Description
Projec ton nabard
THE PRODUCTIVITY OF
AGRICULTURAL CREDIT
ASSESSING THE RECENT ROLE
OF INSTITUTIONAL CREDIT TO AGRICULTURE IN INDIA
USING STATE LEVEL DATA
Final Report submitted to NABARD
August 24, 2014
Sudha Narayanan1
Indira Gandhi Institute of Development Research
Goregaon (East), Mumbai 400065
Assistant Professor, IGIDR , [email protected]
1
1
EXECUTIVE SUMMARY
Amongst recent policy interventions implemented to revive the languishing agricultural
sector in India, those pertaining to agricultural credit have been very much in the forefront. In
particular, three major policy initiatives have shaped the past decade in institutional credit to
agriculture. The policy of doubling of institutional credit to agriculture between 2004-05 and 200607 (over the 2004-05 base year) marked the first attempt to alleviate the financial constraints of
farmers. In 2008-09, the Agricultural Debt Waiver and Debt Relief Scheme (ADWDRS) was
introduced to waive specific outstanding debts for a large number of small farmers; this was
followed by the interest subvention scheme, that sought to remedy the perceived negative impact
of the waiver on loan repayment culture by rewarding timely repayment with loans carrying lower
interest rates. These three schemes combined have, implicitly and explicitly, resulted in an
increasing volume of institutional credit to agriculture. Whereas credit accounted for only 16% of
the total value of paid out inputs in the triennium ending (TE) 1998-99, and 26.3% in TE 2003-04,
by the end of the decade, in TE 2011-12, it had risen to as high as 80.3% of the estimated total paid
out costs of inputs.
Despite its importance, little is known about the effectiveness of credit in supporting
agricultural growth as represented by the GDP and indeed the very nature of the relationship
between formal agricultural credit and agricultural GDP. This research project is a modest effort in
this direction. While acknowledging that questions of the impact of credit on agricultural output or
value addition or productivity are best addressed through a textured understanding of household
behavior and microstudies, this study is based on the premise that aggregate secondary data too
can reveal some of these important relationships. This study uses state-level data to examine the
relationship between institutional credit to the agricultural sector and agricultural GDP at the
national level. Despite serious limitations of aggregation, which typically disregards distributional
issues and often masks more than it reveals, it is also true that any systematic or pervasive
relationship should reflect in aggregate data and offers a level of generalization not available in
small scale surveys. The goal of this study is four-fold: How productive is institutional credit to the
agricultural sector? What has been the trend since mid-1990s? What are the pathways through
which credit impacts agriculture? How, if at all, have these pathways changed over the years? The
analysis covers the period 1995-96 to 2011-12 using data that includes all major states within
India. The study also conducts, data permitting, a disaggregate analysis of two sub-periods – the
first phase denoting the Pre-doubling period (1995-96 to 2003-04) and the second representing the
Post-doubling period (2004-05 to 2011-12). Where feasible, the study replicates the analysis at the
state level.
In this study, credit is conceptualized as an enabling input that influences agricultural GDP
primarily via use of variable inputs and through investments in fixed capital that support
agricultural production. To the extent that credit can also contribute to consumption smoothing of
borrowers or better their capacity for risk bearing, credit could have a non-specific influence on
agricultural GDP via variables that are typically unobserved by the researchers. To parse this
complex relationship given the limitations of data, a combination of three approaches are used. The
first is a simple model that regresses agricultural GDP on current credit flow using state level data.
The second method estimates a hybrid profit-production function that regresses agricultural GDP
2
on a vector of relevant inputs, prices and agricultural credit flow during that year. This is a direct
approach to estimating the relationship between credit and agricultural GDP in reduced form. The
possible endogeneity of credit is addressed by the use of a control function that “controls” for the
estimated endogenous component of observed credit flow. The third method represents the
`pathways approach’ which estimates input demand as a function of credit, among other things and
controlling for endogeneity. The coefficients representing the responsiveness of input use to
institutional credit are then used as components to construct the total impact of agricultural credit
on agricultural GDP. The impact of credit on agricultural output is thus derived as the sum of the
contribution of credit to the use of specific inputs, capital or the cropping pattern, weighted by the
contribution of these to the total value of agricultural production.
The range of estimates obtained from the various methods suggest that the credit elasticity
of agricultural GDP for the entire period 1995-96 to 2011-12 is 0..21, i.e. a 10% increase in
institutional credit flow to agriculture in current prices is associated with a 2.1% increase in
agricultural GDP the following year expressed in current prices. When controlling for prices
represented by the wholesale price index, a 10% increase in nominal credit associated with a
0.97% increase in real GDP, indicating that inflation might be eroding some of gains made in
nominal terms. Compared with these results in the simple one period lag model (method 1), the
estimated credit elasticity is 0.04 when the model controls for the use of inputs and a vector of
input and output prices and for the possible endogeneity of credit through a control function
approach (method 2). The structural model incorporating the pathways through which credit
influences agricultural GDP (method 3) yields estimates of credit elasticity of 0.21. These results
however has weak statistical significance.
The results from a period-wise disaggregate analysis is less conclusive. While the first
model suggests that the elasticity continues to be statistically significant but has weakened in the
post-doubling period, the other two approaches, one that controls for prices and input and the
other the captures the pathways suggest that the relationship between credit and agricultural GDP
may have declined, but none of the estimated credit elasticity coefficients are statistically
significant implying that the hypothesis that the credit has no asociation with agricultural GDP
cannot be rejected.
At the state level, estimates of credit elasticity of agricultural GDP from a simple one-period
lag model, the only feasible option given the data, vary mostly between 0.05 and 0.7. For only a few
exceptions, the credit elasticity turned out to be statistically insignificant. Further, at the state level,
the time trend of elasticity estimates varies across states. In some states the relationship appears to
have strengthened post doubling whereas for others it has weakened. Further clarity and insight
can only be obtained through detailed case studies or primary surveys, owing to the paucity of state
level data that precludes modeling efforts at the state level.
This study goes beyond to understand the precise role of credit, in other words, the
pathways through which it influences or is associated with agricultural GDP. The findings from the
analysis suggest that all the inputs, are highly responsive to an increase in institutional credit to
agriculture, after controlling for input prices, output prices, sectoral composition of agriculture,
area sown and so on. A 10 % increase in credit flow in nominal terms leads to an increase by 1.7%
in fertilizers (N, P, K) consumption in physical quantities, 5.1% increase in the tonnes of pesticides,
10.8% increase in tractor purchases. The credit elasticity of new pumpsets energized is however
not statistically. A disaggregate analysis, for the pre-doubling and post-doubling phases, suggest
3
that the relative importance of the inputs have changed. Whereas in the pre-doubling phase,
fertilizers were statistically significantly responsive, in the post-doubling phase credit appears to
have a strong relationship with tractors.
Overall, it seems quite clear that input use is sensitive to credit flow, whereas GDP of
agriculture is not. This seems to indicate that the ability of credit to engineer growth in agricultural
GDP is impeded by a problem of productivity and efficiency where the increase in input use and
adjustments in the pattern of input use are not (yet) translating into higher agricultural GDP. Credit
seems therefore to be an enabling input, but one whose effectiveness is undermined by low
technical efficiency and productivity.
4
ACKNOWLEDGEMENTS
This study was funded by the Research and Development (R&D) fund of the Department of
Economic Analysis and Research (DEAR), NABARD. We thank Dr.Prakash Bakshi, Dr.R.N.Kulkarni,
Dr.Satyasai and the DEAR team at NABARD for their support advice and inputs at every stage of this
work. Garima Dhir provided valuable research assistance. Andaleeb Rahman, Krushna Ranaware,
Sowmya Dhanraj, Sanjay Prasad, Sumit Mishra, Khaijamang and others assisted in putting together
the dataset used for this study. This work has benefitted immensely from Nirupam Mehrotra’s
inputs and insights into the various issues relating to the role of credit in Indian agriculture. We
also thank all the participants of a seminar on July 4, 2013 at DEAR, NABARD where we shared the
preliminary findings. Their many suggestions have been incorporated in this report. All remaining
errors and omissions are however ours.
5
PRODUCTIVITY OF AGRICULTURAL CREDIT IN INDIA
Assessing the Recent Role of Institutional Credit to Agriculture using State Level Data
1. Introduction
The decades since 1990 have been challenging times for Indian agriculture. Growth rates of
agricultural Gross Domestic Product (GDP) have been languishing and the traditional crop sectors
have seen declining profitability. This has pushed policy makers to direct special attention to
addressing some of the pressing concerns confronting Indian agriculture. Institutional credit has
been an important lever in this effort. Indeed, as many as three major policy initiatives focussed on
institutional credit have been implemented since 2000 to bolster the agricultural sector. The first
policy initiative, introduced in 2004-05, was to double the volume of credit to agriculture over a
period of three years (to 2006-07), relative to the 2004-05 base to expand the reach of formal
finance. Close on its heels came the Agricultural Debt Waiver and Debt Relief Scheme (ADWDRS)
2008, in response to the persistent problem of indebtedness and to alleviate financial pressures
faced by the farmers. The interest rate subvention was then introduced in 2010-11 with the stated
goal of providing incentives for prompt repayment of loans, partly to address the perceived fallout
of the ADWDRS, that it had somehow vitiated the repayment culture. All three measures have
contributed, both explicitly and implicitly, to burgeoning institutional lending to agriculture in the
last decade.2
Despite the significance of these interventions, very little is known regarding the outcomes,
in particular, whether institutional credit has had the intended impact on agricultural growth.
Existing commentaries focussing on this period point out the poor correlation between the two
(Chavan and Ramakumar, 2007 for example). In this context, this research project aims to
understand the extent to which, if at all, growing institutional credit to agriculture supports growth
in the sector. What is its precise role and through what pathways does it support agriculture? These
questions assume particular significance in the context of recent speculation that agricultural credit
might not entirely flow to agriculture or and that there is a significant spillover to other sectors
(Chavan, 2009; Burgess and Pande, 2005 and Binswanger and Khandker, 1992). Some ask if formal
credit a `sensible’ way to support agriculture in India. While answers to this question are perhaps
addressed best through detailed microstudies, it is also possible to elicit patterns and relationships
between agricultural credit and agricultural GDP using secondary data. This study uses secondary
data to examine four specific questions: How productive is institutional credit to the agricultural
sector? What has been the trend since mid-1990s? What are the pathways through which credit
impacts agriculture? How, if at all, have these pathways changed over the years? Using detailed
state level data for the period 1995-2012, we analyze the possible impact of credit on agricultural
GDP using multiple methods, using a control function approach. We also analyse the potential
pathways through which institutional credit can influence agricultural growth, focussing on the
2
There is evidence to suggest that the institutional lending to agriculture might have picked up since 2000,
even before the doubling of credit in 2004-05 (Chavan and Ramakumar, 2007).
6
responsiveness of input demand to institutional credit flow in some detail. An aggregate analysis of
this nature necessarily has severe limitations and needs to be interpreted with care but can serve to
complement our understanding of the productivity of agricultural credit in India.
The paper is organized as follows. The following section (Section 2) provides the context of
agriculture and credit in recent times, underscoring the motivation for this study. Section 3 reviews
the empirical evidence on the productivity of rural credit in India, focussing on secondary data
analysis at the aggregate level (state or national). Section 4 provides a conceptual framework for
the present analysis. Section 5 then discusses the empirical strategies adopted in this work to tackle
some inherent econometric issues that pose problems for establishing a causal relationship
between credit and agricultural GDP. The section discusses the data used for the analysis and also
outlines the scope and limitations of these approaches. Section 6 is devoted to the results from the
econometric exercise and discusses the findings at length. The concluding section 7 closes with
some remarks on the study and the way forward.
2. Characterizing Agriculture and Institutional Credit since the 1990s.
Both the structure of agriculture and the nature of institutional credit have been undergoing a
rapid change since the 1990s but became especially pronounced since 2000. Initiatives to expand
the reach of formal credit has been a goal pursued consistently in the past, even in the 1990s, for
example, the introduction of the Kisan Credit Card (KCC) scheme in 1998-99 aimed to provide
farmers with adequate and timely credit support from the banking system for agriculture and allied
activities in a flexible and cost-effective manner.
Three major policy initiatives in recent years have come to define the context of institutional
credit to agriculture in India, as outlined in the introduction. The first policy introduced in 2004
sought to double the volume of agricultural credit relative to what it was in 2004-05, over a period
of three years. Since then, the actual credit flow has consistently exceeded the target (Government
of India, 2012). Against a credit flow target of Rs.3,25,000 crore during 2009-10, the achievement
was Rs.3,84,514 crore, forming 118 percent of the target. The target for 2010-11 was Rs.3,75,000
crore while the achievement on March, 2011 is Rs.4,46,779 crore (Government of India ,2013). A
second policy involved the waiving of agricultural debtfor small farmers and an opportunity for one
time settlement for others.3 Close on its heels, an interest subvention scheme was introduced to
reward prompt repayment of loans, widely perceived has having been vitiated by the debt waiver
scheme. Under the existing interest subvention scheme, farmers get short-term crop loans at seven
per cent interest. If the loan to the bank is promptly paid then the effective rate of interest to the
farmer works out to four per cent a year due to the additional interest subvention.4 Interest
subvention scheme for short-term crop loans to be continued scheme extended for crop loans
3
As per the provisional figures, a total of 3.01 crore small and marginal farmers and 67 lakh 'other farmers'
have benefited from the Scheme involving debt waiver and debtrelief of Rs.65,318.33crore (Government of
India, 2013).
4
From kharif 2006-07, farmers are receiving crop loans upto a principal amount of 3 lakh at 7% rate of
interest. In the year 2009-10, Government provided an additional 1% interest subvention to those farmers
who repaid their short term crop loans as per schedule. This subvention for timely repayment of crop loans
was raised from 1% to 2% in 2010-11, further 3% from the year 2011-12. Thus the effective rate of interest
for such farmers will be 4 % p.a. (Government of India, 2013).
7
borrowed from private sector scheduled commercial banks. Together, these interventions have
both explicitly and implicitly transferred large amounts to the agricultural sector. Although the
increasing trend of institutional credi flow might have begun in 2000 itself (Ramakumar and
Chavan, 2007), credit flow in recent years have stood out for its magnitude, if not for reversing the
trend of the 1990s.
For example, ground level credit flow as a proportion of the total value of paid out inputs in
agriculture and allied sector has increased from an average of 21% in 1995-96 to 2003-04 to 69%
during the period 2004-05 to 2011-12. As of 2011-12, as much as 85% of paid out inputs were
accounted for by institutional credit. This increasing share is suggestive of credit outstripping the
increasing costs and perhaps crowding out informal borrowing in the aggregate.
Alongside the significant increase in the total credit flow into agriculture, the nature and
source of credit have also seen significant shifts. Indirect finance has, for instance, accounted for an
increasing share of total credit and in terms of the type of institution that provide credit,
commercial banks have been growing in importance as a source (Figure 2). As observers point out
the definition of indirect finance (and what constituted permissible lending avenues) have widened
in scope (Ramakumar and Chavan, 2007, for example) accounting for the apparent increases.
Parallel to transformation in the structure of agricultural credit is the equally rapid
transformation of the agricultural sector. The past two decades have seen an increase in the share
of livestock relative to crops, with extraordinary growth in poultry and dairy sectors. The crop
sector has seen significant diversification, to high value commodities, to horticulture for example.
At the same time, the contextual constraints of agriculture have also been significant issues. Rising
costs of inputs, tightening labour markets, increasing wages and a host of environmental
8
constraints such as water and soil quality degradation have resulted in plateauing yields and
thinning profit margins. The past decade has seen increasing mechanization as evidence by the sale
of machinery (Figure 3) as well as in the machine hours employed per hectare (Figure 4). In
particular, it is evident from state level data on the Cost of Cultivation studies in India that the
mechanization process has replaced animal and draught power while decline in human labour
inputs into agriculture have been less by comparison (Figure 4). 5
5
For more details on cost of cultivation studies in India, seehttp://eands.dacnet.nic.in/Cost_of_Cultivation.htm.
Accessed January, 2014.
9
Source: Cost of Cultivation in India (several years). The All India average is the weighted avearge
across crops and states, where the weights are area sown under each crop within state and the net
sown area across states.
10
There is now limited evidence that despite the widespread notion of a crisis, Indian
agriculture might be more productive but that these improvements as represented by Total Factor
Productivity (TFP), might be coming from certain states (in the south and west) and certain sectors
(such as horticulture and livestock). 6 Other evidence similarly suggests that productivity
improvements is marked only in a few states (Chaudhari, 2013). Improvements in efficiency are
low for a majority of states and might have in fact declined in several states implying the presence
of potential gains in production even with existing technology.
The links that institutional credit has to agricultural productivity and growth are still
somewhat underresearched. Figure 5 plots the ratio of agricultural GDP to credit flow over the
period 1996-2011 for the various states. It is apparent that notwithstanding the variation across
states, the ratio for the country as a whole has been declining over the past 15 years and is now
close to one on average. These patterns appear to indicate that although credit is contributing to a
larger share of the value of purchased inputs, the relationship between agricultural GDP and
agricutlural credit are possibly weak, raising important questions on the role of agricultural credit.
Figure 5: The Ratio of Agricultural GDP and Credit for major states (1996-2011)
Notes: The scatter points represent the ratio of agricultural GDP and credit outflow and the line
represent the lowess fit, i.e. locally weighted scatterplot smoothing.
6See
Rada(2013) for a recent analysis of Total Factor Productivity in India. Results suggest renewed growth
in aggregate TFP growth despite a slowdown in cereal grain yield growth. TFP growth appears to have shifted
to the Indian South and West, led by growth in horticultural and livestock products over the period 19802008.
