SHOUMITRO CHATTERJEE*

Princeton University
DEVESH KAPUR†
University of Pennsylvania

Six Puzzles in Indian Agriculture‡

ABSTRACT A quarter century after India’s historic shift to a more market-oriented

economy, the agriculture sector continues to be racked by severe policy distortions,
despite its importance for the livelihoods of the majority of India’s population and
in addressing major long-term challenges facing the country, from food security to
natural resource sustainability. The paper highlights some key empirical realities and
data anomalies to raise some fundamental questions about Indian agriculture and
anchors the analysis in six major puzzles: prices, procurement, political economy,
trade, productivity, and exit. While the list is neither exhaustive nor original in and
of itself, these puzzles raise important policy issues. We conclude by arguing that
more unified frameworks of analysis are needed for effective policy in a much
neglected but crucial sector for securing India’s future.

Keywords: Agriculture, Market Imperfection, International Trade, Agriculture

and Environment, India, Land Ownership, Political Processes
JEL Classification: D43, D45, O1, Q11, Q12, Q13, Q18

1. Introduction

A quarter century after India’s historic shift to a more market-oriented

economy, with industrial delicensing, trade liberalization, and (more
limited) reforms in factor markets, one sector continues to be plagued
by severely intrusive government regulations in both factor and product
markets, an arbitrary policy and regulatory environment, and low public
investment. Unfortunately, that sector—agriculture—not only accounts
for the livelihoods of the majority of India’s population but also is critical
to multiple long-term challenges facing the country, from food security to
natural resource sustainability, especially soil and water.
* [email protected]
† [email protected]
‡ The authors would like to thank Amarsingh Gawande and Beeban Rai for their excellent

research assistance.

185
186 I N D I A P O L I C Y F O R U M , 2016–17

The challenges facing Indian agriculture and its tens of millions of farmers
have been well recognized, whether by the media attention and hand wring-
ing on farmer suicides, the reports of the National Commission on Farmers
(led by M. S. Swaminathan) or official government documents, such as the
Economic Survey, 2016. Yet academic research has largely not kept pace with
the most pressing challenges facing Indian agriculture, barring certain excep-
tions.1 Given the increasing availability of computing power, novel geospatial
datasets and new quantitative tools in trade, industrial organization, economic
geography, and political science, this might be an opportune moment to focus
on some old and new puzzles in Indian agriculture—all of which have major
welfare implications for the tens of millions of Indian farmers.
The primary purpose of this paper is to raise some fundamental questions
about Indian agriculture by highlighting a number of empirical realities and
anomalies in the data. We discuss six major puzzles in Indian agriculture
related to prices, procurement, political economy, trade, productivity, and
the near-absence of exit. Our list is by no means exhaustive. Nor do we
claim to be the first ones to identify these issues, but these puzzles raise, we
argue, important policy issues, and they are all the more salient when viewed
through the lens of a unified framework. We argue that it is often simplistic
and naïve to provide isolated answers to these questions. For instance, as
we note below, there appears to be little change in the time trends of spatial
price variation in agriculture at both the retail as well as the wholesale level,
and this is despite substantial investments in rural infrastructure in recent
years, especially in rural roads and cell phone connectivity.
More broadly, our goal is to assemble some of the building blocks for
an internally consistent framework and, in doing so, also highlight the need
for more quantitative macro-research in agriculture. Much recent research
on agriculture in developing countries has a micro-focus with limited
generalizability. However, for policy, it is essential to be able to build a
macro-framework that not only enables the incorporation of key actors and
modeling their interaction with each other but also helps analyze meaningful
policy counterfactuals with general equilibrium effects. We discuss this in
greater detail in the concluding section.
We examine six key puzzles in detail below. First, we highlight high and
stable spatial price dispersion (of wholesale prices) in the decade 2005–2014.
Others have found the same in retail prices (Li 2016). This observation is

1. Examples include the following: Allen and Atkin 2016; Banerji, Gupta, and Meenakshi
(2012); Banerji and Meenakshi (2004, 2008); Fafchamps and Minten (2012); Goyal (2010);
Kapur and Krishnamurthy (2014a, 2014b); Krishnamurthy 2011; and Mitra et al. (2015).
 Shoumitro Chatterjee and Devesh Kapur 187

at odds with the law of one price, which would predict a decline in price
dispersion following massive investments in rural roads and communications
(via cell phone penetration) in this period. Second, we show that while in
principle the Minimum Support Price (MSP) policy is a national policy
meant to protect the incomes of all farmers, there is substantial variation (both
across regions and crops) in its implementation. Third, while farmers’ incomes
have languished, they have been unable to leverage their sheer numbers into
pressuring governments to undertake policies and programs that benefit them,
as one might expect in a democracy. Why has this been the case?
Fourth, India produces an excess of commodities and then exports them
even though they have a high factor content of its scarce natural resources—
water and land. This violates one of the central tenets of international trade
theory—the Heckscher–Ohlin theorem. What factors might explain this
puzzle? Fifth, there are large productivity gaps across geography, that is, the
productivity level that can be reached conditional on natural endowments.
Why? And why do these productivity gaps persist, that is, what prevents the
diffusion of efficiency/efficacy norms across space? Finally, why do farmers
not exit agriculture even when agricultural incomes remain abysmally low
and highly uncertain, when they could potentially earn a riskless stream of
income by selling their land, for instance?
We hope that the cataloguing of these puzzles in a unified manner will
help to structure and motivate future research on Indian agricultural markets,
policies, and their consequences.

2. The Price Variation Puzzle

Economic theory predicts that spatial price dispersion is a function of trade

costs (such as transport costs and information asymmetries) and mark-ups
charged by market intermediaries (the law of one price). Since the early
2000s, there have been large public investments in India in constructing high-
ways and rural roads. The Pradhan Mantri Gram Sadak Yojana (PMGSY),
launched in end-2000, aimed at providing all-weather road connectivity to
every rural habitation with a minimum population of 500 in the plains (250-
plus in hill states, tribal districts, and desert areas). Of a total of 178,184
habitations identified, two-thirds were connected by mid-2016 with nearly
half a million km of roads. In the case of telecommunications infrastructure,
the investments were made by the private sector. According to the Socio-
Economic and Caste Census 2011, two out of every three rural Indian house-
holds have access to cell phones (from negligible numbers in 2000), and
188 I N D I A P O L I C Y F O R U M , 2016–17

there have been active efforts by the central and state governments to inform
farmers of mandi prices near them (Tewari 2015). The twin investments in
core connectivity infrastructure—roads and cell phones—have resulted in
substantial reductions of both transport and information costs in India. Yet we
do not see evidence of spatial price convergence in agriculture commodities
during this time period. As we show below using high frequency data, the
variation in prices of commodities between agricultural markets or mandis
has been stable over the past decade and is consistent across different crops.
Fafchamps and Minten (2012) found no evidence of the impact of SMS-
based information on the prices received by farmers in a field experiment
in Maharashtra. Although Goyal (2010) finds mild effects of improvement
in farm gate prices for soybean farmers because of the introduction of elec-
tronic kiosks in villages of Madhya Pradesh (MP) in early 2000 by Indian
Tobacco Company (ITC), the aggregate effects are negligible since the
overall price variation has not gone down.
Before we look for possible answers, we examine the data in detail.
We focus on average monthly price data for eight, major, non-perishable
commodities—paddy, wheat, barley, maize, finger millet, pearl millet, sor-
ghum and soybean—sold in any mandi in India between 2005 and 2014. It
should be noted that the prices we analyze are wholesale prices observed
at Agricultural Produce Market Committee (APMC) mandis. It is very
likely that these are not prices received by farmers in general, since it is
large farmers who sell in mandis, whereas small farmers are more likely to
sell to local village intermediaries (Table 1). However, our price data has

TABLE 1. Percentage of Cereal Output by Farm Size Sold to Various Buyers

Paddy
Farm Size Local Private Mandi Government Input Dealers Processors
0–2 ha 55.44 20.19 11.17 8.72 1.62
2–5 ha 41.89 28.92 5.54 19.44 2.44
5–10 ha 29.58 34.77 6.52 27.46 0.51
>10 ha 14.15 50.43 3.76 15.38 0.65

Wheat
Farm Size Local Private Mandi Government Input Dealers Processors
0–2 ha 41.40 38.71 11.01 8.1 0.14
2–5 ha 25.23 49.97 5.02 19.42 0.24
5–10 ha 16.68 45.68 7.36 29.8 0.3
>10 ha 6.07 40.45 1.67 51.77 0.08
Source: NSS Situation Assessment Survey of Agricultural Households (2012).
 Shoumitro Chatterjee and Devesh Kapur 189

several advantages. They are actual prices recorded at a high frequency and
at a crucial stage in the supply chain: mandis are key points of aggregation.
Other sources of price data are usually recalled estimates of unit values,
geographically aggregated and at a very low frequency.
We use log (real) prices (with the consumer price index for agricul-
tural laborers [CPIAL; food] as the deflator) so that the variance of log
prices is unit-independent and compatible for pan-country comparisons.
The average standard deviation of log (real) prices across mandis in a
given month is 0.17. For comparative purposes, this is higher than the
Philippines: for rice and corn (the country’s principal home-grown food
commodities), Allen (2014) found the standard deviation to be 0.15, and
this in a country comprising a group of over a thousand islands, with high
transport and information costs.
As stated earlier, we do not observe any trend in time-series of the
standard deviation, which implies that the trade cost during this period
does not appear to have had a causal effect on price variation in grains across
India (Figure 1). There is some heterogeneity in the price dispersion across

FIGURE 1. Average Spatial Variation in Mandi Prices

0.2

10yr
Avg
(0.166)
Avg. Standard Deviation

0.15

0.1

0.05
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Year
Source: AgMarknet data.
Notes: Standard deviation computed for each month and each crop. The figure plots the yearly averages.
Commodities in data: Barley, pearl millet, finger millet, maize, paddy, soybean, sorghum, wheat.
190 I N D I A P O L I C Y F O R U M , 2016–17

TABLE 2. Price Variation between Mandis in Commodities

Avg Std Dev
Full Sample All commodities 0.17
Paddy 0.22
No Basmati/Superfine 0.21
Wheat 0.13
All other 0.17
Trimmed Sample All commodities 0.16
Paddy 0.21
Wheat 0.13
Soybean 0.07
Sorghum 0.28
All other 0.16
Source: Authors’ Calculations from AgMarknet data.
Note: The trimmed sample drops the top 5th percentile of the sample in terms of prices. Commodities in
data: Barley, pearl millet, finger millet, maize, paddy, soybean, sorghum, and wheat.

commodities. It is lowest for soybean, which is chiefly grown in MP as well

as parts of Maharashtra, and higher for other cereals (Table 2).
Some of the variation in price might be due to high-value varieties—for
example, basmati rice is more expensive than common rice. To control for
that, we drop the top 5 percentile of our data by crop in terms of prices. That
brings down the average annual variation only mildly. Most importantly,
however, even in the trimmed sample, we do not see any evidence of reduc-
tion in spatial price variation over time (Table 3).
The average standard deviation of log (real) prices across mandis within
states is also high. To the extent that high average standard deviation of log
(real) prices across mandis in the country might be due to different varieties
of wheat and rice grown in agro-ecological zones that vary across different
states, this finding attenuates the concern. High within-state variation pro-
vides additional evidence that the variation is not entirely due to quality.

