Tata AIG General Insurance Company - Insurer Innovation Award 2024
05.16.2013 - Dilip Mookherjee
1. Asymmetric Information and Middleman Margins:
An Experiment with Potato Farmers in India
Sandip Mitra
Indian Statistical Institute, Kolkata
Dilip Mookherjee
Boston University
Maximo Torero
International Food Policy Research Institute
Sujata Visaria
Hong Kong Univ of Science & Technology
May 2013
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
2. Questions
How large are middleman margins relative to prices
received by farmers in LDCs?
To what extent do benefits of market growth trickle down to
farmers?
How are these affected by improvements in external
market price information available to farmers owing to
advances in IT?
More broadly, what role do middlemen play in agricultural
marketing
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
3. Existing Evidence
Considerable evidence of large gaps between retail prices
of LDC exports and prices received by original producers
(Feenstra (1998), Morisset (1998))
Difficult to evaluate how much of these are accounted by
transport and distribution costs of intermediaries
Key problem: difficult to obtain data concerning costs and
markups of intermediaries
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
4. Existing Evidence, contd.
Micro-level evidence of limited pass-through (less than
50%) of commodity export price increases to farmers:
Ugandan coffee 2002-03 (Fafchamps-Hill (2008))
Mozambique cashews in 1990s (McMillan, Rodrik and
Welch (2002))
Fafchamps-Hill show rising middleman margins were not
accompanied by rise in transport or storage costs
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
5. Broad Questions
Possible reasons for high margins and limited
pass-through to farmers:
(a) trader market power, owing to collusion, limited entry
(capital requirements, network/reputational barriers), scale
economies in transport, storage etc.
(b) risk premium for insurance provided by traders to
farmers
(c) asymmetric information between traders and farmers
concerning price movements in wholesale or retail markets
where traders re-sell
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
6. Broad Questions
Possible reasons for high margins and limited
pass-through to farmers:
(a) trader market power, owing to collusion, limited entry
(capital requirements, network/reputational barriers), scale
economies in transport, storage etc.
(b) risk premium for insurance provided by traders to
farmers
(c) asymmetric information between traders and farmers
concerning price movements in wholesale or retail markets
where traders re-sell
Relative importance of (a,b,c)? If (c) is important, scope for
low-cost policy interventions using new IT
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
7. Recent Evidence Concerning Price Effects of
Cellphones
Jensen (2007), Aker (2010), Goyal (2010): increased use
of IT products reduce price dispersion across locations,
raise average returns for farmers (Kerala/Niger/MP)
