Estimation of Agricultural Supply Response by Cointegration Approach

8/14/2019 Estimation of Agricultural Supply Response by Cointegration Approach.

http://slidepdf.com/reader/full/estimation-of-agricultural-supply-response-by-cointegration-approach 1/49

ESTIMATION OF AGRICULTURAL SUPPLY RESPONSE BY

COINTEGRATION APPROACH

A Report Submitted under Visiting Research Scholar Programme 2008

Amarnath Tripathi

Indira Gandhi Institute of Development Research

Gen. A. K. Vaidya Marg, Film City Road

Goregaon (E), Mumbai- 400065



2

CERTIFICATE



3

ACKNOWLEDGMENT

I would like to thank Indira Gandhi Institute of Development Research, Mumbai for

providing me visiting scholarship and all the facilities required for the implementation of

my study. I am extremely grateful to Dr. G. Mythili, Associate professor, Indira Gandhi

Institute of Development Research, Mumbai for her kind guidance and sincere support

during my stay at IGIDR. I extend my warm thanks to Professors J. Sarkar, S. Thamos , and

C. Verramanai for giving me valuable suggestions during the course of my study.

I would like to express my deepest sense of gratitude to Dr. A. R. Prasad, Professor,

Banaras Hindu University, Varanasi my PhD supervisor at Banaras Hindu University

without whose help and encouragement this report would not have been completed.



4

CONTENTS

1. List of Tables ...(05)

2. Abstract..(06)

3. Introduction...(07)

4. A Brief Literature Review on Supply Response.(09)

5. Basic Framework and Approaches in Estimation..(19)

a. Indirect Approach...(19)

b. Direct Approach..(20)

c. Nerlove and ECM Model(21)

6. Empirical Framework.(23)

a. Model..(23)

b. Estimation Procedure.(25)

c. Selection of States..(26)

d. Period of Study(26)

e. Data Sources(27)

7. Results and Discussion(28)

8. Conclusion and Policy Implication(41)

9. References(43)



5

List of Tables

Table 1: Earlier Studies in Indian Context……………………………………………………................(13)

Table 2: Share of Agricultural Income in total income for all major states………………………..(26)

Table 3: The Mean Values of Selected Variables………………………………………………............(27)

Table 4: Descriptive Statistics………………………………………………........................................(28)

Table 5: Correlation Matrix………………………………………………………………………………….(28)

Table 6: Results of ADF Test……………………………………………………………………………….(30)

Table 7: Results of ADF Test……………………………………………………………………………….(31)

Table 8: Bivariate Analysis for All India……………………………………….....................................(32)

Table 9: Bivariate Analysis for High Agricultural Based States……………………………………..(32)

Table 10: Bivariate Analysis for Medium Agricultural Based States………………………………..(33)

Table 11: Bivariate Analysis for Low Agricultural Based States…………….................................(33)

Table 12: Multivariate Analysis for All India……………………………………………………………..(34)

Table 13: Multivariate Analysis for High Agricultural Based States………………………………...(34)

Table 14: Multivariate Analysis for Medium Agricultural Based States……………………………(34)

Table 15: Multivariate Analysis for Low Agricultural Based States…...........................................(35)

Table 16: Results of ECM for All India……………………………………………………………………(36)

Table 17: Results of ECM for High Agricultural Based States……………………………………….(37)

Table 18: Results of ECM for Medium Agricultural Based States……..........................................(37)

Table 19: Results of ECM for Low Agricultural Based States………………………………………..(37)

Table 20: Aggregate Agricultural Elasticities w.r.t. Agricultural TOT………………………………(38)



6

ABSTRACT

The issue of agricultural supply response is a very important one as it has an impact on

growth, poverty, and environment. The size of agricultural supply response is expected to

improve after removing some of the constraints that farmers were facing before. Though

many constraints have been removed from agrarian system and many incentives have been

provided to farmers, still the supply response for Indian agriculture is price inelastic. Hence

the question “why supply response is price inelastic” becomes relevant. The present study is

an attempt to find supply response through cointegration approach and to see if the response

has been better at the all India level in comparison to previous studies. Further, it also focuses

on the question whether there is difference in the supply response among highly agricultural

based, medium agricultural based, and low agricultural based states. The study indicates that

aggregate agricultural output elasticity with respect to agricultural TOT is very low and not

statistically different from zero.

Key Words: Agricultural Supply Response, Cointegration Analysis, and Error Correction

Model

JEL Classification: C22, Q11



7

INTRODUCTION

One of the most important issues in agricultural development economics is supply response

since the responsiveness of farmers to economic incentives*

largely determines agriculture’s

contribution to the economy. Further, the response elasticities are also important for policy

decision regarding agricultural growth. That is why it has been a debatable issue among

economist and policy makers in both India and abroad for the last five decades. Agricultural

Supply Response represents change in agricultural output due to a change in agricultural

output price. The concept of supply response is dynamic and different from supply function

which is the static concept. The supply function describes a price quantity relation, where all

other factors are held constant. The response relation is more general concept; it shows the

change in quantity with changes in prices as well as supply shifters†

and, therefore,

approximates to the long run, dynamic concept of supply theory.

In India, agriculture has made remarkable growth after initiation of new agricultural strategy

in mid-sixties. Indian agriculture has progressed not only in output and yield terms but the

structural changes have also contributed. Government provided more incentive such as

remunerative prices to protect farmers’ interest, availability of credit facilities, improving

irrigation facilities, improving markets of both output and input, more investment in

agricultural research and extension services, etc. to farmers to produce more. After these

developments, it is expected that farmers would become more price responsive. The previous

studies as well as recent studies (Krishna, 1980 Palanivel, 1995, Rao, 2003, etc) have shown

that Indian agriculture is low price responsive. Then, the question arises “why is the supply

response low for Indian agriculture?” The present study is an attempt to reexamine supply

response through better approach as compared to the previous studies. Further, it also

focuses on the question whether there is difference in the supply response among highly

agricultural based, medium agricultural based, and low agricultural based states.

* Economic incentives are offered to encourage people to make certain choices or behave in a certain way.† Except product own price there are several factors that influance supply; as for example, ralative prices of the

substitute products, climatic conditions, technological, progress (improvement in the art of production), changes

in the institutional and policy variables and even attitude of the producers. In economic terminology these

factors are called supply shifters.



8

The previous studies on agricultural supply response in India used time series‡

and panel

data§. Most of the studies applied Nerlovian Framework

**(1958). However, most economic

time series data are trended overtime and regression between trended series may produce

significant results with high R 2’s, but may be spurious (Granger and New Bold, 1974). So,

we use cointegration analysis††

and error correction model (ECM)‡‡

to overcome the problem

of spurious regression§§

.

The ECM takes into account the partial adjustment in production and the mechanism used by

farmers in forming expectations. These are the fundamentals of agricultural supply response

model. So, ECM can be used in this context. We found one study related to India and many

studies for other countries based on application of ECM in estimating agricultural supply

response.

The present study is organized as follows: Section (II) presents review of literature related to

India agriculture. Section (III) discusses fundamentals and approaches of estimation of

agricultural supply response. This section also shows the relevance of ECM in estimating

agricultural supply response. Section (IV) discusses methodology and estimation procedure

used in the study. Section (V) gives the results and findings. Section (VI) makes some

observations on the causes of low supply response. Section (VII) concludes the whole study,

gives limitation of the study and suggests policy implication.

‡ A time series is a set of observations on the values that a variable takes at different times.§ Panel data is a special type of pooled data in which the same cross-section unit is surveyed over time.** Nerlove proposed a dynamic distributed lag model in which he introduced a dynamic element into the

response equation by creating expectation.†† Cointegration is an economectric property of time series variables. If two or more series are themselves non-

stationary, but a linear combination of them is stationary then the series are said to be cointegrated.‡‡ Error correction model is useful for representing the short run relationships between variables. §§ Sometimes we expect no relationship between variables, yet a regression of one on the other variable often

shows a significant relationship. This satiation is called spurious regression.



9

A BRIEF REVIEW OF LITERATURE ON SUPPLY RESPONSE

There are two types of studies found in the existing literature; one is at individual crop level

and the other is at the aggregate output level. First type of studies shows change in the

composition of agricultural output to a change in the relative price of individual agricultural

commodities. On other hand, second type of studies shows change in total agricultural output

due to change in the relative price of agricultural commodities compared with industrial

goods, referred to terms of trade in the literature. A review of relevant literatures follows.

Raj Krishana’s (1963) study is the first study among those studies that used Nerloveframework in Indian Context. He used pre-Independence data for Punjab (Undivided) and

concluded that the Punjab farmers have responded to economic stimuli.

Jai Krishna and Rao (1965, 1967) tried quite a few price formulations to specify the price

variables and those on which farmers base their expectations. They found that traditional

regression models give satisfactory results if not superior results compared with those

obtained by using adjustment lag model as far as proportion of explained variation in wheat

acreage is concerned.

