Date post: | 30-Nov-2023 |
Category: |
Documents |
Upload: | independent |
View: | 0 times |
Download: | 0 times |
האוניברסיטה העברית בירושליםThe Hebrew University of Jerusalem
המחלקה לכלכלה חקלאית ומנהל המרכז למחקר בכלכלה חקלאיתThe Center for Agricultural
Economic Research The Department of Agricultural
Economics and Management
Discussion Paper No. 16.09
Contracting with Smallholders under Joint Liability
by
Jonathan Kaminski
Papers by members of the Department can be found in their home sites:
מאמרים של חברי המחלקה נמצאים :גם באתרי הבית שלהם
http://departments.agri.huji.ac.il/economics/indexe.html
76100רחובות , 12. ד.ת P.O. Box 12, Rehovot 76100
Contracting with Smallholders under Joint Liability ∗
Jonathan Kaminski
Hebrew University of Jerusalem, Dept of Agricultural Economics,
P.O. Box 12, Rehovot 76100, Israel.
Email: [email protected]
December 22, 2009
Abstract
This paper examines the performance of contract farming when agents are groups
of jointly-liable farmers who receive input credit from a monopsonistic agribusiness.
Accounting for group mechanisms in credit repayment through joint liability and
peer monitoring, we derive the optimal monopsonistic contract under moral hazard
on production effort. The principal takes into account price incentives not only
on farmers’ effort but also on peer monitoring. Then, we show that the optimal
pricing rule is not monotonic with respect to the group’s characteristics. Imper-
fect information implies a distortion on pricing for low-efficient groups, which is
Pareto-improving from a social point. Groups of intermediary size and heterogene-
ity provide the best effort and peer-monitoring incentives.
JEL Codes: D82, L14, 013, Q13
Keywords : contract farming, moral hazard, joint liability, peer monitoring
∗I am grateful to participants of Toulouse Lunch Seminar, to Philippe Bontems, Pierre Dubois,
Stéphane Straub, Jean-Paul Azam, Shanta Devarajan, Vincent Requillard and François Salanié for valu-
able comments and useful advices. This work has benefited from the comments of Xavier Gine (World
Bank) at the NEUCD conference in Harvard University for a former version of this paper. I am also
indebted to participants of the conference on agricultural cooperatives held in Rehovot, at the Hebrew
University of Jerusalem, to Ayal Kimhi for financial support, and to Israel Finkelshtain, Ziv Bar-Shira,
Ethan Ligon, and Richard Sexton for technical advices.
1
1 Introduction
Contract farming has become one of the major tools for agricultural development in many
developing countries over the past three decades1. It typically involves an agribusiness
to provide quality inputs, a set of extension and other agricultural services, and pos-
sibly credit to individual or farmers’ groups. In exchange, farmers have a marketing
arrangement with a fixed price and outlet for their output on high-value markets (mostly
cash crops, organic and horticultural products) and must follow a particular production
method. Reciprocal obligations entail the provision of agricultural services and com-
mitment on purchase prices on the firm’s side, and quality requirements and exclusive
purchase rights on the farmers’ side.
The expansion of contract farming in the developing world has been driven by changes
in consumer demands, international trade, technology and policies. Contracting between
farmers and agribusinesses has enabled to tighten commodity chains and strengthen ver-
tical coordination2, which are increasingly required by the supermarket procurement sys-
tems or by manufacturing/retailing importers. In the course of commodity reforms (see
Akiyama et al., 2001), provision of inputs, rural credit, or extension services have become
problematic for small-scale farmers facing several market failures. Contract farming is
often one of the only channel that links smallholders to input and to output markets (Key
and Runsten, 1999) through the interlinking of markets. This is an efficient solution in
an incomplete market environment, as shown by the theoretical literature. Interlinked
agreements among rational economic agents arise in a very fragmented and incomplete
market environment (Mitra, 1983) where agents are isolated with barriers from entry to
some markets (Basu, 1983). Braverman and Stiglitz (1982) show that the creation of
efficient surplus is improved when interlinking occurs. Interlinkages make enforce trans-
actions easier and allow for information, enforcement and monitoring costs savings (see
1For a review and a complete list of references, see Bijman (2008). Little and Watts (1994) provide an
historical survey of contract farming in Sub-Saharan Africa. Glover (1984) focuses on outgrower schemes
in poor countries. Most of the papers in the development literature try to assess the impacts of contract
farming on smallholders’ welfare and when contract farming takes place. They provide an assessment of
advantages and drawbacks for contractors and farmers. This paper does not tackle all of these positive
issues. We are rather interested in efficiency and normative concerns.2Contracts reduce transaction costs through repeated interaction and a better control of the production
process (for instance see Winter-Nelson and Temu, 2002). They also provide better risk allocation for
farmers and lower coordination costs for contractors. The latter can better align with their customers’
demands by a greater regularity of agricultural product supplies and quality (with technical assistance),
bargain for more profitable prices for farmers, benefit from the low production costs of family farms, and
more easily access to credit and subsidies (vertical differenciation).