TABLE 3. Spatial Price Variation in Commodities over Time

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Barley 0.11 0.14 0.11 0.12 0.16 0.16 0.16 0.11 0.11 0.11
Millets 0.19 0.16 0.17 0.18 0.20 0.17 0.17 0.16 0.18 0.20
Maize 0.17 0.17 0.15 0.14 0.16 0.13 0.14 0.15 0.16 0.16
Paddy 0.18 0.19 0.22 0.23 0.23 0.22 0.19 0.20 0.24 0.21
Soybean 0.06 0.06 0.06 0.06 0.09 0.05 0.06 0.09 0.06 0.09
Sorghum 0.25 0.25 0.27 0.25 0.27 0.31 0.38 0.29 0.24 0.32
Wheat 0.14 0.15 0.14 0.11 0.12 0.13 0.14 0.12 0.13 0.12
Source: Authors’ calculations from AgMarknet data.
 Shoumitro Chatterjee and Devesh Kapur 191

TABLE 4. Variation in Real Prices Across Mandis Within States in 2014

State Standard Deviation
Andhra Pradesh 0.15
Chhattisgarh 0.13
Gujarat 0.14
Haryana 0.13
Jharkhand 0.14
Karnataka 0.18
Kerala 0.17
Madhya Pradesh 0.21
Maharashtra 0.16
Odisha 0.70
Punjab 0.26
Rajasthan 0.14
Tamil Nadu 0.21
Uttar Pradesh 0.11
Source: Authors’ calculations from AgMarknet data.

We present the results for 2014 in Table 4. The results for previous years
are similar.
What is the relative weight of different factors in the variation in prices?
To get at this, we performed a Shapley–Shorrocks decomposition. This
procedure considers the various factors that together determine an indicator
(such as the overall variation in prices) and assigns to each factor an average
marginal contribution. This technique ensures that the decomposition is
always exact and that the factors are treated symmetrically. The results from
the Shapley–Shorrocks decomposition found that 37 percent of the variation
in log (real) prices is due to time-invariant district fixed effects (which in this
case could be soil quality effecting yields and hence prices), 20 percent is
due to location-invariant aggregate time shocks (like global demand), 4 per-
cent is due to differences in monthly rainfall across districts, and 39 percent
remains unexplained. The residual time- and location-varying effects could
be due to changes in connectivity, crop choice, or the expansion of welfare
schemes—such as the Public Distribution System (PDS) and National Rural
Employment Guarantee Act (NREGA)—impacting agriculture.
This analysis is revealing because, if the law of one price is valid, then
trade costs alone cannot create the binding constraint. The law of one price
states that the difference in prices at two locations is a function of trade
costs and market power. There is compelling evidence that the PMGSY
has reduced trade costs in rural India (Agarwal 2014), and highways have
done so between regions (Allen and Atkin 2016). Therefore, it must be
192 I N D I A P O L I C Y F O R U M , 2016–17

either that the market power of intermediaries has stayed constant over
the years and/or that it interacts with trade costs in ways that need better
understanding.
In ongoing work, Chatterjee (2017) explores the role of spatial compe-
tition in price variation between mandis. Using a large dataset on prices
in mandis and their geo-locations, the paper finds evidence that additional
markets create more competition and hence increase the prices received by
farmers. A causal border-discontinuity model, which exploits restrictions to
inter-state movement of goods to tease out the competition effect, estimates
price increases of 4 percent for an additional mandi in the neighborhood. The
limitation of this approach is that it is unable to capture any within-mandi
friction that might give rise to market power.2 While we have a good under-
standing of how trade costs impact prices, we know much less about the
sources of market power within a mandi. To that end, understanding the
mandi as an institution, and its internal political economy that gives market
power to traders vis-à-vis farmers, is the logical analytical next step. Then,
ideally, economists can model these forces and should be able to quantify
their effects on price dispersion.

3. The Procurement Puzzle

Two key government interventions that affect agriculture markets and

commodity prices in India are the MSP and procurement by government
agencies (mainly, but not solely, at mandi sites). The MSP sets a floor
price and thereby shapes the investment decisions of farmers. Its rationale
is to ensure that farmers are not compelled to sell their produce below the
support price due to either exploitation by large market players or a bumper
harvest. The MSP is effective mainly for four crops: wheat, paddy, cotton
(albeit only to a modest degree), and sugarcane (for which mills are legally
obligated to buy cane from farmers at government-fixed prices).
The puzzle is that although in principle the MSP procurement policy is a
national policy and should benefit all farmers regardless of their location or
crop choice, in practice, it varies (a) across states, (b) across districts within
states, and (c) across crops. Although we understand some of this variation,
a large part of it is still left unexplained. To highlight this puzzle, we first
consider an indirect measure: farmers’ awareness about MSP. The reason

2. See Krishnamurthy (2011) for an ethnographic study of the multiple ways in which
market power works within a mandi.
 Shoumitro Chatterjee and Devesh Kapur 193

FIGURE 2. Farmer Awareness about Minimum Support Prices 2012–13

Rice Wheat
100
% of farmers aware of MSP
40 6020
0 80
RAJ
GJ
JH
MH
JK
GA
AP
MP
TN
KA
WB
OR
HAR
KE
NG
CH
UP
PUN

OR
CH
GJ
JH
MH
KA
JK
WB
RAJ
BI
UP
MP
HAR
PUN
BI

Source: NSS Situation Assessment Survey of Agricultural Households Round 70.

for choosing this measure over actual procurement is that the quantum of
procurement is a choice of the farmer. If market prices are above MSP, then
even in the presence of efforts by public agencies, farmers may choose not
to sell since the MSP acts like an option. However, the farmers’ awareness
about MSP is more likely to reflect the presence of government agencies
in his neighborhood.
Figure 2 shows that most farmers are not even aware of the existence
of MSPs, and there is considerable variation across states. Whereas most
farmers in Punjab and Haryana are aware of MSP, very few are aware in
other states such as Gujarat, Maharashtra, Jharkhand, or West Bengal. This
is indicative of the absence of active government procurement efforts by
public agencies in many parts of the country.
It follows, therefore, that there are large disparities across states in actual
procurement. In Tables 5 and 6, not surprisingly, one observes that the states
where awareness of MSP is high are also the states where there is heavy pro-
curement of grains—both in absolute terms and relative to total production.
Therefore, awareness is highly correlated to the intensity of procurement in
a state (Figure 3). Notice also that as paddy is more intensely procured than
wheat (as a percentage of total production), the overall level of awareness
is higher for paddy than for wheat.
194 I N D I A P O L I C Y F O R U M , 2016–17

TABLE 5. Production and Procurement of Rice

Procurement by FCI % of Procurement
Production and State Agencies All India as a % of Total
(million tons) (million tons) Procurement Production
State 2013–14 2014–15 2013–14 2014–15 2013–15 2013–15
Andhra Pradesh 6.97 7.23 3.737 3.596 11.65 51.63
Bihar 5.51 6.36 0.942 1.614 4.06 21.55
Chhattisgarh 6.72 6.32 4.29 3.423 12.26 59.16
Gujarat 1.64 1.83 0 0 0.00 0.00
Haryana 4.00 4.01 2.406 2.015 7.03 55.23
Jharkhand 2.81 3.36 0 0.006 0.01 0.10
Karnataka 3.57 3.54 0 0.088 0.14 1.24
Kerala 0.51 0.56 0.359 0.374 1.16 68.42
Madhya Pradesh 2.84 3.63 1.045 0.807 2.94 28.62
Maharashtra 3.12 2.95 0.161 0.1988 0.57 5.93
Odisha 7.61 8.30 2.801 3.357 9.79 38.70
Punjab 11.27 11.11 8.106 7.786 25.26 71.03
Rajasthan 0.31 0.37 0 0 0.00 0.00
Tamil Nadu 5.35 5.73 0.684 1.051 2.76 15.66
Telangana 5.75 4.44 4.353 3.504 12.49 77.06
Uttar Pradesh 14.64 12.17 1.127 1.698 4.49 10.54
West Bengal 15.37 14.68 1.359 2.032 5.39 11.29
Source: Ministry of Agriculture and Farmers Welfare, Government of India.

TABLE 6. Production and Procurement of Wheat

Procurement as
Procurement by FCI % of All a Percentage
Production and State Agencies India of Total
(lakh tons) (lakh tons) Procurement Production
State 2013–14 2014–15 2013–14 2014–15 2013–15 2013–15
Andhra Pradesh 0.04 0 0 0 0.00 0.00
Bihar 47.38 39.87 0 0 0.00 0.00
Chhattisgarh 1.34 1.35 0 0 0.00 0.00
Gujarat 46.94 30.59 0 0 0.00 0.00
Haryana 118.00 103.54 58.73 6.50 23.29 55.8
Jharkhand 3.70 3.30 0 0 0.00 0.00
Karnataka 2.10 2.61 0 0 0.00 0.00
Madhya Pradesh 129.37 171.04 63.55 70.94 25.33 44.8
Maharashtra 16.02 13.08 0 0 0.00 0.00
Odisha 0.01 0.006 0 0 0.00 0.00
Punjab 176.20 150.50 108.97 116.41 42.45 69.0
Rajasthan 86.63 98.24 12.70 21.59 6.46 18.5
Telangana 0 0.07 0 0 0.00 0.00
Uttar Pradesh 298.91 224.17 6.82 6.28 2.47 2.5
Source: Ministry of Agriculture and Farmers Welfare, Government of India.
FIGURE 3. Farmer Awareness versus Procurement

PUN

100

100
PUN

80
80
UP

CH

60
60
HAR
KE MP

OR HAR

40
40
WB
KA TN

% of farmers aware of MSP in rice

UP

% of farmers aware of MSP in wheat

MP AP BI
BI

20

20
WB RAJ
UK
MH KA
JH MH
GJ JH
HP GJ

0
0

0 .2 .4 .6 .8 0 .2 .4 .6 .8
Procurement as a fraction of production Procurement as a fraction of production
Source: NSS–SAS (2012).
Note: Size of the bubble represents state shares in national output. Acronyms refer to state names.
196 I N D I A P O L I C Y F O R U M , 2016–17

While there has been considerable discussion on procurement of food

grains by public agencies for the PDS, the key point that is often missed is
that government procurement is a luxury for most farmers in the country.
There are large differences not only across states but within them as well.
And this variation in procurement has substantial consequences on the price
that farmers receive and the crops they choose to produce.
The disparity in procurement within states can be seen in the maps in
Figures 4 and 5. The figures plot, for each district, paddy and wheat pro-
cured as a percentage of total production for that district. Although there is

FIGURE 4. Average Fraction of Paddy Production Procured 2005–14

Sources: Food Corporation of India and Ministry of Agriculture and Farmers Welfare, Government of India.
Note: This map is not to scale and may not depict authentic boundaries.
 Shoumitro Chatterjee and Devesh Kapur 197

FIGURE 5. Average Fraction of Wheat Production Procured 2005–14

Sources: Food Corporation of India and Ministry of Agriculture and Farmers Welfare, Government of India.
Note: This map is not to scale and may not depict authentic boundaries.

uniformly high procurement across districts in Punjab and Haryana, there is

large variation across districts in UP, MP, Maharashtra, and Odisha.
What explains the variation in procurement across districts? Potential
explanations (all of which require careful research) include the following:

1. In districts where procurement is low, market prices are above MSP

such that it is not in the farmers’ interest to sell to the government.
2. Output per person is relatively low in districts where procurement
is low, indicating that low procurement reflects low marketable
198 I N D I A P O L I C Y F O R U M , 2016–17

surplus. In other words, a larger fraction of production is used for

self-consumption.
3. Variations in procurement could also reflect variations in the nodal
procurement agencies. While the Food Corporation of India (FCI) is
the principal storage and distribution agency for food grains, a sub-
stantial portion of the actual physical procurement is done by state
agencies on behalf of the FCI. All procurement agencies—whether
the FCI or state agencies—face resource constraints. However, since
they must procure a target amount of grains for the PDS, they are
likely to deploy limited resources efficiently by mobilizing them in
spatially concentrated, high-productivity districts.
4. Procurement of grains can also be used by politicians to selectively
channel resources toward their constituencies (or away from their
rivals’ constituencies) and thereby reward or penalize sections of
the electorate. For instance, if the local politician is from the ruling
party at the Center, then he can potentially influence the FCI on the
location and number of purchase centers or, if he is from the ruling
party in the state, the state procurement agencies.