Contrast: Fafchamps-Minten (2012) find no significant
treament effects on prices received by Maharashtra
farmers of offer of free SMS service providing
price/weather information
All of the above contexts: farmers sell directly in wholesale
markets rather than at farmgate
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
8. Effect on Farmgate prices?
Focus on contexts where most sales are at the farmgate:
bilateral contracting between farmer and customary trader
(Ugandan coffee, Mozambique cashews, West Bengal
potatoes)
IT services provide each farmer with information
concerning prices at which their middleman re-sells the
good in markets
Primary advantage to farmer: reduce informational
disadvantage vis-a-vis middlemen (rather than who or
where to sell)
Farmer could negotiate a better farmgate price, or reduce
informational distortions in ex ante bilateral contract
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
9. This Paper
Context: potato farmers in West Bengal
Calculate (lower bound to) average margins earned by
local intermediaries, estimate pass-through
Conduct a field experiment providing farmers with
information about the intermediaries’ resale price in
wholesale and retail markets
Use observed impacts on farmgate prices and trade
volumes to test alternative models of
contracting/bargaining
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
10. Preview of Main Findings
Large middleman margins in WB potato (≥ 50-90% of
farmgate prices); low (6-20%) pass-through from
wholesale to farmgate prices
Results of providing price information to farmers: no
average effects, increased volatility
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
11. Preview of Main Findings
Large middleman margins in WB potato (≥ 50-90% of
farmgate prices); low (6-20%) pass-through from
wholesale to farmgate prices
Results of providing price information to farmers: no
average effects, increased volatility (consistent with pooling
equilibria of bargaining model, inconsistent with ex ante
contracting model)
Hence neither risk-sharing nor asymmetric information play
a significant role
Key cause of high margins: hierarchical marketing
arrangements, market power of middlemen
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
12. Presentation Outline
1. Context: WB Potato Supply Chain, Markets and
Prices
2. Alternative Theories: ex ante contracting, ex
post bargaining
3. Experiment Details
4. Testing Different Theories
5. Implications
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
13. 1. Context: Potato Supply Chain
We study two main potato growing districts of W Bengal:
Hugli, W Medinipur
Farmers harvest potatoes once a year (Feb-April), sell half
at harvest time, store the rest and sell later (May-Dec)
Sell mostly to local traders (phorias) who resell in
neighboring wholesale markets (mandi) to wholesalers
Who in turn sell to traders in city markets (Kolkata for
Hugli, Bhubaneswar for Medinipur) and in neighboring
states (Assam/AP)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
14. Supply Chain
City Markets
/Kolkata,
Bhubneshwar
Other States
Wholesale
Markets / Mandis
Farmers
Village Traders
Local Markets
92% 8%
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
15. Supply Chain: Distances
City Markets
/Kolkata,
Bhubneshwar
Other States
Wholesale
Markets / Mandis
Farmers
Village Traders
Local Markets
80-400 km
3 km
0 km
5 km
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
16. Gross/Net Prices at Various Layers
City Markets
/Kolkata,
Bhubneshwar
Other States
Wholesale
Markets / Mandis
Farmers
Village Traders
Wholesale
(Local) Markets
Rs. 6.38
Rs. 4.85
Rs. 4.52
Rs. 2.24/2.15 Rs. 2.55/2.22
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
17. Average Margins
Key Fact: Less than 1% farmers sell to wholesalers in
mandi directly, so almost invariably farmers must go
through a local middleman (phoria)
When farmers sell in a local market or the mandi, they sell
to some other phoria located there
Prices farmers receive are significantly below resale prices
phorias receive at the mandi
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
18. Average Margins, contd.
In 2008 (year of the experiment), average farmgate (gross)
price was Rs 2.24/kg; farmer sold at Rs 2.55 in local
market/haat
Whereas phorias received average price of Rs 4.85 in the
mandi
For farmers that sell directly to a local market, we know
their selling costs (transport, handling, storage), which
provides an upper bound estimate of these costs incurred
by traders
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
19. Average Margins, contd.
Per Kg: transport costs Rs 0.24, handling & other costs Rs
0.35, storage costs (only for potatoes sold after June) Rs
0.89
Using these as upper bound for corresponding costs
incurred by traders, obtain a lower bound for net phoria
margin of Rs 2.02/1.13 for harvest/post-harvest sales
Equals 42/23% of wholesale price, and 90/50% of
farmgate prices for harvest/post-harvest sales
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
20. Hierarchical Marketing System
What prevents farmers from selling directly to a
wholesaler?
