
Aggarwal, P.K., Kumar, S.N., Pathak, H., 2010. Impacts of Climate Change on Growth
Iizumi, T., Ramankutty, N., 2015. How do weather and climate influence cropping
and Yield of Rice and Wheat in the Upper Ganga Basin. Indian Agricultural
area and intensity? Glob. Food Secur. 4, 46–50.
Research Institute, New Delhi.
IRRI (India International Rice Research Institute), 2013. Cluster Demonstrations of
Aggarwal, P.K., Mall, R.K., 2002. Climate change and rice yields in diverse agro-
Stress Tolerant Rice Varieties in Stress Prone Parts of India. Annual Report
environments of India. II Effect of uncertainties in scenarios and crop models on
submitted to National Food Security Mission, Ministry of Agriculture,
impact assessment. Clim. Change 52, 331–343.
Government of India. IRRI, New Delhi Office.
Aryal, J.P., Bhatia, M., Jat, M.L., Sidhu, H.S., 2014. Impact of laser land levelling in
Jacoby, H., Rabassa, M., Skoufias, E., 2011. Distributional Implications of Climate Change
rice-wheat rotations of the north-western Indo-Gangetic plains of India. In:
in India. Policy Research Working Paper 5623, The World Bank, Washington, DC.
Paper Presented at the Fifth World Congress of Environmental and Resource
Jat, M.L., Chandna, P., Gupta, R., Sharma, S.K., Gill, M.A., 2006. Laser Land Levelling: A
Economists, Istanbul, June 28–July 2.
Precursor Technology for Resource Conservation. Rice-Wheat Consortium
Auffhammer, M., Ramanathan, V., Vincent, J.R., 2011. Climate change, the monsoon
Technical Bulletin Series 7, Rice-Wheat Consortium for the Indo-Gangetic
and rice yield in India. Clim. Change 111 (2), 411–424.
Plains, New Delhi.
Barah, B.C., Narendranath, 2011. Status of SRI in India: upscaling strategy and global
Jenkins, M.W., Lund, J.R., Howitt, R.E., 2003. Economic losses for urban water
experience sharing. In: Proceedings of the Roundtable Discussion, The National
scarcity in California. J. Am. Water Works Assoc. 95 (2), 58–70.
Consortium on SRI, New Delhi.
Joshi, P.K., Jha, A.K., Wani, S.P., Joshi, L., Shiyani, R.L., 2005. Meta-analysis to assess
Birthal, P.S., Negi, D.S., Kumar, S., Agarwal, S., Suresh, A., Khan, Md.T., 2014. How
impact of watershed program and people’s participation. IWMI Research
sensitive is Indian agriculture to climate change? India J. Agric. Econ. 69 (4),
Reports H037467, International Water Management Institute, Colombo.
474–487.
Kostandini, G., Bradford, F.M., Omamo, S.W., Wood, S., 2009. Ex-ante analysis of the
Booker, J.F., Michelsen, A.M., Ward, F.A., 2005. Economic impact of alternative policy
benefits of transgenic drought tolerance research on cereal crops in low-income
responses to prolonged and severe drought in the Rio Grande Basin. Water
countries. Agric. Econ. 40 (4), 477–492.
Resour. Res. 41 (2), W02026.
Kumar, P., Joshi, P.K., Aggarwal, P.K., 2014. Projected effect of droughts on supply,
Cline, W., 1996. The impact of global warming on agriculture: comment. Am. Econ.
demand, and prices of crops in India. Econ. Pol. Weekly 49 (2), 54–63.
Rev. 86 (5), 1309–1311.
Kumar, K.S.K., 2009. Climate Sensitivity of Indian Agriculture: Do Spatial Effects
De Salvo, M., Raffael, R., Moser, R., 2013. The impact of climate change on
Matter? SANDEE Working Paper 45-09. SANDEE, Kathmandu.
permanent crops in an alpine region: a Ricardian analysis. Agric. Syst. 118,
Kumar, K.S.K., Parikh, J., 2001. Indian agriculture and climate sensitivity. Glob.
23–32.
Environ. Change 11 (2), 147–154.
Dono, G., Mazzapicchio, G., 2010. Uncertain water supply in an irrigated
Kumar, S.N., Aggarwal, P.K., Rani, S., Jain, S., Saxena, R., Chauhan, N., 2011. Impact of
Mediterranean area: an analysis of the possible economic impact of climate
climate on crop productivity in Western Ghats, coastal and north-eastern
change on the farm sector. Agric. Syst. 103, 361–370.
regions of India. Curr. Sci. 101 (3), 332–341.
ECA (Economics of Climate Adaptation), 2009. Shaping Climate Resilient
Kumar, V., Ladha, J.K., 2011. Direct-seeding of rice: recent developments and future
Development: A Framework for Decision-making. A Report of the Climate
research needs. Adv. Agron. 111, 297–413.
Adaptation Working Group. Climate Works Foundation, Global Environment
Kurukulasuriya, P., Kala, N., Mendelsohn, R., 2011. Adaptation and climate change
Facility, European Commission, McKinsey & Company, The Rockfeller
impacts: a structural Ricardian model of irrigation and farm income in Africa.