2
also Bell, 1989). Yet, this literature does not explore the endogenous mechanisms through
which contracts are enforced and production is monitored.
In this paper, we consider the interlinked nature of these contracts when the agribusi-
ness has a dominant position and contract with smallholders with no access to credit and
extension services. The outside option of smallholders is supposedly of limited scope,
because of market incompleteness (as for the case of outgrower schemes). We also account
for group mechanisms-namely joint liability3 and peer monitoring-whereby the agribusi-
ness elicit credit repayment of production inputs, due to several advantages. There is a
natural advantage to rely on peer-monitoring and social mechanisms through joint lia-
bility in many rural communities of the developing world, where many smallholders lack
appropriate collateral to secure a loan. Because of local informal institutions and social
ties, joint liability acts as an effective social collateral inducing delegated monitoring that
ensures workable credit repayment rates when groups are well designed4. Then, high
transaction costs are often prohibitive for the contractors who can rely on group mech-
anisms for monitoring, contracting, and enforcement. In addition, since credit is linked
to a productive output on which the contractor gets exclusive purchase rights, then it
provides an additional physical collateral.
Peer monitoring and joint liability are borrowed from the rich literature on group lend-
ing and collective credit agreements. Most of the theoretical literature has shown that
joint liability in credit agreements was efficient in several circumstances and may induce
higher profits for the contractor than under individual liability. First, group formation
allows positive assortative matching by affinities (Ghatak and Guinnane, 1999), thereby
reducing ex ante moral hazard (the risk type of the project or individuals: probability of
success). According to Besley and Coate (1995), the joint liability provides two opposite
repayment incentives, one positive coming from the mutual insurance of the group and
one negative because expected marginal profit depends not only on individual effort but
also on the effort of others (free rider’s effect). This negative effect from joint liability can
be mitigated by the use of credible social sanctions if farmers’ actions can be observable ex
post with a costly monitoring effort from the group. This will enforce cooperative behav-
iors within groups and will discourage opportunistic behaviors5. Armendariz de Aghion
3In case of one member’s defaulting, each member of the groups repays her share of the debt. Other-
wise, the group will be denied access to credit in the future, or the contract could not be renewed.4These collective credit agreements are widespread in the developing world, as witnessed by the suc-
cessful Grameen Bank in Bangladesh, and other microfinance institutions (see Morduch, 1999, for a
review).5They involve strategic defaulting due to ex post moral hazard (after the project’s outcome is realized),
that is, the strategy leading a borrower to default on credit repayment while being able to reimburse.
3
(1999) focuses on ex post moral hazard (strategic defaulting) through the lens of a group-
lending repayment game with a peer-monitoring decision level stage. For a monopolist
lender, it is proven that joint liability brings even more profit than individual creditors,
on condition that social sanctions are sufficiently credible and monitoring costs are not
too high. These insights also apply to the case of contract farming in developing coun-
tries with smallholders, where collective agreements are widespread (e.g. input credit).
This is notably the case for outgrower schemes whereby credit repayment is proceeded by
deducting its value from the output sales to the contractor. In this literature though, the
probability of the project’s success (and hence, of credit repayment) is exogenous and the
moral hazard problem is only explored through the project choice or the ex post behavior
of agents. Yet, peer monitoring must also affect actions of the agents ex ante through
individual and collective incentives for effort and project choice.