Wholesaler does not want to negotiate with small farmers:
‘not worth my time’
Wholesaler transacts with 7-8 phorias per village, who are
‘aggregators’ in a classical hierarchical manner (a la Calvo
Wellisz):
deliver ‘reliable’ supply: stable quantity, good quality (80%
phorias inspect potatoes supplied by farmers)
can accept delayed payment from wholesaler
maintain ‘good relations’ with farmers (risk-sharing role?)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
21. Phoria Market Concentration
Not particularly concentrated: number of phorias in a
village ranges from 2–20, with mode 6 and mean 9.5,
number of farmers selling is 150-200
Median village market share of a phoria is 5%
58% (resp. 28%) selling to same phoria for more than 2
(resp. 5) years; no exclusive dealing or contractual
commitments
Some interlinkage: one out of three farmers borrow and/or
purchase from same phoria
No difference in price between old and new suppliers (but
some difference in non-price services)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
22. Resale Price Information
Farmers say they would like to know price at which phoria
resells the crop at the mandi
Latter decided by private negotiation between phoria and
wholesaler: they don’t like to publicize these prices
Farmers access to phones does not help: ‘we don’t know
who to call’
We paid ‘insider’ market vendors at the mandi to provide
us with this information daily
74% farmers say phoria is their primary source of
information about the resale price
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
23. Price Shocks
Predicting prices on the basis of past price information for
farmers is difficult owing to huge fluctuations over time:
both across and within years
Annual average mandi (resale) price: Rs 7.60 in 2007,
4.85 in 2008, 10.72 in 2011, 12.37 in 2012
High co-movement with city prices: within-year
pass-through of weekly city price to mandi price exceeds
65%
Compared with insignificant (6%) pass through to farmgate
prices in 2008 (making it even harder to predict mandi
prices on the basis of farmgate prices)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
24. Wholesale Potato Prices for Different Years, Agmark
data, deflated by WB food price index
051015
Modalpriceperkg
2005w26 2006w26 2007w26 2008w26 2009w26 2010w26 2011w26
Week
Mandi prices adjusted for inflation in food article prices in West Bengal.
2005−2011
Mandi prices for jyoti potatoes, Agmark data
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
25. Analysis of Variance of Mandi Prices
Table 3: Analysis of Variance of Mandi Prices
Source MSE F
Year 6382.05 2980.75***
Period 1756.82 820.53***
Year x Period 731.69 341.74***
Mandi 165.92 77.49***
Mandi x Year 54.61 25.50***
Observations
R-squared
2717
0.87
Notes: Based on weekly averages of mandi
prices for the months June-November in
2007, January - November in 2008 and
2011, and January-April in 2012. Periods
refer to three "seasons" when potato sales
occur: harvest (weeks 1-12), post-harvest
early (weeks 13-26) and post-harvest late
(weeks 27-52). A mandi is defined as a
variety-year combination.
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
26. Intra-year Pass-through of City Prices
Table 2: Within-year Pass-through of Weekly City
Prices to Farmgate and Mandi Prices
Farmgate Mandi Mandi
price price price
2008 2008 2011
City price 0.06 0.67*** 0.74***
(1.07) (16.85) (13.82)
Constant 2.11*** 0.65 -0.72
(4.66) (1.38) (-1.54)
R-squared 0.50 0.91 0.95
Observations 689 949 1,031
Notes: Observations are market-variety-week
combinations. City is Kolkata for Hugli and
Bhubaneswar for Medinipur. Week and mandi
dummies are included. t-statistics in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
27. Explaining Across-Year Variations in Mandi Prices
Table
1:
Explaining
Varia1ons
in
Mandi
Prices
Pooling
data
across
2007,
2008,
2011,
2012
(1) (2)
Local area yield (tonnes/acre) -0.37 -0.42***
(-1.44) (-7.11)
Distance ('00 km) -0.21
(-0.86)
% households with landline 1.31
(0.25)
% villages with pucca road 1.14
(1.00)
% villages with factory/mill -1.50
(-1.33)
adult male wage rate 0.01
(0.18)
2008 -1.81** -1.13***
(-2.11) (-5.79)
2011 4.52*** 5.68***
(4.10) (22.53)
2012 5.73*** 8.98***
(6.59) (37.99)
Constant 10.62*** 0.81
(3.69) (1.31)
R-squared 0.78 0.81
Observations 86 2,717
Notes: Column 1 observations are mandi-year;
Column 2 observations are mandi-week;
Mandi denotes market-variety combination;
Distance measures distance to nearest city;
t-statistics in parentheses
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
28. Ranking of Mandis by Price: 2008 versus other years
05101520
0 5 10 15 20
Mandi’s rank before 2008
after 2008 2008