Parikh’s (1967) work of supply response through the distributed lag models comes very close

to the Nerlovian framework. He has used both adjustment lag and price expectation models

along with the distributive lags. After analyzing different formulations with the data given in

the appendices of Dharam Narain’s study (1965), Parikh arrived at the result that non-price

factors are quite important as compared to the price factors. It is worth noting here that the

data set covered the period prior to 1939.

Herdt, (1970) has evaluated supply response at aggregate output level for period 1907 to

1964 for Punjab agriculture. He divided the reference period into two sub-periods: 1907 to

1946 and 1947 to 1964. He concluded that the 1907-1946 period tends to support the

hypothesis of a positive, although small, aggregate supply response of agriculture. The

results for 1951-1964 suggest the opposite, or at least do not support the hypothesis. He used

disaggregated approach given by Tweeten and Quance (1963) to measure supply elasticites.



10

Madhavan (1972) utilized a production function framework to arrive at the desired demand

for land. A production function under the assumption of constant ratios of elasticity is

maximized to obtain the factor demand equation. The estimated equation indicated that

actual planted area of a crop in the given period is a function of the log of relative price of the

crop, the expected yield of the crop under consideration as well as that of the competing crop

and the weather index.

Cummimgs (1975) examined supply responsiveness of the Indian farmers for post-

Independent period by using Nerlove model. This work was followed by another study which

is the joint work of Askari and Cummings (1976). They used a slightly modified version of

the Nerlove model with maximum-likelihood estimation procedure to overcome the problem

of serial correlation and non-consistency of the estimates.

Narayana et al . (1981) estimated supply response crop-wise for Indian agriculture. They

formulated expectation function for each crop by isolating stationary and random

components in past prices and attach suitable weights for both in prediction. The method is

based on ARIMA technique combined with BOX-Jenkins procedure.

Bapna et al. (1984) derived a system of output supply and factor demand equation from the

profit function. They began with monotonically increasing profit function and derived output

supply and factor demand curves. Two systems are derived from the maximization of the

profit function, namely, Generalized Leonteif and Normalized Quadratic Systems. The

authors pooled the time series and cross section data by following error component model. It

is confirmed by the authors that a Maximum-Likelihood Estimation procedure would be

better than the three-stage least-squares procedure. The results were obtained for 96 districts

spread over semi-arid tropical regions of the country. The authors indicated that 25 out of the

32 own elasticities had the anticipated sign and also demonstrated remarkable extent of price

responsiveness of the semi-arid tropical farmers. Very high supply elasticity was noted for

sorghum despite the small proportion of its marketed surplus.

Hazell, et al (1995) used national and state level data for estimating the aggregate supply

elasticities for Indian agriculture. The results from the national and state level analyses are

very similar, and show that aggregate supply is inelastic. Both levels of analyses showed that



11

growth in agricultural output in India in recent decades is largely attributable to increased

irrigation.

Palanivel, (1995) estimated the aggregate agricultural supply response by using more

appropriately constructed variables for the period 1951-52 through 1987-88. The model was

developed within the basic Nerlovian partial adjustment framework. He constructed his own

agricultural terms of trade (TOT)***

. He used farm harvest prices and retail prices in

estimating the index for prices received and the index for prices paid by farmers,

respectively. He argued that since farmers sell bulk of their products immediately after the

harvest, farm harvest prices seemed to be a better approximation for the prices received.

Similarly, he also pointed that as farmers purchase their requirements (particularly family

consumption items) from retail markets, rural retail prices seemed to be a better

approximation for the prices paid. His results indicated that the elasticity of aggregate

agricultural output with respect to TOT is positive and statistically significant but non-price

factors are equally, if not more, important.

Mishra, (1998) attempted to assess the impact of economic reforms initiated in 1991 along

with price and non-price factors on aggregate supply in the post-green revolution period. His

results assured that the aggregate supply measured either through aggregate output or

marketable surplus does respond significantly positively to terms of trade.

Surajit Deb, (2003) explored the presence of long-run relationship between TOT and

agricultural output by using cointegration analysis and error correction model. The bivariate

results between TOT and output level in agriculture reflect no statistically significant

cointegration. The non-cointegratedness indicates that no direct long-run relationship exists

between TOT and output level in Indian agriculture. This, in turn, would suggest that a

favourable TOT structure alone may not be effective in sustaining higher agricultural growth.

Further, he included irrigation ratio as a technology between TOT and output. Then he found

that variables are cointegrated. Finally he concludes that growth in agricultural output may

respond better if specific structural variables are suitably combined with the price variables.

Chandrasekhara Rao, (2004) examined agricultural supply response at aggregate level for

Andhra Pradesh by using Nerlove Partial Adjustment Model. He found the short-run

*** The agricultural terms of trade is defined as the ratio of prices received and prices paid by farmers.



12

elasticites of output with respect to TOT for aggregate agriculture vary from 0.20 to 0.29 and

the long-run elasticities vary from 0.21 to 0.31. The results also indicated that non-price

factors are more important determinants in aggregate agricultural supply than price related

factors in the state of Andhra Pradesh.

Mythili, (2008) estimated supply response for major crops during pre-and post-reform

periods using Nerlovian adjustment-cum-adaptive expectation model. Estimation is based on

dynamic panel data approach with pooled cross section - time series data across states for

India. The study found no significant difference in supply elasticities between pre-and post-

reform periods for majority of crops. This study also indicated that farmers increasingly

respond better through non-acreage inputs than shifting the acreage. This includes better

technology, use of better quality of inputs and intensive cultivation.

Many more studies (Thamarajaki, 1977; Krishna, 1982; Mungekar, 1997; Desai and

Namboodiri, 1997) are also available in the context of Indian agriculture. Thamarajakshi,

(1977) reported a statistically significant and effective relation between aggregate farm

output and non-price factors (irrigation) but the author could not detect any statistically

significant impact of TOT on agricultural output. Krishna, (1982) also drew same conclusion

but difference between both the studies was that Krishna observed a marginal impact of TOT

on agricultural output. Mungekar, (1997) and Desai & Namboodiri, (1997) reported negative

impact of TOT on agricultural output. Both studies also observed significant impact of non-

price factors such as irrigation, rainfall, area under high yielding verity seeds, fertilizer use,

rural roads and other infrastructure facilities.

After reviewing these studies, some common inferences can be drawn. The supply response

of individual crop is well documented but study related to aggregate output remains to be a

less researched area. Most of the studies have followed same methodology given by Nerlove

in original form or with some modification. Most of the studies have reported low supply

response. They also indicated that non-price factors are relatively more important than price

factors. However, there has been controversy as to whether aggregate agricultural supply is

really not responsive. Schiff and Montenegro (1997), argued that aggregate agricultural

supply response is, in fact, high but that there are other constraints such as financing that

hinder this response such that a low elasticity is found. A lot of methodological questions



13

have been raised on the previously used model (Nerlove Model) and the estimation

techniques applied. For instance, the Nerlove Model is unlikely to capture the full dynamics

of agricultural supply (Thiele, 2000), this model has inability to give adequate distinction

between short-and long-run elasticities (McKay et.al., 1998), it uses integrated series, which

poses the danger of spurious regression (Granger & Newbold, 1974), etc. Therefore, an

alternative of Nerlove model is required. For this particular case, the suitable techniques are

the general co-integration methodology and the ECM (detailed in the next section). Indeed,

one work (Surjit Dev, 2003) has been done based on cointegration approach but his special

emphasis was on long-run relationship between agricultural output and price. The present

study is an attempt to fill this gap. This study has done the analysis at the aggregate all Indiaas well as at the diaggregated state level. Supply response at the disaggregated level can

throw more light since aggregation at the all India level might have hidden the variation that

could explain the response better. This study has classified states into low, medium and high

agricultural based states and to see if the response varies significantly between them.





S.No. Author Title of the Paper Objective/Objectives of theStudy

Data Base/ Period of Study& Methodology

Estimated Model R

7. Herdt (1970) A Disaggregated Approachto Aggregate Supply

To evaluate supplyresponse of aggregate

agricultural output

i. Time seriesii. 1907-1964

iii. OLS estimation procedure

Multiple RegressionModel using

disaggregatedapproach given byTweeten and Quance

Ts

ir6

8. Maji et.al (1971) Dynamic Supply andDemand Models for Better Estimation and Projections: An Econometric Study for Major Foodgrains in thePunjab Regions

To estimate supplyand demandelasticities

i. Time seriesii. 1948/49 – 1965/67iii. OLS estimation procedure

Nerlovian adjustmentmodel. s

9. Kaul and Sadhu(1972)

Acreage Response to Pricesfor Major Crops in Punjab:

An Econometric Study

To estimate acreageresponse with respect

to change in relativeprices and riskvariable.