In this paper, we solve for a group-repayment game that is linked to a production
activity (the project) under the supervision of a Principal. The production activity is
undertaken according to an endogenous effort which can be only observable by other peers
(at some cost) and that determines the probability of success. Hence, both production
effort (and credit repayment probability) and the level of peer monitoring are endogenous
and simultaneously chosen by farmers. We restrict ourselves to the case of no strategic
defaulting, that is, when it is not possible for agents to contract or to market their
output with another agribusiness ex post. The group repayment game is linked to a prior
contracting stage between a Principal and a group of farmers (the agent).
The main contribution of this paper is to account for group mechanisms in the design
of contract farming, bridging the gap between the theoretical literature on interlinkages,
and the one on collective credit agreements. This is done by treating the problem of ex
ante moral hazard as endogenous on one hand. On the other hand, we consider contract
farming as an interlinked agreement with groups of smallholders within which strategic
interactions occur. We do not examine why contract farming with joint liability exist
or what is optimal scale of these arrangements in agricultural production. Rather, we
study the efficiency of such contracts and their optimal design from the agribusiness
perspective and from a social welfare point of view6. We show that the sufficient level
of peer monitoring needed to yield the cooperative effort equilibrium is less likely to be
undertaken by the group when individual and group production incentives are higher.
However, this is more likely to hold when the group’s marginal return on peer monitoring
6We assume that joint liability is favored by a Principal. We consider indeed the cases where the
Principal faces a disuasive contract monitoring cost, and farmers are endowed with no physical (or not
verifiable) collateral.
4
increases. Hence, the Principal faces a trade-off between price incentives on production
on one hand, and on peer monitoring on the other hand. Then, the optimal pricing rule
is not monotonic with respect to the group’s capacity of enforcing cooperative behaviors.
There are several implications for agricultural development. First, the efficiency of
contracting with smallholders relies on both individual and group characteristics. There
is thus a need for collecting data about individual and their related groups, which means
information on individual incentives and social dynamics. Second, information problems
are double-sided and have direct consequences on contract design. Imperfect information
on the group’s ability to peer monitor induces the Principal to adjust non-cooperative
pricing above or below the optimal level desired by the agribusiness, thereby inducing a
sub-optimal non-cooperative effort. Imperfect information on the group and individual’s
effort costs induce the Principal to move away from efficient pricing for the higher-cost
group types. Last, the efficiency of such contracts is affected by group design (size and
composition), with implications in terms of the regulation of co-operative formation.
The remainder of this paper is organized as follows. Section 2 presents the conceptual
framework with a group of two symmetric farmers (the agent) and an agribusiness (the
principal) under perfect information. Section 3 deals with imperfect information and
efficiency concerns. Section 4 explores the robustness of the model and the implications
of group’s size and heterogeneity. Section 5 concludes and discusses the policy implications
of the model.
2 The model
2.1 Game-theoretical framework and basic assumptions
In the basic framework, a group-the agent-is composed by two symmetric farmers who
contract with a monopsonistic agribusiness (hereafter denoted as the Principal). First,
the Principal sets the terms of the contract by defining the purchase price of the cash
crop p and providing one unit of input credit. The cost of one physical unit of input
is normalized to unity and the Principal faces a marginal receipt at farm gate ep7. Weassume that all farmers who participate to contract farming are cash-constrained and
cannot access rural credit markets, so they have no alternative to access inputs: seeds,
fertilizers, and pesticides. In the basic model, we also assume that the Principal gets
all the bargaining power and is a monopsony over crop purchases (at least at the local
7This is basically the FOB price of crop production from which transport and processing costs are
deducted.
5
scale), making a take-it-or-leave-it offer. Relaxing this assumption entails contractual
enforcement problems and possibility of side-selling. This will be discussed at the end
of the paper. The Principal is assumed to have perfect information about farmers and
group characteristics (relaxed in the next section). Second, the two farmers play a credit-
repayment game in group, that is, they are jointly-liable for the repayment of their peers.