11
3. Empirical Evidence on the Productivity of Institutional Credit in India
The best known study of the impact of formal rural credit in the context of India is by
Binswanger and Khandkher (1992) who found that rural credit has a measurable positive effect on
agricultural output. Cooperative credit advanced has elasticity with respect to output of 0.063. It is
larger than the elasticity of crop output with respect to predicted overall rural credit which is near
0.027, but not precisely estimated. The estimate for the impact of commercial bank branches on
output is more precisely estimated at 0.02. Others suggest that the effect on output is either nonexistent, for example Burgess and Pande (2005) who claim that the increase in output due to formal
credit comes entirely from increases in non-farm output, or have been negligible.7 Others show that
there is a positive association between credit and agricultural output but that this varies cross
states and further that there is a positive association between the number of persons with accounts
and agricultural output suggesting the financial inclusion could impact agricultural output
positively (Das, et al, 2009).8 However a dynamic panel data estimation of this relationship does
not yield a statistically significant relationship at the state level. A district level panel for 2001-06
for four states however reveals that direct agricultural credit has a positive and immediate impact
on agricultural output, and the number of account relating to indirect agricultural credit has a
positive impact but with a year’s lag. More recent work using time series techniques without
modeling the underlying structure indicate that the elasticity of real agricultural GDP with respect
to institutional credit to agriculture (from commercial banks, cooperatives and RRBs) is 0.22 with a
one-year lag (Subbarao, 2012).9 10In contrast to the somewhat ambivalent findings on the
association between agircultural credit and output, there appears to be consensus that formal
agricultural credit has an important effect on the use of inputs. Bhalla and Singh (2010)
demonstrate in their cross sectional analysis using data for 2003-06 that the elasticity of demand
for inputs with respect to credit is quite significant. At the all India level, credit elasticities for use of
fertilisers, tractors and tubewells hovered around 0.85 suggesting that 10 per cent increase in
supply of direct institutional credit to the farmers to leads to 8-9 per cent increase in use of
fertiliser, tractors and tubewells in long run. Their finding comes from a simple model that
regresses thelogarithm of inputs per unit of output on logarithm of institutional credit. They find
that these elasticities vary across regions and credit elasticities are exceptionally very high for
tractors, tubewells and irrigation for the technologically backward eastern region. Bhalla and Singh
7The
estimates suggested that a one percent increase in the number of rural banked locations reduced rural
poverty by roughly 0.4 percent and increased total output by 0.30 percent. The output effects are solely
accounted for by increases in non-agricultural output – a finding which suggests that increased financial
intermediation in rural India aided output and employment diversification out of agriculture.
8
There are two models that have been estimated in the literature - fixed effects and random effects. The fixed
effects model assumes that there is an unobserved time independent effect for each state of India and this
effect could be correlated with other explanatory variables. The Hausman test helps decide whether to
estimate a fixed effects model or a random effects model. The random effects model assumes that the
unobserved effect is uncorrelated with all the explanatory variables.
9The model regresses ln (AGDP) on ln (Acredit(-1)), where AGDP = GDP from agriculture and allied activities
at constant prices and Acredit = Credit for agriculture and allied activities deflated by GDP deflator with one
year lag.
10
Other studies of this type include Ghosh (2010) , Pavaskar, et al. (2011).
12
(2010) then suggest that institutional credit is indispensable for these regions with low input and
investment in agriculture. Binswangerand Khandker (1992) point out that institutional growth and
higher lending volumes lead to modest increases in aggregate crop output but sharp increases in
the use of fertilizers and in investments in physical capital and, substantial reductions in
agricultural employment. They conclude on that basis that expansion of credit has, therefore, led to
the substitution of capital for agricultural labor.
These two studies emphasize the multiple pathways in which formal agricultural credit
impacts production and this is well recognized by now (see Sriram 2007, for example). If one is to
understand this linkage in all its complexity, one needs a detailed construct of these relationships.
4.
Conceptualization of the Role of Formal Credit
The fundamental attribute of credit implies that it serves as an intermediate input and does
not directly enter as an input into agricultural production. It is therefore an enabling input. On
account of this, it plays a complex role in farmers’ production decisions, unlike physical inputs that
have a more transparent relationship with the levels of output.
The impact of agricultural credit on agricultural production, efficiency and productivity
could potentially occur through multiple channels. A simple conceptualization identifies three
pathways through which formal credit can influence outcomes (Figure 6). First, formal credit can be
used to purchase inputs over the cropping season, enabling a farmer to maximize the yield from the
cultivated area, given a level of capital stock. This channel represents a direct and within-season
impact on production. Second, formal credit can be used to make investments in irrigation facilities,
machines and draught animals that represent the use of credit for building up capital stock to
support agricultural production. This second channel typically impacts production with a time lag.
Both of these represent a “liquidity effect” (Binswanger and Khandkher, 1992) since they relieve a
farmer’s credit constraint and enables purchase of critical inputs to support production. Third,
formal credit is often used to replace informal credit associated with high interest burden.
Anecdotal evidence suggests that farmers often borrow from formal sources to pay off high interest
loans taken from money lenders. This has the effect of relieving credit constraints, reducing the
interest burden and indebtedness. Existing economic literature on wealth effects and risk aversion
suggests that this often enables farmers to make decisions that increase profitability and efficiency.
Even when formal credit is diverted to consumption, there could be an implicit wealth effect that
impacts farmer’s production decisions. This last channel, which incorporates a “consumption
smoothing” effect is is often difficult to capture.
Collectively, formal agricultural credit can be regarded as having two kinds of impacts –
first, it could enable a farmer to move to the production frontier so that given prevalent technology,
a farmer is using levels of inputs that enable him/her to produce at the frontier, from among many
feasible combinations of crops. Second, it could enable a farmer to move on to a superior
production frontier, so that given a level of inputs, the farmer is able to produce more of one or
more of the crops. The first is represented as a move from within the production possibility set to
the frontier (constituting efficiency improvement) and the second is represented as a shift of the
frontier itself (constituting productivity improvement). The impact of formal agricultural credit on
agricultural output conflates these two aspects of productivity and efficiency effects.
13
Figure 6 : Schematic Representation of Pathways
“Consumption Smoothing
Effect” Replace usurious loans,
relax consumption constraints,
etc.increasing the risk-bearing
capacity of farmers.
Direct Credit from
formal institutions
“Liquidity Effect” 1:
Working Capital (for purchase
of inputs)
Agricultural
Production
“Liquidity Effect” 2:
Investment credit (for
purchase of `capital stock’ to
support production)
5.
Empirical Strategy
a. The Challenges
Empirically, these effects are difficult to entangle. While a separation of these effects and
pathways are ideally studied at the household level, this logic can be extended to an aggregate level
by choosing empirical counterparts that represent these dimensions at the state or district level,
with important caveats. Aggregation often masks a lot of the heterogeneity and complexity of the
ways in which formal agricultural mediates production processes. The distribution of credit among
farmers or farmer groups is often uneven and is not taken into account when in an aggregate
analysis. Similar problems occur with aggregating over all the crops and commodities, which
masks the differential impacts and relative importance of credit. While this study is cognizant of
these issues, data limitations allow only an aggregate-level analysis.
Several methodological and data challenges persist in estimating the impact of formal
agricultural credit on output, especially at the aggregate level. Firstly, informal credit which forms a
major source of credit is something of a black box with virtualy no data available on its quantum or
how it is used. The fungibility of credit too poses difficult problems for research since it makes
short and long term credit indistinguishable at the farmers’ end. Similarly there could be spillovers
into non-farm sector that are unknown to the lenders and to researchers. Direct and Indirect
14
finance might also not be watertight categories so that it could be the case that direct credit to the
farmer is in fact used for ancillary activities that support agriculture. All of these make it hard to pin
down the precise nature of relationship between credit and agriculture.Further, the dynamic effects
are difficult to capture since credit flow in a particular year might yield cumulative benefits over
several years. This is particularly difficult to model. The other major challenges stem from data
constraints at the state level. Existing data for variables of interest are not often available for all the
states and for all the years, forcing us to confine the analysis to the states for which we have
complete data for the period of focus.
The empirical challenges of studying the relationship between formal agricultural credit
and output at an aggregate level are best described by Binswanger and Khandker (1992). The first
problem is the joint determination of both observed formal credit to agriculture and aggregate
output. The second problem emanates from the absence of data on informal credit, which makes it
difficult to capture the impacts of formal credit that might work through reduced informal
borrowing, and not factoring this might yield the estimates that reflect the true effect of formal
credit. Credit advanced by formal lending agencies such as banks is an outcome of both the supply
of and demand for formal credit. The amount of formal credit available to the farmer, his credit
ration, enters into his decision to make investments, and to finance and use variable inputs such as
fertilizer and labor. The third econometric problem arises because formal agriculture lending is not
exogenously given or randomly distributed across space. Ways to be able to address some of these
issues are central concerns of this project.
b. Methodology
To parse this complex relationship given the limitations of data, a combination of three
approaches are used (Box 1). The first is a simple model that regresses agricultural GDP on credit
flow using state level data. This is a catchall approach that
Box 1: The Three Approaches
cannot comment on the pathways or provide a causal
interpretation. The second method estimates a hybrid profitMethod 1: The Simple Model using
production function that regresses agricultural GDP on a
state level data and dividing into time
vector of relevant inputs, prices and agricultural credit for
periods in nominal terms as well as
the same year. This is a direct approach to estimating the
accounting for prices.
relationship between credit and agricultural GDP in reduced
Method 2: The Direct Approach that
form. The possible endogeneity of credit is addressed by the
regresses agricultural GDP on various
use of a control function approach where a regression
inputs
(fertilizers,
tractors
pumpsets), prices, rainfall, public
function is estimated that identified and then “controls” for
expenditure on agricultural, including
the endogenous component of observed credit flow (Imbens
credit flow and the estimated
and Wooldridge, 2007). The coefficient on credit in this case
endogenous component of credit or
captures one dimension of impact of credit that is not
the “control” variable.
mediated through inputs. The third method is perhaps the
Method 3: The Pathways Approach
most comprehensive and models the pathways approach in
that works on three stages – credit
what is referred to as a mediation analysis framework,
market, input demand functions and
where inputs are regarded as mediating the relationship
value of GDP function, estimated in a
between institutional credit and agricultural GDP (Preacher
SURE framework for panels and
and Hayes, 2008). Here, a set of regressions estimates input
incorporating the control variable.
15
demand as a function of credit, among other things and controlling for endogeneity of credit
(indirect effect), and the hybrid production-profit function as a function of inputs and credit,
recognizing that credit can also have direct effect on GDP. The coefficients representing the
responsiveness of input use to institutional credit are therefore used as components to construct
the total impact of agricultural credit on agricultural GDP (Preacher and Hayes, 2008). The impact
of credit on agricultural output is thus derived as the sum of the contribution of credit to the use of
specific inputs, capital or the cropping pattern, weighted by the contribution of these to the total
value of agricultural production. These are estimated in a Seemingly Unrelated Regression
Equations (SURE) framework that acknowledges the potential interrelationship between these
variables and the fact that they might be jointly determined. Appendix 1 contains a representation
of the models estimated.
For all three methods we use state-level data to estimate the relevant parameters of interest
for India as a whole. The analysis pertains to the time period 1995-96 to 2011-12, for which the
data is complete. Further, we also perform an analysis for two sub-periods pre-doubling (1995-96
to 2003-04) and post doubling (2004-05 to 2011-12). In all the methods, we make the assumption
of constant elasticity of demand, which is in fact a non-trivial assumption, but one that is typical of
studies of this kind.
As mentioned earlier, the chief methodological challenge involves dealing with the the issue
of endogeneity of observed credit. There are several approaches to deal with this. One approach to
tackle the endogeneity of observed volumes of credit is to use the predicted supply of credit at the
state level, following Binswanger and Khandker (1992) or to use lagged credit that is correlated
with current year credit. Each of these involves a set of defensible assumptions. The latter approach
however creates problems because there could typically be lagged response of agricultural GDP
which renders lagged credit an inappropriate instrument. In this study we use a control function
approach to separate out the explained exogenous variation in the credit flow to agriculture from
the unexplained and possibly endogenous component of credit flow and use the predicted residuals
from the control function to control for endogeneity in themain set of regression equations (Imbens
and Wooldridge, 2007). Appendix 1 provides more details on the approach and the regressions
estimated. The standard errors for both models 2 and 3 are bootstrapped to account for the use of
predicted variables as explantory variables.
c. Data Sources
To implement this method, we use a data set that is more detailed than used in the literature till
date. For all the major states in India details on credit, agricultural GDP, composition of the value of
output in the agriculture and allied sector and variables relating to land under cultivation provide
the key variables of interest. Data on physical quantities of Nitrogen, Phosphorus and Potassium
fertilizers have been assembled as also pesticides (technical grade) as also tractors and pumpsets
energized. Use of certified seeds in only available at the national level and is only used in explain
agricultural GDP but not as a separate input since this cannot be done at the state level. Other state
level variables representing the level of development include per capita State Domestic Product,
percentage of villages electrified, the number of commercial bank branches. Prices are typically
available at the all-India level, for the various inputs, power and fuel as well as output (food grains,
16
etc.). State level wage rates are compiled and in the absence of annual data on labour inputs used,
wage rates are expected to proxy labour use. We are also able to account for labour, machine and
animal power intensity per hectare from Cost of Cultivation data at the state level. These are
computed within state as the weighted average across crops (with weights being the area under
different crops), and across states as the weighted average across states (with weights being the
state’s share of gross cropped area). While for labour and animal, we use hours per hectare,
machine use data are in value terms. Appendix 2 provides details of the data used for the analysis
and the sources. While data is not available for all the states for all the years, only those states and
years for which all data was available are used in the analysis. Essentially, the data then consists of
time series data for the major states so that the panel data framework is used to estimate the
impact of credit on agricultural output at the national level, with state fixed effects. The models are
estimated for the major agricultural states, since data is not complete for all the states.
d. Scope and Limitations of the Study
The scope of this effort will be limited to estimating the impact of formal credit from different
institutions – cooperatives, rural and commercial banks – on agricultural output. The spillover
effects of formal credit on the rural non-farm sector will not be addressed specifically, an issue that
research suggests might be quite important (Pande and Burgess, 2005). Neither does this work
address the implications of recent interventions in credit policy such as debt waiver; this is already
studied elsewhere (Kanz, 2012; Cole, 2009). Another important area that is beyond the remit of this
study is the fiscal implication of the system of disbursing formal rural credit. One could argue that
to gauge the true impact of credit, one would have to account for the fiscal burden (or some notion
of net benefit cost ratio) (Binswanger and Khandkerm 1992). In this work, the question of interest
is to gauge whether or not direct formal rural credit impacts agricultural output, the extent to
which it does so and the relative importance of the different pathways through which these effects
occur.
6.
a.
The Results
The Productivity of Credit: Credit Elasticity of Agricultural GDP
The range of estimates obtained from the various methods suggest that the credit elasticity
of agricultural GDP for the entire period 1995-96 to 2011-12 is 0.21, i.e. a 10% increase in
institutional credit flow to agriculture in current prices is associated with a 2.1% increase in
agricultural GDP that year expressed in current prices (Table 1). This model controls for prices and
hence account for inflation.
Compared with these results in the simple model (method 1), the estimated credit elasticity
is 0.04 when the model controls for the use of inputs and a vector of input and output prices and for
the possible endogeneity of credit through a control function approach (method 2). The structural
model incorporating the pathways through which credit influences agricultural GDP (method 3)
yields estimates of credit elasticity of 0.0.2 . But neither method indicates that these coefficients are
statistically significant (Table 1).
The results from a period-wise disaggregate analysis is less conclusive. While the simple
model suggests that the elasticity continues to be statistically significant but has weakened in the
17
post-doubling period, the other two approaches, one that controls for prices and inputs and the
other the captures the pathways suggest that the relationship between credit and agricultural GDP
may have declined, but none of the estimated credit elasticity coefficients are statistically
significant and hence on cannot reject the null that the responsiveness of agricultural GDP to credit
has been zero.
At the state level, estimates of credit elasticity of agricultural GDP from the `simple’ model,
the only feasible option given the data, vary mostly between 0.05 and 0.7 with several states show
statistically insignificant elasticities (Table 2). Further, at the state level, the time trend of elasticity
estimates varies across states. In some states the relationship appears to have strengthened post
doubling (for example, in Tamil Nadu, Maharashtra and Gujarat) whereas for several others it has
weakened (including for Himachal Pradesh, Rajasthan, Uttar Pradesh, Karnataka, Kerala,
Chhattisgarh Madhya Pradesh, etc.). Punjab appears to show a consistently strong relationship
between agricultural GDP and credit.Notwithstanding these varitions, a strinking feature in the
relationship between agricultural GDP and credit flow is the pronounced convergence in the
agricultural GDP-credit flow ratio suggesting that perhaps the marginal returns to credit might be
equalizing across states (Figure 7).
Further clarity and insight can only be obtained through detailed case studies or primary
surveys, owing to the paucity of state level data that precludes modeling efforts at the state
level.This underscores the potential problems with aggregation and that observations on trends
cannot be generalized.
Table 1: Summary Results of the three models
Method 2 (Direct
Approach using
Control Function
methods.)
Method 3
(Pathways
approach using
Control Function
for credit)
(2)
(3)
(4)
0.214***
0. .036
0.210
Pre-doubling
0.266***
-0.010
0.102
Post-doubling
0.099***
0.138
-0.030
Time period for which
elasticity of GDP with
respect to credit is
computed.
(1)
The Entire Period
Method 1(Simple
Model Controlling
for Prices)
Notes:
(1) The Hausman Test suggests that the fixed effects model is appropriate. For Model 1, the Hausman chi-sqaured
(1)=15.51***
(2) Granger Causality tests indicate that agricultural credit Granger-causes agricultural GDP and not the other way.
(3) The Chow test for Method 1 indicates that the post-doubling coefficient is not statistically significantly different
from that from the pre-doubling period.
(4) Detailed results are available with the authors and can be provided on request. See also Appendix 3 and 4.
(5) Standard errors for the control function approach are bootstrapped 200 times.
18
(6) The states included in this regression are Andhra Pradesh, Bihar, Chhattisgarh, Gujarat, Haryana, Himachal
Pradesh, Jharkhand, Karnataka, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar
Pradesh, Uttarakhand, West Bengal. For new states, data since their inception are included.
Table 2: State-wise Credit Elasticity of Agricultural GDP under the Lag Model
Census
Code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
State
Jammu and Kashmir
Himachal Pradesh
Punjab
Chandigarh
Uttaranchal
Haryana
Delhi
Rajasthan
Uttar Pradesh
Bihar
Sikkim
Arunachal Pradesh
Nagaland
Manipur
Mizoram
Tripura
Meghalaya
Assam
West Bengal
Jharkhand
Orissa
Chhattisgarh
Madhya Pradesh
Gujarat
Daman and Diu
Dadra Nagar Haveli
Maharashtra
Andhra Pradesh
Karnataka
Goa
Lakshadweep
Kerala
Tamil Nadu
Pondicherry
Andaman and Nicobar
Islands
Whole
Period
0.053*
0.303**
0.340***
0.047
0.019
0.406***
0.029
0.171
0.288***
-0.062
0.092***
0.035
-0.022
-0.048*
0.072*
0.013
-0.000
-0.006
0.015
0.255**
0.321***
0.376*
0.490***
0.574***
0.016
0.448***
0.278*
-0.095
0.325***
0.355***
0.187**
-0.006
Predoubling
0.151
0.232**
0.522**
0.016
-0.103***
1.900***
-0.024
0.853***
-0.960**
-0.427***
0.029
0.044
-0.217
-0.073**
0.092
-0.147
-0.048*
-0.018
-0.866***
0.252
0.421
0.205***
0.555*
0.826
-0.007
1.046*
1.591***
-0.300
0.395***
0.493
0.069
0.040*
Post
Doubling
0.002
0.112
0.129**
-0.011
-0.086
-0.440
-0.109
0.060
0.040
0.574*
-0.030
-0.004
-0.023
-0.061
-0.119
-0.001
0.065***
-0.005
0.005
-0.420
0.028
-0.054
-0.049
0.567***
0.209*
0.143
0.039
0.017
-0.033
0.298***
0.133
-0.070
19
Notes: The state level elasticities are the slope coefficient from a regression of agricultural GDP on
credit flow to agriculture, controlling for wholesale price index. States have been arranged
according to census code.