Mandis’ relative price rank
Table: Price RankingsS. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
29. Summary: Farmers’ Price Information
Farmers have no access to direct information regarding
mandi prices at which phoria will be able to re-sell potatoes
Predicting this on the basis of past price information of
farmers is difficult (esp in 2008) because:
farmgate prices co-move little with mandi or city prices
huge year-to-year fluctuations in mandi prices, of which
small part explained by local potato yields
2008 intra-year and across-mandi price patterns were so
different from other years
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
30. 2.Theoretical Predictions
When farmers sell to middlemen rather than to markets
directly, effects of information provision depend crucially on
whether trade is governed by ex ante contracts or ex post
bargaining
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
31. Theoretical Predictions: Implications of Ex Ante
Contracting
With ex ante contracts (necessary for risk-sharing),
asymmetric information results in screening distortions
Traders tempted to claim resale price is low in order to pay
low farmgate price; counteracted by lowering trade
volumes below efficient level
Providing information to farmers should reduce these
distortions: hence information provision should raise
average and reduce volatility of trade volumes
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
32. Theoretical Predictions: Implications of Ex Ante
Contracting
With ex ante contracts (necessary for risk-sharing),
asymmetric information results in screening distortions
Traders tempted to claim resale price is low in order to pay
low farmgate price; counteracted by lowering trade
volumes below efficient level
Providing information to farmers should reduce these
distortions: hence information provision should raise
average and reduce volatility of trade volumes
(assuming competition among middlemen)
If traders have monopoly power, information provision will
have no effect
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
33. EAC Prediction
!" !" !" !"
#" $"
#%!&"
$'%!&"
$%!&"
#'%!&"
Figure: Bertrand competition among traders, risk neutrality
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
34. EAC Prediction
!" !" !" !"
#" $"
#%!&"
$'%!&"
$%!&"
#'%!&"
Figure: Bertrand competition among traders, risk aversion
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
35. Testable Predictions of EAC
Robust prediction of EAC: reducing asymmetric
information should increase volume traded
Information treatment should expand average quantity
traded
Extent of increase typically larger, lower the mandi price
Expect heterogenous treatment effect: quantity expansion
should be larger, lower the realized mandi price
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
36. Alternative Theory: Ex Post Bargaining
Field interviews confirm absence of any ex ante
agreements: stress the role of ex post bargaining
Middlemen make price offers (after observing v): signaling
rather than screening
Outside option: farmer can take his potatoes to the local
market/mandi and sell to another trader there (who re-sells
to a wholesaler at same v); 11% of farmer transactions
(4.5% of annual output) were in local markets in 2008
Collusion amongst village traders, and amongst market
traders, but no collusion between village and market
traders: sequential duopoly
Sunk transport cost at the local market: scope for hold-up
by market trader
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
37. Separating Equilibrium: Graphical Illustration
v
p(v)
m(v)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
38. Partially Pooling Equilibrium: Graphical Illustration
v
p(v)
m(v)
v1
v2
r1
r2
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
39. Comparative Statics of Partially Pooling Equilibrium
v
p(v)
m(v)
v1
v2
r1
r2
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
40. Pooling Equilibria: Testable Predictions
Starting with a pooling equilibrium, improving the farmers
information lowers p, q for low v and raises p, q for high v
Average treatment effect will be zero, since m(v) is
unaffected and price offers made so that farmers are
indifferent between selling to VT or MT
Higher transport costs: lowers price offers and probability
of farmer accepting village trader’s offer
Higher v raises offer and lowers probability of farmer
accepting
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
41. 4. Information Provision Experiment
Experiment involving 72 villages and 25 farmers per village
in Hugli and Medinipur
Two information treatments in 24 villages each:
Public (noticeboards)
Private (‘blocked’ cell phones for 4 farmers/village)
