Chakrabarty r 20150709_1730_upmc_jussieu_-_room_101_(building_14-24)

Climatic Change and Indian Agriculture: A Case Study of West Bengal

Ranajit Chakrabarty

Department of Business Management, Calcutta University

Smwarajit Lahiri Chakravarty

St. Xavier’s College, Calcutta

Somarata Chakrabarty Cammelia Institute of Technology, West Bengal

Introduction Global Warming

The temperature of the earth’s surface has increased by about 0.60C in the 20th century.

By 21st century, the Earth’s temperature is expected to increase by 1.60 C to 2.00C.

Effects

intense rainfall,

melting of glaciers,

rise in the sea level and

increase in the rise of the extreme weather phenomena.

Impact on the productivity of crops:

effect of climatic changes on the output of rice and wheat

Effect of changes of both temperature and precipitation rise

2

Outline

1) Analysis of role of agriculture on Indian and West Bengal economy

2) Agriculture and Climate change

3) Climate Sensitivity

4) Future projections and worries (through Cline’s method)

5) Conclusion and policy implications

3

Stationary test results- Indian GDP vs West Bengal SDP

We consider India and one of her states which is West Bengal.

We consider the data set from 1970 to 2010- the data source being the CSO data for GDP of India and SDP of West Bengal; the source of climatic data being Indian Institute of Tropical Meteorology.

To study the nature of the series we use: i. The data of agricultural output and GDP of India along with

agricultural output and SDP of West Bengal are transformed into constant prices at 1993/94 prices.

ii. Then, logarithm values of GDP (LG), SDP (LS) and the agricultural output of India (LAI) and West Bengal (LAW) are considered. The following are observed

4

We carried out the unit root tests to test the Stationarity of the series and find that the variables LG, LS, LAI and LAW are all I(1).

Based on unit root tests, one can go for cointegration but after determining the optimum lag length.

Here the endogenous variables are LG, LS, LAI and LAW and it is observed that the optimal lag length is one (1).

Hence, in the analysis from now on, following SIC, the lag length of one is considered.

5

Stationary test results- Indian GDP vs West Bengal SDP

Cointegration Tests

The cointegration tests are generally carried out to test whether the selected variables exhibit same stochastic trends or not, given that the variables are LG, LS, LAI and LAW.

Using the trace test (Johansen, 1988) and the maximum eigen value test (Johansen and Juselius, 1990) we obtained:

i. There is cointegration between LG and LAI as indicated by the trace test but not the maximum eigen value test. Hence, a long run dynamic relation is visible for these two variables. The system may be represented in a VECM.

ii. There is no cointegration between LS and LAW as indicated by both the trace test and the maximum eigen value test. The system may be represented in a VAR, given that LS and LAW are not cointegrated.

6

Granger Causality Tests

LS and LAW inspite of being of I(1), are not cointegrated, and one may now test for the Granger Causality within first difference vector autoregressive (VARD) models (Konya, 2004).

As LG and LAI are of I(1), and are cointegrated, the test for the Granger Causality may be carried out at level.

The results indicate the following conclusions –

i. The null hypothesis that LG does not Granger cause LAI is

accepted. However, the null hypothesis that LAI does not Granger cause LG is rejected, at 10% level of significance. Here, the Granger Causality is unidirectional.

ii. The null hypothesis that LAW does not Granger cause LS is accepted. However, at 10% level of significance, the null hypothesis that LS does not Granger cause LAW is rejected. Hence, unidirectional Granger Causality is observed.

7

Agriculture and Climate Change- model specification

The economic impacts on agriculture due to climatic change have been studied in great detail.

We introduce a variant of the Ricardian model which is based on the availability of the data and have used net productivity (NPR) as a social indicator to evaluate the changes.

The effect and changes in the environment is neatly captured through NPR.

So, the functional relation is-

NPR = f (Tj, Rj, h, pden, a, cr) … …… (1)

where Tj temperature in C in jth season,

Rj precipitation in mm in jth season.

Equation (1) also has the control variables which are –

h altitude of that region given in mm,

pden population density,

a net area under cultivation,

cr credit available which captures the purchasing power to buy inputs like

tractors, cows, seeds, etc.

8

Agriculture and climate change

The data source is Indian Institute of Tropical Meteorology (IITM), Pune. We have subdivided the entire year into 4 parts following the Indian Metrological departments’ process of subdivision.

Jan, Feb, Mar --1st Part

April, May, June --2nd Part

July, Aug, Sep -- 3rd Part

Oct, Nov, Dec -- 4th Part

The Hausman test(1978) tries to find out whether random effects estimation

would be almost as good as the fixed effects if in a model and data we have fixed

effects estimation as appropriate.

Estimation procedure, following Hausman test, states we are using fixed effect

criteria against the random effects.

9

Hausman test, Estimation, Results and Analysis

The dependent variable in each case is net productivity(NPR) which is expressed at constant prices by considering 1993/ 94 as the base year. The interesting results are --

1) In all the three periods the control variables have the expected sign.

2) F tests rejected null hypothesis that climate coefficients do not affect agriculture.

3) Apart from a few exceptions, the climate coefficients are all significant

4) The climate coefficients are affecting agriculture in an adverse direction.

5) In a dynamic scenario, we see that over time the intensity of these effects are increasing which can lead us to conclude, especially with respect to rice and potato, the increasing negative climatic changes of the agricultural output. This phenomenon is uniform across India.

10

Case of West Bengal

In this analysis, the effects of climatic changes are

measured through the changes in net productivity.