If the group repays its whole credit at the end of the game, then contract farming can
be repeated in the future. Otherwise, the group is denied credit access in the repeated
stages of the game. The structure of this sub-game is the same as Armendariz de Aghion
(1999) except that the probability of repayment is endogenous in our model, and depends
upon an effort production variable. In addition, there is no strategic defaulting (ex post
moral hazard). Assuming no parallel market, there is no opportunity for side-selling the
cash crop in the case of contract farming with a monopsony. While peer monitoring is
a device to deter strategic defaulting in Armendariz de Aghion (1999), we use it here to
reduce the scope of ex ante moral hazard. The timing of the game is represented in the
below figure. Note that this one-shot game can be repeated n times, which would lead
to endogenize the value of contract renewal. Implication of repeated interactions will be
discussed at the end of the paper.
Principal-agent Simultaneous monitoring and effort Returns Learning Social sanctions contracting decisions: realizations: peer’s effort and refunding p*, and input credit γ*(e) and ec(γ)/ enc(γ) /Y Y ec/ enc W, V
Credit repayment sub-game
Figure 1: Principal-group game under joint-liability
After the contracting stage, farmers simultaneously choose a level of individual pro-
duction effort e and of peer monitoring γ, which are assumed to lie on the interval [0, 1].
The individual effort is private information but can be observed ex post with probability
γ. The equilibrium choice of γ is the minimal desired level of peer monitoring among the
group’s members. The level of effort represents a labor effort and the quality of input
application (that can be diverted or resold on the black market). Effort is assumed to
have a quadratic cost C(e) = ce2/2 and peer monitoring a linear one C(γ) = dγ.
We model effort as a moral hazard variable, that is, the observed production outcome
is realized with a probability increasing with effort. For sake of simplicity, we assume that
farmers can reach two levels of production, when using one unit of input:
6
Y with probability e (1)
Y with probability 1− e
Note that although production outcome is random, there is no exogenous source of risk
in the model. Adding exogenous risk would not add much to our analysis, except when
considering risk-averse farmers.
The lower production outcome does not allow the principal to recover the input loan,
while this is the case for the higher one. In the latter case, the farmer gets a positive
profit, while he is defaulting in the second case:
(p− p)Y > 1 > (p− p)Y and pY > 1 > pY (2)
Since farmers have a limited liability, they get zero profit in case of defaulting (no side-
selling). If the group defaults, we assume that accessing credit is lost for the future and
any produced cash crop is seized8 Since we are interested in the case of joint liability, we
allow for a group of two symmetric farmers to be able to repay if one of the two farmers
defaults, such that:
pY + pY > 2 (3)
When playing in groups, farmers can choose their individual-maximizing effort, which
drives an individual optimum level of effort. However, they can attain a better level of
effort that would make every farmers better off. This cooperative effort ec that they want
to induce (maximizing the joint profit of the group) cannot be supported as an equilibrium
since everybody has individual incentives to deviate. Deviation involves savings on cost
effort while benefiting from other’s higher efforts (less expected individual and group’s
defaulting). Let us now formally describe how cooperative effort can be enforced through
peer monitoring.
With peer monitoring, farmers’ efforts are now possibly observable with some proba-
bility. Setting the cooperative level of effort as a group’s objective (or an informal rule or
norm), anyone who is observed cheating can be punished. If the group observes ex post
that a farmer has not respected this commitment and played its individual maximizing-
profit level of effort enc(non-cooperative), she will be imposed a social sanction W. This
social sanction can be a reputational loss in the farmer’s community or an exclusion from
the group for instances. Incentives to invest in peer monitoring thus depend on its effi-
ciency, that is, the relative level ofW compared to d, and on the additional profit moving8We assume limited liability of farmers as no more than the produced cash crop can be seized by the
monopsony.
7
from non-cooperative to cooperative production efforts. Simultaneous monitoring and
effort decisions are thus endogenous and related to one another. According to Armen-
dariz de Aghion (1999), full joint liability and implementation of the whole social sanction
are always optimal for the group. We then derive the optimal solutions under full joint
liability and punishment.
Once production occurs, farmers learn their peers’ effort with some probability and
punish cheaters. If the group repays its global debt, then it will be refunded for the next
period. The exogenous value of accessing credit in the next periods is denoted V > 09.
All parameters are common knowledge within the group.