24 villages served as control
Villages at least 8 km from each other
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
42. Experiment, contd.
Control-treatment villages paired to ensure balance within
each local area
Expect no effect on mandi prices: less than 1% supply to
mandi came from treated villages
Started experiment June-Nov 2007 but many problems
with cell-phones needed to be sorted out
Sorted out by end-2007, so we use data for one full year:
Jan-Nov 2008
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
43. Table 4: Descriptive Statistics, by intervention group (baseline survey, 2007)
(1) (2) (3)
Village-level
Distance to mandi (km) 8.07 8.56 8.93
(1.01) (1.65) (0.88)
Has a PCO box (2007) 0.46 0.42* 0.67
(0.10) (0.10) (0.09)
Household-level
Owned land (2008) (acres) 1.19 1.09 1.09
(0.05) (0.05) (0.05)
Fraction land planted with potatoes (2007) 0.51 0.51* 0.53
(0.01) (0.01) (0.02)
Has a landline phone (2007) 0.24 0.23 0.25
(0.02) (0.02) (0.02)
Has a cell phone (2007) 0.34 0.31 0.34
(0.02) (0.02) (0.02)
ControlPrivate InfoPublic Info
Notes:***, ** and * denote difference from control is significant at 1%, 5%, 10% resp.
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
44. Effects of Interventions on Source of Farmer Price
Information
Table 5: Effect of Interventions on Farmers Price Information
Probability of
tracking Days
since
tracked
Friends/Nbrs Other
(1) (2) (3) (5)
Private info 0.80 0.71*** 1.12 0.50 1.82
(0.47) (2.88) (0.19) (1.05) (0.79)
Phone recipient 1.71* 0.86*** 1.43 2.31 20.33***
(1.65) (2.78) (0.82) (1.47) (4.42)
Public info 8.00*** 0.80** 1.03 0.49 30.19***
(3.03) (2.20) (0.04) (1.10) (4.72)
Observations 10771 10771
Prob > χ2
0.00 0.00
Notes: In the fortnightly trading surveys (March — December 2008) a randomly selected 50%
sample (stratified by village) was asked if they kept track of retail or wholesale potato prices,
and if they answered yes, were asked to list up to 3 markets (2 varieties per market) where
they tracked prices, how long ago they last tracked the price, the price when they last
tracked it and the source of their information. Each observation is a household-variety-market
0.00
Information source
(4)
9302
Trader
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
45. 5. Treatment Effects on Quantity and Prices
Predicted average effects:
contracting model predicts traded volumes should rise
bargaining model predicts zero average effect on prices
and quantities
Also examine heterogeneity of treatment effects w.r.t.
realized mandi price:
contract model predicts treatment will raise trade volumes
more in low v states
bargaining model predicts trade volumes will fall in low v
states, and rise in high v states
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
46. Table 6: Average Effects of Intervention on Total Quantity Sold and Net Price Received
(1) (2) (3) (4)
Private info 484.65 16.09 -0.07 0.02
(0.88) (0.03) (-0.55) (0.15)
Phone 524.55 454.09 0.10 0.08
(1.18) (0.99) (1.05) (0.98)
Public info 229.40 -275.17 -0.10 -0.05
(0.44) (0.54) (-0.82) (0.47)
Land 2252.24*** 2218.82*** -0.09*** -0.07***
(12.89) (12.42) (-5.03) (4.82)
Constant 2844.90*** 3022.93*** 2.21*** 2.37***
(5.16) (6.69) (18.28) (26.40)
Mandi fixed effects No Yes No Yes
R-squared 0.35 0.39 0.37 0.44
Observations 2318 2318 2318 2318
Notes: The unit of observation is a farmer-variety-quality. A mandi is
defined as a market-variety combination. In columns (1) and (3) a variety
dummy for chandramukhi potatoes, a dummy for low-quality potatoes and a
dummy for Medinipur district are included; therefore the constant refers to
high-quality jyoti potatoes in Hugli district. All columns control for the
farmer's landholding. Columns (1) and (3) include controls for variety and
quality of potatoes, and a dummy for Medinipur district. Columns (2) and
(4) include mandi dummies. t-statistics are in parentheses. Standard errors
are clustered at the village level.