Foundation, Standard Chartered Bank and Swiss Re.
Clim. Change Econ. 2, 149–174.
Gautam, A., 2009. Impact Evaluation of Drought Tolerant Rice Technologies through
Lal, P.N., Mitchell, T., Aldunce, P., Auld, H., Mechler, R., Miyan, A., Romano, L.E., Zakaria, S.,
Participatory Approaches in Eastern India. Masters’ dissertation. The State
2012. National systems for managing the risks from climate extremes and disasters.
University of New Jersey, Rutgers.
In: Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J., Ebi, K.L., Mastrandrea,
GoI (Government of India), 2009. Manual for Drought Management. Department of
M.D., Mach, K.J., Plattner, G.K., Allen, S.K., Tignor, M., Midgley, P.M. (Eds.), Managing
Agriculture and Cooperation, Ministry of Agriculture, New Delhi.
the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. A
GoI (Government of India), 2013. Agricultural Statistics at a Glance. Department of
Special Report of Working Groups I and II of the Intergovernmental Panel on Climate
Agriculture and Cooperation, Ministry of Agriculture, New Delhi.
Change. Cambridge University Press, Cambridge and New York, pp. 339–392.
12 P.S. Birthal et al. / Food Policy 56 (2015) 1–12
Li, Y.P., Ye, W., Wang, M., Yan, X.D., 2009. Climate change and drought: a risk Rathore, B.M.S., Sud, R., Saxena, V., Rathore, L.S., Rathore, T.S., Subrahmanyam, V.G.,
assessment of crop-yield impacts. Clim. Res. 39 (1), 31–46. Roy, M.M., 2014. Drought conditions and management strategies in India. In:
Lybbert, T.J., Bell, A., 2010. Stochastic benefit streams, learning, and technology Country Report Presented at the Regional Workshop for Asia-Pacific on Capacity
diffusion: why drought tolerance is not the new Bt. AgBioForum 13 (1), 13–24. Development to Support National Drought Management Policies Hanoi, UN-
Martin-Ortega, J., Berbel, J., 2010. Using multi-criteria analysis to explore non- Water Initiative, May 6–9.
market monetary values of water quality changes in the context of the Water Sanghi, A., Mendelsohn, R., 2008. The impacts of global warming on farmers in
Framework Directive. Sci. Total Environ. 408 (19), 3990–3997. Brazil and India. Glob. Environ. Change 18 (4), 655–665.
Mendelsohn, R., Dinar, A., 2009. Climate Change and Agriculture: An Economic Sanghi, A., Mendelsohn, R., Dinar, A., 1998. The climate sensitivity of Indian
Analysis of Global Impacts, Adaptation, and Distributional Effects. Edward Elgar agriculture. In: Dinar, A., Mendelsohn, R., Evenson, R., Parikh, J., Sanghi, A.,
Publishing, Cheltenham. Kumar, K., McKinsey, J., Lonergon, S. (Eds.), Measuring the Impact of Climatic
Mendelsohn, R., Dinar, A., Sanghi, A., 2001. The effect of development on the climate Change on Indian Agriculture. World Bank Technical Report No. 409, World
sensitivity of agriculture. Environ. Dev. Econ. 6 (2001), 85–101. Bank, Washington.
Mottaleb, K.A., Rejesus, M.R., Mohanty, S., Murty, M.V.R., Li, Tao. Valera, H.G., Schlenker, W., Roberts, M.J., 2006. Estimating the Impact of Climate Change on Crop
Gumma, M.K., 2012. Ex-ante impact assessment of a drought tolerant rice Yields: The Importance of Non-linear Temperature Effects. Working Paper
variety in the presence of climate change. In: Paper Presented at the 13799, National Bureau of Economic Research, Cambridge.
Agricultural & Applied Economics Association Annual Meeting, Seattle, August Selvaraj, K.N., Ramasamy, C., 2006. Drought, agricultural risk and rural income: case
12–14. of a water limiting rice production environment, Tamil Nadu. Econ. Pol. Weekly
Munshi, K., 2004. Social learning in a heterogeneous population: technology 41 (26), 2739–2746.
diffusion in the Indian Green Revolution. J. Dev. Econ. 73 (1), 185–213. Sharma, P.K., Bhushan, L., Ladha, J.K., Naresh, R.K., Gupta, R.K., Balasubramanian,
Nelson, G.C., Rosegrant, M.W., Koo, J., Robertson, R., Sulser, T., Zhu, T., Ringler, C., B.V., Bouman, B.A.M., 2002. Crop water relations in rice–wheat cropping under
Msangi, S., Palazzo, A., Batka, M., Magalhaes, M., Valmonte-Santos, R., Ewing, M., different tillage systems and water management practices in a marginally sodic,
Lee, D., 2009. Climate Change: Impact on Agriculture and Costs of Adaptation. Food medium-textured soil. In: Bouman, B.A.M., Hengsdijk, H., Hardy, B., Toung, T.P.,
Policy Report. International Food Policy Research Institute, Washington, DC. Ladha, J.K. (Eds.), Water-wise Rice Production. International Rice Research
O’Brien, D., Hayenga, M., Babcock, B., 1996. Deriving forecast probability distributions Institute, Manila.
of harvest-time corn futures prices. Rev. Agric. Econ. 18 (2), 167–180. Sivakumar, S., Kerbart, E., 2004. Drought, sustenance and livelihoods: ‘Akal’ survey
O’Toole, T.C., 2004. Rice water: the final frontier. In: Paper Presented in the First in Rajasthan. Econ. Pol. Weekly 39 (3), 285–294.