2.2 The two-symmetric credit-repayment game
Assuming risk-neutrality, we can write the objective function of the group of two sym-
metric farmers under a full joint-liability agreement (if same efforts are undertaken by
players). This contains individual and group incentives:
Eπ = e2[pY − 1] + e(1− e)[p(Y + Y )− 2] + V e(2− e)− ce2/2 (4)
where e2 is the probability that both farmers repay their credit and e(1− e) is the prob-
ability that the individual farmer repays but has to finance the debt of her peer. Finally,
e(2− e) is the probability that at least one farmer repays its credit, so that the group can
be refunded in the next period. The two first terms in the RHS are individual profit incen-
tives, and the third one is a group incentive: the whole group repayment and motivation
for contract renewal or reputation-building.
We solve the game by backward induction. First, we compute the cooperative, and
then, the non-cooperative production effort levels (maximizing individual objective), then
we look at equilibria, according to peer-monitoring decisions. We notably define parame-
ter intervals where cooperative and non-cooperative effort levels are equilibria. Second,
according to the cooperative and non-cooperative effort levels, we derive the optimal
peer-monitoring investment. We finally derive simultaneous effort and peer-monitoring
equilibria according to the model’s parameters.
To get the optimal cooperative effort level, we maximize (4) with respect to effort
under the participation constraint that expected profit is non-negative and treating the
9V can be interpreted as a reputational loss for the farmer with respect to the Principal, but also as
earnings losses from not accessing future credits. In the latter case, V can be endogenized in a repeated
game framework, as mentioned earlier. However, an exogenous V is robust to our results (see final
discussion).
8
peer’s effort as endogenous. Note that this is sufficient since farmers are symmetric.
First-order conditions entail:
ec =p(Y + Y )− 2 + 2Vc+ 2pY − 2 + 2V (5)
To have interior solutions, we assume that
c > p∆Y = p(Y − Y ) => c/2 > 1− pY (6)
because of (3). To make sure that this assumption is restrictive, we need to consider cases
where c/2 < 1. The non-cooperative effort is the optimal individual effort, treating the
peer’s effort as exogenous, such as:
maxe
ee0[pY − 1] + e(1− e0)[p(Y + Y )− 2] + V (e+ e0(1− e))− ce2/2 (7)
where e0 stands for the exogenous effort of the other player. First-order conditions yield
the best-response function of each farmer:
enc(e0) =p(Y + Y )− 2 + V
c− e0
(pY − 1 + V )
c(8)
Note that for low values of V , efforts are strategic complements, while they become
strategic substitutes when V is higher.
Now, we look at the effort equilibria according to peer-monitoring levels. In the credit
repayment game, there are two pure strategies: play cooperatively or not. Here is the
table of all possible effort equilibria
Strategies Cooperative Non-cooperative
Cooperative (ec, ec) (ec, enc(ec))
Non-cooperative (enc(ec), ec) (enc(enc), enc(enc))
Table 1: Strategies of the credit repayment game
(ec, ec) is an equilibrium if the expected profit of playing cooperatively is greater than
deviating minus the expected social sanction, that is Eπ(ec, ec) > Eπ(enc(ec), ec) − γW.
Similarly, (enc(enc), enc(enc)) is an equilibrium if and only if the return from cooperative
behavior is not offset by the additional profit of deviation minus the expected social
sanction. In other terms, Eπ(enc(enc), enc(enc)) − γW > Eπ(ec, enc(enc)). After some
calculations, we find that the cooperative effort level is a symmetric equilibrium if and
only if:
γ > γc = min(c(∆e)2
2W, 1) (9)
9
and that the non-cooperative one is a symmetric equilibrium as soon as:
γ < γnc = (min(c(δe)2
2W, 1) (10)
where ∆e = ec−enc(ec) > 0 and δe = ec−enc(enc) > 0 because of (5) and (7) when effortsare interior solutions. When V < 1 − pY , δe > ∆e => γnc > γc, and this is the reverse
for the other case. For large V , when the peer-monitoring level is between these two
thresholds, only (ec, enc(ec)) can occur10 because γnc < γc. However, it can be shown that
in-between levels of peer monitoring cannot support this equilibrium. Two conditions are
necessary:
Eπ(ec, ec) < Eπ(enc(ec), ec)− γW => γ < γc
Eπ(ec, enc(ec)) > Eπ(enc(enc), enc(enc))− γW => γ > γnc,c
Only the second condition is relevant for in-between levels of peer monitoring. The equi-
librium exists for in-between level of peer monitoring whenever γnc < γnc,c = γnc + (δe−∆e)(ec(pY − 1 + V ) − V )/W < γc. Because V is large, then δe < ∆e and this con-
dition cannot be ever satisfied. Asymmetric equilibrium cannot be supported by given
peer-monitoring levels. The asymmetric equilibrium only exists for peer-monitoring lev-
els larger than γnc,c, that is, in the region where cooperative equilibrium exists and is
Pareto-dominating.