Total Quantity Sold (kg)
Net Price Received
(Rs/kg)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
47. Table 7: Heterogeneous Effects of Intervention on Total Quantity Sold and Net Price Received
(1) (2) (3) (4)
Private info -3027.18* -3066.96** -0.75** -0.61**
-1.67 (2.22) (2.11) (2.03)
Private info x Mandi price 782.06* 695.64** 0.15* 0.14**
(1.85) (2.15) (1.95) (2.02)
Phone 1221.77 1715.92 0.06 0.05
(0.94) (1.23) (0.18) (0.18)
Phone x Mandi price -159.73 -293.64 0.01 0.00
(0.50) (0.87) (0.09) (0.05)
Public info -2064.80 -2845.59** -0.21 0.02
(1.30) (2.26) (0.62) (0.08)
Public info x Mandi price 496.83 581.60** 0.02 -0.01
(1.33) (2.04) (0.28) (0.17)
Mandi price (annual average) -519.22*** 65.40 0.18** 0.18**
(2.93) (0.27) (2.38) (2.60)
Land 2249.33*** 2184.05*** -0.09*** -0.07***
(12.72) (11.96) (5.05) (5.55)
Constant 5197.65*** 2837.50** 1.73*** 1.58***
(5.37) (2.63) (5.63) (5.34)
Mandi fixed effects No Yes No Yes
R-squared 0.36 0.39 0.41 0.46
Observations 2299 2299 2299 2299
Quantity sold (kg) Net price received (Rs/kg)
Notes: The unit of observation is a farmer-variety-quality. A mandi is defined as a market-
variety combination. All columns include the farmer's landholding. In columns (1) and (3) a
variety dummy for chandramukhi potatoes, a dummy for low-quality potatoes and a dummy
for Medinipur district are included; therefore the constant refers to high-quality jyoti potatoes
in Hugli district. Columns (2) and (4) include mandi dummies. The average mandi price for
each individual farmer-variety is calculated as the average price of potatoes in the relevant
mandi, averaged over the weeks in which the farmer made a potato sale. Thus the average
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
48. Predicted Treatment Effects at 10th/90th percentile
Table 8: Total Effects of the Intervention on Quantity Sold and Net Price Received, 10th and 90th Percentile
(1) (2) (3) (4)
Private information, no phone
10th percentile -746.16 -1038.01* -0.32* -0.20
(1.04) (1.70) (1.98) (1.57)
90th percentile1736.71* 1170.48 0.14 0.25
(1.80) (1.37) (0.86) (1.46)
Private information, with phone
10th percentile 9.72 -178.55 -0.25 -0.13
(0.01) (0.21) (1.30) (0.77)
90th percentile1985.47* 1097.70 0.23 0.33
(1.76) (1.04) (1.06) (1.47)
Public information
10th percentile -615.73 -1149.27* -0.15 -0.01
(0.97) (1.97) (0.98) (0.08)
90th percentile 961.58 697.16 -0.08 -0.04
(1.09) (0.94) (0.45) (0.26)
Mandi fixed effects No Yes No Yes
Notes: Total effects are computed at the 10th percentile and 90th percentile of mandi prices
using results from the heteregenous effects regressions.