The first and second explanations are in line with government-mandated

policies. The last two, however, can have important welfare and policy
implications. If we regress fraction of output procured in each district on log
yields and log population, we indeed find support for (2) and (3) (Table 7).
Districts with larger populations have relatively lower procurement, and
districts with higher productivity have relatively greater procurement.
Nonetheless, these regressions can only explain about 30 percent of the
cross-sectional variation in procurement, which leaves much scope for
future research.

TABLE 7. Regression Table

Dep var: Fraction of Output Procured
Log yield 28.40 32.22
(2.68)*** (5.58)***
Log population –16.69 –20.78
(1.40)*** (2.76)***
N 445 188
R2 0.32 0.35
Crop Rice Wheat
Source: District population is from 2011 Census of India.
Notes: All observations are district averages between 2011 and 2014. Crops include paddy and wheat.
* p < 0.1, ** p < 0.05, *** p < 0.01.
 Shoumitro Chatterjee and Devesh Kapur 199

4. Political Economy Puzzles

Historically, India has been seen as a rural society, with farming the dominant
occupation and economic mainstay for the vast majority of the population.
In the 1950s, agriculture accounted for more than half of India’s gross
domestic product (GDP), and cultivators accounted for half of the workforce.
From that position, agriculture’s share of GDP had declined by two-thirds
(to about 16 percent) by 2011, while the percentage of cultivators in the
workforce had declined from nearly 50 percent in 1951 to 24 percent in
2011. Nonetheless, the absolute numbers are very large (118.6 million as
per the 2011 Census), and if we add agricultural laborers who account for
30 percent of the labor force (144 million in 2011 up from 27.3 million and
19.5 percent of the workforce in 1951), the two still constitute more than
half of India’s labor force.
Furthermore, as per National Sample Survey (NSS) data from the 70th
Round, during the agricultural year July 2012–June 2013, rural India had
an estimated total of 90.2 million agricultural households, which constituted
about 57.8 percent of all rural households and 37 percent of all households.
In a democracy, these massive numbers suggest that farmers should be a
potent political force. But their economic travails suggest otherwise. Why
have Indian farmers not leveraged their sheer numbers into pressuring
governments to undertake policies and programs that benefit them more?
An obvious counter to this question is that farmers do enjoy substantial
political influence. Many of India’s agriculture policies are in place precisely
because of farmers’ political clout. Subsidies on inputs—whether electricity,
water, fertilizer, or credit—or on outputs (especially price supports) are in
place precisely because farmers matter politically. While there are many
questions on the incidence and distributional impact of these subsidies, the
reality that the vast majority of Indian farmers have low and volatile incomes
is not in doubt. And yet, the reality is that famers’ movements were much
more politically salient in the 1970s and 1980s than in the last two decades.
In the early decades following India’s independence, the Indian state
enjoyed substantial autonomy from powerful rural interests, which, while
they could thwart rural programs (such as land reforms or community devel-
opment programs), had weak impact on agriculture policies and programs.
This changed dramatically after the Green Revolution. Initially there were
fears that the Green Revolution would “turn red,” based on apprehensions
that the gains would accrue disproportionally to large farmers, widen-
ing the gap with marginal farmers and landless laborers (Frankel 1978).
Instead, as the Green Revolution started empowering the landed peasantry
200 I N D I A P O L I C Y F O R U M , 2016–17

(owner-cultivators—the so-called “bullock capitalists”), the focus shifted

to rural farmer-led “demand groups” that became an important political
constituency shaping public spending on agriculture subsidies and pitting
rural India against urban India (Rudolph and Rudolph 1987). The “Bharat
versus India” tension was especially manifest in pressures for higher MSPs,
since these benefited producers (farmers) at the expense of consumers (or
the public exchequer, if consumer subsidies rose). This drove farmers’
movements to demand higher support prices (Varshney 1995), but farm-
ers’ incomes continued to be dismal, as we note below. This is even more
puzzling since the 1990s saw marked growth in regional parties dominated
by farmers. Why have farmers been unable to translate their large numbers
into more pro-farmer policies and programs?
An obvious explanation of the failure of collective action by farmers
goes back to Olson (1965), who famously argued that large and dispersed
groups are hard to organize anywhere. And, given the fragmented nature of
Indian society, they are more susceptible to cross-cutting religious, caste,
and ethnic/linguistic appeals. But two developments of the 1990s appear to
have further weakened collective action by farmers. The sharpening of caste-
based political cleavages since the late 1980s (following the implementation
of reservations under the Mandal Report in 1990) may have caused the glue
of common economic interests to come unstuck. A second possibility is the
73rd Amendment to the Indian Constitution, which made statutory provisions
for Panchayati Raj as the third level of administration in villages. This was
a much-needed effort at administrative decentralization. But by mandating
elections at the base of rural society (villages), it appears to have shifted
political energies inward—in the form of electoral competition within
villages to capture increasing resources coming to villages—and thereby
diluted efforts to project collective farming interests outward and upward.
These factors, however, only explain the demand side of the story. They
do not explain why politicians have been willing to spend vast resources
and efforts on programs aimed at rural India, for which there was pressure
from civil society groups, but little from rural voters (for example, NREGA),
while remaining unwilling to commit resources and efforts to improve
farmer productivity and incomes, whether through rural markets, agriculture
research and extension, or better credit and irrigation systems.
Under the Indian Constitution, agriculture is a state subject. One might,
therefore, expect considerable inter-state variation in agriculture policies
stemming from differences in the politics and political economies of states.
But, in fact, differences in the policy approaches of politicians across states
in agriculture—be it taxes, subsidies, or trade laws—are quite limited. Why?
 Shoumitro Chatterjee and Devesh Kapur 201

To take one example: reforms to the APMC Act. In 2003, recognizing

that the role of the APMCs and the state agriculture marketing boards
needed to change from market regulation to market development, which
required removing trade barriers and creating a common market, the Central
Government formulated a model APMC Act for adoption by the states.
The reality of the reforms carried out by the different states paint a
different picture. Maharashtra, for example, went back on the reforms soon
after they were announced.3 UP has not yet adopted any of the main features
of the model APMC Act except giving permission to some big players for
direct procurement of food grains (primarily wheat), on the condition that their
total procurement in a season should be greater than 50,000 tons. Crucially,
this notification is issued year to year, and no changes have been made
in the legislation, thereby ensuring little private investment (and possibly
substantial annual rent seeking). In other states, by putting in large up-front
license fees to set up new markets or insisting that traders outside the market
still pay market fees, the reforms have been effectively stymied. Even when
licenses are awarded, they are for selected crops. For example, in Karnataka,
of the 36 licenses awarded at the time of writing, all were either for cotton
or fruits and vegetables.4
A research agenda on the limited inter-state variation in agriculture
policies could investigate two possibilities. First, whether multiple
interventions by the Center on different issues have important structural
consequences for agriculture that limit what states can do. The National
Food Security Act, 2013, for instance, mandated legal entitlements for
existing food security programs at a fixed price for specific food grains.
Inevitably, this sets certain structural incentives for the states’ own pro-
curement efforts and thereby farmers’ crop choices, with cascading effects
on ecologically sustainable agriculture practices. A second explanation
might stem from certain common features of agriculture markets across
rural India, namely the layer of intermediaries constituting the trad-
ing-cum-moneylending classes. The persistence of interlinked markets in
credit and outputs enhances the bargaining power of these intermediaries
vis-à-vis farmers, as well as giving them more influence in local politics,
thereby making it easier to undermine reform efforts. Research into the
political economy of the structure of rents that accrue under the status quo,
and whether and why these might affect the incentives of governments to
implement reforms, would be valuable.

3. http://tinyurl.com/maharashtra-apmc-reform
4. Data based on interviews with Karnataka state government officials by authors.
202 I N D I A P O L I C Y F O R U M , 2016–17

FIGURE 6. Per Capita Availability of Renewable Freshwater Resources

70,000
4,000
cu meters per person (India; China)

60,000
cu meters per person (Brazil)
3,000

40,000 50,000
2,000

30,000
1,000

1960 1970 1980 1990 2000 2010

Year
India China Brazil
Source: FAOSTAT, http://faostat3.fao.org/

5. The Trade Puzzle: Exporting Scarce Resources

One of India’s scarcest natural resources is water. Figure 6 provides an

international comparison of availability of fresh water resources. The
scarcity of fresh water in India is much worse than even what is depicted
in the picture since Brazil and China use only 60 percent of their stock of
fresh water availability, while India uses a little over 90 percent. In an era
of globalization, as the Hecksher–Ohlin trade model would also predict,
India should produce relatively more of the commodities that use its scarce
natural resources less intensively and should export them. And it should
produce relatively less of commodities whose production requires intense
use of its scarce resources and should import them.
In aggregate, India’s production patterns are the converse. Hoekstra and
Mekonnen (2012) compute the water footprint (amount of water used for
production) of crop production around the world from 1996 to 2005. Not
only is India’s water footprint the highest (followed by those of China and
Brazil) but also, at 1,047 cu gm per year, it is thrice that of Brazil (329 cu
gm per year), despite India having fewer water resources than other countries
to begin with. Furthermore, India’s production technology and crop choice
amplifies the amount of water it uses for production. This can also be seen in
Figure 7. Further, globally among all provinces or states, the water footprint
FIGURE 7. Water Footprint of Agricultural Production

Agricultural water footprint

[mm/yr]

0–10

10–50

50–100

100–200

200–500

500–1,000

>1,000

Source: Hoekstra and Mekonnen (2012).