Figure 7: State-wise ratio of Agricultural Gross Domestic Product and Credit Flow (19962011)
Notes: Only the major states have been included.
20
b.
Pathways of “productivity” : Input Demand and Credit Flow
If credit is an enabling or mediating input, its impact on output and productivity operates
through its influence on the level of purchased inputs, variable and fixed. A system of input demand
functions is estimated as a Seemingly Unrelated Regression Equations (SURE), with credit as one of
the explanatory variables along with the predicted residuals from the control function to account
for the endogeneity of credit (Table 3). The inputs included are fertilizers (a total of Nitrogen,
Phosphate and Potassic fertilizers), pesticides, tractors purchased and pumpsets energized
annually. The other inputs include labour and animal power intensity as well as expenditure per
hectare on machine use. Controls include other inputs like land, distinguished by type of irrigation,
prices of inputs, prices of food articles, lagged wages of unskilled labour, government expenditure,
lagged variable accounting for the structure of agriculture. Due to paucity of state level annual data,
detailed information on other equipments are not available for inclusion; tractors are therefore a
coarse proxy for equipment. So too with pumpsets, which represent one type of irrigation
investment. Investments in drip and sprinkler irrigation, etc. are hard to capture for lack of data.
The inclusion of government expenditure likely captures the subsidies offered for these irrigation
investments. These results need to be interpreted with caution. The results (Table 1) suggest that
over the entire period, institutional credit has a strong association with all inputs excepting
pumpsets energized. A 10 % increase in credit flow in nominal terms leads to an increase by 1.7%
in fertilizers (N, P, K) consumption in physical quantities, 5.1% increase in the tonnes of pesticides,
10.8% increase in tractor purchases. The credit elasticity of new pumpsets energized is however
not statistically significant.
Interestingly, there appears to be a marked shift in the pathways between the pre-doubling
and post-doubling phases. Whereas in the pre-doubling phase, institutional credit seems to have
been channelled into purchase of variable inputs such as fertilizers, in the post-doubling phase,
credit seems to be directed to investments in tractors. This is consistent with the popular
perception that high labour costs and a shortage of farm hands is prompting mechanization and it
appears that credit is aiding and enabling this transition. The absence of a strong relationship with
pumpsets could be on account of the variable representing irrigated land and perhaps government
expenditure, which might include subsidies for pumpsets. There might thus be a conflation of the
many explanatory variables.
It is apparent that availability of credit also reduces the labour intensity of agriculture by
2%. However there is no evidence that could be is consistent with the idea of labour substituting
mechanization. One possible interpretation is that increasingly some operations such as manual
weeding are being replaced by the use of chemical weedicides and so on. Likewise greater
ownership of tractors reflects this mechanization rather than just the paid out cost for machine use.
Alternatively it could be that mechanization as represented by the responsiveness of tractors to
credit flow substitutes animal power (rather than labour use).
21
Usually, this weak relationship especially of capital equipment such as tractor and pumpset
is strongly suggestive that mechanization is preserving productivity or agricultural growth rather
than enhancing it (Binswanger and Khandkher, 1992). In these contexts, credit can be interpreted
as performing two roles the preservation of productivity levels by supporting mechanization of
certain kinds and contributing to the growth of agricultural GDP through the purchase of variable
inputs. All these results collectively suggest that credit indeed appears to have played a role in
supporting the changing face of agriculture in India.
Overall, it seems quite clear that input use is sensitive to credit flow, whereas GDP of
agriculture is not. This seems to indicate that the ability of credit to engineer growth in agricultural
GDP is impeded by a problem of productivity and efficiency where the increase in input use and
adjustments in the pattern of input use are not (yet) translating into higher agricultural GDP. Credit
seems therefore to be an enabling input, but one whose effectiveness is undermined by low
technical efficiency and productivity.
Table 3: Input Demand System : The Credit Elasticity of Input Demand from a SURE Model
Input or Agricultural
GSDP
All
(1995/96 to 2011/12)
Phase 1
(1995/96 to 2003/04)
Pre-doubling
Fertilizers
0.17*
0.33**
0.06
Chemicals
0.51***
0.83
0.26
Tractors bought
1.08***
0.10
1.67***
-0.84
0.04
-0.83
Pumpsets Energized
Phase 2
(2004/05 to 2011/12)
Post-doubling
Labour hours per
-0.20**
-0.28
-0.16
hectare
Animal hours per
0.18
-0.07
-0.04
hectare
Machine use (Rs. Per
-0.67**
-1.13
-0.17
hectare)
Agricultural Gross State
0.083
-0.1
0.13
Domestic product
NOTES:
(1) This is estimated as a SURE (Seemingly Unrelated Regression Equation). Breusch-Pagan test of
independence: chi2(28) = 48.303, Pr = 0.0099 suggests that the null hypothesis of independence is rejected and
that these euqations need to be estimated as a system.
(2) The standard errors were bootstrapped with 200 repetitions to account for the inclusion of the predicted
variable from the control function.
(3) The regression was run in deviation form to allow the direct use of SUREG command in STATA 13.
(4) The coefficient of the control function variable is significant only in the case of pumpset suggesting that
endogeneity of credit is only true of pumpsets.
(5) The detailed regressions are available with the author. See Appendix 3 and 4
22
7.
Concluding Remarks
This report sought to investigate the relationship between institutional credit to agriculture
and agricultural Gross Domestic Product (GDP). Collectively, the results suggest that the fears that
credit might be ineffective are perhaps misplaced. There is strong evidence that credit is indeed
playing its part of supporting the purchase of inputs and perhaps even aiding the agricultural sector
respond to its contextual constraints.
The evidence of the impact of credit on agricultural GDP is however weak at best ,
irrespective of the approach used, assuming a constant credit elasticity of agricultural GDP.
Empirical patterns suggest that the relationship between credit and agricultural GDP is somewhat
weak in the post-doubling phase. Further, as is evident from the regression of agricultural GDP on
inputs and prices, other than fertilizers and labour, few inputs are strong drivers of GDP. In fact it
appears that the sectoral composition and output prices are important determinants of agricultural
GDP, apart from certain types of government expenditure and the irrigated area. Usually, this weak
relationship especially of capital equipment such as tractor and pumpset is strongly suggestive that
mechanization is preserving productivity or agricultural growth rather than enhancing it. In these
contexts, credit can be interpreted as performing two roles the preservation of productivity levels
by supporting mechanization of certain kinds and contributing to the growth of agricultural GDP
through the purchase of variable inputs. All these results collectively suggest that the success of
credit in enabling the increase in use of purchased inputs and effecting changes in input mix,
supporting the changing face of agriculture in India has not translated fully into agricultural GDP
growth as such.
REFERENCES
Bhalla, G.S. andGurmail Singh (2010) Growth of Indian Agriculture: A District Level Study, Planning
Commission, Government of India. Available athttp://planningcommission.nic.in/reports/sereport/ser/ser_gia2604.pdf
Binswanger, Hans.P. and Shahidur Khandker (1992): ‘The Impact of Formal Finance on Rural
Economy of India’, World Bank, Working Paper No. 949. (also appeared in The Journal of
Development Studies Volume 32, Issue 2, 1995)
Burgess, Robin and RohiniPande (2005) Do Rural Banks Matter? Evidence from the Indian Social
Banking Experiment, American Economic Review, American Economic Association, vol. 95(3), pages
780-795, June.
23
Chaudhary, Shilpa (2013) Trends in Total Factor Productivity in Indian Agriculture: State-level
Evidence using non-parametric Sequential Malmquist Index, Working Paper.
Chavan, P. (2009). How Rural is India’s Agricultural Credit. The Hindu.
Cole, S. (2009). Fixing market failures or fixing elections? Agricultural credit in India.American
Economic Journal: Applied Economics, 219-250.
Das, Abhiman, Manjusha Senapati, Joice John (2009): 'Impact of Agricultural Credit on Agriculture
Production: An Empirical Analysis in India', Reserve Bank of India Occasional Papers Vol. 30, No.2,
Monsoon 2009
De, Sankarand and SiddharthVij (2012): Are Banks Responsive to Exogenous Shocks in Credit
Demand? District – level Evidence from India, Research Paper, CAE, ISB, Hyderabad
Ghosh, Nilanjan (2010) Incredulity of Irresponsiveness: Is Agricultural Credit Productive?
Commodity Vision, Volume 4, Issue 1, July 2010 Takshashila Academia of Economic Research Ltd,
2010
Golait, R. (2007): Current Issues in Agriculture Credit in India: An Assessment, RBI Occasional
Papers, 28: 79-100.
Government of India (2013) Status of Indian Agriculture 2011-2012, Ministry of Agriculture,
Government of India.
Imbens, Guido and Jeffrey Wooldridge (2007) Control Function and Related Methods, What’s New
in Econometrics? Lecture Notes 6, National Bureau of Economic Research (NBER), Summer 2007.http://www.nber.org/WNE/lect_6_controlfuncs.pdf. Accessed July, 2013.
Kanz, M. (2012). What does debt relief do for development? Evidence from India's bailout program for
highly-indebted rural households.World Bank Policy Research Working Paper, 6258, Washington D.C.
Pavaskar Madhoo, Sarika Rachuri, Aditi Mehta (2011) Agricultural Credit Productivity in India
Commodity Vision Volume 4, Issue 5, March 2011 Takshashila Academia of Economic Research Ltd,
2011
Preacher, K.J. and Hayes, A.F. (2008).Asymptotic and resampling strategies for assessing and
comparing indirect effects in multiple mediator models.Behavioral Research Methods, 40, 879-891.
Rada, Nicholas E., 2013. Agricultural Growth in India: Examining the Post-Green Revolution
Transition 2013 Annual Meeting, August 4-6, 2013, Washington, D.C. 149547, Agricultural and
Applied Economics Association.
Ramakumar, R. and Chavan, P. (2007). Revival of agricultural credit in the 2000s: An Explanation.
Economic and Political Weekly, 57-63.
24
Sriram, M.S.(2007): ‘Productivity of Rural Credit: A Review of Issues and Some Recent Literature’,
Indian Institute of Management Ahmedabad, Working Paper No.2007-06-01.
Subbarao Duvvuri (2012) Agricultural Credit - Accomplishments and Challenges, Speech delivered
at NABARD, July 12, 2012.
25
APPENDIX 1: Empirical Strategy: The Methods Described
Method 1: Time Series Simple Model
The first method is a simple model, where agricultural GDP is regressed on the current time
period’s credit to agriculture. This model is estimated as a panel model with fixed effects, based on
Hausman test for choice of models. This model is also run separately for the Pre-doubling phase
(1995-96 to 2003-04) and post-doubling phase (2004-05 to 2010-11) and for individual states.
wherei refers to the state and t the financial year.
is the `lagged’ credit elasticity of agricultural GDP,
Method 2: Reduced Form Control Function Approach
Credit equation/ Control Function
The first step in this method is to address the endogeneity of credit. Since the demand for credit
itself could be a result of agricultural GDP, a control function approach is adopted to separate out
that part of credit that could be due to exogenous factors and that part which might represent the
endogenous component. IN this regression, we use variables that are hypothesized to exogenously
influence the level of credit. This includes the previous year’s rainfall, per capita income, structure
of agriculture and the number of branches of commercial banks in the state.
where
represents the total credit flow to agriculture (all sources and short and long term) and
. can be regarded as the endogenous part of credit, the estimated values of which are used in
the next stage regression.
Outcome function
The outcome function is essentially a hybrid production-profit function that maps a set of inputs to
outputs, controlling for exogenous factors such as the weather, market prices of output and inputs,
public infrastructure. Since the aggregate value of output is likely sensitive to the composition of
crops or the cropping pattern, the regression will control for the proportion of area under the major
groups of crops – foodgrains (cereals and pulses), oilseeds, fibre, horticulture, spices and plantation
crops (such as tea, rubber and coffee). In lieu of private and public capital stock and investment in
agriculture which capture capital inputs into production but are not available for all the states,
select machinery and equipment are included. A key component of this regression is the estimated
“control” variable from the Control Function described above that serves to control for endogeneity
26
and thereby allow us to interpret the coefficient on credit as a causal effect rather than mere
correlation.
where K is the credit flow, Z is the vector of inputs (including N, P K fertilizers, pesticides, tractors
and pumpsets) and other factors (O) such as rainfall, per capita state domestic product, percentage
of villages in the state that are electrified and so on, P is the vector of prices, etc.
is the credit elasticity of agricultural GDP, with associated bootstrapped standard errors for
hypothesis testing.
Method 3: Reduced Form Control Function Approach
(a) Credit equation/ Control Function
(b) Input / Capital Demand Equations
In order to retrieve the coefficients that represent the different pathways, we will estimate the set
of structural equations to understand the relative contribution of credit to different components of
the agricultural production-profit function.
The input demand functions depend on credit (among other things). We then estimate input
demand equations as a system, where the inputs are measured in physical units, and explanatory
variables include both the “control” variable and credit. The rapid changes in the cropping pattern
in India in the past two decades is both in response to the growing market opportunities as well as
the growth of processing sectors which in turn are likely impacted by indirect credit. So the
composition of the agricultural sector and the growing importance of livestock, poultry and
fisheries would need to be accounted for. Credit for purchase of milch animals as well as
construction of broiler sheds for contract growing are important components of agricultural output.
Due to paucity of detailed annual data on draught animals, share of livestock output in total
agricultural output is used as a proxy. The inputs used include fertilizers such as Nitrogen (N),
Phosphorous (P) and Potassium (K), pesiticides. Standard errors are computed through
bootstrapping procedures to account for the fact that these regressions use predicted values at
different stages.
27
(c) Outcome function
We then estimate the function explaining agricultural GDP in monetary terms as a hybrid profit
function. Compute the total impact of credit as the sum of the impacts on inputs weighted by the
impact of the input in question on agricultural GDP.
(d) Credit Elasticity of Agricultural GDP
The impact of credit on agricultural GDP can then be derived as the sum of the contribution of
credit to the use of specific inputs, capital or the cropping pattern, weighted by the contribution of
these to the total value of agricultural production. Standard errors reflect bootstrapped estimates.
28
APPENDIX 2: Data and Sources
Stata variable
name
Variable label (units)
Mean
Min
80.72
animalhoursha
Animal (hours/ha)
57.42
1.15
241.16
areanonfoodtotal
2905.47
1
52398.00
cagalliedtotal
Total area unde nonfood crops
(`000 hectares)
total cagallied
13384.02
541032.00
cladvecoservtotal
total cladvecoserv
64155.42
commercial
Number of commercial bank
branches
Production(in Lakhs)
4165.49
67499
18091
7
12404.66
1
321068.00
7835.36
0
72344.00
gca
No. of Fertilizer Sale Points
(Total)
Gross Cropped Area (`000 ha)
11606.41
2
195357.00
Ministry of Agriculture,
Agricultural Statistics at a
Glance
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture
gsdp_agri
GSDP(in Rs. Lakh)
3943
127000000.00
Ministry of Agriculture
irrcanal
Canal Irrigated area (`000 ha)
2943807.0
0
2178.10
0
17995.00
Ministry of Agriculture
irrtank
Tank Irrigated area (`000 ha)
312.70
0
3343.00
Ministry of Agriculture
k
175.45
0.02
3632.40
Fertiliser Association of India
labhoursha
Potassic fertilizers (`000
tonnes)
Labour (hours/ha) wtd avg
568.69
1691.17
machinersperha
Machine (Rs./ha)
1417.22
211.6
6
0
milk_prod
Production(in '000 MT)
2766.42
1
21031.00
n
Nitrogenous fertilizers (`000
tonnes)
Net Sown Area (`000 ha)
1092.02
0.6
17300.25
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture,
Agricultural Statistics at a
Glance
Fertiliser Association of India
8631.22
1
142960.00
Ministry of Agriculture
byproduct: Value of Output(Rs.
Lakhs)
Cereals: Value of Output(Rs.
Lakhs)
drugs: Value of Output(Rs.
Lakhs)
fibres: Value of Output(Rs.
Lakhs)
fish: Value of Output(Rs. Lakhs)
176028.30
0
3222156.00
National Accounts Statistics
810692.80
0
16300000.00
National Accounts Statistics
81793.53
0
1434137.00
National Accounts Statistics
94783.26
0
2756026.00
National Accounts Statistics
183344.90
19
3906563.00
National Accounts Statistics
fruits: Value of Output(Rs.
Lakhs)
607584.60
0
13800000.00
National Accounts Statistics
egg_prod
fert_salepoint_total
nsa
output_byproduct
output_cereals
output_drugs
output_fibres
output_fish
output_fruits
512.88
Source
Power tariff to agriculture
(paise/kWh)
agelectariff
0
Max
Central Electricity Authority
EPWRF time series
1881901.00
101261.00
6028.95
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture
State Finances : A Study of
Budgets, RBI
State Finances : A Study of
Budgets, RBI
CMIE
29
Stata variable
name
Variable label (units)
Mean
Min
Max
Source
output_livestoc
k
output_oilseeds
livestock: Value of Output(Rs.
Lakhs)
924819.20
479
21200000.00
National Accounts Statistics
oilseeds: Value of Output(Rs.
Lakhs)
othercrops: Value of Output(Rs.
Lakhs)
257429.60
0
5449477.00
National Accounts Statistics
140840.20
0
3097888.00
National Accounts Statistics
Pulses: Value of Output(Rs.
Lakhs)
spices: Value of Output(Rs.
Lakhs)
sugar: Value of Output(Rs.