Note also that for low V, the intermediary peer-monitoring levels induce the occur-
rence of both cooperative and non-cooperative equilibria. The cooperative one Pareto-
dominates the latter. This is represented in Figure A.1 (see appendix). From (9) and
(10), we can discuss the pattern of peer-monitoring thresholds that define the areas of
equilibrium occurrence. The slope and shape of these thresholds are examined in the
appendix. The main insights can be summed up in the following lemma.
Lemma 1 When 2 symmetric farmers play a credit repayment game under full joint
liability and moral hazard on production effort, the peer-monitoring level γc that is re-
quired to enforce cooperative equilibrium increases with V. The level of peer-monitoring
γnc needed to deter the non-cooperative equilibrium first increases then decreases with V.
Mixed-strategy equilibria with the probability to play cooperatively
β =γnc,c − γ
γnc,c − γc
are relevant in the regions where γ belongs to [γnc,c, γc] and [γc, γnc,c]. In [γnc, γc], neither
pure-strategy nor mixed equilibria exist.10This equilibrium would be an asymmetric one.
10
Proof. See appendix.
We found how effort equilibria are related to levels of peer-monitoring. Now, we need
to derive what are the levels of peer-monitoring equilibria, according to effort levels and
to the parameters of the model. Peer monitoring is profitable when it enables farmers
to undertake cooperative effort instead of non-cooperative one. As it is costly, optimal
peer monitoring is the minimum level of peer monitoring that can trigger a profitable
shift in effort equilibria. The solutions of the credit-repayment game can be stated in the
following proposition and represented in the below graph.
Proposition 2 In the credit repayment sub-game with two symmetric farmers under
moral hazard, simultaneous individual effort and collective peer-monitoring decisions en-
tail:
• Farmers play cooperatively with γ∗ = γc when V and d/W are low
• Farmers play non-cooperatively with γ∗ = 0 when V and d/W are high
• Farmers play mixed strategies with γ∗ = γ < γc when V is low and d/W is inter-
mediary
• All effort levels increase with V and p and decrease with c
Proof. See appendix for solving, computational details, and comparative statics.
V
d/W
1-pY
1
Non-cooperative effort and no peer monitoring
Cooperative effort and peer monitoring γc
Mixed strategy with
γ and β
Figure 2: Effort equilibria and optimal peer-monitoring decisions
Intuitively, when V becomes large, cooperative and non-cooperative effort become so
close that it is no longer profitable to peer monitor since the marginal profit is too low.
11
Even a lower peer-monitoring cost (associated with mixed strategies) is not profitable to
invest for intermediate expected profits. In contrast, mixed strategies are profitable when
V is low since it is associated to lower peer-monitoring costs and intermediate expected
profits.
Note that the occurrence of cooperative equilibrium first increases with V for its lowest
values since efforts are strategic complements when V is low (see figure 2). Figure 3 below
displays effort equilibria and their related levels along V and according to d/W . Note
also that an increase in p will correspond to a leftward shift of the curves of Figure 2.
Cooperative equilibrium can be more likely for low values of V but then is less likely. The
likelihood of non-cooperative equilibrium increases for larger values of V. In brief, a price
change has non-monotonic effects and should be accounted for by the Principal. That is
the aim of the next subsection. For sake of simplicity, we will only consider pure-strategy
equilibrium occurrence in what follows.