We find out the impacts by considering the six districts

(South 24 Parganas, Howrah, Hoogly, Nadia, Burdwan and

North 24 Parganas)separately.

The average effects for all the years are mentioned in

detail in Table E. We select the output of 1991/92 at

1993/94 prices as the base years. Thus, the NP values are

expressed at constant prices.

11

TABLE E- Dynamic climatic changes on Agriculture

Situation 1970 – 1980

% change NP

wrt 1991/92

1980-1991

% change NP

wrt 1991/92

1991-2010

%-change NP

wrt 1991/92

+ 1.5C/5% - 1.9% - 6.8% - 10.9%

+ 3.0% / 10% - 11.1% - 16.2% -18.8%

West Bengal

In a dynamic scenario, these effects are increasing which can lead us to conclude the increasing negative climatic changes of the agricultural output especially with respect to rice and potato.

We find that the modulus (absolute) value of the effects is increasing. But in the period of green revolution, these effects are subsequently low which may be due better use of fertilizers, HYV seeds and implementation of better techniques of production.

In the first and second periods, we find the impacts are high which may indicate the vulnerability and weakness of Indian agriculture that is growing over time. But in the last period we find that the effects are much higher due to the neglect of agriculture in the post liberalization period.

13

Climate Sensitivity-Spatial Autocorrelation

To get a good estimate of the climate sensitivity of agriculture, it is necessary to account for the spatial autocorrelation.

We have considered a linear model. We study the effects on climate sensitivity due to spatial autocorrelation for the period 1970 – 2010.

In the NP vs Wy plot, the scatter points are clustered in the north-east and south-

west areas with Moran’s I being 0.459.

positive spatial autocorrelation.

The equations are estimated for time period 1950 – 2010.

Two types of climate coefficients are observed-- one that accounts for the

autocorrelation while the other set is devoid of this autocorrelation.

We find that the latter is higher in numerical terms than the former, in most of the

case.

The analysis indicates that there is sufficient presence of climatic sensitivity on

Indian agriculture. 14

We consider the effects with or without autocorrelation under two hypothetical

situations as shown in Table 3.

1) We have 1.50C increase in temperature and a 5% increase in precipitation.

2) We have 3.00C increase in temperature and 10% increase in precipitation

Table 3: Impacts of temperature and precipitation

Temp/rainfall Without Autocorrelation With Autocorrelation (lag

model)

Impacts % Change NP Impacts % Change--NP

1.5C/5% - 69.8 - 11. 24 11.1 1. 22

3.0C/10% - 87.8 - 11. 2 7.86 0. 69

Climate Sensitivity-Spatial Autocorrelation

Climate Sensitivity-Spatial Autocorrelation

The impacts speak about the effects that are generated and are observed as the percentage change in NP.

Under autocorrelation, that the values are positive but if autocorrelation is not considered, the values are negative.

We find that the lag model gives us positive changes though in numerical terms these are not very strong.

16

Future Projections And Worries

Precipitation (%) Temperature Rise (C)

NE NW SE SW NE NW SE SW

Jun – Mar -5.2 5.6 -12.3 16.4 3.87 2.16 3.11 2.41

Apr -Jun 11.2 8.7 21.8 20.8 2.15 2.02 1.89 2.18

Jul – Sep 13.4 9.1 8.2 5.3 1.86 1.63 1.12 1.65

Oct - Dec 6.4 31.2 1.1 5.2 2.98 2.13 1.62 1.43

17

Following Cline (2007), climate projections for West Bengal can now be generated.

Our analysis highlights the temperature changes region wise and rainfall changes

season wise for the period 2020–2050.

Table 4: Temperature and precipitation changes in West Bengal

Note: NE – North East; NW – North West; SE – South East; SW – South West

Future Projections And Worries

Results tells us that there is a tendency for the overall temperature to increase.

In July – September, the overall temperaures have a tendency to decline due to the rainfalls in those months.

Also, in all the four regions, the temperatures decline from Jan-March to April – June only to fall drastically in July – Sep.

After the monsoons, the temperature again picks up in the months of Oct- Dec. But this rise is not so prominent due to the winter seasons. Also, in the months of Oct – Dec, the temperature rise of all regions is less prominent due to the influence of the seas.

The projected figures on precipitation clearly tell us that all the four regions more or less would experience an increase. However, during the season of Jan- March, the regions of NE and SE would experience a negative precipitation.

18

Conclusions and policy implications

For India as a whole the Agriculture output influences the GDP but SDP influences the WB agricultural

output and it is unidirectional in both the cases.

In dynamic scenario, the impacts of climatic changes have an upward trend of rising.

Indian agriculture is very climate sensitive, especially with respect to the basic food items of rice,

wheat and potato.

Spatial autocorrelation becomes an important test of analysis and more so for Indian agriculture.

If the autocorrelation effects are introduced in the model, the intensity of the climatic sensitivity on

Indian and WB agriculture is much lower.

Thus public investment in agriculture, particularly in irrigation and research and extension, is needed

for higher growth.

The policy makers have to answer the question of adaptation- how and when to adapt the

appropriate policy is important.

The rate at which adaptation procedure selected and implemented, called the speed of adjustment

and the cost benefit analysis of each adjustment must be analyzed.

The shift in consumption patterns indicates that in allied activities like livestock and fisheries there is

need for crop diversification and improvements which would improve employment opportunities.

19

Thank you !

Ranajit Chakrabarty

[email protected]

[email protected]