204 I N D I A P O L I C Y F O R U M , 2016–17

is the highest for UP, Maharashtra, Karnataka, Andhra Pradesh, and MP.
This, to some extent, reflects the large size of these provinces. Within India,
per 5 arc minute by 5 arc minute grid cell, the water footprint is the highest
in the Punjab–Haryana region, as can been seen in Figure 8.

FIGURE 8. Water Footprint of Agricultural Production in India

Source: Hoekstra and Mekonnen (2012).

Notes: The map reflects water footprint for every 5 arc minute by 5 arc minute surface of the earth. This
is roughly a cell of 100 sq km area.
This map is not to scale and may not depict authentic boundaries.
 Shoumitro Chatterjee and Devesh Kapur 205

To understand the trade patterns of water, Goswami and Nishad (2015)

estimate water content embedded in crops at the time of trade. This is dif-
ferent from water used in production, which is much higher and partially
recoverable. Water “embedded” in crops, on the other hand, is the water
content of each crop which, once the crop is exported, cannot be recovered.
In 2010, India exported about 25 cu km of water embedded in its agricultural
exports. This is equivalent to the demand of nearly 13 million people. India
was a net importer of water until the 1980s. With increasing food grain
exports, India has now become a net exporter of water—exporting about 1
percent of its total available water every year. The ratio of export to import
of virtual water is about 4 for India and 0.1 for China. Thus, China remains
a net importer of water.
What might have led to this trading pattern? India’s major agriculture
exports (by weight or value) are rice and wheat. India exported 11.9 million
tons of paddy and 2.9 million tons of wheat in the year 2014–15. In the
same year, India’s major agriculture import was 4.6 million tons of pulses
and 11.5 million tons of edible oils. This pattern has been consistent in the
last two decades, ever since the country became self-sufficient in rice and
wheat. Since the water content by weight of paddy and wheat is 15 percent
whereas that of pulses is 10.5 percent, India’s agriculture trade has become
a net exporter of water, even as domestic water supplies dwindle.5
Thus, India’s trade pattern reveals that it is massively over-producing
rice and under-producing pulses. The root cause is the distorted relative
prices of paddy and pulses, which in this case operates through the MSP.
Free (or underpriced) water for irrigation, coupled with a guaranteed MSP
for paddy and wheat, incentivizes farmers to overproduce these crops.
India’s water exports rapidly increased around 1990, coinciding with
sharp increases in the MSPs for paddy and wheat (Goswami and Nishad
2015). If prices were to reflect market demand and incorporate the social
externalities of water, India’s trade pattern in food commodities would
reflect the relative scarcities of its natural endowments. If we look at
China, for example, which is also a water-scarce country, major imports

5. The water footprint (water required for production) is much higher for pulses (5,354 m3
per ton) than paddy (~3,000 m3 per ton). But much of this water is recoverable. In addition,
water used for the production of pulses does not use flood irrigation (unlike rice) and comes
from closer to the top soil, thereby protecting the water table. The production of pulses also
helps fix nitrogen (thereby reducing the consumption of urea). More importantly, the water
content in the final grain is 10.5 percent of the weight in pulses, compared to 15 percent in
paddy and wheat.
206 I N D I A P O L I C Y F O R U M , 2016–17

are soybeans, cotton, meat, and cereal grains,6 and exports are vegetables,
fruits, and processed food.
The pernicious consequences of India’s MSP policy for water sus-
tainability and availability have been evident for a while, as water tables
decline across the country. Analytically, it is important to disentangle
the relative roles of MSP and free power and water on declining water
tables. In principle, this could be done by comparing regions which have
similar water policies but different procurement policies. However, this
would require controlling for the underlying bedrock system, which is
analytically difficult. Policies meant to help farmers in the short run, such
as the MSP for wheat and rice, are having deeply negative long-term con-
sequences for the same farmers, due to the shrinking availability of water.
One can argue that both farmers and politicians have high discount rates
and are myopic. Hence, they cannot internalize the devastating effects of
the MSP in the long run and consequently persist with the current MSP
policy. Furthermore, even if this were not the case, high switching costs
(from more water-intensive crops to less water-intensive crops) may make
them loath to advocate a change in MSP policies. But such an equilibrium
persisted for many years in the case of fuel subsidies, and it was eventu-
ally addressed by incremental increases in price over a protracted period
of time. Under what conditions might something similar be possible in
the case of MSP policies is unclear. For instance, what political economy
conditions would make it possible to implement annual increases in MSP
inversely proportional to the water intensity of the crop, for example, mod-
est increases in MSP for paddy and sugarcane, and more rapid increases
in MSP for pulses and oilseeds?

6. The Productivity Gap Puzzle

Geography, availability of natural resources, and endowment of fertile soil

make Punjab, Haryana, and the flood plains of the Indo-Gangetic Plain one
of the most fertile lands in the world. This can be easily seen in Figures 9
and 10. However, comparing the actual productivity of land in this belt to
its own potential raises several important questions.
The Food and Agriculture Organization’s (FAO’s) Global Agro-
Ecological Zones (GAEZ) dataset provides data on actual agricultural

6. See http://www.ers.usda.gov/media/1784488/eib136.pdf and http://www.fao.org/filead-

min/templates/est/meetings/wto_comm/Trade_policy_brief_China_final.pdf
FIGURE 9. Paddy Yields in 2013–14

US

Turkey

8,000
China

6,000
Punjab

Haryana
Brazil

Yield (kg/ha)
Tamil Nadu
Andhra PradeshBangladesh
West Bengal
Karnataka

4,000 World
India
Uttar Pradesh

Odisha
Chhattisgarh
Converted Rice yields for Indian States to Paddy yields by a factor of 0.67

2,000
0 10 20 30
% of World Production
Sources: FAO STAT for the World and Agricultural Statistics at a Glance, 2014, GoI, for India. All estimates for 2014–15.
Note: Size of the bubble is proportional to the quantity of total production in the state/country.
FIGURE 10. Wheat Yields in 2013–14

China

5,000
Punjab

Haryana

4,000
Rajasthan

Yield (kg/ha)
US
India
Bangladesh

3,000 World
Uttar Pradesh
Brazil
Gujarat

Madhya Pradesh
Bihar

2,000
0 5 10 15 20
% of World Production
Sources: FAO STAT for the World and Agricultural Statistics at a Glance, 2014, GoI, for India. All estimates for 2014–15.
Note: Size of the bubble is proportional to the quantity of total production in the state/country.
 Shoumitro Chatterjee and Devesh Kapur 209

productivity on any piece of land as well as the potential productivity that

could be realized if inputs were used optimally, conditional on the crop
that is being currently grown there. The data are available for every 5 arc
minute by 5 arc minute grid cell of the earth, roughly equal to 10 km by
10 km blocks.
Therefore, if a district in Punjab and another in eastern UP grow paddy,
we can estimate their current and potential productivity. Since the potential
productivity takes into account the crop choice and local agro-environmental
conditions, the gap between actual and potential productivity should only be
due to spatially specific distortions. In other words, in the absence of local
distortions, all districts should produce at their true potential and there should
be no difference between the actual realized and potential productivity.
From the data, however, it is clear that this is not the case. Table 8 notes
the fraction of all districts (in percent) whose productivity gap is less than
10 percent, 25 percent, 50 percent, and more than 80 percent for rice, wheat,
sorghum, and millet. The productivity gap is calculated as (1 – actual
productivity/potential productivity) × 100: if actual productivity was equal
to potential productivity, the gap would be zero. The larger the number for
the fraction of districts in the < 10 percent productivity gap column, the
more India’s districts are closer to their true potential for that particular crop.
There are about 27 percent paddy-producing districts and 39 percent
wheat-producing districts where the productivity gap is less than 25 per-
cent. Similarly, about 5 percent paddy-producing districts and 6 percent
wheat-producing districts have a productivity gap of more than 80 percent.
These are the districts with the potential for the largest gains because they
are so below their potential. Comparing across crops, whereas in paddy and
wheat about 18 percent and 32 percent, respectively, of all districts operate
within 10 percent of their potential, this number is starkly low for sorghum
(4.2 percent), a dryland crop.

TABLE 8. Distribution of the Agricultural Productivity Gap of Indian

Districts
<10% <25% <50% >80%
Rice 18.5% 27.3% 62.6% 5.1%
Wheat 31.9% 38.6% 80.9% 5.7%
Sorghum 4.2% 5.0% 6.7% 79.5%
Millet 30.5% 31.5% 34.4% 2.9%
Source: FAO Global Agro-Ecological Zones Database.
Notes: % of districts within × % of their potential.
Productivity gap = (1 – actual productivity/potential productivity)× 100.
210 I N D I A P O L I C Y F O R U M , 2016–17

What is probably happening is that Punjab, Haryana, and parts of UP are

operating close to their potential, and these are the regions that dominate
paddy and wheat production and have also been seen to optimize their pro-
ductivity potential. In other parts of the country, farms are operating far from
their productivity potential and there is substantial scope for improvement
through better resource allocation alone.
The evidence also strongly suggests that the use of inputs is distorted.
Table 9 provides evidence on inter-state variation in the use of nitrogen,
phosphorus, and potash per hectare. The proportions in which the fertilizers
are used are also skewed. A lower absolute consumption does not neces-
sarily imply that fertilizer use is sub-optimal. The proportions also need to
be balanced. While the correct proportions are context-specific, measuring
deviations from the accepted rule of thumb of N:P:K in the ratio of 4:2:1
is revealing.
Table 9 also presents deviations from the optimal N:P and N:K ratios
(with respect to N). It is clear that fertilizer use is disproportionately tilted
toward the use of urea (the principal source of N) as compared to other
fertilizers. These distortions—more in N:K than N:P, indicating extremely
low application of potash—are significantly higher as compared to those in
China, Bangladesh, Vietnam, and Thailand.
Most states use almost twice the amount of nitrogen as compared to
phosphorus than they should using the rule of thumb. These include the most
productive states such as Punjab, Haryana, UP, and Gujarat. The distortions
are less in Maharashtra, Karnataka, and Kerala, which consume close to
the optimal ratio, although this could simply be the result of different crop
mixes. The overuse of urea as compared to potash is even more perverse.
Relative to the accepted rule of thumb, Bangladesh uses only 4 percent more
urea than potash, whereas Indian states such as Punjab and Haryana use over
1,300 percent more urea as compared to potash.
Understanding the reason for such large distortions in fertilizer use is
important not only because of their impact on current productivity but
also because of their impact on future productivity due to their long-term
effects on soil health. While the structure of fertilizer subsidies might be the
obvious suspect, the fact is that fertilizer subsidies are a national policy, and
prices are broadly similar across the country. While variances in fertilizer
inputs may be due to varying agricultural production functions, we do not
have careful estimates to see if this is the case. It is also important to note
that the productivity gap across space is at least in part due to differences
in the intensity of use of inputs, but further work is needed to understand
why that is the case.
 Shoumitro Chatterjee and Devesh Kapur 211