i. Time seriesii. 1960/61 – 1969/70

iii. OLS estimation procedure

Nerlovian adjustmentmodel r

s

10. Madhavan (1972) Acreage Response of IndianFarmers: A Case Study of Tamil Nadu

To examine whether and to what extentIndian farmersinfluenced in their production decisionsby economic stimuli

i. Time seriesii. 1947/48 – 1964/65iii. Production function

approachiv. OLS estimation

procedure

Multiple regressionmodel

Trrdf



S.No. Author Title of the Paper Objective/Objectives of theStudy

Data Base/ Period of Study& Methodology

Estimated Model R

11. Cummings (1975) The Supply Responsivenessof Indian Farmers in the

Post-Independence Period

To examine supplyresponsiveness of

Indian farmers

i. Cross-sectionii. Maximum likelihood

estimation procedure

Nerlovian adjustmentmodel

Td

rTtie

12. Narayana et.al. (1981)

Estimation of Farm SupplyResponse and Acreage Allocation

To investigate how dofarmers form their expectations and howdo their expectationsaffect their crucialdecisions about landallocation?

i. Time seriesii. 1950 -1974iii. Box-Jenkins estimation

procedure

Autoregressiveintegrated movingaverage model

Tscm

13. Bapna et.al.(1984) Systems of Output Supplyand Factor DemandEquations for Semi-AridTropical India

To estimate supplyelasticities

i. Panel Dataii. Error component modeliii. Maximum likelihood

estimation procedure

Output supply andinput demand equationderived from profitunction

Ders

14. Hazell et.al. (1995) Role of Terms of Trade inIndian Agricultural Growth: A

national and State level Analysis

To analyze relativecontributions of terms

of trade and non-pricevariables in explainingagricultural growth

i. Time series (for nationallevel analysis) and panel

data (for state levelanalysis)ii. 1952 – 1988 (for national

level) 1971-1988 (for state level

iii. OLS and 2SLSestimation procedure

Multiple regressionmodel

Tn

aasi







BASIC FRAMEWORK AND APPRAOCHES USED IN ESTIMATION

In agriculture the observed prices are known after the production has occurred, while

planting decisions are based on the prices expected to prevail later at the harvest time.

Because of this time lag, producer price expectations play a key role in the analysis. Three

alternative agricultural producer price expectations hypotheses commonly found in the

literature are naïve expectation, adaptive expectation, and rational expectation. Naïve

expectation means expected price is the actual price in the previous period. Adaptive

expectations means that people form their expectations about what will happen in the future

based on what has happened in the past. According to rational expectation an economic agent

forecast his future event on the basis of all available information. Farmers, especially in the

developing countries, are mostly low literate and hence it is hard to obtain all the relevant

information. Therefore, rational expectation behavior is not relevant. The present study

assumed that farmers learn from their past mistakes, and so in this situation, adaptive

expectation becomes more relevant. Further, it is also important in the analysis of supply

response that observed quantities may differ from desired ones because of the adjustment lag

of variable factors.

There are two approaches to estimation of agricultural supply response; indirect structural

form approach and direct reduced form approach.

(A) Indirect structural form approach

This approach involves derivation of input demand function and supply function from the

available data, information relation to production function, and individuals’ behaviours. This

approach is more theoretically rigorous but fails to take into account the partial adjustment in

production and the mechanism used by farmers in forming expectations. This approach

requires detailed information on all the input prices. Moreover, the agricultural input markets

are not functioning in a competitive environment in India, particularly land and labour

markets. Market intervention in delivering material inputs to the farmers is a common

practice. It is difficult to get information on price at which the inputs are supplied to the

farmers. Keeping in view these aspects, most of the previous studies have chosen second

approach; “Direct reduced form approach”.



20

(B) Direct reduced form approach

In this approach the supply response is directly estimated by including Partial Adjustment

and Expectations Formation. This is also known as Nerlovian Model. Most of the existing

studies on the agricultural supply response in India and abroad have applied the Nerlove

method. According to Nerlove, desired output can be expressed as a function of expected

price and exogenous shifters:

where, Q*t is desired output, P*

t is expected price, Z t is a set of exogenous shifters such as

technology change, weather condition, etc.

Actual output may differ from the desired ones because of the adjustment lags of variable

factors. Therefore, it is assumed that actual output would only be a fraction δ of the desired

output.

where, Qt is actual output in period t, Qt-1 is actual output in period t-1, and δ is adjustment

coefficients. Its value lies between 0 and 1.

The farmers’ expected price at harvest time can be observed. So, we have to formally

describe how decision makers form expectations based on the knowledge of actual and past

price and other observable information. We may think that farmers maintain in their memory

the magnitude of the mistake they made in the previous period and learn by adjusting the

difference between actual and expected price in t-1 by a fraction λ

Putting the value of P*t and Q* t from equation 2 and 3, in equation 1, the equation 1

becomes

where, A0 is a d λ, A1 is b d λ, A2 is (1- d) +(1- λ), and A3 is c d.



21

This model is estimated from variables measured in levels and, should the suspicion of non-

stationarity be confirmed, as it happens with most economic series, the statistical significance

of the t tests of the estimated regressions lose sense. It is here where the use of advanced

econometric methodologies becomes relevant. For this particular case, the suitable

techniques are the general co-integration methodology and the ECM. The ECM allows

examining existence of long-run and short-run relationship among variables. All the used

variables are stationary and so the results are consistent. Instead of consistent results, the

ECM also considers the partial adjustment and the mechanism used by farmers in farming

expectations which are the fundamental in the analysis of agricultural supply response. In the

next section we show that partial adjustment model is nested within the ECM.

(C) Error Correction Model and Partial Adjustment Model

Let us take two variables Q and P and assume that both are integrated of order 1. The long-

run relationship between them can be expressed as:

Given that Qt and Pt are cointegrated, say of order 1, the term ut will be stationary and an

ECM-type representation exists for these two variables (Granger´s Representation Theorem),

which is expressed as:

Putting the value of ut-1 from equation 5,



22

Compare equations (9) and (4). They are identical after ignoring exogenous additional

variables - only if the term involving Pt is omitted from the ECM specification, which shows

that the partial adjustment model is nested within the ECM model, which constitutes a more

general specification of the problem.



23

EMPIRICAL METHODOLOGY

(A) Model

Agricultural supply depends on both output and input prices. An increase in output prices

increases profit and increase in profit provides incentive to farmers to produce more.

Similarly, an increase in input prices increases cost of production and increase in cost is

disincentive to produce more. Thus, real output price, that is, output over input price ratio

positively affects the agricultural production. Except for real output prices, there are several

factors that affect agricultural production. Manmingi (1997) subsumed these factors under

four different categories: rural infrastructure, human capital, technology, and agroclimatic

conditions. All rural infrastructure services such as irrigation facilities, accessibility of roads,

markets facilities, farmers access to credit, agro extension services, availability of fertilizer,

high yields variety seeds and pesticides, communication, and transport facilities are expected

to positively affect agricultural output through its effect on productivity and cost of

production. Education, agricultural extension, agricultural research and other technological

progress are supposed to have positive effect on agricultural production by improving

efficiency and reducing cost. Among the agroclimatic factors, soil quality and the intensity

and regularity of rainfall are likely to be most decisive for the supply response.

We cannot incorporate all of these factors due to non-availability of data and quantification

problems (some variables such as soil quality cannot be quantified). Considering these

problems, long-run supply response is estimated using variables indicated in equation 10

below:

where,

Yt = is the dependent variable in year t. Agricultural output was used as dependent

variables in this study. Agricultural output is measured by agricultural value added

output(AGDP) at constant prices (1999-2000) in crores of rupees

TOTt = represents real output price in year t. Agricultural terms of trade was used to measure

the relative prices. Agricultural terms of trade is ratio of prices received by farmers to

prices paid by farmers. It captures changes in producer prices, intermediate input



24

costs, real exchange rates, and world market prices. Agricultural terms of trade was

calculated using the following formula:

TOT= IPD for agricultural sector / IPD for non-agricultural sector

where,

IPD for agricultural sector = implicit price deflator for agricultural sector.

IPD for non- agricultural sector = implicit price deflator for non-agricultural sector.

IRR t = is irrigation ratio in year t. Irrigation ratio is the share of total irrigated land in the

total cropped area. It was calculated as a proxy for rural infrastructure.

Techt = indicates technology in year t. Previous studies have used either trend variable or

irrigation ratio or total factor productivity index to capture technological improvement.

Use of time trend as a technology is very old pattern. It also catches up effects of all

other variables so when we already incorporated other variables, use of time trend is not

viable. Likewise, productivity index captures the effect of improvement in technology,

efficiency, rural infrastructure and weather. Use of high yield variety of seeds per

hectare is the most appropriate variables for indicating technology but data related to

this is not available for a long period so we used consumption of fertilizers (measured

in kilogram) per hectare as a technology variable in this study.

AAR t = represents annual average rainfall in year t. To incorporate weather condition in the

model average annual rainfall (measured in millimeter) was used.

et = is error term

Cointegration and error-correction techniques are applied in this study. These techniques are

believed to overcome the problem of spurious regressions and to give consistent and distinctestimates of long-run and short-run elasticities that satisfy the properties of the classical

regression procedure. This is because all variables in an ECM are integrated of order zero, I

(0). Spurious regression and inconsistent and indistinct short-run and long-run elasticity

estimates are major problems exhibited by traditional Adaptive Expectation and Partial

Adjustment models (Hallam & Zanoli, 1993; McKay et al , 1998). Co-integration and ECMs

have been used in agricultural supply response analysis in other countries by a number of

researchers, namely Hallam & Zanoli, (1993); Townsend (1997); Schimmelpfennig et al.