Lakhs)
Phosphatic fertilizers (`000
tonnes)
Per capita State Domestic
Product
Percentage of villages electrified
115964.70
0
2315486.00
National Accounts Statistics
79445.64
0
1838363.00
National Accounts Statistics
206510.30
0
3665781.00
National Accounts Statistics
429.72
0.38
8049.70
Fertiliser Association of India
144345.00
3037
47200000.00
National Accounts Statistics
85.16
0
285.87
Central Electricity Authority
output_othercr
ops
output_pulses
output_spices
output_sugar
p
pcsdpcurrent
pcvillageselectri
c
pesticide
EPWRF time series
2998.05
0.39
63651.00
Fertiliser Association of India
100.50
70.42
148.50
RBI
pfodder
Pesticide Technical Grade
Consumption (MT)
Price Index for Fertilisers and
Pesticides
Price Index for Fodder
114.77
63.17
245.64
RBI
pfoodarticles
Price Index for Food Articles
121.15
67.9
215.20
RBI
pfuelpower
Price Index for Fuel and Power
120.95
66.1
195.53
RBI
ptractor
Price Index for Tractors
101.75
65.54
144.21
RBI
pumpset
Pumpsets energized
835496.10
0
16300000.00
Central Electiricty Authority
ragrialliedtotal
total ragriallied
132352.00
415
4339185.00
rainavgann
Rainfall Annual (mm)
995.41
0
2630.00
State Finances : A Study of
Budgets, RBI
Ministry of Agriculture
seeds
132.78
62.2
283.85
Fertiliser Association of India
total
Certified seeds distributed (lakh
quintals)
Total Disbursment (in Rs. lakh)
824263.20
0
46700000.00
DEAR, NABARD
tractorsale
Sale of Tractors
28179.76
36
545109.00
Agricultural research data
book IASRI
tractorstock
Stock of Tractors (CMIE?)
312343.40
4629
4547080.00
CMIE
unskilledlabour
ers
villelectric
Wage Index for unskilled
labourers
85.46
32.33
393.82
RBI
Percentage of villages electrified
85.16
0
285.87
EPWRF time series
wpi_all_avg
Wholesale Price Index
(Financial Year averages)
Wholesale Price Index
(Calendar Year averages)
102.51
63.58
152.33
RBI
97.05
58.28
164.93
RBI
pfertpest
wpi_allcalavg
Observations with missing values are excluded from the regression
30
APPENDIX 3 : Regression Results for SURE Model 3.
Variables (in deviation, log form)
Fertilizers
Both
Phases
Phase 1
Phase 2
b/t
b/t
b/t
Credit
0.169
0.326*
0.058
(1.88)
(2.03)
(0.51)
"Error" from Control function
0.003
0.025
-0.103
(0.05)
(0.11)
(-0.86)
Net Sown Area
0.632*
0.441
0.255
(2.01)
(0.88)
(0.50)
Area under non-food Crops
0.000
0.000
0.000
(1.04)
(0.25)
(1.24)
Fertilizer Sale points
0.000
0.000
0.000
(0.73)
(1.30)
(0.27)
Index of Fertilizer/Pesticide prices
-0.002
0.036*
0.021
(-0.58)
(1.96)
(1.14)
Wholesale Price Index
-0.002
0.015
0.002
(-0.46)
(0.92)
(0.15)
Lagged Food Articles Price Index
0.006**
-0.074
-0.001
(2.74)
(-1.83)
(-0.13)
(LAGGED) SHARE OF TOTAL VALUE OF AGRICULTURAL OUTPUT
Fibres
0.003***
0.002
0.002*
(3.71)
(1.52)
(2.06)
Spices
-0.000
-0.001
0.000
(-0.09)
(-0.53)
(0.23)
Cereals
0.000
0.001
-0.000
(0.69)
(1.06)
(-0.27)
Fruits
-0.000
-0.000
0.000
(-0.04)
(-0.28)
(0.12)
Oilseeds
0.000
-0.000
0.000
(0.53)
(-0.23)
(0.49)
Pulses
0.001
-0.000
0.000
(0.72)
(-0.03)
(0.06)
Sugar
-0.000
0.000
-0.000
(-0.37)
(0.03)
(-0.01)
Per Capita State Domestic Product
-0.000
-0.000
0.000
(-0.32)
(-0.46)
(0.36)
Annual Average Rainfall
0.016
0.024
0.004
(0.36)
(0.26)
(0.06)
Constant
-0.182
2.855
-2.394
(-0.49)
(1.92)
(-1.38)
Pesticides
Both
Phases
b/t
0.510**
(2.69)
0.077
(0.46)
1.261
(1.39)
-0.000
(-0.36)
0.000
(0.15)
-0.011
(-0.80)
-0.026*
(-2.00)
0.016*
(2.32)
Phase 1
b/t
0.827
(1.54)
-0.035
(-0.05)
0.762
(0.57)
0.000
(1.51)
0.000
(0.93)
-0.006
(-0.09)
-0.079
(-1.17)
0.068
(0.45)
Phase 2
b/t
0.263
(1.30)
0.530
(1.58)
0.288
(0.22)
-0.000
(-1.50)
-0.000
(-0.50)
-0.069
(-1.47)
-0.008
(-0.28)
0.018
(0.80)
-0.002
(-0.73)
0.001
(0.21)
-0.001
(-0.67)
0.000
(0.03)
0.001
(0.42)
-0.000
(-0.17)
-0.001
(-0.53)
-0.000
(-0.07)
-0.165
(-1.09)
2.154*
-0.002
(-0.36)
0.008
(0.75)
-0.002
(-0.64)
-0.001
(-0.34)
0.003
(0.58)
-0.001
(-0.27)
0.001
(0.25)
0.000
(0.19)
-0.041
(-0.09)
1.419
-0.002
(-0.66)
-0.004
(-0.69)
0.001
(0.47)
0.000
(0.14)
-0.004
(-1.21)
-0.007
(-1.28)
-0.001
(-0.21)
-0.000
(-0.39)
0.078
(0.32)
6.199
(1.98)
(0.27)
(1.60)
b refers to coefficient and t refers to t value (in parenthesis)
31
APPENDIX 3 (Continued)
Variables (in deviation, log form)
Credit
"Error" from Control function
Net Sown Area
Area under non-food Crops
Fertilizer Sale points
Index of Fertilizer/Pesticide prices
Wholesale Price Index
Lagged Food Articles Price Index
Animal hours per ha
Both
Phases
Phase 1
Phase 2
Machine expenses per hectare
Both
Phases
Phase 1
Phase 2
Labour hours per hectare
Both
Phases
Phase 1
Phase 2
b/t
b/t
b/t
b/t
b/t
b/t
b/t
b/t
b/t
0.176
(0.83)
-0.070
-0.038
-0.671*
-1.125
-0.172
-0.200*
-0.282
-0.157
(-0.21)
(-0.09)
(-2.02)
(-1.34)
(-0.77)
(-2.36)
(-1.47)
(-1.07)
0.147
0.090
0.231
-0.633
-0.062
-0.297
-0.078
-0.052
0.036
(0.62)
(0.19)
(0.46)
(-0.94)
(-0.05)
(-1.09)
(-1.01)
(-0.21)
(0.24)
0.865
0.249
2.563
0.679
2.608
-0.259
0.252
0.442
-0.116
(1.46)
(0.26)
(1.45)
(0.72)
(1.06)
(-0.22)
(1.15)
(0.87)
(-0.18)
0.000
-0.000
0.000
-0.000
0.000
-0.000
0.000
-0.000
0.000
(0.82)
(-0.68)
(0.31)
(-0.05)
(0.47)
(-0.43)
(0.54)
(-0.70)
(0.53)
0.000
0.000
0.000
-0.000
-0.000
0.000
-0.000
-0.000
-0.000
(0.68)
(0.64)
(0.34)
(-0.65)
(-1.40)
(0.59)
(-1.67)
(-0.83)
(-1.11)
-0.020
-0.041
0.048
0.032
0.021
0.030
-0.007
-0.014
0.012
(-1.19)
(-0.79)
(0.58)
(1.82)
(0.20)
(0.74)
(-1.44)
(-0.53)
(0.46)
0.007
-0.020
0.045
0.056*
0.072
-0.005
0.011*
0.005
0.019
(0.52)
(-0.55)
(0.97)
(2.43)
(0.76)
(-0.25)
(2.03)
(0.31)
(1.18)
-0.007
0.053
-0.036
-0.030**
0.015
0.004
-0.001
0.019
-0.012
(-0.93)
(0.51)
(LAGGED) SHARE OF TOTAL VALUE OF AGRICULTURAL OUTPUT
(-1.12)
(-2.96)
(0.07)
(0.28)
(-0.30)
(0.41)
(-1.20)
Fibres
Spices
Cereals
Fruits
Oilseeds
Pulses
Sugar
0.001
-0.003
0.000
-0.003
-0.004
0.001
0.002*
-0.000
0.005***
(0.43)
(-0.75)
(0.01)
(-1.34)
(-0.47)
(0.38)
(2.17)
(-0.11)
(3.98)
0.001
-0.006
0.007
-0.013
-0.028
-0.002
-0.002
-0.004
0.001
(0.23)
(-1.05)
(0.63)
(-1.31)
(-1.20)
(-0.47)
(-1.27)
(-1.32)
(0.18)
0.003*
0.002
0.006
0.002
0.008
-0.002
0.001*
0.001
0.002*
(2.46)
(1.01)
(1.85)
(1.24)
(1.77)
(-1.25)
(2.19)
(0.92)
(2.13)
0.002
0.001
0.004
-0.002
0.001
-0.003
0.001
0.000
0.001
(1.31)
(0.45)
(0.98)
(-0.88)
(0.29)
(-1.12)
(1.15)
(0.35)
(0.75)
-0.000
0.002
-0.004
0.001
-0.000
0.002
0.001
0.001
0.000
(-0.08)
(0.75)
(-0.80)
(0.54)
(-0.01)
(0.75)
(0.85)
(0.67)
(0.19)
-0.004
-0.002
-0.009
0.006
0.009
0.003
-0.003***
-0.002
-0.003
(-1.26)
(-0.50)
(-1.15)
(1.79)
(1.12)
(0.69)
(-3.41)
(-1.01)
(-1.16)
0.003
-0.000
0.009
0.000
-0.004
-0.002
0.001
-0.000
0.002
(1.09)
(-0.02)
(1.81)
(0.15)
(-0.41)
(-0.67)
(1.33)
(-0.04)
(1.03)
32
Variables (in deviation, log form)
Per Capita State Domestic Product
Annual Average Rainfall
Constant
Animal hours per ha
Both
Phases
Phase 1
Phase 2
Machine expenses per hectare
Both
Phases
Phase 1
Phase 2
Labour hours per hectare
Both
Phases
Phase 1
Phase 2
b/t
-0.000
b/t
b/t
b/t
b/t
b/t
b/t
b/t
b/t
0.000
-0.000*
-0.000
-0.000
0.000
-0.000
-0.000
-0.000
(-1.21)
(0.75)
(-2.06)
(-0.46)
(-0.86)
(0.64)
(-0.98)
(-0.33)
(-0.75)
-0.153
-0.104
-0.039
-0.108
-0.518
-0.135
0.110
0.182
0.056
(-0.70)
(-0.49)
(-0.10)
(-0.49)
(-0.68)
(-0.84)
(1.41)
(1.02)
(0.57)
1.847
0.421
-6.334
-6.280**
-11.712
-2.720
-0.363
-1.401
-2.127
(1.33)
(0.12)
(-0.95)
(-3.23)
(-1.72)
(-0.84)
(-0.74)
(-0.87)
(-1.04)
b refers to coefficient and t refers to t-value (in parenthesis)
33
APPENDIX TABLE 3 (Continued)
Variables (in deviation, log form)
Credit
"Error" from Control function
Net Sown Area
Area under non-food Crops
Tractor sale
Both
Phases
Phase 1
b/t
b/t
1.079***
0.096
(3.32)
(0.18)
-0.025
0.347
(-0.07)
(0.45)
-0.094
0.297
(-0.13)
(0.20)
0.000
-0.000
(0.72)
(-0.62)
Phase 2
b/t
1.671**
(3.11)
0.040
(0.08)
0.562
(0.26)
0.000
(0.92)
Index of Fertilizer/Pesticide prices
Wholesale Price Index
-0.054**
-0.013
-0.000
(-2.85)
(-0.09)
(-0.00)
Lagged Food Articles Price Index
0.012
-0.129
0.040
(1.33)
(-0.70)
(0.60)
Tractor price index
-0.007
0.133
-0.109
(-0.33)
(0.92)
(-1.74)
Unskilled workers wage index
0.005
0.013
0.004
(1.02)
(0.89)
(0.31)
Fule and Power price index
0.004
0.010
-0.055
(1.56)
(0.94)
(-0.87)
(LAGGED) SHARE OF TOTAL VALUE OF AGRICULTURAL OUTPUT
Fibres
-0.002
-0.002
-0.004
(-0.59)
(-0.30)
(-0.59)
Spices
0.009
0.013
0.004
(1.51)
(1.03)
(0.32)
Cereals
-0.001
-0.002
-0.003
(-1.01)
(-0.69)
(-0.82)
Fruits
0.000
0.003
-0.003
(0.21)
(1.11)
(-0.72)
Oilseeds
0.001
-0.000
-0.003
(0.66)
(-0.10)
(-0.66)
Pulses
0.004
0.005
-0.000
(1.45)
(0.74)
(-0.02)
Sugar
-0.008*
-0.009
-0.012
(-2.54)
(-1.24)
(-1.87)
Per Capita State Domestic Product
0.000*
0.000
0.000
(2.02)
(0.08)
(1.18)
Annual Average Rainfall
0.226
0.195
-0.217
(1.21)
(0.39)
(-0.45)
Constant
4.835*
1.170
12.919*
(2.54)
(0.11)
(2.17)
Pumpsets
Both
Phases
b/t
-0.842
(-1.52)
-0.067
(-0.08)
0.778
(0.38)
-0.000
(-0.36)
0.048
(0.83)
0.007
(0.15)
-0.029
(-1.13)
Phase 1
b/t
0.038
(0.03)
-1.084
(-0.62)
-1.698
(-0.56)
-0.000
(-0.37)
0.456
(1.58)
-0.477
(-1.74)
0.310
(0.79)
Phase 2
b/t
-0.832
(-0.82)
-0.557
(-0.48)
5.586
(1.04)
-0.000
(-0.23)
0.009
(0.05)
0.041
(0.18)
-0.040
(-0.31)
-0.002
(-0.16)
-0.001
(-0.16)
0.025
(0.76)
-0.076
(-1.91)
-0.010
(-0.37)
0.007
(0.09)
-0.002
(-0.44)
-0.009
(-0.51)
-0.001
(-0.23)
-0.003
(-0.76)
-0.001
(-0.32)
-0.008
(-1.17)
0.015*
(2.39)
0.000**
(3.03)
-0.393
(-0.95)
-1.789
0.004
(0.40)
-0.003
(-0.14)
0.001
(0.16)
-0.007
(-0.84)
-0.002
(-0.44)
-0.002
(-0.20)
0.015
(1.18)
0.000*
(2.03)
0.529
(0.63)
-20.078
-0.005
(-0.38)
-0.016
(-0.48)
-0.003
(-0.47)
-0.000
(-0.00)
0.001
(0.05)
-0.005
(-0.22)
0.017
(1.31)
0.000
(1.32)
-0.484
(-0.64)
-1.878
(-0.48)
(-1.19)
(-0.12)
34
APPENDIX 4: Results for Agricultural GDP as part of SURE Model 3
Variable (in log devation form)
Both Phases
b/t
Phase 1
b/t
Phase 2
b/t
Credit
0.083
(1.00)
-0.028
(-0.42)
0.000
(0.52)
0.000
(1.43)
0.193
(0.92)
0.010
(0.41)
0.011
(0.36)
-0.000
(-0.63)
0.010
(0.31)
0.492**
(2.82)
-0.073
(-1.32)
-0.210
(-1.16)
0.265**
(3.12)
-0.040
(-1.20)
-0.001
(-0.02)
0.001
(0.96)
-0.016
(-0.33)
0.020**
(3.28)
0.002
(1.42)
-0.006
(-0.74)
-0.003**
(-2.69)
-0.095
(-0.39)
-0.130
(-0.37)
0.000
(0.95)
0.000
(0.47)
0.186
(0.44)
0.044
(0.65)
0.002
(0.03)
-0.000
(-0.07)
0.032
(0.32)
0.745*
(2.42)
-0.005
(-0.04)
0.033
(0.07)
0.216
(0.99)
-0.043
(-0.32)
0.014
(0.14)
-0.000
(-0.09)
-0.032
(-0.24)
0.071
(0.99)
0.003
(1.23)
-0.019
(-0.56)
-0.012
(-0.77)
0.133
(0.77)
-0.042
(-0.37)
-0.000
(-0.18)
0.000
(0.10)
-0.124
(-0.18)
-0.017
(-0.21)
0.030
(0.32)
0.000
(0.11)
-0.045
(-0.43)
0.184
(0.66)
-0.101
(-0.69)
-0.074
(-0.18)
0.350
(1.95)
-0.017
(-0.26)
-0.003
(-0.02)
0.002
(0.49)
-0.025
(-0.26)
0.161
(0.58)
-0.019
(-0.52)
0.011
(0.24)
-0.011
(-0.58)
"Error" from Control Function
Milk production
Egg production
Gross cropped area
Land irrigated by canals
Land irrigated by tanks
Area under Non-food crops
Pesticide
Fertilizers (N, P and K)
Tractors
Pumpsets
Labour hours per hectare
Aimal hours per hectare
Machine (Rs. Per hectare)
Lagged unskilled wage index
Average Annual Rainfall
Lagged Fertilizer and Pesticide price index
Lagged Fodder price index
Lagged tractor price index
Lagged Fuel and poer price index
35
Variable (in log devation form)
Both Phases
b/t
Phase 1
b/t
Phase 2
b/t
Wholesale Price Index
-0.006
(-0.89)
-0.015
(-0.55)
-0.028
(-0.50)
dlncladvecoservtotal
-0.009
(-0.80)
0.000
(0.06)
-0.019
(-0.64)
-0.001
(-0.14)
-0.005
(-0.25)
0.001
(0.06)
N
R-squared
BIC
0.076
(0.87)
-0.000
(-0.03)
-0.105
(-0.80)
121.000
0.739
603.766
-0.052
(-0.31)
0.008
(0.78)
0.000
(.)