T A B L E 9 . Fertilizer Consumption and Distortion in Use of Urea (N),

Phosphorus (P), and Potash (K), 2014
Nitrogen Phosphorous Potash (K) Total N:P N:K
(N)(kg/ha) (P)(kg/ha) (kg/ha) Distortion Distortion
Delhi 93.33 0 0 93.33
Nagaland 2.26 1.48 0.72 4.45 –47% –22%
Mizoram 30.31 2.37 2.99 35.67 1079% 153%
Meghalaya 10.35 3.07 0.8 14.22 137% 223%
Manipur 22.66 3.75 3.18 29.59 404% 78%
Himachal Pradesh 35.03 8.69 9.03 52.74 203% –3%
Assam 36.24 9.99 19.19 65.41 163% –53%
Rajasthan 37.97 11.51 0.21 49.69 130% 4420%
Jharkhand 65.72 13.79 2.94 82.45 277% 459%
Thailand 68.4 19.2 22.5 110.2 156% –24%
Uttarakhand 118.89 19.24 5.67 143.8 418% 424%
Tripura 28.73 21.27 11.73 61.73 –65% –39%
J&K 57.83 21.87 9.37 89.07 64% 54%
Odisha 63.05 23.71 11.37 98.13 66% 39%
Gujarat 88.52 24.09 6.92 119.52 167% 220%
The United States 74.1 24.8 26.7 125.6 99% –31%
World 69.5 26.3 17.9 113.7 64% –3%
Kerala 51.06 26.42 43.55 121.03 –7% –71%
Bihar 124.88 27.44 12.55 164.87 255% 149%
West Bengal 74.09 28.11 28.97 131.17 64% –36%
Madhya Pradesh 53.76 28.17 2.5 84.43 –9% 438%
Chhattisgarh 63.08 28.72 8.43 100.22 20% 87%
India 85.79 28.85 10.75 125.39 97% 100%
Uttar Pradesh 115.18 29.63 4.06 148.86 189% 609%
Haryana 146.49 30.58 2.41 179.48 279% 1420%
Tamil Nadu 93.27 33.47 27.03 153.76 79% –14%
Maharashtra 72.41 35.29 19.37 127.07 5% –7%
Karnataka 78.37 36.67 21.01 136.06 14% –7%
Punjab 172.55 41.14 3.04 216.73 219% 1319%
Brazil 42.6 48.8 56.1 147.5 –113% –81%
Bangladesh 140.7 53.9 33.9 228.5 61% 4%
Andhra Pradesh 153.55 54.32 18.85 226.72 83% 104%
Vietnam 127.5 66.7 43.1 237.3 –9% –26%
China 267.6 91 41.2 399.8 94% 62%
Source: Agricultural Statistics at a Glance, 2014, Ministry of Agriculture, Government of India.
Note: Distortion is relative to rule of thumb of using N:P:K in the ratios of 4:2:1.

7. The Exit Puzzle

Adamopoulos and Restuccia (2014) document that the typical farm size
in the 20 richest countries is 34 times greater than that in the 20 poorest
countries. The average farm size in rich countries is 54 hectares, while in
poor countries, it is only 1.6 hectares. The median farm size in India is
212 I N D I A P O L I C Y F O R U M , 2016–17

1.5 hectares, while the mean is much lower. Diego Restuccia, in a series of
papers with different co-authors, has estimated the gains of approximately
3 to 4 times in productivity if land is reallocated to the most productive
farmers in developing counties such as Malawi, Ethiopia, and China.7
A critical reason for the underlying misallocation of land is distortions in
land markets. Agricultural land in India is a peculiar asset. Its rental value
for farming purposes (a proxy for farming income that may be derived from
it) is a fraction of a risk-free stream of income that would accrue from its
sale (that is, its capital value), which in turn is a fraction of the price if it is
converted to non-agricultural land (for commercial use) if in proximity to
an urban area. Yet farmers are reluctant to sell their land. Understanding
the effects of these distortions in land markets on agriculture productivity
and the reasons why farmers seem reluctant to follow economic logic are
two fruitful areas for further research.
There are several potential reasons why agriculturalists may not be exiting
farming. First, there might be high opportunity costs of exiting. Figure 11
shows that this is not likely to be the case. Our analysis of NSS data suggests
that the average annual income from cultivation of the median farmer (net
of production costs) is less than `20,000 in 17 states. This includes produce
that farmers did not sell valued at local market prices. Given the large wedge
between retail and farm gate price, this might underestimate income, but it is
still very low. Some farmers who own cattle, for example, might substitute
this income with allied-agriculture income. Even then, a vast majority of
farmers live very close to the poverty line.
A second—and perhaps key—factor may be the paucity of other live-
lihood options. India’s failure to industrialize has meant that the classic
ladder of occupational mobility—from farms to factories, as has been the
case with China—has been foreclosed, leaving low-end service jobs in
severe urban environments or modest self-employment opportunities as the
only options facing someone who exits farming. With 72 percent of India’s
farms below 1 hectare, the investment required to increase productivity in
these tiny farms is unlikely to be readily forthcoming. Exit would allow for
land consolidation, and increases in farm size would, in turn, allow for more
investments and productivity. But that requires the sort of labor-intensive
manufacturing growth that has simply not been taking place. Research on
rural–urban linkages in factor and product markets is needed to improve our
understanding of its effects on farm productivity.

7. For a comprehensive survey, see Restuccia (2016).

 FIGURE 11.

Net income in thousand `

0 19.25 50 100
HIMACHAL PRADESH
DELHI
Source: Data from NSS SAS Round 70.

WEST BENGAL
UTTARAKHAND
KERALA
JHARKHAND
Income

India Avg

ORISSA
Land Holding

BIHAR
GOA
UTTAR PRADESH
SIKKIM
ANDHRA PRADESH
J&K
TAMIL NADU
RAJASTHAN
Net Annual Income for Indian Farmers in 2012–13

MANIPUR
TRIPURA
GUJARAT
MADHYA PRADESH
MIZORAM
NAGALAND
ASSAM
MAHARASHTRA
CHHATTISGARH
KARNATAKA
TELENGANA
MEGHALAYA
ARUNACHAL P
HARYANA
PUNJAB
0 .5 1 1.5
Land Holding (in ha)

Notes: Sample restricted to households surveyed in both Rabi and Kharif. Income only from cultivation net of costs. Unsold produce valued at local market rate.
214 I N D I A P O L I C Y F O R U M , 2016–17

8. Conclusion and Directions for Future Research

Sectoral policies need to draw on research that is generalizable and offers
feasible policy options. Even if the research is micro, it is imperative that
we uncover mechanisms that are more general, at least if the goal is to help
formulate national level policies. Although the random and quasi-random
experimental literature has been insightful, it has its limitations for policy
counterfactuals when formulating sector-level policies owing to exter-
nal validity, general equilibrium effects, and other statistical challenges
(Cartwright 2007; Deaton 2010; Deaton and Cartwright 2016; Monte,
Redding, and Rossi-Hansberg 2015; Redding and Rossi-Hansberg 2016).
Here we illustrate this point with specific examples.
The literature analyzing the effects of information provision has mostly
relied on randomization and quasi-randomization techniques and remains
inconclusive (Fafchamps and Minten 2012; Goyal 2010; Mitra et al. 2015).
First, it is unclear whether the literature suggests traders are already offering
competitive prices conditional on trade costs, or whether farmers are unable
to utilize this information to improve their bargaining position. Second, the
provision of price information seems to “work” in certain regions and not
in others, and the literature has not reconciled what underlying factors may
explain these differences. Third, we do not know what happens in general
equilibrium. For example, once farmers have more information, how do
traders respond? Do they exit because their margins shrink? If some traders
quit, does that eventually concentrate market power further? Therefore, we
do not know the macro-implications of information frictions in the Indian
agriculture sector. Consequently, we need studies like Allen (2014) for the
Philippines, who, through a more macro-trade approach, found that half of
the price dispersion in that country was due to information frictions. This
approach, however, requires time series and regionally disaggregated data
on (a) cellphone penetration and (b) time series data on efforts made by the
government to relay price information to farmers. We tried to obtain this
data for India from the Telecom Regulatory Authority of India but were told
that this data does not exist.
We know from micro-experiments that traders in some mandis of west-
ern UP collude (Banerji and Meenakshi 2004, 2008). However, we do not
know whether we are likely to find collusion even in other mandis and, if
so, whether this would necessarily require a similar set of economic factors.
This limits generalizability. Furthermore, even when there is collusion
within a mandi, we do not know if other mandis can be a source of com-
petition, and how spatial (distance) or other considerations (for example,
 Shoumitro Chatterjee and Devesh Kapur 215

transport) would affect the results. The latter channel would be particularly
important because that helps us better estimate the welfare effects of new
mandi construction. Again, this requires studying forces that have been
present for some time when the economy is in equilibrium. Introducing
random variation in micro-regions could lead us to false conclusions when
in reality, the local economy might just be in transition. Moreover, studying
spatial competition requires variation of market power in space, which is
inherently a macro-problem.
As a final example, consider the policy of making India “one market.”
This is an important and pivotal policy that the Central Government has
been trying to implement through its initiative on the National Agricultural
Market. However, the policy has faced severe road blocks from states. Since
agriculture is a state subject, implementation requires states to reform their
respective APMC Acts. However, for various reasons, states have not been
interested and instead appear intent on trying to protect the monopoly power
of the mandis. There are three points to note here. First, even though this is a
critical policy issue with huge potential benefits, there is dearth of academic
research (especially in economics).8 Second, understanding the issue requires
unearthing the political economy of the mandi structure in the internal politics
of the states. Third, general equilibrium counterfactuals that policy makers
would be interested in, like changes in economic welfare due to the removal
of inter-state trade barriers, require understanding the sources of monopoly
power of the mandis, how their market power would change once border
restrictions are removed and then a model to estimate welfare changes.
A notable exception in the literature is Allen and Atkin (2016). Using four
decades of micro-data for the entire country, they find that the creation of
highways integrated Indian districts and made farmer revenues more volatile
by breaking the negative correlation between local prices and productivity
shocks. Increase in farmers’ revenue volatility made them switch toward
crops with less risky yields. This shift in production amplified the gains
from integration for farmers. We can draw many lessons (without having to
worry about external validity and missing general equilibrium effects) from
such research. For example, greater market integration is likely to negatively
impact the production of riskier crops like pulses unless supported with
insurance policies. Many more studies of this type are needed to understand
the many puzzles and principal challenges facing Indian agriculture.