25

(1996); Townsend & Thirtle (1994); Almu et al. (2003); McKay et al.(1998); Thiele (2003);

and Avasola (2006).

One major use of the co-integration technique is to establish long-run equilibrium

relationships between variables. However, two conditions must be met for co-integration to

hold. First, individual variables should be integrated of the same order. Second, the linear

combination of these variables must be integrated of an order one less than the original

variables (Engle & Granger, 1987). In other words, if the variables under consideration are

integrated of order one, or I (1), the error term from the co-integrating relationship should be

integrated of order zero, I (0), implying that any drift between variables in the short run is

temporary and that equilibrium holds in the long run.

If deviation from the long-run equilibrium path is bounded or co-integration is confirmed,

Engle & Granger (1987) show that the variables can be represented in a dynamic error-

correction framework. Therefore, in this paper, like similar studies elsewhere, supply

response is modeled in two stages. First, a static co-integrating regression given by equation

(10) is estimated for Indian agriculture and tests for co-integration are conducted. Second, if

the null for no co-integration is rejected, the lagged residuals from the co-integrating

regression are imposed as the error correction term in error correction model. An error

correction model is shown below:

where, represents first differencing, λ measures the extent of correction of errors by

adjustment in Yt. βi

measures the short-run effect on supply or short-run elasticities when the

variables are measured in logarithm, while αimeasure the long-run price elasticities. ut is

error term.

(B) Estimation Procedure

Each of the series is tested for the presence of a unit root by estimating an Augmented

Dickey Fuller (ADF) equation both with and without the deterministic trend. The number of

lags in the ADF equation is chosen by using LR, AIC, and SBIC tests. After verifying that

variables are stationary or not, we took first lag difference of all series and again estimated

ADF equation both with and without the deterministic trend. The final stage is to test for



26

cointegration. We test for cointegration by using Engel Granger two-step procedure. In this

approach, first we estimated long-run relationship if all variables are integrated in same order

and obtained residuals. The residuals of this relationship is tested for the presence of a unit

root. If the test reported presence of unit root in residuals, the variables used in long-run

relationship are not cointegrated and if the test rejected null hypothesis, the variables used in

long-run relationship are cointegrated.

(C) Selection of States

The study also focuses on whether there is difference in the supply response among highly

agricultural based, medium agricultural based, and low agricultural based states. For this

purpose, we divided all states into three categories. For classification, first we estimated the

share of agricultural income in total state’s income for each state10

for the 3 base years 1980-

81, 1990-91, and 2000-01, respectively (see Table 2). After then we categorized these states

into three categories as mentioned above. The criterion used was, those states which have

more than 30 percent agricultural share are highly agricultural based, which have less than 30

and more than 20 percent are medium agricultural based, and which have less than 20 percent

agricultural share are low agricultural based states. Through this procedure finally we have

chosen 14 states in which four states (Bihar, Hariyana, Punjab, and Uttar Pradesh) belong to

the first category, six states (Andhara Pradesh, Karnataka, Madhya Pradesh, Orissa,

Rajasthan, and West Bangal) belong to second category, and four states (Gujarat, Kerala,

Maharashtra, and Tamil Nadu) belong to the third category.

(D) Period of Study

The study covered time span from 1970-71 to 2004-05 for all India level analysis and from

1980 to 2005 for regional analysis. The year 1970-71 is chosen as base year because Indian

agriculture has made significant growth since late 60s. Further, during seventies agricultural

and rural development was given more importance. Despite these, one more reason to leave

50s and 60s is during these periods Indian economy faced many structural changes. Due to

these changes agricultural output has more variation in this period than study period (1970-

71 to 2004-05). The year 2004-05 was chosen as the end period of the study because the

published data on which this study is based is available up to 2004-05.

10 Northeast, small such as Goa etc, and new states were excluded in this procedure.



27

At the beginning period of Green Revolution, there were substantial inequalities in

agricultural production among states of India (Das and Barua, 1996). Due to these

inequalities, there might have occurred differences in supply response among states and

results may be biased. To avoid this biasedness, the year 1980-81 is chosen as base for

regional analysis.

(E) Data Source

Data related to agricultural gross value of output was collected from various issues of

National Account Statistics published by Central Statistical Organization, Government of

India, New Delhi. Gross sown area, and gross irrigated area were taken from Agricultural

Statistics at a Glance (2007) published from Directorate of Economics and Statistics,

Ministry of Agriculture, Government of India. Consumption of fertilizers was also collected

from Agricultural Statistics at a Glance (2007). Average annual rainfall was calculated from

monthly journal Agricultural Situation of India published by Directorate of Economics and

Statistics, Ministry of Agriculture, Government of India. All the state level data for each

variable are taken from various issues of Fertilizers Statistics published by Fertilizers

association in India, New Delhi, and www.indiastat.com.

Table 2: Share of Agricultural Income in total income for all major states

Name of state 1980-81 1990-91 2000-01

ANDPSNDP ANDP/SNDP*100 ANDP SNDP ANDP/SNDP*100 ANDP SNDP ANDP/SNDP*100

Andhra Pradesh 315705 732395 43.11 441689 1358012 32.52 2141924 7707692 27.79Bihar 302911 634922 47.71 396613 1025327 38.68 1347880 3136296 42.98Gujarat 243408 654742 37.18 268421 1083915 24.76 921900 6257500 14.73Haryana 163026 303195 53.77 257704 571921 45.06 936612 2888511 32.43Karnataka 239327 558736 42.83 303450 911210 33.30 1852057 6213241 29.81Kerala 129384 382273 33.85 176135 526234 33.47 544822 3396268 16.04Madhya Pradesh 306072 701244 43.65 478591 1110721 43.09 1108250 4309917 25.71Maharashtra 374908 1516258 24.73 546770 2722382 20.08 2125238 13671250 15.55

Orissa 161736 344269 46.98 148966 434470 34.29 521155 2027161 25.71Punjab 215618 444925 48.46 357434 750493 47.63 1516910 3663590 41.41Rajasthan 202801 412571 49.16 385419 847260 45.49 1111879 4566369 24.35Tamil Nadu 177319 721816 24.57 271417 1242299 21.85 1393387 8045255 17.32Uttar Pradesh 699614 1401182 49.93 957778 2277965 42.05 3317322 9168983 36.18West Bengal 264060 959400 27.52 392967 1445781 27.18 1902846 7825403 24.32

Note: ANDP = State net domestic product from agriculture, and SNDP = State net domestic product.Source: State domestic product, EPW research foundation



28

RESULTS AND DISCUSSION

(A) Descriptive Statistics

Table 3 shows the mean value of all variables (used in absolute form) for all India and each

category of states; high, medium, and low agricultural based states. From this table, the

situation of agriculture in each category of states can be compared. The mean value of

agricultural TOT is higher for medium agricultural based states in comparison both high and

low agricultural based states. The mean value of agricultural TOT is almost same for high

and low agricultural based states. The mean values of technology and irrigation ratio are the

highest for highly agricultural based states. This shows that high agricultural based states are

better in terms of infrastructure and technology than the other two groups of states. Tables 4

and 5 describe the descriptive statistics (all variables are used in natural logarithmic form in

order to be consistent with the variables used in the statistical analysis) and correlation

matrix, respectively.

Table 3: The Mean Values of Selected Variables

Group AGDP (In RsCrore)

TOT IRR TECH (in Kgper Hectare)

AAR (in mm)

All India 290805.1 95.12 32.78 53.75 1257.9Category of States

High 63813.98 96.05 65.78 100.26 876.2Medium 73381.17 103.35 29.83 65.01 1277.7

Low 44758.65 93.07 25.32 72.75 1285.3

Note: 1. All variables are in absolute measure.