59.000
0.468
411.019
0.056
(0.49)
-0.002
(-0.34)
-0.488
(-0.37)
62.000
0.685
324.758
AIC
170.418
91.079
-4.948
Capital expenditure on Agricultre and allied services
Revenue expenditure on agriculture and allied
services
Percetage of villages electrified
Constant
b refers to coefficient and t refers to t-value (in parenthesis)
36
37
doc_404597537.pdf
Projec ton nabard
THE PRODUCTIVITY OF
AGRICULTURAL CREDIT
ASSESSING THE RECENT ROLE
OF INSTITUTIONAL CREDIT TO AGRICULTURE IN INDIA
USING STATE LEVEL DATA
Final Report submitted to NABARD
August 24, 2014
Sudha Narayanan1
Indira Gandhi Institute of Development Research
Goregaon (East), Mumbai 400065
Assistant Professor, IGIDR , [email protected]
1
1
EXECUTIVE SUMMARY
Amongst recent policy interventions implemented to revive the languishing agricultural
sector in India, those pertaining to agricultural credit have been very much in the forefront. In
particular, three major policy initiatives have shaped the past decade in institutional credit to
agriculture. The policy of doubling of institutional credit to agriculture between 2004-05 and 200607 (over the 2004-05 base year) marked the first attempt to alleviate the financial constraints of
farmers. In 2008-09, the Agricultural Debt Waiver and Debt Relief Scheme (ADWDRS) was
introduced to waive specific outstanding debts for a large number of small farmers; this was
followed by the interest subvention scheme, that sought to remedy the perceived negative impact
of the waiver on loan repayment culture by rewarding timely repayment with loans carrying lower
interest rates. These three schemes combined have, implicitly and explicitly, resulted in an
increasing volume of institutional credit to agriculture. Whereas credit accounted for only 16% of
the total value of paid out inputs in the triennium ending (TE) 1998-99, and 26.3% in TE 2003-04,
by the end of the decade, in TE 2011-12, it had risen to as high as 80.3% of the estimated total paid
out costs of inputs.
Despite its importance, little is known about the effectiveness of credit in supporting
agricultural growth as represented by the GDP and indeed the very nature of the relationship
between formal agricultural credit and agricultural GDP. This research project is a modest effort in
this direction. While acknowledging that questions of the impact of credit on agricultural output or
value addition or productivity are best addressed through a textured understanding of household
behavior and microstudies, this study is based on the premise that aggregate secondary data too
can reveal some of these important relationships. This study uses state-level data to examine the
relationship between institutional credit to the agricultural sector and agricultural GDP at the
national level. Despite serious limitations of aggregation, which typically disregards distributional
issues and often masks more than it reveals, it is also true that any systematic or pervasive
relationship should reflect in aggregate data and offers a level of generalization not available in
small scale surveys. The goal of this study is four-fold: How productive is institutional credit to the
agricultural sector? What has been the trend since mid-1990s? What are the pathways through
which credit impacts agriculture? How, if at all, have these pathways changed over the years? The
analysis covers the period 1995-96 to 2011-12 using data that includes all major states within
India. The study also conducts, data permitting, a disaggregate analysis of two sub-periods – the
first phase denoting the Pre-doubling period (1995-96 to 2003-04) and the second representing the
Post-doubling period (2004-05 to 2011-12). Where feasible, the study replicates the analysis at the
state level.
In this study, credit is conceptualized as an enabling input that influences agricultural GDP
primarily via use of variable inputs and through investments in fixed capital that support
agricultural production. To the extent that credit can also contribute to consumption smoothing of
borrowers or better their capacity for risk bearing, credit could have a non-specific influence on
agricultural GDP via variables that are typically unobserved by the researchers. To parse this
complex relationship given the limitations of data, a combination of three approaches are used. The
first is a simple model that regresses agricultural GDP on current credit flow using state level data.
The second method estimates a hybrid profit-production function that regresses agricultural GDP
2
on a vector of relevant inputs, prices and agricultural credit flow during that year. This is a direct
approach to estimating the relationship between credit and agricultural GDP in reduced form. The
possible endogeneity of credit is addressed by the use of a control function that “controls” for the
estimated endogenous component of observed credit flow. The third method represents the
`pathways approach’ which estimates input demand as a function of credit, among other things and
controlling for endogeneity. The coefficients representing the responsiveness of input use to
institutional credit are then used as components to construct the total impact of agricultural credit
on agricultural GDP. The impact of credit on agricultural output is thus derived as the sum of the
contribution of credit to the use of specific inputs, capital or the cropping pattern, weighted by the
contribution of these to the total value of agricultural production.
The range of estimates obtained from the various methods suggest that the credit elasticity
of agricultural GDP for the entire period 1995-96 to 2011-12 is 0..21, i.e. a 10% increase in
institutional credit flow to agriculture in current prices is associated with a 2.1% increase in
agricultural GDP the following year expressed in current prices. When controlling for prices
represented by the wholesale price index, a 10% increase in nominal credit associated with a
0.97% increase in real GDP, indicating that inflation might be eroding some of gains made in
nominal terms. Compared with these results in the simple one period lag model (method 1), the
estimated credit elasticity is 0.04 when the model controls for the use of inputs and a vector of
input and output prices and for the possible endogeneity of credit through a control function
approach (method 2). The structural model incorporating the pathways through which credit
influences agricultural GDP (method 3) yields estimates of credit elasticity of 0.21. These results
however has weak statistical significance.
The results from a period-wise disaggregate analysis is less conclusive. While the first
model suggests that the elasticity continues to be statistically significant but has weakened in the
post-doubling period, the other two approaches, one that controls for prices and input and the
other the captures the pathways suggest that the relationship between credit and agricultural GDP
may have declined, but none of the estimated credit elasticity coefficients are statistically
significant implying that the hypothesis that the credit has no asociation with agricultural GDP
cannot be rejected.
At the state level, estimates of credit elasticity of agricultural GDP from a simple one-period
lag model, the only feasible option given the data, vary mostly between 0.05 and 0.7. For only a few
exceptions, the credit elasticity turned out to be statistically insignificant. Further, at the state level,
the time trend of elasticity estimates varies across states. In some states the relationship appears to
have strengthened post doubling whereas for others it has weakened. Further clarity and insight
can only be obtained through detailed case studies or primary surveys, owing to the paucity of state
level data that precludes modeling efforts at the state level.
This study goes beyond to understand the precise role of credit, in other words, the
pathways through which it influences or is associated with agricultural GDP. The findings from the
analysis suggest that all the inputs, are highly responsive to an increase in institutional credit to
agriculture, after controlling for input prices, output prices, sectoral composition of agriculture,
area sown and so on. A 10 % increase in credit flow in nominal terms leads to an increase by 1.7%
in fertilizers (N, P, K) consumption in physical quantities, 5.1% increase in the tonnes of pesticides,
10.8% increase in tractor purchases. The credit elasticity of new pumpsets energized is however
not statistically. A disaggregate analysis, for the pre-doubling and post-doubling phases, suggest
3
that the relative importance of the inputs have changed. Whereas in the pre-doubling phase,
fertilizers were statistically significantly responsive, in the post-doubling phase credit appears to
have a strong relationship with tractors.
Overall, it seems quite clear that input use is sensitive to credit flow, whereas GDP of
agriculture is not. This seems to indicate that the ability of credit to engineer growth in agricultural
GDP is impeded by a problem of productivity and efficiency where the increase in input use and
adjustments in the pattern of input use are not (yet) translating into higher agricultural GDP. Credit
seems therefore to be an enabling input, but one whose effectiveness is undermined by low
technical efficiency and productivity.
4
ACKNOWLEDGEMENTS
This study was funded by the Research and Development (R&D) fund of the Department of
Economic Analysis and Research (DEAR), NABARD. We thank Dr.Prakash Bakshi, Dr.R.N.Kulkarni,
Dr.Satyasai and the DEAR team at NABARD for their support advice and inputs at every stage of this
work. Garima Dhir provided valuable research assistance. Andaleeb Rahman, Krushna Ranaware,
Sowmya Dhanraj, Sanjay Prasad, Sumit Mishra, Khaijamang and others assisted in putting together
the dataset used for this study. This work has benefitted immensely from Nirupam Mehrotra’s
inputs and insights into the various issues relating to the role of credit in Indian agriculture. We
also thank all the participants of a seminar on July 4, 2013 at DEAR, NABARD where we shared the
preliminary findings. Their many suggestions have been incorporated in this report. All remaining
errors and omissions are however ours.
5
PRODUCTIVITY OF AGRICULTURAL CREDIT IN INDIA
Assessing the Recent Role of Institutional Credit to Agriculture using State Level Data
1. Introduction
The decades since 1990 have been challenging times for Indian agriculture. Growth rates of
agricultural Gross Domestic Product (GDP) have been languishing and the traditional crop sectors
have seen declining profitability. This has pushed policy makers to direct special attention to
addressing some of the pressing concerns confronting Indian agriculture. Institutional credit has
been an important lever in this effort. Indeed, as many as three major policy initiatives focussed on
institutional credit have been implemented since 2000 to bolster the agricultural sector. The first
policy initiative, introduced in 2004-05, was to double the volume of credit to agriculture over a
period of three years (to 2006-07), relative to the 2004-05 base to expand the reach of formal
finance. Close on its heels came the Agricultural Debt Waiver and Debt Relief Scheme (ADWDRS)
2008, in response to the persistent problem of indebtedness and to alleviate financial pressures
faced by the farmers. The interest rate subvention was then introduced in 2010-11 with the stated
goal of providing incentives for prompt repayment of loans, partly to address the perceived fallout
of the ADWDRS, that it had somehow vitiated the repayment culture. All three measures have
contributed, both explicitly and implicitly, to burgeoning institutional lending to agriculture in the
last decade.2
Despite the significance of these interventions, very little is known regarding the outcomes,
in particular, whether institutional credit has had the intended impact on agricultural growth.
Existing commentaries focussing on this period point out the poor correlation between the two
(Chavan and Ramakumar, 2007 for example). In this context, this research project aims to
understand the extent to which, if at all, growing institutional credit to agriculture supports growth
in the sector. What is its precise role and through what pathways does it support agriculture? These
questions assume particular significance in the context of recent speculation that agricultural credit
might not entirely flow to agriculture or and that there is a significant spillover to other sectors
(Chavan, 2009; Burgess and Pande, 2005 and Binswanger and Khandker, 1992). Some ask if formal
credit a `sensible’ way to support agriculture in India. While answers to this question are perhaps
addressed best through detailed microstudies, it is also possible to elicit patterns and relationships
between agricultural credit and agricultural GDP using secondary data. This study uses secondary
data to examine four specific questions: How productive is institutional credit to the agricultural
sector? What has been the trend since mid-1990s? What are the pathways through which credit
impacts agriculture? How, if at all, have these pathways changed over the years? Using detailed
state level data for the period 1995-2012, we analyze the possible impact of credit on agricultural
GDP using multiple methods, using a control function approach. We also analyse the potential
pathways through which institutional credit can influence agricultural growth, focussing on the
2
There is evidence to suggest that the institutional lending to agriculture might have picked up since 2000,
even before the doubling of credit in 2004-05 (Chavan and Ramakumar, 2007).
6
responsiveness of input demand to institutional credit flow in some detail. An aggregate analysis of
this nature necessarily has severe limitations and needs to be interpreted with care but can serve to
complement our understanding of the productivity of agricultural credit in India.
The paper is organized as follows. The following section (Section 2) provides the context of
agriculture and credit in recent times, underscoring the motivation for this study. Section 3 reviews
the empirical evidence on the productivity of rural credit in India, focussing on secondary data
analysis at the aggregate level (state or national). Section 4 provides a conceptual framework for
the present analysis. Section 5 then discusses the empirical strategies adopted in this work to tackle
some inherent econometric issues that pose problems for establishing a causal relationship
between credit and agricultural GDP. The section discusses the data used for the analysis and also
outlines the scope and limitations of these approaches. Section 6 is devoted to the results from the
econometric exercise and discusses the findings at length. The concluding section 7 closes with
some remarks on the study and the way forward.
2. Characterizing Agriculture and Institutional Credit since the 1990s.
Both the structure of agriculture and the nature of institutional credit have been undergoing a
rapid change since the 1990s but became especially pronounced since 2000. Initiatives to expand
the reach of formal credit has been a goal pursued consistently in the past, even in the 1990s, for
example, the introduction of the Kisan Credit Card (KCC) scheme in 1998-99 aimed to provide
farmers with adequate and timely credit support from the banking system for agriculture and allied
activities in a flexible and cost-effective manner.
Three major policy initiatives in recent years have come to define the context of institutional
credit to agriculture in India, as outlined in the introduction. The first policy introduced in 2004
sought to double the volume of agricultural credit relative to what it was in 2004-05, over a period
of three years. Since then, the actual credit flow has consistently exceeded the target (Government
of India, 2012). Against a credit flow target of Rs.3,25,000 crore during 2009-10, the achievement
was Rs.3,84,514 crore, forming 118 percent of the target. The target for 2010-11 was Rs.3,75,000
crore while the achievement on March, 2011 is Rs.4,46,779 crore (Government of India ,2013). A
second policy involved the waiving of agricultural debtfor small farmers and an opportunity for one
time settlement for others.3 Close on its heels, an interest subvention scheme was introduced to
reward prompt repayment of loans, widely perceived has having been vitiated by the debt waiver
scheme. Under the existing interest subvention scheme, farmers get short-term crop loans at seven
per cent interest. If the loan to the bank is promptly paid then the effective rate of interest to the
farmer works out to four per cent a year due to the additional interest subvention.4 Interest
subvention scheme for short-term crop loans to be continued scheme extended for crop loans
3
As per the provisional figures, a total of 3.01 crore small and marginal farmers and 67 lakh 'other farmers'
have benefited from the Scheme involving debt waiver and debtrelief of Rs.65,318.33crore (Government of
India, 2013).
4
From kharif 2006-07, farmers are receiving crop loans upto a principal amount of 3 lakh at 7% rate of
interest. In the year 2009-10, Government provided an additional 1% interest subvention to those farmers
who repaid their short term crop loans as per schedule. This subvention for timely repayment of crop loans
was raised from 1% to 2% in 2010-11, further 3% from the year 2011-12. Thus the effective rate of interest
for such farmers will be 4 % p.a. (Government of India, 2013).
7
borrowed from private sector scheduled commercial banks. Together, these interventions have
both explicitly and implicitly transferred large amounts to the agricultural sector. Although the
increasing trend of institutional credi flow might have begun in 2000 itself (Ramakumar and
Chavan, 2007), credit flow in recent years have stood out for its magnitude, if not for reversing the
trend of the 1990s.
For example, ground level credit flow as a proportion of the total value of paid out inputs in
agriculture and allied sector has increased from an average of 21% in 1995-96 to 2003-04 to 69%
during the period 2004-05 to 2011-12. As of 2011-12, as much as 85% of paid out inputs were
accounted for by institutional credit. This increasing share is suggestive of credit outstripping the
increasing costs and perhaps crowding out informal borrowing in the aggregate.
Alongside the significant increase in the total credit flow into agriculture, the nature and
source of credit have also seen significant shifts. Indirect finance has, for instance, accounted for an
increasing share of total credit and in terms of the type of institution that provide credit,
commercial banks have been growing in importance as a source (Figure 2). As observers point out
the definition of indirect finance (and what constituted permissible lending avenues) have widened
in scope (Ramakumar and Chavan, 2007, for example) accounting for the apparent increases.
Parallel to transformation in the structure of agricultural credit is the equally rapid
transformation of the agricultural sector. The past two decades have seen an increase in the share
of livestock relative to crops, with extraordinary growth in poultry and dairy sectors. The crop
sector has seen significant diversification, to high value commodities, to horticulture for example.
At the same time, the contextual constraints of agriculture have also been significant issues. Rising
costs of inputs, tightening labour markets, increasing wages and a host of environmental
8
constraints such as water and soil quality degradation have resulted in plateauing yields and
thinning profit margins. The past decade has seen increasing mechanization as evidence by the sale
of machinery (Figure 3) as well as in the machine hours employed per hectare (Figure 4). In
particular, it is evident from state level data on the Cost of Cultivation studies in India that the
mechanization process has replaced animal and draught power while decline in human labour
inputs into agriculture have been less by comparison (Figure 4). 5
5
For more details on cost of cultivation studies in India, seehttp://eands.dacnet.nic.in/Cost_of_Cultivation.htm.
Accessed January, 2014.
9
Source: Cost of Cultivation in India (several years). The All India average is the weighted avearge
across crops and states, where the weights are area sown under each crop within state and the net
sown area across states.
10
There is now limited evidence that despite the widespread notion of a crisis, Indian
agriculture might be more productive but that these improvements as represented by Total Factor
Productivity (TFP), might be coming from certain states (in the south and west) and certain sectors
(such as horticulture and livestock). 6 Other evidence similarly suggests that productivity
improvements is marked only in a few states (Chaudhari, 2013). Improvements in efficiency are
low for a majority of states and might have in fact declined in several states implying the presence
of potential gains in production even with existing technology.
The links that institutional credit has to agricultural productivity and growth are still
somewhat underresearched. Figure 5 plots the ratio of agricultural GDP to credit flow over the
period 1996-2011 for the various states. It is apparent that notwithstanding the variation across
states, the ratio for the country as a whole has been declining over the past 15 years and is now
close to one on average. These patterns appear to indicate that although credit is contributing to a
larger share of the value of purchased inputs, the relationship between agricultural GDP and
agricutlural credit are possibly weak, raising important questions on the role of agricultural credit.
Figure 5: The Ratio of Agricultural GDP and Credit for major states (1996-2011)
Notes: The scatter points represent the ratio of agricultural GDP and credit outflow and the line
represent the lowess fit, i.e. locally weighted scatterplot smoothing.
6See
Rada(2013) for a recent analysis of Total Factor Productivity in India. Results suggest renewed growth
in aggregate TFP growth despite a slowdown in cereal grain yield growth. TFP growth appears to have shifted
to the Indian South and West, led by growth in horticultural and livestock products over the period 19802008.