International Conference on Rice for the Future. Bangkok, Thailand August 31– Stern, N., 2006. The Economics of Climate Change: The Stern Review. H.M. Treasury,
September 2. London.
Palanisami, K., Mohan, K., Kakumanu, K.R., Raman, S., 2011. Spread and economics TAC (Technical Advisory Committee of the CGIAR), 1994. A Review of CGIAR
of micro-irrigation in India: evidence from nine states. Econ. Pol. Weekly 46 Priorities and Strategies. TAC Secretariat, FAO, Rome.
(26/27), 81–86. Udmale, P., Ichikawa, Y., Manandhar, S., Ishidaira, H., Kiem, A.S., 2014. Farmers’
Pandey, S., Bhandari, H., Hardy, B., 2007. Economic Costs of Drought and Rice perception of drought impacts, local adaptation and administrative mitigation
Farmers’ Coping Mechanisms: A Cross-country Comparative Analysis from Asia. measures in Maharashtra State, India. Int. J. Disaster Risk Reduction 10, 250–
International Rice Research Institute, Manila. 269.
Pandey, S., Gauchan, D., Malabayabas, M., Bool-Emerick, M., Hardy, B. (Eds.), 2012. Virk, S.D., Witcombe, J.R., 2007. Trade-offs between on-farm varietal diversity and
Patterns of Adoption of Improved Rice Varieties and Farm-level Impacts in highly client-oriented breeding: a case study of upland rice in India. Genet.
Stress-prone Rainfed Areas in South Asia. International Rice Research Institute, Resour. Crop Evol. 54, 823–835.
Manila. Waddington, S.R., Li, X., Dixon, J., Hyman, G., de Vivente, M.C., 2010. Getting the
Pandey, S., Pal, S., 2007. Are less-favored environments over-invested?: the case of focus right: production constraints for six major food crops in Asian and African
rice research in India. Food Policy 32 (5–6), 606–623. farming systems. Food Secur. 2 (1), 27–48.
Pathak, H., Tewari, A., Sankhyan, S., Dubey, D.S., Mina, U., Singh, V.K., Jain, N., Bhatia, Wassmann, R., Dobermann, A., 2007. Climate change adaptation through rice
A., 2011. Direct-seeded rice: potential, performance and problems. Curr. Adv. production in regions with high poverty levels. SAT e-J. 4 (1), 1–24.
Agric. Sci. 3 (2), 77–88. Wassmann, R., Jagdish, S.V.K., Heuer, S., Ismail, A., Redona, E., Serraj, R., Singh, R.K.,
Peck, D.E., Adams, R.M., 2010. Farm-level impacts of prolonged drought: is a Howell, G., Pathak, H., Sumfleth, K., 2009. Climate change affecting rice
multiyear event more than the sum of its parts? Aust. J. Resour. Econ. 54 (1), production: the physiological and agronomic basis for possible adaptation
43–60. strategies. Adv. Agron. 101, 59–122.
Pérez y Pérez, L., Barreiro-Hurlé, J., 2009. Assessing the socioeconomic impacts of World Bank, 2013. Turn Down the Heat: Climate Extremes, Regional Impacts, and
drought in the Ebro River Basin. Span. J. Agric. Res. 7 (2), 269–280. the Case for Resilience. A Report for the World Bank by the Potsdam Institute for
Pray, C., Nagarajan, L., Li, L., Huang, J., Hu, R., Selvaraj, K.N., Napasintuwong, O., Babu, Climate Impact Research and Climate Analytics. World Bank, Washington, DC.
R.C., 2011. Potential impact of biotechnology on adaption of agriculture to Xiao-Jun, W., Jian-Yun, Z., Shahid, S., El Mahdi, A., Rui-Min, H., Zhen-Xin, B., Ali, M.,
climate change: the case of drought tolerant rice breeding in Asia. Sustainability 2012. Water resources management strategy for adaptation to droughts in
3 (10), 1723–1741. China. Mitigation Adapt. Strat. Glob. Change 17 (8), 1–15.
Quiroga, S., Iglesias, A., 2009. A comparison of the climate risks of cereal, citrus, Yu, T., Babcock, B.A., 2010. Are U.S. corn and soybeans becoming more drought
grapevine and olive production in Spain. Agric. Syst. 101, 91–100. tolerant? Am. J. Agric. Econ. 92 (5), 1310–1323.