29

Table 4: Descriptive Statistics

Variables AGDP TOT IRR TECH ARR

All India (Period: 1970 – 2004)Mean 12.54 4.55 3.47 3.82 7.13

Std. Dev. 0.30 0.075 0.20 0.63 .09Maximum 12.99 4.77 3.73 4.57 7.28Minimum 12.05 4.44 3.14 2.61 6.89

High Agricultural Based States (Period: 1980-2004) Mean 11.05 4.56 4.18 4.56 6.77

Std. Dev. 0.18 0.05 0.106 0.31 0.14Maximum 11.33 4.66 4.31 4.93 6.94Minimum 10.73 4.47 3.99 3.81 6.51

Medium Agricultural Based States (Period: 1980-2004) Mean 11.18 4.64 3.38 4.09 7.15

Std. Dev. 0.22 0.053 0.17 0.429 0.08

Maximum 11.48 4.76 3.67 4.65 7.31Minimum 10.77 4.55 3.05 3.23 6.97

Low Agricultural Based States (Period: 1980-2004)Mean 10.69 4.53 3.22 4.24 7.15

Std. Dev. 0.21 0.05 0.12 0.32 0.13Maximum 10.97 4.62 3.40 4.65 7.50Minimum 10.35 4.44 3.02 3.56 6.88

Table 5: Correlation Matrix

Variables AGDP TOT IRR TECH ARR

All India (Period: 1970 – 2004)AGDP 1.0000TOT -0.19 1.0000IRR 0.98 -0.23 1.0000

TECH 0.96 -0.33 0.98 1.0000ARR 0.020 -0.004 -0.09 -0.08 1.0000

High Agricultural Based States (Period: 1980-2004)

AGDP 1.0000TOT 0.38 1.0000IRR 0.95 0.46 1.0000

TECH 0.96 0.38 0.95 1.0000ARR 0.003 -0.09 -0.05 0.04 1.0000

Medium Agricultural Based States (Period: 1980-2004)

AGDP 1.0000TOT 0.31 1.0000IRR 0.96 0.27 1.0000

TECH 0.96 0.33 0.97 1.0000ARR 0.13 0.27 -0.02 0.09 1.0000

Low Agricultural Based States (Period: 1980-2004)AGDP 1.0000TOT 0.25 1.0000IRR 0.86 0.29 1.0000

TECH 0.85 0.36 0.86 1.0000ARR 0.33 -0.10 0.13 0.09 1.0000



30

(B) Test of Stationarity

Before doing estimation, we test for the presence of stationarity of all the variables in the

model for both India and regional level. We used Augmented Dickey Fuller test both with

trend and without trend. Table (6) shows the results of ADF test, we found all variables are

non-stationary for each category; all India level, highly agricultural based states, medium

agricultural based states, and low agricultural based states.

(C) Order of Integration

The test for the order of integration is the next step in co-integration analysis. If a series is

integrated, it accumulates past effects. This means that perturbation to the series does notreturn to any particular mean value. Therefore, an integrated series is non-stationary. Order

of integration of such a series is determined by the number of times that it must be

differenced before it is actually made stationary. It follows that if two or more series are

integrated of the same order then a linear relationship can be estimated. Examining the order

of integration of this linear relationship is similar to testing for the null hypothesis that there

is no co-integration against its alternative that there is co-integration. In this section, an

attempt is made to determine the order of integration of the variables. Table (7) showed that

all variables are integrated of order 1 or I (1). The first lag difference of each of the series is

tested for the presence of a unit root by estimating an Augmented Dickey Fuller (ADF)

equation both with and without the deterministic trend. The results are given in tables from 5

to 8. All the tables show that first lag difference of each series is stationary. This means that

all the series are integrated of order one for all India, and also at the regional level, high,

medium, and low agricultural based states.



31

Table 6: Results of ADF Test

Variables With Trend Without Trend

All India LevelAGDP -3.427 (-4.325) -0.565 (-3.709)TOT -2.108 (-4.306) -2.138 (-3.696)IRR -0.364 (-4.316) -2.077 (-3.702)TECH -1.312 (-4.306) -1.306 (-3.696)AAR

-4.253 (-4.306) -4.331 (-3.696)

High Agricultural Based StatesAGDP -3.427 (-4.325) -0.565 (-3.709)TOT -2.108 (-4.306) -2.138 (-3.696)IRR -0.364 (-4.316) -2.077 (-3.702)TECH -1.312 (-4.306) -1.306 (-3.696)AAR

-4.253 (-4.306) -4.331 (-3.696)

Medium Agricultural Based States

AGDP -2.020 (-4.380) -1.366 (-3.750)TOT -1.791 (-4.380) -1.786 (-3.750)IRR -3.839 (-4.380) -1.469 (-3.750)TECH -1.825 (-4.380) -2.226 (-3.750)AAR

-4.268 (-4.380) -4.250 (-3.750)

Low Agricultural Based StatesAGDP -1.817 (-4.380) -0.979 (-3.750)TOT -2.368 (-4.380) -2.611 (-3.750)IRR -1.697 (-4.380) -0.997 (-3.750)TECH -2.190 (-4.380) -1.877 (-3.750)AAR

-3.019 (-4.380) -3.063 (-3.750)

Note: The value given in parentheses indicates t-statistics



32

Table 7: Results of ADF Test

Variables With Trend Without Trend

All India LevelAGDP -4.446 (-4.325) -4.528 (-3.709)TOT -5.285 (-4.306) -5.378 (-3.696)IRR -5.634 (-4.316) -4.983 (-3.702)TECH -4.135 (-3.572) -3.780 (-3.702)AAR*

-7.273 (-4.316) -7.295 (-3.702)

High Agricultural Based StatesAGDP -5.029 (-4.380) -4.822 (-3.750)TOT -4.538 (-4.380) -4.722 (-3.750)IRR -5.071 (-4.380) -4.683 (-3.750)TECH -4.870 (-4.380) - 3.856 (-3.750)AAR

-4.721 (-4.380) -4.799 (-3.750)

Medium Agricultural Based States

AGDP -4.112 (-3.600) -4.025 (-3.750)TOT -4.931 (-4.380) -5.068 (-3.750)IRR -4.295 (-3.600) -4.239 (-3.750)TECH -5.885 (-4.380) -5.385 (-3.750)AAR

-5.210 (-4.380) -5.102 (-3.750)

Low Agricultural Based StatesAGDP -3.822 (-3.600) -3.825 (-3.750)TOT -5.108 (-4.380) -4.881 (-3.750)IRR -4.512 (-4.380) -4.573 (-3.750)TECH -6.090 (-4.380) -6.004 (-3.750)AAR

-3.849 (-3.600) -3.963 (-3.750)

Note: The value given in parentheses indicates t-statistics

(D)Test of Cointegration

The main objective of this section is to determine whether the variables are integrated of

order zero, or in short, whether they are co-integrated. If co-integration is confirmed, a non-

spurious long-run equilibrium relationship exists. When this is combined with ECM, whose

variables are I (0), consistent estimates of both long-run and short-run elasticities is evident.

Residual-based, proposed by Engle & Granger (1987) was employed to test for co-

integration. The residual-based procedure is known as a single-equation approach. We have

seen in Table (7) all selected variables are integrated of same order, I (1). So, we can

estimate long run relationship among variables represented by equation (10) and if residual

of estimated equation becomes stationary then we can say that equation (10) or long run

relationship among variables is cointegrated.

In this study first we examine the long-run relationship or cointegration separately between

pairs of variables (further it is indicated by bivariate analysis in this study); for example

agricultural output with agricultural terms of trade, agricultural output with irrigation ratio,



33

agricultural output with technology, and agricultural output with annual rainfall. The

bivariate analysis has also been done for each category. The results are given in tables from

8 to 11. Table 8 shows that agricultural output is cointegrated with only irrigation ratio for

all India level. This means that there is a long-run relation between agricultural output and

irrigation. Tables 9, 10, and 11 indicate that agricultural output is separately cointegrated

with irrigation ratio and technology for high, medium, and low agricultural based states

respectively. This reveals that the long-run relationship between agricultural output and

irrigation ratio and between agricultural output and technology exist in each category of

states.

Table 8: Bivariate Analysis for All India

Relationbetweenvariables

Estimated Equation R-

dw-stat ADF testof residuals

Interpretation

AGDP&TOT

Agdp = 15.99 - .76 tot(5.14) (-1.11)

0.01 0.049 -0285(-2.649)

No Cointegration

AGDP&IRR

Agdp = 7.15 + 1.55 irr*(34.46) (26.01)

0.95 1.775 -5.691(-2.646)

Cointegration

AGDP

&TECH

Agdp = 10.79 + .46 tech*

(114.29) (18.76)

0.91 0.656 -0.844

(-2.649)

No Cointegration

AGDP&AAR

Agdp = 11.85 + .07 arr -0.02 0.0466 -0.592(-2.650)

No Cointegration

Table 9: Bivariate Analysis for High Agricultural Based States


Estimated Equation R-2


Interpretation

AGDP

&TOT

agdp = 4.71+1.39tot***

(1.45) (1.95)

0.10 0.269 -1.516

(-2.660)

No Cointegration

AGDP&IRR

agdp = 4.31 + 1.61irr*(9.39) (14.67)

0.89 2.41 -5.815(-2.660)

Cointegration

AGDP&TECH

agdp = 8.56 +.54 tech*(56.91) (16.58)

0.91 1.89 -4.564(-2.660)

Cointegration

AGDP&AAR

agdp = 11.49 - .04 aar (3.53) (-0.14)

0.04 0.116 -1.208(-2.660)

No Cointegration



34

Table 10: Bivariate Analysis for Medium Agricultural Based States

Relation

betweenvariables

Estimated Equation R-

dw-stat ADF test

of residuals

Interpretation

AGDP&TOT

agdp = 5.14 + 1.30 tot(1.32) (1.55)

0.05 0.236 -1.346(-2.660)

No Cointegration

AGDP&IRR

agdp = 7.01 + 1.23 irr*(27.78) (16.52)

0.91 2.145 -5.179(-2.660)

Cointegration

AGDP&TECH

agdp = 9.14 + 0.49 tech*(71.80) (16.08)