11
3. Empirical Evidence on the Productivity of Institutional Credit in India
The best known study of the impact of formal rural credit in the context of India is by
Binswanger and Khandkher (1992) who found that rural credit has a measurable positive effect on
agricultural output. Cooperative credit advanced has elasticity with respect to output of 0.063. It is
larger than the elasticity of crop output with respect to predicted overall rural credit which is near
0.027, but not precisely estimated. The estimate for the impact of commercial bank branches on
output is more precisely estimated at 0.02. Others suggest that the effect on output is either nonexistent, for example Burgess and Pande (2005) who claim that the increase in output due to formal
credit comes entirely from increases in non-farm output, or have been negligible.7 Others show that
there is a positive association between credit and agricultural output but that this varies cross
states and further that there is a positive association between the number of persons with accounts
and agricultural output suggesting the financial inclusion could impact agricultural output
positively (Das, et al, 2009).8 However a dynamic panel data estimation of this relationship does
not yield a statistically significant relationship at the state level. A district level panel for 2001-06
for four states however reveals that direct agricultural credit has a positive and immediate impact
on agricultural output, and the number of account relating to indirect agricultural credit has a
positive impact but with a year’s lag. More recent work using time series techniques without
modeling the underlying structure indicate that the elasticity of real agricultural GDP with respect
to institutional credit to agriculture (from commercial banks, cooperatives and RRBs) is 0.22 with a
one-year lag (Subbarao, 2012).9 10In contrast to the somewhat ambivalent findings on the
association between agircultural credit and output, there appears to be consensus that formal
agricultural credit has an important effect on the use of inputs. Bhalla and Singh (2010)
demonstrate in their cross sectional analysis using data for 2003-06 that the elasticity of demand
for inputs with respect to credit is quite significant. At the all India level, credit elasticities for use of
fertilisers, tractors and tubewells hovered around 0.85 suggesting that 10 per cent increase in
supply of direct institutional credit to the farmers to leads to 8-9 per cent increase in use of
fertiliser, tractors and tubewells in long run. Their finding comes from a simple model that
regresses thelogarithm of inputs per unit of output on logarithm of institutional credit. They find
that these elasticities vary across regions and credit elasticities are exceptionally very high for
tractors, tubewells and irrigation for the technologically backward eastern region. Bhalla and Singh
7The
estimates suggested that a one percent increase in the number of rural banked locations reduced rural
poverty by roughly 0.4 percent and increased total output by 0.30 percent. The output effects are solely
accounted for by increases in non-agricultural output – a finding which suggests that increased financial
intermediation in rural India aided output and employment diversification out of agriculture.
8
There are two models that have been estimated in the literature - fixed effects and random effects. The fixed
effects model assumes that there is an unobserved time independent effect for each state of India and this
effect could be correlated with other explanatory variables. The Hausman test helps decide whether to
estimate a fixed effects model or a random effects model. The random effects model assumes that the
unobserved effect is uncorrelated with all the explanatory variables.
9The model regresses ln (AGDP) on ln (Acredit(-1)), where AGDP = GDP from agriculture and allied activities
at constant prices and Acredit = Credit for agriculture and allied activities deflated by GDP deflator with one
year lag.
10
Other studies of this type include Ghosh (2010) , Pavaskar, et al. (2011).
12
(2010) then suggest that institutional credit is indispensable for these regions with low input and
investment in agriculture. Binswangerand Khandker (1992) point out that institutional growth and
higher lending volumes lead to modest increases in aggregate crop output but sharp increases in
the use of fertilizers and in investments in physical capital and, substantial reductions in
agricultural employment. They conclude on that basis that expansion of credit has, therefore, led to
the substitution of capital for agricultural labor.
These two studies emphasize the multiple pathways in which formal agricultural credit
impacts production and this is well recognized by now (see Sriram 2007, for example). If one is to
understand this linkage in all its complexity, one needs a detailed construct of these relationships.
4.
Conceptualization of the Role of Formal Credit
The fundamental attribute of credit implies that it serves as an intermediate input and does
not directly enter as an input into agricultural production. It is therefore an enabling input. On
account of this, it plays a complex role in farmers’ production decisions, unlike physical inputs that
have a more transparent relationship with the levels of output.
The impact of agricultural credit on agricultural production, efficiency and productivity
could potentially occur through multiple channels. A simple conceptualization identifies three
pathways through which formal credit can influence outcomes (Figure 6). First, formal credit can be
used to purchase inputs over the cropping season, enabling a farmer to maximize the yield from the
cultivated area, given a level of capital stock. This channel represents a direct and within-season
impact on production. Second, formal credit can be used to make investments in irrigation facilities,
machines and draught animals that represent the use of credit for building up capital stock to
support agricultural production. This second channel typically impacts production with a time lag.
Both of these represent a “liquidity effect” (Binswanger and Khandkher, 1992) since they relieve a
farmer’s credit constraint and enables purchase of critical inputs to support production. Third,
formal credit is often used to replace informal credit associated with high interest burden.
Anecdotal evidence suggests that farmers often borrow from formal sources to pay off high interest
loans taken from money lenders. This has the effect of relieving credit constraints, reducing the
interest burden and indebtedness. Existing economic literature on wealth effects and risk aversion
suggests that this often enables farmers to make decisions that increase profitability and efficiency.
Even when formal credit is diverted to consumption, there could be an implicit wealth effect that
impacts farmer’s production decisions. This last channel, which incorporates a “consumption
smoothing” effect is is often difficult to capture.
Collectively, formal agricultural credit can be regarded as having two kinds of impacts –
first, it could enable a farmer to move to the production frontier so that given prevalent technology,
a farmer is using levels of inputs that enable him/her to produce at the frontier, from among many
feasible combinations of crops. Second, it could enable a farmer to move on to a superior
production frontier, so that given a level of inputs, the farmer is able to produce more of one or
more of the crops. The first is represented as a move from within the production possibility set to
the frontier (constituting efficiency improvement) and the second is represented as a shift of the
frontier itself (constituting productivity improvement). The impact of formal agricultural credit on
agricultural output conflates these two aspects of productivity and efficiency effects.
13
Figure 6 : Schematic Representation of Pathways
“Consumption Smoothing
Effect” Replace usurious loans,
relax consumption constraints,
etc.increasing the risk-bearing
capacity of farmers.
Direct Credit from
formal institutions
“Liquidity Effect” 1:
Working Capital (for purchase
of inputs)
Agricultural
Production
“Liquidity Effect” 2:
Investment credit (for
purchase of `capital stock’ to
support production)
5.
Empirical Strategy
a. The Challenges
Empirically, these effects are difficult to entangle. While a separation of these effects and
pathways are ideally studied at the household level, this logic can be extended to an aggregate level
by choosing empirical counterparts that represent these dimensions at the state or district level,
with important caveats. Aggregation often masks a lot of the heterogeneity and complexity of the
ways in which formal agricultural mediates production processes. The distribution of credit among
farmers or farmer groups is often uneven and is not taken into account when in an aggregate
analysis. Similar problems occur with aggregating over all the crops and commodities, which
masks the differential impacts and relative importance of credit. While this study is cognizant of
these issues, data limitations allow only an aggregate-level analysis.
Several methodological and data challenges persist in estimating the impact of formal
agricultural credit on output, especially at the aggregate level. Firstly, informal credit which forms a
major source of credit is something of a black box with virtualy no data available on its quantum or
how it is used. The fungibility of credit too poses difficult problems for research since it makes
short and long term credit indistinguishable at the farmers’ end. Similarly there could be spillovers
into non-farm sector that are unknown to the lenders and to researchers. Direct and Indirect
14
finance might also not be watertight categories so that it could be the case that direct credit to the
farmer is in fact used for ancillary activities that support agriculture. All of these make it hard to pin
down the precise nature of relationship between credit and agriculture.Further, the dynamic effects
are difficult to capture since credit flow in a particular year might yield cumulative benefits over
several years. This is particularly difficult to model. The other major challenges stem from data
constraints at the state level. Existing data for variables of interest are not often available for all the
states and for all the years, forcing us to confine the analysis to the states for which we have
complete data for the period of focus.
The empirical challenges of studying the relationship between formal agricultural credit
and output at an aggregate level are best described by Binswanger and Khandker (1992). The first
problem is the joint determination of both observed formal credit to agriculture and aggregate
output. The second problem emanates from the absence of data on informal credit, which makes it
difficult to capture the impacts of formal credit that might work through reduced informal
borrowing, and not factoring this might yield the estimates that reflect the true effect of formal
credit. Credit advanced by formal lending agencies such as banks is an outcome of both the supply
of and demand for formal credit. The amount of formal credit available to the farmer, his credit
ration, enters into his decision to make investments, and to finance and use variable inputs such as
fertilizer and labor. The third econometric problem arises because formal agriculture lending is not
exogenously given or randomly distributed across space. Ways to be able to address some of these
issues are central concerns of this project.
b. Methodology
To parse this complex relationship given the limitations of data, a combination of three
approaches are used (Box 1). The first is a simple model that regresses agricultural GDP on credit
flow using state level data. This is a catchall approach that
Box 1: The Three Approaches
cannot comment on the pathways or provide a causal
interpretation. The second method estimates a hybrid profitMethod 1: The Simple Model using
production function that regresses agricultural GDP on a
state level data and dividing into time
vector of relevant inputs, prices and agricultural credit for
periods in nominal terms as well as
the same year. This is a direct approach to estimating the
accounting for prices.
relationship between credit and agricultural GDP in reduced
Method 2: The Direct Approach that
form. The possible endogeneity of credit is addressed by the
regresses agricultural GDP on various
use of a control function approach where a regression
inputs
(fertilizers,
tractors
pumpsets), prices, rainfall, public
function is estimated that identified and then “controls” for
expenditure on agricultural, including
the endogenous component of observed credit flow (Imbens
credit flow and the estimated
and Wooldridge, 2007). The coefficient on credit in this case
endogenous component of credit or
captures one dimension of impact of credit that is not
the “control” variable.
mediated through inputs. The third method is perhaps the
Method 3: The Pathways Approach
most comprehensive and models the pathways approach in
that works on three stages – credit
what is referred to as a mediation analysis framework,
market, input demand functions and
where inputs are regarded as mediating the relationship
value of GDP function, estimated in a
between institutional credit and agricultural GDP (Preacher
SURE framework for panels and
and Hayes, 2008). Here, a set of regressions estimates input
incorporating the control variable.
15
demand as a function of credit, among other things and controlling for endogeneity of credit
(indirect effect), and the hybrid production-profit function as a function of inputs and credit,
recognizing that credit can also have direct effect on GDP. The coefficients representing the
responsiveness of input use to institutional credit are therefore used as components to construct
the total impact of agricultural credit on agricultural GDP (Preacher and Hayes, 2008). The impact
of credit on agricultural output is thus derived as the sum of the contribution of credit to the use of
specific inputs, capital or the cropping pattern, weighted by the contribution of these to the total
value of agricultural production. These are estimated in a Seemingly Unrelated Regression
Equations (SURE) framework that acknowledges the potential interrelationship between these
variables and the fact that they might be jointly determined. Appendix 1 contains a representation
of the models estimated.
For all three methods we use state-level data to estimate the relevant parameters of interest
for India as a whole. The analysis pertains to the time period 1995-96 to 2011-12, for which the
data is complete. Further, we also perform an analysis for two sub-periods pre-doubling (1995-96
to 2003-04) and post doubling (2004-05 to 2011-12). In all the methods, we make the assumption
of constant elasticity of demand, which is in fact a non-trivial assumption, but one that is typical of
studies of this kind.
As mentioned earlier, the chief methodological challenge involves dealing with the the issue
of endogeneity of observed credit. There are several approaches to deal with this. One approach to
tackle the endogeneity of observed volumes of credit is to use the predicted supply of credit at the
state level, following Binswanger and Khandker (1992) or to use lagged credit that is correlated
with current year credit. Each of these involves a set of defensible assumptions. The latter approach
however creates problems because there could typically be lagged response of agricultural GDP
which renders lagged credit an inappropriate instrument. In this study we use a control function
approach to separate out the explained exogenous variation in the credit flow to agriculture from
the unexplained and possibly endogenous component of credit flow and use the predicted residuals
from the control function to control for endogeneity in themain set of regression equations (Imbens
and Wooldridge, 2007). Appendix 1 provides more details on the approach and the regressions
estimated. The standard errors for both models 2 and 3 are bootstrapped to account for the use of
predicted variables as explantory variables.
c. Data Sources
To implement this method, we use a data set that is more detailed than used in the literature till
date. For all the major states in India details on credit, agricultural GDP, composition of the value of
output in the agriculture and allied sector and variables relating to land under cultivation provide
the key variables of interest. Data on physical quantities of Nitrogen, Phosphorus and Potassium
fertilizers have been assembled as also pesticides (technical grade) as also tractors and pumpsets
energized. Use of certified seeds in only available at the national level and is only used in explain
agricultural GDP but not as a separate input since this cannot be done at the state level. Other state
level variables representing the level of development include per capita State Domestic Product,
percentage of villages electrified, the number of commercial bank branches. Prices are typically
available at the all-India level, for the various inputs, power and fuel as well as output (food grains,
16
etc.). State level wage rates are compiled and in the absence of annual data on labour inputs used,
wage rates are expected to proxy labour use. We are also able to account for labour, machine and
animal power intensity per hectare from Cost of Cultivation data at the state level. These are
computed within state as the weighted average across crops (with weights being the area under
different crops), and across states as the weighted average across states (with weights being the
state’s share of gross cropped area). While for labour and animal, we use hours per hectare,
machine use data are in value terms. Appendix 2 provides details of the data used for the analysis
and the sources. While data is not available for all the states for all the years, only those states and
years for which all data was available are used in the analysis. Essentially, the data then consists of
time series data for the major states so that the panel data framework is used to estimate the
impact of credit on agricultural output at the national level, with state fixed effects. The models are
estimated for the major agricultural states, since data is not complete for all the states.
d. Scope and Limitations of the Study
The scope of this effort will be limited to estimating the impact of formal credit from different
institutions – cooperatives, rural and commercial banks – on agricultural output. The spillover
effects of formal credit on the rural non-farm sector will not be addressed specifically, an issue that
research suggests might be quite important (Pande and Burgess, 2005). Neither does this work
address the implications of recent interventions in credit policy such as debt waiver; this is already
studied elsewhere (Kanz, 2012; Cole, 2009). Another important area that is beyond the remit of this
study is the fiscal implication of the system of disbursing formal rural credit. One could argue that
to gauge the true impact of credit, one would have to account for the fiscal burden (or some notion
of net benefit cost ratio) (Binswanger and Khandkerm 1992). In this work, the question of interest
is to gauge whether or not direct formal rural credit impacts agricultural output, the extent to
which it does so and the relative importance of the different pathways through which these effects
occur.
6.
a.
The Results
The Productivity of Credit: Credit Elasticity of Agricultural GDP
The range of estimates obtained from the various methods suggest that the credit elasticity
of agricultural GDP for the entire period 1995-96 to 2011-12 is 0.21, i.e. a 10% increase in
institutional credit flow to agriculture in current prices is associated with a 2.1% increase in
agricultural GDP that year expressed in current prices (Table 1). This model controls for prices and
hence account for inflation.
Compared with these results in the simple model (method 1), the estimated credit elasticity
is 0.04 when the model controls for the use of inputs and a vector of input and output prices and for
the possible endogeneity of credit through a control function approach (method 2). The structural
model incorporating the pathways through which credit influences agricultural GDP (method 3)
yields estimates of credit elasticity of 0.0.2 . But neither method indicates that these coefficients are
statistically significant (Table 1).
The results from a period-wise disaggregate analysis is less conclusive. While the simple
model suggests that the elasticity continues to be statistically significant but has weakened in the
17
post-doubling period, the other two approaches, one that controls for prices and inputs and the
other the captures the pathways suggest that the relationship between credit and agricultural GDP
may have declined, but none of the estimated credit elasticity coefficients are statistically
significant and hence on cannot reject the null that the responsiveness of agricultural GDP to credit
has been zero.
At the state level, estimates of credit elasticity of agricultural GDP from the `simple’ model,
the only feasible option given the data, vary mostly between 0.05 and 0.7 with several states show
statistically insignificant elasticities (Table 2). Further, at the state level, the time trend of elasticity
estimates varies across states. In some states the relationship appears to have strengthened post
doubling (for example, in Tamil Nadu, Maharashtra and Gujarat) whereas for several others it has
weakened (including for Himachal Pradesh, Rajasthan, Uttar Pradesh, Karnataka, Kerala,
Chhattisgarh Madhya Pradesh, etc.). Punjab appears to show a consistently strong relationship
between agricultural GDP and credit.Notwithstanding these varitions, a strinking feature in the
relationship between agricultural GDP and credit flow is the pronounced convergence in the
agricultural GDP-credit flow ratio suggesting that perhaps the marginal returns to credit might be
equalizing across states (Figure 7).
Further clarity and insight can only be obtained through detailed case studies or primary
surveys, owing to the paucity of state level data that precludes modeling efforts at the state
level.This underscores the potential problems with aggregation and that observations on trends
cannot be generalized.
Table 1: Summary Results of the three models
Method 2 (Direct
Approach using
Control Function
methods.)
Method 3
(Pathways
approach using
Control Function
for credit)
(2)
(3)
(4)
0.214***
0. .036
0.210
Pre-doubling
0.266***
-0.010
0.102
Post-doubling
0.099***
0.138
-0.030
Time period for which
elasticity of GDP with
respect to credit is
computed.
(1)
The Entire Period
Method 1(Simple
Model Controlling
for Prices)
Notes:
(1) The Hausman Test suggests that the fixed effects model is appropriate. For Model 1, the Hausman chi-sqaured
(1)=15.51***
(2) Granger Causality tests indicate that agricultural credit Granger-causes agricultural GDP and not the other way.
(3) The Chow test for Method 1 indicates that the post-doubling coefficient is not statistically significantly different
from that from the pre-doubling period.
(4) Detailed results are available with the authors and can be provided on request. See also Appendix 3 and 4.
(5) Standard errors for the control function approach are bootstrapped 200 times.
18
(6) The states included in this regression are Andhra Pradesh, Bihar, Chhattisgarh, Gujarat, Haryana, Himachal
Pradesh, Jharkhand, Karnataka, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar
Pradesh, Uttarakhand, West Bengal. For new states, data since their inception are included.
Table 2: State-wise Credit Elasticity of Agricultural GDP under the Lag Model
Census
Code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
State
Jammu and Kashmir
Himachal Pradesh
Punjab
Chandigarh
Uttaranchal
Haryana
Delhi
Rajasthan
Uttar Pradesh
Bihar
Sikkim
Arunachal Pradesh
Nagaland
Manipur
Mizoram
Tripura
Meghalaya
Assam
West Bengal
Jharkhand
Orissa
Chhattisgarh
Madhya Pradesh
Gujarat
Daman and Diu
Dadra Nagar Haveli
Maharashtra
Andhra Pradesh
Karnataka
Goa
Lakshadweep
Kerala
Tamil Nadu
Pondicherry
Andaman and Nicobar
Islands
Whole
Period
0.053*
0.303**
0.340***
0.047
0.019
0.406***
0.029
0.171
0.288***
-0.062
0.092***
0.035
-0.022
-0.048*
0.072*
0.013
-0.000
-0.006
0.015
0.255**
0.321***
0.376*
0.490***
0.574***
0.016
0.448***
0.278*
-0.095
0.325***
0.355***
0.187**
-0.006
Predoubling
0.151
0.232**
0.522**
0.016
-0.103***
1.900***
-0.024
0.853***
-0.960**
-0.427***
0.029
0.044
-0.217
-0.073**
0.092
-0.147
-0.048*
-0.018
-0.866***
0.252
0.421
0.205***
0.555*
0.826
-0.007
1.046*
1.591***
-0.300
0.395***
0.493
0.069
0.040*
Post
Doubling
0.002
0.112
0.129**
-0.011
-0.086
-0.440
-0.109
0.060
0.040
0.574*
-0.030
-0.004
-0.023
-0.061
-0.119
-0.001
0.065***
-0.005
0.005
-0.420
0.028
-0.054
-0.049
0.567***
0.209*
0.143
0.039
0.017
-0.033
0.298***
0.133
-0.070
19
Notes: The state level elasticities are the slope coefficient from a regression of agricultural GDP on
credit flow to agriculture, controlling for wholesale price index. States have been arranged
according to census code.