8. We could only find and cite news articles. As an example, see: http://indianexpress.
com/article/business/business-others/farm-marketing-reforms-not-in-name-alone-2884499/
(accessed June 15, 2017).
216 I N D I A P O L I C Y F O R U M , 2016–17

References
Adamopoulos, T. and D. Restuccia. 2014. “The Size Distribution of Farms and
International Productivity Differences,” The American Economic Review, 104(6):
1667–1697.
Agarwal, S. 2014. “Do Rural Roads Create Pathways out of Poverty? Evidence
from India,” Unpublished. Santa Cruz: University of California. http://cegadev.
org/assets/cega_events/61/3A_Program_Evaluation_in_Developing_Countries.
pdf (accessed May 1, 2017).
Allen, T. 2014. “Information Frictions in Trade,” Econometrica, 82(6): 2041–2083.
Allen, T. and D. Atkin. 2016. “Volatility and the Gains from Trade,” NBER Working
Paper 22276, Cambridge, MA: National Bureau of Economic Research, http://
www.nber.org/papers/w22276 (accessed May 1, 2017).
Banerji, A., N. Gupta, and J. V. Meenakshi. 2012. “Auctions in Grain Markets and
Farmer Welfare,” Economic and Political Weekly, 47(52): 65.
Banerji, A. and J. V. Meenakshi. 2004. “Buyer Collusion and Efficiency of
Government Intervention in Wheat Markets in Northern India: An Asymmetric
Structural Auctions Analysis,” American Journal of Agricultural Economics,
86(1): 236–253, https://doi.org/10.1111/j.0092-5853.2004.00575.x (accessed
May 1, 2017).
———. 2008. “Millers, Commission Agents and Collusion in Grain Markets:
Evidence from Basmati Auctions in North India,” The BE Journal of Economic
Analysis & Policy, 8(1): 1–25
Cartwright, N. 2007. “Are RCTs the Gold Standard?” BioSocieties, 2(1): 11–20.
Chatterjee, S. 2017. “The Tiller’s Travails: Borders, Misallocation and Welfare,”
Princeton University Dissertation (in progress).
Deaton, A. 2010. “Instruments, Randomization, and Learning about Development,”
Journal of Economic Literature, 48(2): 424–455.
Deaton, A. and N. Cartwright. 2016. “Understanding and Misunderstanding
Randomized Controlled Trials,” NBER Working Paper 22595, Cambridge, MA:
National Bureau of Economic Research, http://www.princeton.edu/~deaton/
downloads/Deaton_Cartwright_RCTs_with_ABSTRACT_August_25.pdf
(accessed May 17, 2017).
Fafchamps, M. and B. Minten. 2012. “Impact of SMS-based Agricultural Information
on Indian Farmers,” The World Bank Economic Review, 26(3): 383–414.
Frankel, F. R. 1978. India’s Political Economy, 1947–1977: The Gradual Revolution.
Princeton, NJ: Princeton University Press.
Goswami, P. and S. N. Nishad. 2015. “Virtual Water Trade and Time Scales for
Loss of Water Sustainability: A Comparative Regional Analysis,” Scientific
Reports, 5, Article No. 9306.
Goyal, A. 2010. “Information, Direct Access to Farmers, and Rural Market
Performance in Central India,” American Economic Journal: Applied Economics,
2(3): 22–45.
 Shoumitro Chatterjee and Devesh Kapur 217

Hoekstra, A. Y. and M. M. Mekonnen. 2012. “The Water Footprint of Humanity,”

Proceedings of the National Academy of Sciences, 109(9): 3232–3237.
Kapur, D. and M. Krishnamurthy. 2014a. “Market Towns and the Dynamics of
India’s Rural and Urban Transformations,” CASI Working Paper Series 14-02,
http://www.theigc.org/wp-content/uploads/2014/10/Kapur-Krishnamurthy-2014-
Working-Paper.pdf (accessed May 1, 2017).
———. 2014b. “Understanding Mandis: Market Towns and the Dynamics of India’s
Rural and Urban Transformations,” http://casi.sas.upenn.edu/sites/casi.sas.upenn.
edu/files/research/Understanding%20Mandis%20-%20D.%20Kapur,%20M.%20
Krishnamurthy.pdf (accessed May 1, 2017).
Krishnamurthy, M. 2011. “Harda Mandi: Experiencing Change in an Agricultural
Market in Central India (1980–2010),” PhD Dissertation, University College
London, http://discovery.ucl.ac.uk/1333230/ (accessed May 1, 2017).
Li, N. 2016. “Marketization Costs and Household Specialization: Evidence from
Indian Farmers,” Macro, Trade, and Finance Seminar Series, Washington, D.C.:
The World Bank.
Mitra, S., D. Mookherjee, M. Torero and S. Visaria. 2015. “Asymmetric Information
and Middleman Margins: An Experiment with Indian Potato Farmers,” SSRN
Scholarly Paper No. ID 2639972. Rochester, NY: Social Science Research
Network, https://papers.ssrn.com/abstract=2639972 (accessed May 1, 2017).
Monte, F., S. J. Redding, and E. Rossi-Hansberg. 2015. “Commuting, Migration and
Local Employment Elasticities,” NBER Working Paper 21706, Cambridge, MA:
National Bureau of Economic Research, http://www.nber.org/papers/w21706
(accessed May 2, 2017).
Olson, M. 1965. The Logic of Collective Action: Public Goods and the Theory of
Groups. Cambridge, MA: Harvard University Press.
Redding, S. J. and E. Rossi-Hansberg. 2016. “Quantitative Spatial Economics,”
NBER Working Paper 22655, Cambridge, MA: National Bureau of Economic
Research, http://www.nber.org/papers/w22655 (accessed May 20, 2017).
Restuccia, D. 2016. “Resource Allocation and Productivity in Agriculture,” Mimeo,
University of Toronto.
Rudolph, L. I. and S. H. Rudolph. 1987. In Pursuit of Lakshmi: The Political
Economy of the Indian State. Chicago: University of Chicago Press, http://www.
press.uchicago.edu/ucp/books/book/chicago/I/bo5972167.html (accessed May
20, 2017).
Tewari, R. 2015. “Socio Economic & Caste Census 2011: A Mobile in 2 of Every 3
Rural Homes, a Salaried Job in 1 of 10,” The Indian Express, New Delhi,
July 6, http://indianexpress.com/article/india/india-others/socio-economic-caste-
census-2011-a-mobile-in-2-of-every-3-rural-homes-a-salaried-job-in-1-of-10/
(accessed May 19, 2017).
Varshney, A. 1995. Democracy, Development, and the Countryside: Urban-Rural
Struggles in India. New York: Cambridge University Press.
 Comments and Discussion*

N. Chandrasekhara Rao
Institute of Economic Growth

Variations in agriculture prices impact the livelihoods of many people in

India, as in other countries (Swinnen and Riera 2013). Chatterjee and Kapur
employ an innovative methodology to understand these price swings. The
existing literature uses structural models that show how prices are determined
(see, for example, Gopakumar and Pandit 2014; Gupta 2013). This paper, for
the first time, uses high frequency price and quantity data (from the AgMarknet
portal of the Government of India), in contrast to the annual data used earlier.
This has not been attempted to my knowledge in India, except by Dasgupta,
Dubey, and Sathish (2011), and I hope others will follow. The paper tries to
develop a general theory of price formation in Indian agriculture and employs
advanced econometric tools to control for unobserved heterogeneity. The
significance of this paper is in the innovation in the conceptual framework
and methodology and the use of high frequency data.
This analysis brings forth some interesting results that can be of immense
significance, though some of them can be questioned on the basis of the
existing known or stylized facts. There are four key results. First, the vari-
ation in prices both within and among states is very high as compared to
that in other developing countries. Second, this substantial price variation
persists despite the adoption of new information and communication tech-
nologies that should have led to price signals converging during the years
2005 to 2014 being studied. This is a cause for concern. The authors have
tried to show how much of the variation is due to district-fixed effects and
that 39 percent of it is unexplained, which could be attributed to the time-
and location-varying factors.

* The paper to which these comments relate was originally presented at the IPF 2016

under the title, “Understanding Price Variation in Agricultural Commodities in India: MSP,
Government Procurement, and Agriculture Markets.” The version published in the IPF jour-
nal is a substantially rewritten paper based on discussant and editorial feedback and hence
also the change in title. To preserve the sense of the discussions at the India Policy Forum
(IPF), these discussants’ comments reflect the views expressed at the IPF and do not take into
account the revisions made in preparing the version published in this volume. The original
conference version of the paper is available on NCAER’s website on http://www.ncaer.org/
events/ipf-2016/IPF-2016-Paper-Chatterjee-Kapur.pdf.
 Shoumitro Chatterjee and Devesh Kapur 219

Third, a counter-intuitive result is that procurement increases the

price of paddy by 4 percent above minimum support price (MSP) while
it reduces the price of wheat by 2 percent below MSP. Fourth, the paper
finds that lack of marketing infrastructure reduces prices by as much as
5 percent. These could all be important findings against the backdrop of
the government’s intention to deepen the national agricultural market,
but as the authors themselves note, they are also puzzling and difficult to
understand. A possible answer might have to do with the much greater
international trade after the easing of quantitative restrictions, forcing the
Government of India since about 2006 to procure wheat at very high prices
in order to build stocks. This was a factor creating a wedge between the
MSP, cost of production, and therefore wholesale prices (Chand 2009;
Dev and Rao 2010).
I have a wish list of variables that would be useful to consider in the
modeling of agricultural price variations and the underlying conceptual
framework, not that this would be easy. First, the model could control
for the overall growth momentum of the economy, which several studies
have argued is percolating to the rural and agriculture sectors (Sharma and
Rodgers 2015; World Bank 2014). The NSS data also show the growth
of rural non-farm opportunities and the diversification of farm incomes.
Second, issues such as lagged incomes, production lags, prices of substi-
tute commodities, productivity differences, may all need to be examined
in explaining price formation. As the Economic Surveys have pointed out,
variation in taxes across states may also have an important bearing on prices,
particularly if they are acting as barriers to free trade.
Third, as soon as agricultural prices rise, the government’s first reaction
is to ban future markets (Chand 2016). Although the Abhijit Sen Committee
found that future markets have no significant impact on agricultural prices,
several studies, including Dasgupta, Dubey, and Sathish (2011), suggest that
there is a large impact. Fourth, while mandis do exist, many offer hardly
any physical or marketing facilities. Do price variations have anything to
do with this or with the level of regulation? Purohit (2016) recently came
up with an index of regulation on which states were ranked. These issues
should be looked at in explaining price variations.
States like Chhattisgarh are pursuing a “procurement plus” approach,
giving crop-specific bonuses beyond the MSP. What is its impact on price
variation and the prices farmers are getting? Market and monopsony power
may not merely arise across states but is also seen within states as farmers
are not allowed to sell in nearby mandis if they are in adjoining states or
have to incur a cost even if allowed to sell. As a recent Economic Survey
220 I N D I A P O L I C Y F O R U M , 2016–17

of the government points out, there are 2,200 agriculture markets in India;
Europe effectively has one market.
Examining the impact of private sector participation and its scaling up
on price variation would be very useful for policy. I would also suggest
the authors going beyond the exclusive focus on paddy and wheat and
considering price formation in other crops, such as vegetables and fruits.
Before ending, it is worth mentioning that the geographical concentration
of procurement centers is not a puzzle and can be explained by the manner
in which the Green Revolution unfolded in India.