0.91 1.479 -3.667(-2.660)

Cointegration

AGDP&

AAR

agdp = 10.61 + .05 aar (1.46) (0.08)

-.0.04 0.125 -0.760(-2.660)

No Cointegration

Table 11: Bivariate Analysis for Low Agricultural Based States


Estimated Equation R-2


Interpretation

AGDP&TOT

agdp = 5.68 + 1.10 tot(1.43) (1.26)

0.02 0.575 -0.451(-2.660)

No Cointegration

AGDP&

IRR

agdp = 5.92 + 148 irr*(9.94) (8.01)

0.72 1.325 -2.772(-2.660)

Cointegration

AGDP&TECH

agdp = 8.26 + 0.57 tech*(26.19) (7.72)

0.70 1.17 -2.700(-2.660)

Cointegration

AGDP&AAR

agdp = 4.65 + 0.53 aar (1.36) (1.76)

0.08 0.185 -1.171(-2.660)

No Cointegration

Note: 1. The value given in parentheses indicates t-statistics.2. Superscript *, **, and *** show level of significant at 1 %, 5 %, and 10 % respectively

After examining the bivariate relationship, we moved to examine multivariate relationship

showed by equation (10). In this equation we used irrigation ratio as a proxy for infrastructure and consumption of fertilizer as a proxy of technology. But both are highly

correlated with each other. This casts doubt of multicollinearty in equation (10). So we

estimated equation (10) without incorporating consumption of fertilizer and again estimated

equation (10) without incorporating irrigation ratio for removing the problem of

multicollinearty. Thus we estimated three equations for each category. The tables from 12

to 15 show results. They indicated that model 1(original equation 10), model 2 (without



35

incorporating technology variable in equation 10), and model 3 (without incorporating

irrigation ratio in equation 10) are cointegrated for each category.

Table 12: Multivariate Analysis for All India

Variables Model 1 Model 2 Model 3

Constant 2.13 (1.49) 2.22 (1.65) 4.59 (2.55)TOT 0.14 (0.84) 0.16 (1.19) 0.57** (3.10)IRR 1.65* (5.02) 1.58* (30.10)TECH -.02 (-.21) 0.48* (21.89)AAR 0.39* (1.49) 0.39* (3.61) 0.34** (2.27)R

-0.96 0.96 0.93

dw-stat 1.245 1.232 0.87ADF test for

residuals-2.953(-2.647)

-2.895(-2.647)

-2.164(-2.647)

Interpretation Cointegration Cointegration No Cointegration

Table 13: Multivariate Analysis for High Agricultural Based States


Constant 7.33 (4.35) 4.66 (2.86) 9.01 (5.91)TOT -0.12 (-0.50) -0.27 (-0.99) 0.04 (0.15)IRR 0.69*** (1.92) 1.67* (13.38)TECH 0.33** (2.87) 0.54* (14.75)AAR -.01 (-0.14) .06 (0.61) -0.06 (-0.65)

R- 0.92 0.89 0.91dw-stat 2.34 2.43 1.91ADF test for residuals

-5.809 (-2.660) -5.883 (-2.660) -4.627 (-2.660)

Interpretation Cointegration Cointegration Cointegration

Table 14: Multivariate Analysis for Medium Agricultural Based States


Constant 3.55 (1.52) 1.86 (0.98) 8.16 (3.67)TOT -0.07 (-0.30) -0.04 (-0.15) -0.11 (-0.36)

IRR 0.92* (3.22) 1.26* (17.67)TECH 0.14 (1.21) 0.50* (14.64)AAR 0.35** (2.04) 0.43** (2.65) 0.12 (0.63)R

-0.93 0.93 0.90

dw-stat 2.04 2.19 1.41ADF test for residuals

-4.915(-2.660)

-5.295(-2.660)

-3.528(-2.660)




36

Table 15: Multivariate Analysis for Low Agricultural Based States


Constant 2.09 (0.79) 0.12 (0.05) 3.75 (1.27)TOT 0.04 (0.10) 0.26 (0.59) -0.001 (-0.00)IRR 0.83** (2.80) 1.40* (8.12)TECH 0.27** (2.26) 0.55* (7.58)AAR 0.41** (2.93) 0.43** (2.86) 0.41** (2.55)R

-0.81 0.78 0.75

dw-stat 2.149 1.90 1.54ADF test for residuals

-5.185(-2.660)

-4.686(-2.660)

-3.879(-2.660)


Note: 1. AGDP is dependent variable in each Model2. The value given in parentheses indicates t-statistics.3. Superscript *, **, and *** show level of significant at 1 %, 5 %, and 10 % respectively

From the bivariate analysis, we have seen that agricultural output is not cointegrated with

agricultural TOT. This means no long run equilibrium relationship could exist between

agricultural output and agricultural TOT. And from multivariate analysis we have seen that a

long run equilibrium relationship showed by equation 10 exists. This shows that the impact

of agricultural TOT on agricultural output works in combination with non-price factors

(technology, irrigation and annual rainfall).

(E) Error correction model

After long-run relationships between agricultural output /gross sown area and the variables

predicting it are confirmed, ECM is developed. Results are reported in Tables from 16 to 19.

According to Hallam & Zanoli (1993), a high R 2

in the long-run regression equation is

necessary to minimize the effect of a small sample bias on the parameter estimates of the co-

integrating regression, which may otherwise be carried over to the estimates of the error-

correction model. From Table 16 (All India level), agricultural output is impacted positively

by technology, irrigation ratio and annual rainfall. Agricultural terms of trade gives a

negative relation but the coefficient is not statistically different from ‘zero’ at 5% level. All

variables except agricultural TOT have expected sign. Only coefficient of annual rainfall is

statistically significant at one percent level and other coefficients are not statistically

significant at 5 percent level. All selected variables together explained 69 percent of variation

in agricultural output. The error-correction term in this equation is also significant and has

the required sign. It shows that 0.48 of the deviation of the agricultural output from its long



37

run equilibrium level is corrected each year. Similarly, in test of cointegration in multivariate

case, here also we estimated ECM without incorporating technology (model 2) and irrigation

ratio (model 3). Model 2 also shows that agricultural supply elasticity is not statistically

different from ‘0’. This model also indicated that remaining factors (irrigation ratio and

annual rainfall) have positive effect on agricultural output. The error-correction term in the

model is significant and has the required sign. It shows that 48 percent of the deviation of the

agricultural output from its long run equilibrium level is corrected each year.

From Table 17 (High Agricultural Based States), it is inferred that only ‘technology’ is

significantly explaining output. Here also, the supply elasticity is insignificant. The negative

sign of coefficients of irrigation may be due to multicolinearity between technology and

irrigation ratio. This argument is strongly supported by model 2 where we found positive and

statistically significant coefficient of irrigation ratio. The models 1, 2 and 3, explained 42, 32,

and 35 percent variation in agricultural output, respectively. The error correction term in each

model is statistically significant and has required sign.

The results of the ECM for medium and low agricultural based states are given in Table 18

and 19, respectively. The results from these two groups also indicated that irrigation ratio

and annual rainfall are significant variables affecting positively. Supply elasticity is

significant and negative. The error correction term in each model for both categories is

statistically significant and has the required sign.

Table 16: Results of ECM for All India

Variables Model 1 Model 2

Constant 0.02 (2.37) 0.02 (2.86)TOT -0.18 (-1.25) -0.18 (-1.31)IRR 0.29 (0.08) 0.05 (0.15)TECH 0.03 (0.35)AAR 0.32* (5.52) 0.33* (5.81)Error correction -0.48 (-3.68) -0.48* (-3.74)R

-0.69 0.70

dw-stat 2.600 2.582



38

Table 17: Results of ECM for High Agricultural Based States


Constant 0.01 (1.39) 0.02 (1.97) 0.01 (1.08)TOT -0.33 (-1.28) -0.14 (-0.50) -0.46 (-1.71)IRR -0.08 (-0.26) 0.21 (0.61)TECH 0.21***(1.83) 0.25** (2.11)AAR -0.02 (-0.33) 0.02 (0.46) -0.02 (-0.38)Error correction -0.80* (-3.56) -0.67** (-2.97) -0.70* (3.43)R

-0.42 0.32 0.35

dw-stat 2.059 2.055 2.363

Table 18: Results of ECM for Medium Agricultural Based States


Constant 0.003 (0.20) 0.01 (0.550) 0.02 (1.11)TOT -0.59** (-2.27) -0.64** (-2.65) -0.37 (-1.35)IRR 0.81*** (2.15) 0.90** (2.48)TECH 0.10 (0.90) 0.22 (1.56)AAR 0.42* (4.10) 0.49* (5.13) 0.35** (2.73)Error correction -1.05* (-4.58) -1.10* (-5.04) -0.62* (-3.11)R

-0.60 0.64 0.44

dw-stat 1.68 1.749 2.036

Table 19: Results of ECM for Low Agricultural Based States


Constant 0.02 (1.51) 0.02 (1.22) 0.02 (1.73)TOT -0.54 (-1.84) -0.41 (-1.30) -0.49*** (-1.90)IRR 0.08 (0.28) 0.16 (0.51)TECH -0.06 (-0.36) -0.10 (-0.75)AAR 0.44* (3.96) 0.38* (4.03) 0.44* (4.70)Error correction - 0.55** (-2.28) -0.42*** (-2.06) -.42** (-2.84)R

-0.61 0.58 0.67

dw-stat 1.982 2.005 2.186

Note: 1. AGDP is dependent variable in each Model2. The value given in parentheses indicates t-statistics.