Figure 7: State-wise ratio of Agricultural Gross Domestic Product and Credit Flow (19962011)
Notes: Only the major states have been included.
20
b.
Pathways of “productivity” : Input Demand and Credit Flow
If credit is an enabling or mediating input, its impact on output and productivity operates
through its influence on the level of purchased inputs, variable and fixed. A system of input demand
functions is estimated as a Seemingly Unrelated Regression Equations (SURE), with credit as one of
the explanatory variables along with the predicted residuals from the control function to account
for the endogeneity of credit (Table 3). The inputs included are fertilizers (a total of Nitrogen,
Phosphate and Potassic fertilizers), pesticides, tractors purchased and pumpsets energized
annually. The other inputs include labour and animal power intensity as well as expenditure per
hectare on machine use. Controls include other inputs like land, distinguished by type of irrigation,
prices of inputs, prices of food articles, lagged wages of unskilled labour, government expenditure,
lagged variable accounting for the structure of agriculture. Due to paucity of state level annual data,
detailed information on other equipments are not available for inclusion; tractors are therefore a
coarse proxy for equipment. So too with pumpsets, which represent one type of irrigation
investment. Investments in drip and sprinkler irrigation, etc. are hard to capture for lack of data.
The inclusion of government expenditure likely captures the subsidies offered for these irrigation
investments. These results need to be interpreted with caution. The results (Table 1) suggest that
over the entire period, institutional credit has a strong association with all inputs excepting
pumpsets energized. A 10 % increase in credit flow in nominal terms leads to an increase by 1.7%
in fertilizers (N, P, K) consumption in physical quantities, 5.1% increase in the tonnes of pesticides,
10.8% increase in tractor purchases. The credit elasticity of new pumpsets energized is however
not statistically significant.
Interestingly, there appears to be a marked shift in the pathways between the pre-doubling
and post-doubling phases. Whereas in the pre-doubling phase, institutional credit seems to have
been channelled into purchase of variable inputs such as fertilizers, in the post-doubling phase,
credit seems to be directed to investments in tractors. This is consistent with the popular
perception that high labour costs and a shortage of farm hands is prompting mechanization and it
appears that credit is aiding and enabling this transition. The absence of a strong relationship with
pumpsets could be on account of the variable representing irrigated land and perhaps government
expenditure, which might include subsidies for pumpsets. There might thus be a conflation of the
many explanatory variables.
It is apparent that availability of credit also reduces the labour intensity of agriculture by
2%. However there is no evidence that could be is consistent with the idea of labour substituting
mechanization. One possible interpretation is that increasingly some operations such as manual
weeding are being replaced by the use of chemical weedicides and so on. Likewise greater
ownership of tractors reflects this mechanization rather than just the paid out cost for machine use.
Alternatively it could be that mechanization as represented by the responsiveness of tractors to
credit flow substitutes animal power (rather than labour use).
21
Usually, this weak relationship especially of capital equipment such as tractor and pumpset
is strongly suggestive that mechanization is preserving productivity or agricultural growth rather
than enhancing it (Binswanger and Khandkher, 1992). In these contexts, credit can be interpreted
as performing two roles the preservation of productivity levels by supporting mechanization of
certain kinds and contributing to the growth of agricultural GDP through the purchase of variable
inputs. All these results collectively suggest that credit indeed appears to have played a role in
supporting the changing face of agriculture in India.
Overall, it seems quite clear that input use is sensitive to credit flow, whereas GDP of
agriculture is not. This seems to indicate that the ability of credit to engineer growth in agricultural
GDP is impeded by a problem of productivity and efficiency where the increase in input use and
adjustments in the pattern of input use are not (yet) translating into higher agricultural GDP. Credit
seems therefore to be an enabling input, but one whose effectiveness is undermined by low
technical efficiency and productivity.
Table 3: Input Demand System : The Credit Elasticity of Input Demand from a SURE Model
Input or Agricultural
GSDP
All
(1995/96 to 2011/12)
Phase 1
(1995/96 to 2003/04)
Pre-doubling
Fertilizers
0.17*
0.33**
0.06
Chemicals
0.51***
0.83
0.26
Tractors bought
1.08***
0.10
1.67***
-0.84
0.04
-0.83
Pumpsets Energized
Phase 2
(2004/05 to 2011/12)
Post-doubling
Labour hours per
-0.20**
-0.28
-0.16
hectare
Animal hours per
0.18
-0.07
-0.04
hectare
Machine use (Rs. Per
-0.67**
-1.13
-0.17
hectare)
Agricultural Gross State
0.083
-0.1
0.13
Domestic product
NOTES:
(1) This is estimated as a SURE (Seemingly Unrelated Regression Equation). Breusch-Pagan test of
independence: chi2(28) = 48.303, Pr = 0.0099 suggests that the null hypothesis of independence is rejected and
that these euqations need to be estimated as a system.
(2) The standard errors were bootstrapped with 200 repetitions to account for the inclusion of the predicted
variable from the control function.
(3) The regression was run in deviation form to allow the direct use of SUREG command in STATA 13.
(4) The coefficient of the control function variable is significant only in the case of pumpset suggesting that
endogeneity of credit is only true of pumpsets.
(5) The detailed regressions are available with the author. See Appendix 3 and 4
22
7.
Concluding Remarks
This report sought to investigate the relationship between institutional credit to agriculture
and agricultural Gross Domestic Product (GDP). Collectively, the results suggest that the fears that
credit might be ineffective are perhaps misplaced. There is strong evidence that credit is indeed
playing its part of supporting the purchase of inputs and perhaps even aiding the agricultural sector
respond to its contextual constraints.
The evidence of the impact of credit on agricultural GDP is however weak at best ,
irrespective of the approach used, assuming a constant credit elasticity of agricultural GDP.
Empirical patterns suggest that the relationship between credit and agricultural GDP is somewhat
weak in the post-doubling phase. Further, as is evident from the regression of agricultural GDP on
inputs and prices, other than fertilizers and labour, few inputs are strong drivers of GDP. In fact it
appears that the sectoral composition and output prices are important determinants of agricultural
GDP, apart from certain types of government expenditure and the irrigated area. Usually, this weak
relationship especially of capital equipment such as tractor and pumpset is strongly suggestive that
mechanization is preserving productivity or agricultural growth rather than enhancing it. In these
contexts, credit can be interpreted as performing two roles the preservation of productivity levels
by supporting mechanization of certain kinds and contributing to the growth of agricultural GDP
through the purchase of variable inputs. All these results collectively suggest that the success of
credit in enabling the increase in use of purchased inputs and effecting changes in input mix,
supporting the changing face of agriculture in India has not translated fully into agricultural GDP
growth as such.
REFERENCES
Bhalla, G.S. andGurmail Singh (2010) Growth of Indian Agriculture: A District Level Study, Planning
Commission, Government of India. Available athttp://planningcommission.nic.in/reports/sereport/ser/ser_gia2604.pdf
Binswanger, Hans.P. and Shahidur Khandker (1992): ‘The Impact of Formal Finance on Rural
Economy of India’, World Bank, Working Paper No. 949. (also appeared in The Journal of
Development Studies Volume 32, Issue 2, 1995)
Burgess, Robin and RohiniPande (2005) Do Rural Banks Matter? Evidence from the Indian Social
Banking Experiment, American Economic Review, American Economic Association, vol. 95(3), pages
780-795, June.
23
Chaudhary, Shilpa (2013) Trends in Total Factor Productivity in Indian Agriculture: State-level
Evidence using non-parametric Sequential Malmquist Index, Working Paper.
Chavan, P. (2009). How Rural is India’s Agricultural Credit. The Hindu.
Cole, S. (2009). Fixing market failures or fixing elections? Agricultural credit in India.American
Economic Journal: Applied Economics, 219-250.
Das, Abhiman, Manjusha Senapati, Joice John (2009): 'Impact of Agricultural Credit on Agriculture
Production: An Empirical Analysis in India', Reserve Bank of India Occasional Papers Vol. 30, No.2,
Monsoon 2009
De, Sankarand and SiddharthVij (2012): Are Banks Responsive to Exogenous Shocks in Credit
Demand? District – level Evidence from India, Research Paper, CAE, ISB, Hyderabad
Ghosh, Nilanjan (2010) Incredulity of Irresponsiveness: Is Agricultural Credit Productive?
Commodity Vision, Volume 4, Issue 1, July 2010 Takshashila Academia of Economic Research Ltd,
2010
Golait, R. (2007): Current Issues in Agriculture Credit in India: An Assessment, RBI Occasional
Papers, 28: 79-100.
Government of India (2013) Status of Indian Agriculture 2011-2012, Ministry of Agriculture,
Government of India.
Imbens, Guido and Jeffrey Wooldridge (2007) Control Function and Related Methods, What’s New
in Econometrics? Lecture Notes 6, National Bureau of Economic Research (NBER), Summer 2007.http://www.nber.org/WNE/lect_6_controlfuncs.pdf. Accessed July, 2013.
Kanz, M. (2012). What does debt relief do for development? Evidence from India's bailout program for
highly-indebted rural households.World Bank Policy Research Working Paper, 6258, Washington D.C.
Pavaskar Madhoo, Sarika Rachuri, Aditi Mehta (2011) Agricultural Credit Productivity in India
Commodity Vision Volume 4, Issue 5, March 2011 Takshashila Academia of Economic Research Ltd,
2011
Preacher, K.J. and Hayes, A.F. (2008).Asymptotic and resampling strategies for assessing and
comparing indirect effects in multiple mediator models.Behavioral Research Methods, 40, 879-891.
Rada, Nicholas E., 2013. Agricultural Growth in India: Examining the Post-Green Revolution
Transition 2013 Annual Meeting, August 4-6, 2013, Washington, D.C. 149547, Agricultural and
Applied Economics Association.
Ramakumar, R. and Chavan, P. (2007). Revival of agricultural credit in the 2000s: An Explanation.
Economic and Political Weekly, 57-63.
24
Sriram, M.S.(2007): ‘Productivity of Rural Credit: A Review of Issues and Some Recent Literature’,
Indian Institute of Management Ahmedabad, Working Paper No.2007-06-01.
Subbarao Duvvuri (2012) Agricultural Credit - Accomplishments and Challenges, Speech delivered
at NABARD, July 12, 2012.
25
APPENDIX 1: Empirical Strategy: The Methods Described
Method 1: Time Series Simple Model
The first method is a simple model, where agricultural GDP is regressed on the current time
period’s credit to agriculture. This model is estimated as a panel model with fixed effects, based on
Hausman test for choice of models. This model is also run separately for the Pre-doubling phase
(1995-96 to 2003-04) and post-doubling phase (2004-05 to 2010-11) and for individual states.
wherei refers to the state and t the financial year.
is the `lagged’ credit elasticity of agricultural GDP,
Method 2: Reduced Form Control Function Approach
Credit equation/ Control Function
The first step in this method is to address the endogeneity of credit. Since the demand for credit
itself could be a result of agricultural GDP, a control function approach is adopted to separate out
that part of credit that could be due to exogenous factors and that part which might represent the
endogenous component. IN this regression, we use variables that are hypothesized to exogenously
influence the level of credit. This includes the previous year’s rainfall, per capita income, structure
of agriculture and the number of branches of commercial banks in the state.
where
represents the total credit flow to agriculture (all sources and short and long term) and
. can be regarded as the endogenous part of credit, the estimated values of which are used in
the next stage regression.
Outcome function
The outcome function is essentially a hybrid production-profit function that maps a set of inputs to
outputs, controlling for exogenous factors such as the weather, market prices of output and inputs,
public infrastructure. Since the aggregate value of output is likely sensitive to the composition of
crops or the cropping pattern, the regression will control for the proportion of area under the major
groups of crops – foodgrains (cereals and pulses), oilseeds, fibre, horticulture, spices and plantation
crops (such as tea, rubber and coffee). In lieu of private and public capital stock and investment in
agriculture which capture capital inputs into production but are not available for all the states,
select machinery and equipment are included. A key component of this regression is the estimated
“control” variable from the Control Function described above that serves to control for endogeneity
26
and thereby allow us to interpret the coefficient on credit as a causal effect rather than mere
correlation.
where K is the credit flow, Z is the vector of inputs (including N, P K fertilizers, pesticides, tractors
and pumpsets) and other factors (O) such as rainfall, per capita state domestic product, percentage
of villages in the state that are electrified and so on, P is the vector of prices, etc.
is the credit elasticity of agricultural GDP, with associated bootstrapped standard errors for
hypothesis testing.
Method 3: Reduced Form Control Function Approach
(a) Credit equation/ Control Function
(b) Input / Capital Demand Equations
In order to retrieve the coefficients that represent the different pathways, we will estimate the set
of structural equations to understand the relative contribution of credit to different components of
the agricultural production-profit function.
The input demand functions depend on credit (among other things). We then estimate input
demand equations as a system, where the inputs are measured in physical units, and explanatory
variables include both the “control” variable and credit. The rapid changes in the cropping pattern
in India in the past two decades is both in response to the growing market opportunities as well as
the growth of processing sectors which in turn are likely impacted by indirect credit. So the
composition of the agricultural sector and the growing importance of livestock, poultry and
fisheries would need to be accounted for. Credit for purchase of milch animals as well as
construction of broiler sheds for contract growing are important components of agricultural output.
Due to paucity of detailed annual data on draught animals, share of livestock output in total
agricultural output is used as a proxy. The inputs used include fertilizers such as Nitrogen (N),
Phosphorous (P) and Potassium (K), pesiticides. Standard errors are computed through
bootstrapping procedures to account for the fact that these regressions use predicted values at
different stages.
27
(c) Outcome function
We then estimate the function explaining agricultural GDP in monetary terms as a hybrid profit
function. Compute the total impact of credit as the sum of the impacts on inputs weighted by the
impact of the input in question on agricultural GDP.
(d) Credit Elasticity of Agricultural GDP
The impact of credit on agricultural GDP can then be derived as the sum of the contribution of
credit to the use of specific inputs, capital or the cropping pattern, weighted by the contribution of
these to the total value of agricultural production. Standard errors reflect bootstrapped estimates.
28
APPENDIX 2: Data and Sources
Stata variable
name
Variable label (units)
Mean
Min
80.72
animalhoursha
Animal (hours/ha)
57.42
1.15
241.16
areanonfoodtotal
2905.47
1
52398.00
cagalliedtotal
Total area unde nonfood crops
(`000 hectares)
total cagallied
13384.02
541032.00
cladvecoservtotal
total cladvecoserv
64155.42
commercial
Number of commercial bank
branches
Production(in Lakhs)
4165.49
67499
18091
7
12404.66
1
321068.00
7835.36
0
72344.00
gca
No. of Fertilizer Sale Points
(Total)
Gross Cropped Area (`000 ha)
11606.41
2
195357.00
Ministry of Agriculture,
Agricultural Statistics at a
Glance
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture
gsdp_agri
GSDP(in Rs. Lakh)
3943
127000000.00
Ministry of Agriculture
irrcanal
Canal Irrigated area (`000 ha)
2943807.0
0
2178.10
0
17995.00
Ministry of Agriculture
irrtank
Tank Irrigated area (`000 ha)
312.70
0
3343.00
Ministry of Agriculture
k
175.45
0.02
3632.40
Fertiliser Association of India
labhoursha
Potassic fertilizers (`000
tonnes)
Labour (hours/ha) wtd avg
568.69
1691.17
machinersperha
Machine (Rs./ha)
1417.22
211.6
6
0
milk_prod
Production(in '000 MT)
2766.42
1
21031.00
n
Nitrogenous fertilizers (`000
tonnes)
Net Sown Area (`000 ha)
1092.02
0.6
17300.25
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture,
Agricultural Statistics at a
Glance
Fertiliser Association of India
8631.22
1
142960.00
Ministry of Agriculture
byproduct: Value of Output(Rs.
Lakhs)
Cereals: Value of Output(Rs.
Lakhs)
drugs: Value of Output(Rs.
Lakhs)
fibres: Value of Output(Rs.
Lakhs)
fish: Value of Output(Rs. Lakhs)
176028.30
0
3222156.00
National Accounts Statistics
810692.80
0
16300000.00
National Accounts Statistics
81793.53
0
1434137.00
National Accounts Statistics
94783.26
0
2756026.00
National Accounts Statistics
183344.90
19
3906563.00
National Accounts Statistics
fruits: Value of Output(Rs.
Lakhs)
607584.60
0
13800000.00
National Accounts Statistics
egg_prod
fert_salepoint_total
nsa
output_byproduct
output_cereals
output_drugs
output_fibres
output_fish
output_fruits
512.88
Source
Power tariff to agriculture
(paise/kWh)
agelectariff
0
Max
Central Electricity Authority
EPWRF time series
1881901.00
101261.00
6028.95
Ministry of Agriculture, Cost
of Cultivation Studies
Ministry of Agriculture
State Finances : A Study of
Budgets, RBI
State Finances : A Study of
Budgets, RBI
CMIE
29
Stata variable
name
Variable label (units)
Mean
Min
Max
Source
output_livestoc
k
output_oilseeds
livestock: Value of Output(Rs.
Lakhs)
924819.20
479
21200000.00
National Accounts Statistics
oilseeds: Value of Output(Rs.
Lakhs)
othercrops: Value of Output(Rs.
Lakhs)
257429.60
0
5449477.00
National Accounts Statistics
140840.20
0
3097888.00
National Accounts Statistics
Pulses: Value of Output(Rs.
Lakhs)
spices: Value of Output(Rs.
Lakhs)
sugar: Value of Output(Rs.