References
Chand, Ramesh. 2009. “The Wheat Market: Distortions Caused by Government
Interventions,” Economic and Political Weekly, 44(12): 41–46.
_____. 2016. “e-Platform for National Agricultural Market,” Economic and Political
Weekly, 60(28): 15–18.
Dasgupta, Dipak, R. N. Dubey and R. Sathish. 2011. “Domestic Wheat Price
Formation and Food Inflation in India: International Prices, Domestic Drivers
(Stocks, Weather, Public Policy), and the Efficacy of Public Policy Interventions
in Wheat Markets,” Working Paper 2/2011-DEA, New Delhi: Department of
Economic Affairs, Minister of Finance, Government of India, https://mpra.
ub.uni-muenchen.de/31564/1/MPRA_paper_31564.pdf (accessed May 1, 2017).
Dev, S. Mahendra and N. Chandrasekhara Rao. 2010. “Agricultural Price Policy,
Farm Profitability and Food Security,” Economic and Political Weekly, 55 (26
and 27): 174–182.
Gopakumar, K. U. and Vishwanath Pandit. 2014. “Price Movements for Rice and
Wheat: A Structuralist Policy Perspective,” Working Paper 240, New Delhi:
Centre for Development Economics, Delhi School of Economics.
Gupta, Neha. 2013. “Government Intervention in Grain Markets in India: Rethinking
the Procurement Policy,” Working Paper No. 231. New Delhi: Centre for
Development Economics, Delhi School of Economics.
Purohit, Purnima. 2016. “Measurement of Regulations of the Agricultural Produce
Markets,” Economic and Political Weekly, 60(28): 36–45.
Sharma, Alakh N. and Gerry Rodgers. 2015. “Structural Change in Bihar’s Rural
Economy: Findings from a Longitudinal Study,” Economic and Political Weekly,
50(52): 45–53.
Swinnen, J. and O. Riera. 2013. “The Global Bio-economy,” Agricultural
Economics, 44(1s): 1–5.
World Bank. 2014. “Republic of India: Accelerating Agricultural Productivity
Growth,” Report No 88093 -IN, Washington, D. C.: Agriculture, Irrigation and
Natural Resources Sustainable Development, South Asia, World Bank, http://
hdl.handle.net/10986/18736 (accessed May 1, 2017).
 Shoumitro Chatterjee and Devesh Kapur 221

Elumalai Kannan*
Jawaharlal Nehru University

The paper by Shoumitro Chatterjee and Devesh Kapur analyzes the spa-
tial variation in the prices of wholesale rice and wheat at the first point of
farmer sale, the primary wholesale markets called mandis, and focuses on
two important government interventions, the minimum support price (MSP)
at which the government procures, and the regulation and physical loca-
tion of mandis and their influence on price formation. The paper uses high
frequency data available on the Government of India’s AgMarknet portal
which provides data on weekly and monthly prices and market arrivals by
commodity across all Agricultural Produce Market Committees (APMCs).
Since the authors mention that it is work in progress, I will suggest a few
important analytical issues to be considered while revising the paper.
Let me start with some data issues and problems in the descriptive
analysis of the impact of government procurement on prices and in seeing
APMCs that run the mandis as local monopsonies.
The authors have selected wholesale rice and wheat markets in 16 major
states for their analysis. It is not clear why they have selected these 16
states as not all of them produce rice and wheat, so that in some states there
would be limited or no government procurement operations. This could
be a factor in the contradictory regression analysis results they get for rice
and wheat. Some criteria should have been adopted in selecting states that
would adequately capture government interventions in procurement and
MSP, and thereafter help in analyzing the impact of market location on
wholesale prices. A combination of factors such as the level of production,
marketable surplus, and procurement should have been used as the key
factors in selecting states.
Why is the selection of states important for addressing the research
issues raised in this paper? In the case of rice, 10 states—Andhra Pradesh,
Punjab, Chhattisgarh, Odisha, Haryana, West Bengal, Uttar Pradesh, Bihar,
Tamil Nadu, and Karnataka—accounted for about 86 percent of India’s

* The paper to which these comments relate was originally presented at the IPF 2016
under the title, “Understanding Price Variation in Agricultural Commodities in India: MSP,
Government Procurement, and Agriculture Markets”. The version published in the IPF journal
is a substantially rewritten paper based on discussant and editorial feedback, and hence also
the change in title. To preserve the sense of the discussions at the India Policy Forum, these
discussants’ comments reflect the views expressed at the IPF and do not take into account
the revisions made in preparing the version published in this volume. The original confer-
ence version of the paper is available on NCAER’s website on http://www.ncaer.org/events/
ipf-2016/IPF-2016-Paper-Chatterjee-Kapur.pdf.
222 I N D I A P O L I C Y F O R U M , 2016–17

rice production (Government of India 2016). However, the first five states
taken together account for about 80 percent of rice procurement. This will
influence the overall results from the sample. Some states in the sample may
not even have participated in government rice procurement. Furthermore,
even in states where rice procurement takes place, the procurement is not
necessarily evenly distributed across all districts and mandis. This is because
marketable surplus and consumption patterns vary across regions and also
within states (Government of India 2016). In the case of wheat, 90 per cent
of the production comes from six states—Punjab, Madhya Pradesh, Haryana,
Uttar Pradesh, Rajasthan, and Bihar. The first three states account for 91
percent of procurement (Government of India 2016).
These figures imply that procurement taken as a dummy variable in the
regression model at the district level will have few ones and many zeroes.
This may be the reason why the regression model gives contradictory
results for rice and wheat. Local production patterns, marketable surplus,
and consumption will affect price formation at the mandi level. Production
may be high in some places but the entire quantity may not reach the
market because local consumption might be very high. For instance, West
Bengal is the highest rice-producing state, but its marketable surplus is low
(Government of India 2016), and hence the arrival of food grains in the
market is an important measure to be considered in thinking about what
influences local price formation.
The paper uses an interesting equation to analyze the differential effect
of government procurement on market prices. It hypothesizes that procure-
ment operations create a wedge between farm gate, mandi, and retail prices.
Procurement is a dummy variable, which shows the presence or absence
of procurement. Unfortunately, the regression model provides inconsistent
results; the coefficient of procurement is positive and significant for rice,
while it is negative and significant for wheat. Problems in the data and pos-
sible omitted variables may be responsible for this inconsistency. It may
be useful to measure the procurement variable in a quantitative form with
procurement measured as a ratio of market arrivals, which could have been
a better explanatory variable.
The location of mandis relative to nearby villages can be important. By
design, APMC mandis are generally located near tehsil (district) headquar-
ters, with jurisdiction over a specific number of villages or a specific geo-
graphical area. The AgMarknet portal does not provide complete information
about the geographical coverage of mandis. So it is not clear how the authors
have geocoded villages with mandis. It is important to let the readers know
 Shoumitro Chatterjee and Devesh Kapur 223

this because geocoded villages have been used for the subsequent analysis to
understand the degree of competition and the price variation within markets
and between markets in neighboring states. Further, it would be useful to
provide the number of rice and wheat markets covered across the 16 states in
the study. The characteristics of the states and markets selected for analysis
will certainly influence the results considerably. In fact, studies have shown
that wholesale markets for rice and wheat are integrated to some extent
within the states, while integration between regions is not very significant
(Acharya et al. 2012; Jha, Bhanu Murthy, and Sharma 2005).
Many studies have analyzed the geographical spread of APMCs, poor
infrastructure facilities, problems of marketing inefficiency, and so on
(Acharya 2004). I am puzzled to see the paper mention that “the number
of mandis grew commensurately as the Green Revolution took off, but
investments in market infrastructure slackened in more recent decades
even as output continued to grow.” It is very hard to get investment data on
agricultural markets; it is usually not available, and there is no supporting
citation in the paper.
The AgMarknet data does provide the stock of infrastructure facilities
built across mandis and APMCs, and one can build an infrastructure index
across the mandis and correlate it with market performance indicators such
as commodity arrival, number of market participants, timely payment, and so
on. Further, the effect of telecommunication technology on spatial variation
in price also warrants empirical testing based on field level data.
A major contribution of this paper could be in explaining the factors that
contribute to spatial variation in wholesale agricultural prices. However,
the decomposition analysis in the paper does not provide many insights
and it also needs to be explained in context. There are alternative analytical
approaches, such as those followed by Chand (2003) and Liefert (2011),
to decompose producer prices with useful results that have a lot of policy
relevance.
Figure 8 shows a positive relationship between MSP awareness and
procurement, while Figures 9 and 10 show a negative association between
procurement and production. As I have noted, procurement need not nec-
essarily be evenly spread across districts and mandis. It is quite possible
that farmers are aware of the MSP in a few states because of procurement
operations taking place in those states. The Food Corporation of India (FCI)
has the mandate of procuring a fixed quantity of the grains for buffer stock
operations. But when the prices fall below the MSP, some state governments
intervene under the market intervention scheme to buy whatever quantity is
224 I N D I A P O L I C Y F O R U M , 2016–17

offered by the farmers. It may be useful to relate procurement to the market-

able surplus to understand why some districts have high procurement and
others have low procurement. Marketable surplus and market arrivals are
highly correlated and tend to influence procurement operations.
There is a strong political economy operating in the food grain procure-
ment. It would be useful to analyze how farmers’ lobbies in states such
as Punjab and Haryana influence their state governments, who in turn put
pressure on the central government to procure whatever quantity and quality
the farmers would like to offer in the mandis. It would be interesting, for
example, to examine quantitatively how the “fair average quality” specifica-
tions against which the FCI procures grain has been tweaked periodically
under the influence of these state governments. Taking these considerations
into account in the analysis would enrich the discussion on the political
economy of grains procurement.
The functioning of APMCs has been a subject matter of intense debate
in recent years for reasons such as poor service delivery, lack of adequate
infrastructure, high transaction cost, and collusion among traders (Acharya
2004). With the implementation of a Model APMC Act in 2003, many
states have amended their state agricultural marketing acts to facilitate
greater transparency, freedom to sell outside the mandi, establishment of
private markets, and so on. With the push toward greater liberalization of
domestic markets, APMCs can no longer be considered local monopso-
nies. Despite the introduction of market reforms, collusion among trad-
ers continues to exist, which tends to affect the efficient price discovery
in the markets (for details of the effects of collusion, see Banerji and
Meenakshi 2004).
In the paper, the authors maintain that there is collusive behavior among
traders within the mandis, but assume the absence of such collusion between
the markets. With improvements in telecommunications, it is difficult to
assume that the traders do not collude between markets, and a potential
problem in the formulation of an econometric model to explain price varia-
tions through local market power and distance to the mandi. Further, dense
markets need not necessarily result in better prices for farmers.
To conclude, this is a very interesting paper. However, there is much that
can be done to improve its analytical approach. There is also need to discuss
the policy implications of the results for the current debate on shifting away
from price support to an income support policy for farmers. It would also
be useful to throw light on the unified National Agriculture Market and its
likely impact on farmers’ income and variation in prices.
 Shoumitro Chatterjee and Devesh Kapur 225

References
Acharya, S. S. 2004. Agricultural Marketing, Volume 17, State of the Indian Farmer:
A Millennium Study, New Delhi: Academic Foundation.
Acharya, S. S., R. Chand., P. S. Birthal, S. Kumar, and D. S. Negi. 2012. “Market
Integration and Price Transmission in India: A Case of Rice and Wheat with
Special Reference to the World Food Crisis of 2007/08,” Rome: Food and
Agriculture Organisation.
Banerji, A. and J. V. Meenakshi. 2004. “Buyer Collusion and Efficiency of
Government Intervention in Wheat Markets in Northern India: An Asymmetric
Structural Auctions Analysis,” American Journal of Agricultural Economics,
86(1): 236–253.
Chand, R. 2003. “Government Intervention in Food Grain Market in the New
Context,” Policy Paper 19, New Delhi: National Centre for Agricultural
Economics and Policy Research.
Government of India. 2016. Agricultural Statistics at a Glance 2015, New Delhi:
Directorate of Economics and Statistics, Ministry of Agriculture and Farmers
Welfare.
Jha, R., K. V. Bhanu Murthy and A. Sharma. 2005. “Market Integration in Wholesale
Rice Markets in India,” ASARC Working Paper 2005/03, Canberra: The Australia
South Asia Research Centre, Australian National University.
Liefert, W. M. 2011. “Decomposing Changes in Agricultural Producer Prices,”
Journal of Agricultural Economics, 62(1): 119–136.