3. Superscript *, **, and *** show level of significant at 1 %, 5 %, and 10 % respectively.

(F) Short-Run and Long-Run Elasticities

The coefficients of ECM model indicate short-run elasticites and the coefficients of estimated

equation (10) represent long-run elasticities. The short-run and long-run agricultural output

elasticities with respect to terms of trade are shown in Table 20. Both short-run and long-run

supply elasticities with respect to TOT are not statistically significantly different from zero.

However, the supply elasticity with respect to TOT is positive and statistically significant in



39

the study of Palanivel (1996). The results also indicate that elasticities of technology,

irrigation ratio, and annual rainfall are positive for both short-and long-run period. Turning to

regional level analysis, the short-run and long-run supply elasticities are insignificant for low

and high agriculture based states, whereas it is significant and negative for medium

agricultural based states. The non-price factors’ elasticities for these three groups of states

are also positive and relatively more than agricultural TOT.

Table 20: Aggregate Agricultural Elasticities w.r.t. Agricultural TOT

Level Short Run Long Run

All India -0.18 0.14 High

-0.34 -0.12 Medium -0.59** -0.02 Low -0.54 -0.04

Note: Superscript** indicates 5 % significant level

From this analysis, it is inferred that price is not a significant factor explaining output

growth. The analysis confirms that both irrigation and technology are relatively more

important for higher agricultural growth.



40

THE CAUSES OF LOW SUPPLY RESPONSE

The issue of agricultural supply response is a very important one as it has an impact on

growth, poverty, and environment (Mamingi, 1997). The success of any structural adjustment

programs depends on size of agricultural supply response. The size of agricultural supply

response should improve after removing some of the constraints that farmers were facing

before. India also has removed many constraints from agrarian system and provided many

incentives to farmers. Land reform, massive public investments in irrigation and rural

electrification, strengthening of agricultural research and extension services, expansion and

improvement of the transport network, especially rural roads, development of rural creditinstitutions are some steps which were initiated by Indian Government for improving

agriculture since Independence. But still the supply response for Indian agriculture is low.

Hence the question is “why is supply response low” becomes a big issue. In this section, we

summarize some of the discussion from the earlier literature to support low or no response.

Low supply response means that farmers’ aim is not to maximize profit. One reason

attributable to it is that farmers in India are largely subsistence farmers. This is supported by

a fact that 78 percent farmers are small and marginal farmers in India. So, finally we can say

that subsistence farming is one of the causes of low supply response.

Despite this there are many constraints, which caused low supply response in Indian

agriculture. These are lack of an integrated farming approach, weak agriculture research and

extension network, restrictions on agricultural trade and processing, inadequate public

expenditure on rural infrastructure. Due to these constraints Indian farmers do not like to stay

in farming. The 59th

round NSSO report reported that 40 percent farmers do not like to stay

in farming but they have no other alternatives.



41

CONCLUSION AND SUGGESTIONS

The study reexamined agricultural supply response for Indian agriculture. The study also

verified that whether there is difference in the supply response among highly agricultural

based, medium agricultural based, and low agricultural based states. In this study, we found

that all selected variables have unit root and integrated of order one. The bivariate analysis

confirmed that no cointegration between agricultural TOT and agricultural output. The non-

cointegratedness indicates that no direct long-run relationship exists between TOT and output

level in Indian agriculture. The multivariate analysis confirmed that relationship between

agricultural output and all other selected variable is cointegrated. This shows that the impactof agricultural TOT on agricultural output works in combination with non-price factors

(technology, irrigation and annual rainfall). The study also indicated that aggregate

agricultural output elasticity with respect to agricultural TOT is very low and not statistically

different from 0. This means supply response is price inelastic. There are several reasons for

this. One, small and marginal farmers constitute higher proportion; two, lack of an integrated

farming approach; three, weak agriculture research and extension network; four, restrictions

on agricultural trade and processing; and five, inadequate public expenditure on rural

infrastructure. This short-run supply response is lower than the long-run response and the

reason could be that the use of primary factors that account for about 70 to 85 percent of the

cost of agricultural production in developing countries cannot be changed in the short run

(Binswanger, 1993; Thiele, 2000). The study also confirmed that there is a difference in the

supply response among high agricultural based, medium agricultural based, and low

agricultural based states. The results of this study also suggested that non-price factors are

comparatively more important for higher agricultural growth.

The study confirmed that price and non-price factors are not substitute for each other. They

are complementary. So, a strategy that combines both price and non-price factors is crucial

for policies promoting agricultural development. Besides, farmers’ accessibility to output and

input markets, strengthening of agricultural research and extension removing restrictions on

agricultural trade within country, and improving rural infrastructure especially irrigation,

transport, education, and health facilities, are some of the factors that would lead to increase

in response.



42

Certain limitations of this study should be borne in mind. This study relies on a partial

equilibrium analysis for the agricultural sector. In the long run, the dynamics of agricultural

supply are likely to depend to a large extent on the ability of the sector to attract additional

production factors from other sectors, an effect that cannot be captured in a partial

equilibrium framework. Despite this, another limitation is that the study covered a short time

period especially for state level analysis. To account for the effects of intersectoral factor

movements on agricultural supply, an analysis based on dynamic general equilibrium

framework would give more meaningful results.



43

REFERENCES

Alemu, Z.G., Oosthuizen, K. & Schalkwyk, H. D. V. (2003). “Grain-Supply Response In

Ethiopia: An Error-Correction Approach”, Agrekon, Vol 42, (4): 389-404.

Bapna, S. L., Binswanger, H. P. & Quizon, J. B. (1981). “System of Output Supply and

Factor Demand Equations for the Semi-Arid Tropical India”, Indian Journal of

Agricultural Economics, Vol. 39(2): 179-213.

Binswanger, H.P. (1993). “Determinants of Agricultural Supply and Adjustment Policies. In:

F. Heidhues and B. Knerr (ed.)”, Food and Agricultural Policies under Structural Adjustment . Proceedings of the 29th Seminar of the European Association of

Agricultural Economists. Frankfurt a.M.

Chibber, A. (1988). “Raising Agricultural Output: Price and Non-Price Factors”, Finance

and Development , Vol. 25 (2): 44-47.Das, S. K. & Barua A. (1996). “ Regional Inequalities, economic growth and Libralisation: A

Study of the Inidan Economy”, Journal of Development Studies, Vol 32(3): 360-390

Engle, R. F. & Granger, C.W.J. (1987). “Cointegration and Error Correction: Representation,

Estimation and Testing”, Econometrica Vol. 55: 251-275.

Granger, C. & Newbold, P. (1974). “Spurious Regressions in Economics”, Journal of Econometrics 2(1): 227-238.

Gulati, Ashok (1998). “Indian Agriculture in An Open Economy: Will be Prosper”, In I.J.

Ahluwallia and I.M.D. Little (eds.) Indian Economic Reforms and Development Essays, Oxford University Press.

Hallam, D. & Zanoli, R. (1993). “Error Correction Models and Agricultural Supply

Response”, European Review of Agricultural Economics Vol. 2: 151-166.

Hossein, A. & Cummings, J. T. (1977). “Estimating Agricultural Supply Response with the

Nerlove Model: A Survey”, International Economic Review, Vol. 18 (2): 257-292.

Krishana J. & Rao, M.S. (1965). “Dynamic Acreage Allocation for Wheat in Uttar Pradesh”,

Indian Journal of Agricultural Economics, Vol. 22 (1).

Krishana, J. & Rao, M.S. (1967). “Price Expectation and Acreage Response for Wheat in

Uttar Pradesh”, Indian Journal of Agricultural Economics, Vol. 20 (1): 20-25.

Krishna, R. (1963). “Farm Supply Response in India-Pakistan: A Case Study of the Punjab

Region”, The Economic Journal , Vol. 73 (291): 477-487

Krishna, R. (1982). “Growth, Price Policy and Equity”, Food Research Institute Studies, 18

(3).

Krueger, A.O., Schiff, M. & Valdés, A. (1992). The Political Economy of Agricultural Pricing Policies: A World Bank Comparative Study. Baltimore.

Madhavan, M. (1972). “Acreage response of Indian farmers: A case study of Tamil Nadu”,

Indian Journal of Agricultural Economics, Vol. 27(1): 67–86.



44

Maeshiro, A. (1996). “Teaching Regression with a Lagged Dependent Variable and

Autocorrelated Disturbances”, The Journal of Economic Education, Vol. 27 (1): 72-84.

Mamingi, N. (1997). “The Impact of Prices and Macroeconomic Policies on AgriculturalSupply: A Synthesis of Available Results”, Agricultural Economics Vol.16: 17–34.