Lakhs)
Phosphatic fertilizers (`000
tonnes)
Per capita State Domestic
Product
Percentage of villages electrified
115964.70
0
2315486.00
National Accounts Statistics
79445.64
0
1838363.00
National Accounts Statistics
206510.30
0
3665781.00
National Accounts Statistics
429.72
0.38
8049.70
Fertiliser Association of India
144345.00
3037
47200000.00
National Accounts Statistics
85.16
0
285.87
Central Electricity Authority
output_othercr
ops
output_pulses
output_spices
output_sugar
p
pcsdpcurrent
pcvillageselectri
c
pesticide
EPWRF time series
2998.05
0.39
63651.00
Fertiliser Association of India
100.50
70.42
148.50
RBI
pfodder
Pesticide Technical Grade
Consumption (MT)
Price Index for Fertilisers and
Pesticides
Price Index for Fodder
114.77
63.17
245.64
RBI
pfoodarticles
Price Index for Food Articles
121.15
67.9
215.20
RBI
pfuelpower
Price Index for Fuel and Power
120.95
66.1
195.53
RBI
ptractor
Price Index for Tractors
101.75
65.54
144.21
RBI
pumpset
Pumpsets energized
835496.10
0
16300000.00
Central Electiricty Authority
ragrialliedtotal
total ragriallied
132352.00
415
4339185.00
rainavgann
Rainfall Annual (mm)
995.41
0
2630.00
State Finances : A Study of
Budgets, RBI
Ministry of Agriculture
seeds
132.78
62.2
283.85
Fertiliser Association of India
total
Certified seeds distributed (lakh
quintals)
Total Disbursment (in Rs. lakh)
824263.20
0
46700000.00
DEAR, NABARD
tractorsale
Sale of Tractors
28179.76
36
545109.00
Agricultural research data
book IASRI
tractorstock
Stock of Tractors (CMIE?)
312343.40
4629
4547080.00
CMIE
unskilledlabour
ers
villelectric
Wage Index for unskilled
labourers
85.46
32.33
393.82
RBI
Percentage of villages electrified
85.16
0
285.87
EPWRF time series
wpi_all_avg
Wholesale Price Index
(Financial Year averages)
Wholesale Price Index
(Calendar Year averages)
102.51
63.58
152.33
RBI
97.05
58.28
164.93
RBI
pfertpest
wpi_allcalavg
Observations with missing values are excluded from the regression
30
APPENDIX 3 : Regression Results for SURE Model 3.
Variables (in deviation, log form)
Fertilizers
Both
Phases
Phase 1
Phase 2
b/t
b/t
b/t
Credit
0.169
0.326*
0.058
(1.88)
(2.03)
(0.51)
"Error" from Control function
0.003
0.025
-0.103
(0.05)
(0.11)
(-0.86)
Net Sown Area
0.632*
0.441
0.255
(2.01)
(0.88)
(0.50)
Area under non-food Crops
0.000
0.000
0.000
(1.04)
(0.25)
(1.24)
Fertilizer Sale points
0.000
0.000
0.000
(0.73)
(1.30)
(0.27)
Index of Fertilizer/Pesticide prices
-0.002
0.036*
0.021
(-0.58)
(1.96)
(1.14)
Wholesale Price Index
-0.002
0.015
0.002
(-0.46)
(0.92)
(0.15)
Lagged Food Articles Price Index
0.006**
-0.074
-0.001
(2.74)
(-1.83)
(-0.13)
(LAGGED) SHARE OF TOTAL VALUE OF AGRICULTURAL OUTPUT
Fibres
0.003***
0.002
0.002*
(3.71)
(1.52)
(2.06)
Spices
-0.000
-0.001
0.000
(-0.09)
(-0.53)
(0.23)
Cereals
0.000
0.001
-0.000
(0.69)
(1.06)
(-0.27)
Fruits
-0.000
-0.000
0.000
(-0.04)
(-0.28)
(0.12)
Oilseeds
0.000
-0.000
0.000
(0.53)
(-0.23)
(0.49)
Pulses
0.001
-0.000
0.000
(0.72)
(-0.03)
(0.06)
Sugar
-0.000
0.000
-0.000
(-0.37)
(0.03)
(-0.01)
Per Capita State Domestic Product
-0.000
-0.000
0.000
(-0.32)
(-0.46)
(0.36)
Annual Average Rainfall
0.016
0.024
0.004
(0.36)
(0.26)
(0.06)
Constant
-0.182
2.855
-2.394
(-0.49)
(1.92)
(-1.38)
Pesticides
Both
Phases
b/t
0.510**
(2.69)
0.077
(0.46)
1.261
(1.39)
-0.000
(-0.36)
0.000
(0.15)
-0.011
(-0.80)
-0.026*
(-2.00)
0.016*
(2.32)
Phase 1
b/t
0.827
(1.54)
-0.035
(-0.05)
0.762
(0.57)
0.000
(1.51)
0.000
(0.93)
-0.006
(-0.09)
-0.079
(-1.17)
0.068
(0.45)
Phase 2
b/t
0.263
(1.30)
0.530
(1.58)
0.288
(0.22)
-0.000
(-1.50)
-0.000
(-0.50)
-0.069
(-1.47)
-0.008
(-0.28)
0.018
(0.80)
-0.002
(-0.73)
0.001
(0.21)
-0.001
(-0.67)
0.000
(0.03)
0.001
(0.42)
-0.000
(-0.17)
-0.001
(-0.53)
-0.000
(-0.07)
-0.165
(-1.09)
2.154*
-0.002
(-0.36)
0.008
(0.75)
-0.002
(-0.64)
-0.001
(-0.34)
0.003
(0.58)
-0.001
(-0.27)
0.001
(0.25)
0.000
(0.19)
-0.041
(-0.09)
1.419
-0.002
(-0.66)
-0.004
(-0.69)
0.001
(0.47)
0.000
(0.14)
-0.004
(-1.21)
-0.007
(-1.28)
-0.001
(-0.21)
-0.000
(-0.39)
0.078
(0.32)
6.199
(1.98)
(0.27)
(1.60)
b refers to coefficient and t refers to t value (in parenthesis)
31
APPENDIX 3 (Continued)
Variables (in deviation, log form)
Credit
"Error" from Control function
Net Sown Area
Area under non-food Crops
Fertilizer Sale points
Index of Fertilizer/Pesticide prices
Wholesale Price Index
Lagged Food Articles Price Index
Animal hours per ha
Both
Phases
Phase 1
Phase 2
Machine expenses per hectare
Both
Phases
Phase 1
Phase 2
Labour hours per hectare
Both
Phases
Phase 1
Phase 2
b/t
b/t
b/t
b/t
b/t
b/t
b/t
b/t
b/t
0.176
(0.83)
-0.070
-0.038
-0.671*
-1.125
-0.172
-0.200*
-0.282
-0.157
(-0.21)
(-0.09)
(-2.02)
(-1.34)
(-0.77)
(-2.36)
(-1.47)
(-1.07)
0.147
0.090
0.231
-0.633
-0.062
-0.297
-0.078
-0.052
0.036
(0.62)
(0.19)
(0.46)
(-0.94)
(-0.05)
(-1.09)
(-1.01)
(-0.21)
(0.24)
0.865
0.249
2.563
0.679
2.608
-0.259
0.252
0.442
-0.116
(1.46)
(0.26)
(1.45)
(0.72)
(1.06)
(-0.22)
(1.15)
(0.87)
(-0.18)
0.000
-0.000
0.000
-0.000
0.000
-0.000
0.000
-0.000
0.000
(0.82)
(-0.68)
(0.31)
(-0.05)
(0.47)
(-0.43)
(0.54)
(-0.70)
(0.53)
0.000
0.000
0.000
-0.000
-0.000
0.000
-0.000
-0.000
-0.000
(0.68)
(0.64)
(0.34)
(-0.65)
(-1.40)
(0.59)
(-1.67)
(-0.83)
(-1.11)
-0.020
-0.041
0.048
0.032
0.021
0.030
-0.007
-0.014
0.012
(-1.19)
(-0.79)
(0.58)
(1.82)
(0.20)
(0.74)
(-1.44)
(-0.53)
(0.46)
0.007
-0.020
0.045
0.056*
0.072
-0.005
0.011*
0.005
0.019
(0.52)
(-0.55)
(0.97)
(2.43)
(0.76)
(-0.25)
(2.03)
(0.31)
(1.18)
-0.007
0.053
-0.036
-0.030**
0.015
0.004
-0.001
0.019
-0.012
(-0.93)
(0.51)
(LAGGED) SHARE OF TOTAL VALUE OF AGRICULTURAL OUTPUT
(-1.12)
(-2.96)
(0.07)
(0.28)
(-0.30)
(0.41)
(-1.20)
Fibres
Spices
Cereals
Fruits
Oilseeds
Pulses
Sugar
0.001
-0.003
0.000
-0.003
-0.004
0.001
0.002*
-0.000
0.005***
(0.43)
(-0.75)
(0.01)
(-1.34)
(-0.47)
(0.38)
(2.17)
(-0.11)
(3.98)
0.001
-0.006
0.007
-0.013
-0.028
-0.002
-0.002
-0.004
0.001
(0.23)
(-1.05)
(0.63)
(-1.31)
(-1.20)
(-0.47)
(-1.27)
(-1.32)
(0.18)
0.003*
0.002
0.006
0.002
0.008
-0.002
0.001*
0.001
0.002*
(2.46)
(1.01)
(1.85)
(1.24)
(1.77)
(-1.25)
(2.19)
(0.92)
(2.13)
0.002
0.001
0.004
-0.002
0.001
-0.003
0.001
0.000
0.001
(1.31)
(0.45)
(0.98)
(-0.88)
(0.29)
(-1.12)
(1.15)
(0.35)
(0.75)
-0.000
0.002
-0.004
0.001
-0.000
0.002
0.001
0.001
0.000
(-0.08)
(0.75)
(-0.80)
(0.54)
(-0.01)
(0.75)
(0.85)
(0.67)
(0.19)
-0.004
-0.002
-0.009
0.006
0.009
0.003
-0.003***
-0.002
-0.003
(-1.26)
(-0.50)
(-1.15)
(1.79)
(1.12)
(0.69)
(-3.41)
(-1.01)
(-1.16)
0.003
-0.000
0.009
0.000
-0.004
-0.002
0.001
-0.000
0.002
(1.09)
(-0.02)
(1.81)
(0.15)
(-0.41)
(-0.67)
(1.33)
(-0.04)
(1.03)
32
Variables (in deviation, log form)
Per Capita State Domestic Product
Annual Average Rainfall
Constant
Animal hours per ha
Both
Phases
Phase 1
Phase 2
Machine expenses per hectare
Both
Phases
Phase 1
Phase 2
Labour hours per hectare
Both
Phases
Phase 1
Phase 2
b/t
-0.000
b/t
b/t
b/t
b/t
b/t
b/t
b/t
b/t
0.000
-0.000*
-0.000
-0.000
0.000
-0.000
-0.000
-0.000
(-1.21)
(0.75)
(-2.06)
(-0.46)
(-0.86)
(0.64)
(-0.98)
(-0.33)
(-0.75)
-0.153
-0.104
-0.039
-0.108
-0.518
-0.135
0.110
0.182
0.056
(-0.70)
(-0.49)
(-0.10)
(-0.49)
(-0.68)
(-0.84)
(1.41)
(1.02)
(0.57)
1.847
0.421
-6.334
-6.280**
-11.712
-2.720
-0.363
-1.401
-2.127
(1.33)
(0.12)
(-0.95)
(-3.23)
(-1.72)
(-0.84)
(-0.74)
(-0.87)
(-1.04)
b refers to coefficient and t refers to t-value (in parenthesis)
33
APPENDIX TABLE 3 (Continued)
Variables (in deviation, log form)
Credit
"Error" from Control function
Net Sown Area
Area under non-food Crops
Tractor sale
Both
Phases
Phase 1
b/t
b/t
1.079***
0.096
(3.32)
(0.18)
-0.025
0.347
(-0.07)
(0.45)
-0.094
0.297
(-0.13)
(0.20)
0.000
-0.000
(0.72)
(-0.62)
Phase 2
b/t
1.671**
(3.11)
0.040
(0.08)
0.562
(0.26)
0.000
(0.92)
Index of Fertilizer/Pesticide prices
Wholesale Price Index
-0.054**
-0.013
-0.000
(-2.85)
(-0.09)
(-0.00)
Lagged Food Articles Price Index
0.012
-0.129
0.040
(1.33)
(-0.70)
(0.60)
Tractor price index
-0.007
0.133
-0.109
(-0.33)
(0.92)
(-1.74)
Unskilled workers wage index
0.005
0.013
0.004
(1.02)
(0.89)
(0.31)
Fule and Power price index
0.004
0.010
-0.055
(1.56)
(0.94)
(-0.87)
(LAGGED) SHARE OF TOTAL VALUE OF AGRICULTURAL OUTPUT
Fibres
-0.002
-0.002
-0.004
(-0.59)
(-0.30)
(-0.59)
Spices
0.009
0.013
0.004
(1.51)
(1.03)
(0.32)
Cereals
-0.001
-0.002
-0.003
(-1.01)
(-0.69)
(-0.82)
Fruits
0.000
0.003
-0.003
(0.21)
(1.11)
(-0.72)
Oilseeds
0.001
-0.000
-0.003
(0.66)
(-0.10)
(-0.66)
Pulses
0.004
0.005
-0.000
(1.45)
(0.74)
(-0.02)
Sugar
-0.008*
-0.009
-0.012
(-2.54)
(-1.24)
(-1.87)
Per Capita State Domestic Product
0.000*
0.000
0.000
(2.02)
(0.08)
(1.18)
Annual Average Rainfall
0.226
0.195
-0.217
(1.21)
(0.39)
(-0.45)
Constant
4.835*
1.170
12.919*
(2.54)
(0.11)
(2.17)
Pumpsets
Both
Phases
b/t
-0.842
(-1.52)
-0.067
(-0.08)
0.778
(0.38)
-0.000
(-0.36)
0.048
(0.83)
0.007
(0.15)
-0.029
(-1.13)
Phase 1
b/t
0.038
(0.03)
-1.084
(-0.62)
-1.698
(-0.56)
-0.000
(-0.37)
0.456
(1.58)
-0.477
(-1.74)
0.310
(0.79)
Phase 2
b/t
-0.832
(-0.82)
-0.557
(-0.48)
5.586
(1.04)
-0.000
(-0.23)
0.009
(0.05)
0.041
(0.18)
-0.040
(-0.31)
-0.002
(-0.16)
-0.001
(-0.16)
0.025
(0.76)
-0.076
(-1.91)
-0.010
(-0.37)
0.007
(0.09)
-0.002
(-0.44)
-0.009
(-0.51)
-0.001
(-0.23)
-0.003
(-0.76)
-0.001
(-0.32)
-0.008
(-1.17)
0.015*
(2.39)
0.000**
(3.03)
-0.393
(-0.95)
-1.789
0.004
(0.40)
-0.003
(-0.14)
0.001
(0.16)
-0.007
(-0.84)
-0.002
(-0.44)
-0.002
(-0.20)
0.015
(1.18)
0.000*
(2.03)
0.529
(0.63)
-20.078
-0.005
(-0.38)
-0.016
(-0.48)
-0.003
(-0.47)
-0.000
(-0.00)
0.001
(0.05)
-0.005
(-0.22)
0.017
(1.31)
0.000
(1.32)
-0.484
(-0.64)
-1.878
(-0.48)
(-1.19)
(-0.12)
34
APPENDIX 4: Results for Agricultural GDP as part of SURE Model 3
Variable (in log devation form)
Both Phases
b/t
Phase 1
b/t
Phase 2
b/t
Credit
0.083
(1.00)
-0.028
(-0.42)
0.000
(0.52)
0.000
(1.43)
0.193
(0.92)
0.010
(0.41)
0.011
(0.36)
-0.000
(-0.63)
0.010
(0.31)
0.492**
(2.82)
-0.073
(-1.32)
-0.210
(-1.16)
0.265**
(3.12)
-0.040
(-1.20)
-0.001
(-0.02)
0.001
(0.96)
-0.016
(-0.33)
0.020**
(3.28)
0.002
(1.42)
-0.006
(-0.74)
-0.003**
(-2.69)
-0.095
(-0.39)
-0.130
(-0.37)
0.000
(0.95)
0.000
(0.47)
0.186
(0.44)
0.044
(0.65)
0.002
(0.03)
-0.000
(-0.07)
0.032
(0.32)
0.745*
(2.42)
-0.005
(-0.04)
0.033
(0.07)
0.216
(0.99)
-0.043
(-0.32)
0.014
(0.14)
-0.000
(-0.09)
-0.032
(-0.24)
0.071
(0.99)
0.003
(1.23)
-0.019
(-0.56)
-0.012
(-0.77)
0.133
(0.77)
-0.042
(-0.37)
-0.000
(-0.18)
0.000
(0.10)
-0.124
(-0.18)
-0.017
(-0.21)
0.030
(0.32)
0.000
(0.11)
-0.045
(-0.43)
0.184
(0.66)
-0.101
(-0.69)
-0.074
(-0.18)
0.350
(1.95)
-0.017
(-0.26)
-0.003
(-0.02)
0.002
(0.49)
-0.025
(-0.26)
0.161
(0.58)
-0.019
(-0.52)
0.011
(0.24)
-0.011
(-0.58)
"Error" from Control Function
Milk production
Egg production
Gross cropped area
Land irrigated by canals
Land irrigated by tanks
Area under Non-food crops
Pesticide
Fertilizers (N, P and K)
Tractors
Pumpsets
Labour hours per hectare
Aimal hours per hectare
Machine (Rs. Per hectare)
Lagged unskilled wage index
Average Annual Rainfall
Lagged Fertilizer and Pesticide price index
Lagged Fodder price index
Lagged tractor price index
Lagged Fuel and poer price index
35
Variable (in log devation form)
Both Phases
b/t
Phase 1
b/t
Phase 2
b/t
Wholesale Price Index
-0.006
(-0.89)
-0.015
(-0.55)
-0.028
(-0.50)
dlncladvecoservtotal
-0.009
(-0.80)
0.000
(0.06)
-0.019
(-0.64)
-0.001
(-0.14)
-0.005
(-0.25)
0.001
(0.06)
N
R-squared
BIC
0.076
(0.87)
-0.000
(-0.03)
-0.105
(-0.80)
121.000
0.739
603.766
-0.052
(-0.31)
0.008
(0.78)
0.000
(.)
59.000
0.468
411.019
0.056
(0.49)
-0.002
(-0.34)
-0.488
(-0.37)
62.000
0.685
324.758
AIC
170.418
91.079
-4.948
Capital expenditure on Agricultre and allied services
Revenue expenditure on agriculture and allied
services
Percetage of villages electrified
Constant
b refers to coefficient and t refers to t-value (in parenthesis)
36
37
doc_404597537.pdf