General Discussion*

Sandip Sukhtankar asked the authors to explain their finding on variance

decomposition that suggested that rainfall shocks contributed only about 4
percent to price variation, since rainfall shocks were used as an instrument
for basically all kinds of regressions in India.
Rajnish Mehra asked if the formation of mandis was actually not endog-
enous, since the paper had found that every additional mandi led to a higher

* The paper to which these comments relate was originally presented at the IPF 2016

under the title, “Understanding Price Variation in Agricultural Commodities in India: MSP,
Government Procurement, and Agriculture Markets”. The version published in the IPF journal
is a substantially rewritten paper based on discussant and editorial feedback, and hence also
the change in title. To preserve the sense of the discussions at the India Policy Forum, these
discussants’ comments reflect the views expressed at the IPF and do not take into account
the revisions made in preparing the version published in this volume. The original confer-
ence version of the paper is available on NCAER’s website on http://www.ncaer.org/events/
ipf-2016/IPF-2016-Paper-Chatterjee-Kapur.pdf.
226 I N D I A P O L I C Y F O R U M , 2016–17

price: if the prices fell in some region, mandis would come up there and raise
prices, which would then tend to limit price dispersion. Second, he said that
futures markets were among the most efficient in the US, and wondered if
they were actually quite efficient in India as well. Mehra pointed to the tight
links between storage costs, current prices, spot prices, and futures prices,
implying that arbitrage would clear markets whenever an opportunity arose.
He felt that what the paper was observing had more to do with regulation and
institutional factors. If there was free entry of mandis, and well-operating
futures markets, one would still get a certain dispersion in prices. Properly
anticipated prices do fluctuate randomly, and in stock markets the rule of
thumb is a standard deviation in prices of about 20 percent, and a little less
in bond markets, mainly to do with the arrival and processing of information.
Dilip Mookherjee first noted that procurement is endogenous, so that the
paper’s finding that wheat prices go down with higher procurement may
simply be procurement kicking in when prices are low. So procurement itself
should be a primary topic of analysis. The paper looks at spatial variation in
the levels of prices, but the huge list of omitted variables—distance, storage
facilities, transport costs, and so on—makes that not very credible. Much
better to look at the spatial variation in price changes over time. Industrial
organization and international finance provide nice methodologies that can
help gauge market imperfections by looking at the covariation of prices
across locations rather than in the level of prices. Transport costs and storage
costs may still move in tandem with other factors such as oil prices, bur the
level of omitted variable bias would be much less.
Second, Mookherjee noted that while the paper was focusing on spatial
variation between mandis, his work on agriculture in West Bengal suggested
that there was huge price variation within mandis. And if one was interested
in the prices farmers got, then his work had revealed that the farm gate price
was less than half the mandi price, and only about 20 percent of this gap
was accounted for by transportation and storage costs. The rest was margins
going to traders who buy from farmers and sell in the mandis. When you
look at the pass through, mandi prices co-vary very closely with retail prices,
but farm gate prices do not. Hence, the real imperfections seem to be within
the mandis, rather than between mandis.
Third, while Mookherjee agreed that competition is a key issue for
APMCs, going into questions—such as the desirable number of mandis and
competition between them—raised in the paper entail very complicated cost-
benefit analysis since we know so little about the cost of creating markets.
What would be much easier to do would be to focus on regulation and com-
petition within mandis: for example, who has access to the mandis? He cited
 Shoumitro Chatterjee and Devesh Kapur 227

the very interesting ethnographies that Mekhala Krishnamurthy and Jeffrey

Witsoe had done that showed the differences between Madhya Pradesh and
Bihar, and the role that politics played in the working of mandis. He felt
therefore that more work needed to be done on the politics of procurement.
Karthik Muralidharan commented that researchers can write two types of
papers: a “dependent variable” paper, where the authors discuss a left-hand-
side variable they care about and just document the facts and evidence about
outcomes in that variable, without making any causal claims. Or they could
write a “right-hand-side variable” paper and pin down the identification of one
causal factor, even if it cannot explain much of the variation. Currently, the
paper was struggling between the two types, and ending up not doing either
satisfactorily. The paper could perhaps be more of the first kind, an agenda-
setting paper for research, as opposed to one seeking to answer questions about
causality, where it runs into a lot of problems with omitted variables. Since
the authors have chosen price variation as a theme, they could concentrate
on documenting the facts about this, including variation within states, within
districts, and within other tiers for which data exists. The paper could then
set out proximate right-hand-side variables that might help explain the price
variations at different levels, and provide a road map for future research.
Mekhala Krishnamurthy said that there was no question about private
sector buyers entering into the markets since they have been there for a very
long time. Everybody in the mandi is a private actor, except for the govern-
ment that is procuring. From her ethnographic work, she reported that it
does not help the analysis if we think of the private sector as only corporate
commercial entities buying outside mandis, because large corporates like
ITC or Cargill are wherever they choose to be, buying directly from farmers
through commission agents with a license, entering mandis, using their own
single market yards, and entering the futures market. They employ diversi-
fied buying strategies. So thinking of them as suddenly entering a market
is a bit of a misunderstanding of these markets.
Again, based on her work in Madhya Pradesh and Chhattisgarh,
Krishnamurthy pointed to the interesting temporal and seasonal variation in
prices for crops that the government procures, and for others that it doesn’t,
and how competition between mandis comes in. The government bought
wheat in Madhya Pradesh and Chhattisgarh only in the immediate post-
harvest season, and at topped-up bonus prices. This affected prices in the
off-season. In the market procurement season, all the mandis bought at about
the same price, but in the off season, when there was no procurement effect,
prices varied. Competition between mandis became much more important in
the off season, with mandis with greater access to flour mills offering higher
228 I N D I A P O L I C Y F O R U M , 2016–17

prices due to the higher offtake than the other mandis. In sharp contrast to
wheat, in soybean, where there was no government procurement, farmers
staggered their sales across the year, effectively hedging for higher prices.
There are also interesting instances of possible natural experiments when
mandis switch licensing or auctioning procedures that could be used to
study how competition works: she cited the example of the Gulbarga mandi.
Karthik Muralidharan followed up by noting that this showed how impor-
tant it was to understand the behavior of intermediaries because most of the
farmers were not directly selling to the mandis.
Indira Rajaraman asked a question about how the authors had dealt with
spatial variation in mandi fees and taxes, since these revenues don’t accrue
to the consolidated fund but are supposed to be used locally for improving
market infrastructure.
Rohini Somanathan referred to a 1971 paper by the British historian E. T.
Thomson, called “The Moral Economy of the Crowd in the 18th Century”,
which talks about a period when people were protesting the rapid rise in
grain prices in England. This was because not enough grain was coming on
to the markets due to the signing of forward contracts for grain exports to
America with farmers selling the grain before it was harvested. Thomson’s
paper talked about people blocking grain wagons and forcing sales in
the spot market in protest against the violation of laws that stipulated the
bringing of all produce to the market after harvest. All this suggested how
important it was to model the choice set of the farmer and the different actors
involved. But given how complex that could be, it would be hard to come up
with a causal story. Rather than looking for causality, she suggested simpler
approaches looking at prices and quantities, so that if both were going down
then it was a demand effect, and if only prices were going down then it was
a supply effect. Using such simpler approaches would still lead to a very
provocative paper, even if it did not address causality.
Rajesh Chadha noted that state governments impose different levies on
rice, and asked if this had been taken into consideration in looking at price
variations. In recent years, in some states procurement had been decentral-
ized to state agencies, with FCI buying from these agencies. He wondered
if this had been taken into account in looking at price variation.
The Chair Ramesh Chand concluded by congratulating both authors for
taking the research on agriculture marketing in India to a higher level since
most other studies were largely descriptive and not analytical, and they did
not use quantitative techniques. However, in light of the points made by
discussants and participants he suggested using different data sets since the
AgMarknet data may not be entirely appropriate. The key question in the
 Shoumitro Chatterjee and Devesh Kapur 229

paper, as he saw it, was whether government procurement prevented the mar-
ket price from falling below MSP, i.e., what was the impact of procurement
on price variations. The troubling aspect of the paper was that it seemed not
to matter for wheat, but mattered for rice. He suggested looking at surplus
markets (with procurement) and deficit markets (with little procurement)
separately, with procurement keeping prices at MSP only in surplus areas.
Wheat prices in the harvest season in the deficit districts of Maharashtra
invariably did not fall below MSP even in the months of April and May but
they did fall below MSP in Uttar Pradesh. Therefore, procurement affected
prices only in surplus districts and not in deficit districts.
Second, Chand emphasized that procurement mattered for prices only in
the harvest season. During the lean season, prices were generally half the
MSP; otherwise arbitrage would not take place. If after the harvest grain was
bought from Punjab and Haryana, market charges would be 14 percent, plus
loading and unloading and transport charges, so that there would be a price
spread. He referred to a 2003 study by the National Centre for Agricultural
Economics and Policy Research that had nicely calculated the normal price
spread between a surplus state and a deficit state over all the 12 months of
the year by taking into account transportation costs, marketing costs, and
normal profit margins for intermediaries.
Chand suggested the use of the farm harvest price rather than the market
price when comparing with MSP. Farm harvest prices were available district-
wise for major production areas and referred to a particular period, not all the
months of a year. Using market prices could be misleading. The market price
spread for wheat could be much higher in Punjab than in a deficit state since
lean season prices in the latter would in any case be generally higher.
He pointed to a large number of studies, many by Raghavendra Jha, on
India’s agriculture markets, and Jha’s conclusion that wholesale markets
were integrated across India and had developed to a point that precluded
the need for government intervention.
Finally, Chand advised extreme caution in the use of NSSO data for situ-
ation assessments, rich and reliable as the data was for household consump-
tion expenditures and employment. For instance, the NSSO data quoted in
Table 1 in the paper suggested that the highest figures for the percentage of
wheat and paddy sold by farmers to government agencies was somewhere
between the numbers for small and large agricultural holdings of 11 per-
cent and 4 percent, respectively. But, it is a well-known fact that at least 30
percent of paddy and 35 percent of wheat was procured by the government.
This clearly highlighted the serious limitations with NSSO data, which is
why he felt the authors needed to utilize data from other sources.