McKay, A., Morissey, O. & Vaillant, C. (1999). “Aggregate Supply Response in Tanzanian

Agriculture”, The Journal of International Trade and Economic Development Vol. 8:

107–124.

Mishra, V. N & Hazell, P.B.R (1996). “Terms of Trade, Rural Poverty, Technology and

Investment: The Indian Experience’, 1952/53 to 1990/91, Economic and Political Weekly, 31 (13): A2 – A13.

Mishra, V.N. (1998). “Economic Reforms, Terms of Trade, Supply and Private Investment inAgriculture: Indian Experience”, Economic and Political Weekly, Vol.33 (31).

Mythili, G. (2008), Acreage and Yield Response for Major Crops in the Pre-and Post-Reform Periods in India: A Dynamic Panel Data Approach, PP Series 061, Indira Gandhi

Institute of Development of Research, Mumbai.

Narayana, N.S.S. & Parikh, K. S. (1981). “Estimation of Farm Supply Response and Acreage

Allocation: A Case Study of Indian Agriculture”, Research Report – 81-1, InternationalInstitute for Applied systems Analysis.

Nerlove, M. (1958). The Dynamic of Supply: Estimation of Farmers’ Response to Prices,The Johns Hopkins University press, London.

OO Olubode-Awosola, Oyewumi, O.A. & Jooste, A. (2006). “Vector Error Correction

Modelling of Nigerian Agricultural Supply Response”, Agrekon, Vol 45(4): 421-436.Pal S. P. (1975). “Supply Response and Optimal Agricultural Policy in a Planned Under –

developed Economy”, Indian Journal of Agricultural Economics, Vol. 30 (2): 61-72.

Palanivel, T (1995). “Aggregate Agricultural Supply Response in Indian Agriculture Some

Empirical Evidence and Policy Implementation”, Indian Economic Review, Vol.30 (2):

251-263.

Parikh, A. (1967). “Farm Supply Response: A Distributive Lag Analysis”, Oxford Institute of

Statistics Bulletin, Vol. 33: 57-72.

Schimmelpfennig, D., Thirtle, C. & Van Zyl, I. (1996). “Crop level supply response in South

African Agriculture: An error-correction approach”, Agrekon Vol. 35(3): 111-122.

Singh, M. (1995). “Inauguration Address”, Indian Journal of Agricultural Economics, Vol.50

(1).

Singh, R. D., Singh, D. & Rao, P. R. (1994). “Estimation of agricultural acreage response

relationships: Methodological Issue”, Indian Journal of Agricultural Economics, Vol.29 (1).

Thiele, R. (2000). Estimating the Aggregate Agricultural Supply Response: A Survey of Techniques and Results for Developing Countries, Kiel Institute for World Economics,

Working Paper 1016. Kiel Germany.



45

Thiele, R. (2002). Price Incentives, on-Price Factors, and Agricultural Production in Sub-Saharan Africa: A Cointegration Analysis, Kiel Institute for World Economics,Working Paper 1123. Kiel Germany.

Townsend, R. & Thirtle, C. (1994). Dynamic acreage response: An error-correction model for maize and tobacco in Zimbabwe, Occasional Paper No. 7, International Association

of Agricultural Economics.

Townsend, R. (1997), Policy distortions and agricultural performance in the South Africaneconomy, Development Bank of Southern Africa Development Information Business

Unit. Discussion Paper No 138.



46

Point-by-Point Reply to Comments of Reviewers

Reviewers' comments Reply/action taken

Reviewer #1:

1. Relevance:1.1 The subject of research is relevant to

India. Policy makers need to know whether

the policies they frame are having the

desired impact. Though previous

researchers have gone into the issue of how

agriculture has responded to policyinitiatives, the use of Cointegration

analysis is an addition to knowledge.

Cointegration analysis brings out the longterm relationships and takes care of the

problem of spurious regression in timeseries data. As such the study has

implications for policy formulations.

2. General

2.1 There are a number of typos. The

author should be asked to eliminate these.

Necessary change has been made in the revisedms.

2.2 The paper does not have a separate

section on the limitations of the study. This

is important as findings would have to beseen within these constraints. Future

researchers using better analytical toolsdeveloped in future will be able to

appreciate the work

As suggested, the limitation of study has

been mention in separate section.

2.3. The paper does not have a section on

indications for further research work. Thisis necessary as no researcher is able to

address all the issues in one work. Such anindication can help further knowledge in

this area.

As suggested, scope of further study has

been mention in the end of the revised ms.

3. Abstract3.1.1. It is stated that the "present

study…. is a better approach" . It has not been elaborated as to better than which

approach/s and why.

The change has been made accordingly.

3.1.2. "But still the supply response for

Indian agriculture is low." This statement

assumes that there is standard for supplyresponse and so the actual is lower thanthat. May be it would be better to take the

position that increasing the supply response

The sentence has been corrected

accordingly to reviewer.



47

is desirable. Or, the author should statewhat it is lower than.

3.1.3. "..to see if the response has been

better at the all India level". Perhaps the

same could be seen at the state levels, as

the work does study state level data.

Yes, same could be seen at the state levels.

But the study emphasis that whether there

is difference in the supply response among

highly agricultural based, medium

agricultural based, and low agricultural

based states.

3.1.4. Only one of the results of the

study has been given in the abstract. Is it

the most important according to the author?

Ideally all the major findings should be

mentioned in the abstract. It is not clear whether this is because of a word

constraint.

All major findings have been mentioned in

the revised ms.

3.1.5 "the study indicated that

aggregate agricultural output elasticity with

respect to agricultural TOT is very low andnot statistically different from 0". Some

explanation why this might be so wouldadd to the research finding.

Some explanation have been added in

revised ms.

4. Para (I) Introduction:

4.1. The objective of the paper is not very

clearly stated. "The present study is anattempt to find supply response through

better approach….. and to see if the

response has been better at the all India

level". Is it another attempt to prove what is

the starting point ie., the supply response is

low? And, if the supply response is low at

the state level, would the all India level

supply response be any different?Objectives of the study should be clearlystated; and the observations should be

classified accordingly to map with the

objectives.

It has now been clearly explained in the revised

ms.

4.2. " This casts doubts on the validity of their results." Casting aspersions about the

validity of previous studies without a

complete discussion and elaboration is out

of place in academic writings. It is possible

the authors of these papers would have

indicated the limitations

The sentence has been correctedaccordingly to reviewer.



48

4.3. The author found one study related toIndia using the ECM but has not taken the

trouble to discuss this.

This study has been discussed in literaturereview so we have not taken the trouble to

discuss this here. This is also because of

reason that word limit constraint.

5. Para (II) Literature review

5.1. The author has expressed his"surprise" that Jai Krishna and Rao (1965,

1967) found traditional regression models

give satisfactory results "if not superior

results …….". Such criticisms should be

justified as academic papers should avoid

casting aspersions at previously publishedwork. The author should also remember

that as the methods of analysis improve,

the researches work also improves.

The sentence has been corrected

accordingly to reviewer.

5.2. Palanivel (1995) found that theelasticity of aggregate agricultural output

with respect to TOT is positive and

statistically significant. Since the author'swork has revealed the opposite, a detailed

discussion of this paper in the section of

discussions of results is necessary

This study have been discussed in therevised ms.

5.3. The literature review, inter alia, is a

discussion to indicate the gap in the

existing literature. The concluding lines

seem to indicate that the Cointegration

analysis has been adopted only because

"most of the other studies have followed

the same methodology given by

Nerlov….." !

The necessary correction has been done in

revised ms.

5.4. The study also seems only to "probe"

earlier results instead of trying to fill gaps

in existing literature

This study is not probing earlier result

while trying to fill gaps in existing

literature. To reflect it in ms the necessary

correction has been done in revised ms.

5.5. The main reason for the present study

is that Cointegration analysis is a better

tool than the methodology used in earlier

work; especially the problem of spurious

regression. But, there is no discussion on

the methodology of previous work.

In the section of theoretical framework and

approaches used, we discussed about the

methodology of previous work.

6. Methodology

6.1. The study of the all India level is based on data for 1970-71 to 2004-05. It is

not clear why regional level study is

restricted to 1980-2005.

This has been cleared in the revised ms.



7. Results and discussions:

7.1. The findings show that agricultural

output at the All India level is cointegrated

with only irrigation ratio. But at the

regional level, there is a long runrelationship between agricultural output

and irrigation ratio and between

agricultural output and technology. Since

the All India level is an aggregation of the

regional level data, this should be discussed

in detail.

This has been incorporated in the revised msfollowing the suggestion of the reviewer.

8. Conclusions:

8.1. The conclusion of the study is that

the "aggregate agricultural output elasticitywith respect to agricultural TOT is very

low and not statistically different from 0".

There is no discussion of why this is so.

The author should give his views on this.

Some explanations have been added in the

revised ms.

Recommendations:

The paper can be accepted subject the

author addressing the points raised.

Documents

Estimation of Agricultural Supply Response by Cointegration Approach