Net price received (Rs/kg)Quantity sold (kg)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
49. Predicted Treatment Effects at 10th/90th percentile,
Using Deviation from ‘Expected’ Price
Quantity Quantity Net Price Net Price
Private information, no phone
10th percentile -2449.89 *** -2955.29 *** -0.25 -0.09
(3.08) (3.26) (1.03) (0.53)
90th percentile 2777.22 *** 2123.19 *** 0.06 0.11
(3.79) (3.20) (0.41) (0.73)
Private information, with phone
10th percentile -1808.98 * -2621.14 ** -0.14 0.01
(1.94) (2.50) (0.52) (0.05)
90th percentile 3542.95 *** 3039.67 ** 0.19 0.20
(2.82) (2.60) (1.10) (1.13)
Public information
10th percentile -1830.98 ** -2465.10 *** -0.14 -0.09
(2.39) (2.90) (0.60) (0.58)
90th percentile 1679.93 ** 1037.57 -0.06 0.07
(2.19) (1.56) (0.39) (0.42)
Mandi fixed effects No Yes No Yes
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
50. Summary of Treatment Effects
Absence of any significant average treatment effects
(analogous to Fafchamps-Minten (2011))
Masked significant heterogeneous impacts
Reject contracting model, results consistent with pooling
equilibria of bargaining model
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
51. Additional Predictions of Partial Pooling in Ex Post
Bargaining Model
1. Farmers exposed to information treatments track local
market prices with less error
2. Farmgate (gross) price is lower than gross price
received by a farmer in a mandi sale (p(v) < m(v))
3. More market sales when v is higher (α(v) decreasing)
4. More market sales where transport costs are higher
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
52. Testing Additional Prediction no. 1 of Bargaining Model
Table 9: Effect of Intervention on Normalized Error Variance
in Tracked Price relative to Actual Local Market Price
Public info Private info Control
(1) (2) (3)
Standard deviation
of normalized error 0.33 0.32 0.40
Observations 2957 1335 1802
(3)/(1) (3)/(2) (2)/(3)
1.541*** 1.418*** 0.920*
_______________________________ ___________________________ ______________________
is defined as (tracked price - actual market price)/actual market price.
Variance-comparison test F-statistic
Note: Half the farmers in the sample were asked each fornight
if they tracked the prices in local markets, and if so which markets, and what the price
when they last checked. The date when they checked was matched to the actual
was average price in that market in that fortnight. The normalized error
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
53. Testing Additional Prediction no. 2 of Bargaining Model
Table 10: Farmers Gross Price Difference between Farmgate and Local-Market Sales
Overall Hugli Medinipur
(1) (2) (3)
Sold to market 0.36*** 0.56*** 0.36*
(2.78) (2.99) (1.99)
Mandi price 0.23*** 0.47*** 0.19***
(5.07) (6.92) (4.65)
Constant 1.44*** 0.73*** 1.30***
(9.06) (3.11) (5.78)
R-squared 0.37 0.48 0.29
Observations 3919 2002 1917
Notes: The unit of observation is a farmer-quality-variety-week
when a transaction occurred. Landholding, a variety dummy for
chandramukhi potatoes, a quality dummy for low quality
potatoes and a district dummy for Medinipur are included;
therefore the constant term is the coefficient on high-quality
jyoti potatoes in Hugli. t-statistics are in parentheses. Standard
errors are clustered at the village level.
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
54. Testing Additional Predictions no. 3,4 of Bargaining
Model
Table 11: Likelihood of Market Sales Regressed on Mandi Price
Overall Hugli Medinipur West
(1) (2) (3)
Mandi price 1.61*** 0.85 1.68***
(3.55) (0.45) (3.37)
Land owned 0.93 0.86 0.96
(1.10) (0.76) (0.58)
Medinipur dummy 4.72**
(2.41)
Prob > χ2
0.00 0.58 0.00
Observations 3919 2002 1917
Notes: The unit of observation is a farmer-variety-
quality combination in a week when a positive quantity
is sold. The results show odds-ratios and z-statistics
for logit regressions. Landholding, a variety dummy for
chandramukhi potatoes and a dummy for low-quality
potatoes are included. Standard errors are clustered at
the village level.
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets
55. Implications
Failure of information treatments to help farmers secure
higher prices reflect weaknesses of both formal and
informal institutions
Weakness of formal institutions: no centralized wholesale
markets permitting farmers to sell directly
Weaknesses of informal trading relationships: no long term
contracts or insurance; ex post bargaining with highly
restricted competition between traders
Policy makers need to focus on ways of enhancing
competition (eg encouraging entry of modern retailers that
contract directly with farmers)
S. Mitra, D. Mookherjee, M. Torero, S. Visaria
Asymmetric Info & Middleman Margins in Potato Markets