V0
1
1/2
1 -p Y
ec
enc (enc)
d/W=γcc/γc
Note : The black line is the effort equilibrium when d/W is very low. This trajectory is partly replaced by the grey portions when d/W is of intermediary values and below 1. For higher values, this is replaced by both grey and red portions
d/W=γcc/γc
Figure 3: Sequence of effort equilibria along V
2.3 The contracting stage
Assuming that the Principal is a monopsony having perfect information about farmers’
characteristics, it will optimize its own profit accounting for farmers’ reaction within
their groups, as it has been solved for the credit-repayment game. In a first step, the
12
monopsony maximizes its own profit, according to effort levels. We first solve this op-
timization problem for both effort levels, cooperative and non-cooperative ones. Then,
we look at the output price for which farmers are indifferent between playing coopera-
tively or not. Comparing the indifferent price bp to the optimal pricing under cooperativeand non-cooperative farmer’s behavior enables us to define an optimal pricing rule for a
monopsonistic Principal.
According to the effort reaction of farmers ex, then the monopsony
maxp
Πx(p) = (ep− p)(Y + ex∆Y )− 1 + (e2 + 2e(1− e)) (11)
where the x upperscript stands for the type of effort equilibrium (either x = C, or x =
NC). We assume that the participation constraint of farmers is not binding at optimum
(ep is sufficiently high). We obtain the following first-order condition:[ep− p∗
p∗]x =
1
εEY x/p− 2(1− ex)
p∗∆Y(12)
which is the optimal margin for the monopsonistic Principal, and where EY x is the
expected production level when effort is x and εEY x/p is the individual price elasticity of
supply. Note that the monopsonistic margin is below the standard Ramsey rule since the
monopsony internalizes farmers’ credit repayment and the associated complementarity
between interlinked input and output markets. When V is not too small, the optimal
margin is unambiguously higher under cooperative effort since effort is less reactive to
price. Then, under lower price elasticity and higher effort, the monopsony can propose
a lower price to cooperative farmers compared to non-cooperative ones. For low V the
above expression is also lower under cooperative effort reaction though ∂ec
∂p> ∂enc
∂p.
We then look at the indifferent price level bp such that farmers are indifferent betweencooperative and non-cooperative strategies. According to our computations in the previ-
ous subsection, it means that dW= γcc
γc. This must satisfy the following equation:
(c+ bpY − 1 + V )2d
W= c(c+ 2(bpY − 1 + V ) (13)
such that 1 > c/2 > 1− bpY > 0 (14)
(13) admits solutions and at least one when dW
< 1. This is a necessary condition to have
a non-zero probability of playing cooperatively among farmers, as already derived in the
previous subsection. These solutions are such that :
1− bpY = V − c
d/W[1−d/W ±
p1− d/W ] = {V −V2, V −V1} = {1−pY , 1− pY } (15)
13
where V1 = 1 − d/W −p1− d/W > V2 = 1 − d/W +p1− d/W . (15) satisfies the
constraints stated in (14) if {V − V1, V − V2} ⊂ [0, c/2]2 for the two solutions to be
feasible. For at least one to exist, one of the two elements of {V − V1, V − V2} must becontained in [0, c/2]. We then derive the ranges of parameters (V, d/W ) where solutions
exist. This can be represented in the below figure:
V
d/W
c/2
1
Play NC: p < p
Play C: : p > p
Play C: p < p
Play NC: p > p
Play NC: p < p
Play C: p <p< p
Play NC: p > p
Play C always
Play NC always
Play NC always
Figure 4: Existence of p and p and farmers’ behaviour
The optimal solution for the monopsony would then account for the change in the
effort type of producers x, according to the parameters of the group: d/W and V . The
decision set can be illustrated in the below figure. We denote pc < pc the two bounds on
the Principal profit curve under cooperative effort such that:
maxΠnc(p) = Πnc(pnc∗) = Πc(pc) = Πc(pc) (16)
Since Πnc is always lower that Πc for every p, then pc < pc∗ < pnc∗ < pc. From Figure 4
and our previous computations, we know that the optimal pricing rule for the Principal
is pc∗ when d/W is low and pnc∗ when d/W > 1 or V is very large. The intermediary
case is displayed below. Note that an increase in ep makes shift all the four price curvesupwards. Then the sequence of optimal prices with respect to V would shift leftward. We
see that the sequence of optimal prices is discontinuous with respect to V , decreasing by
parts but non-monotonic. Then, the following proposition can be stated:
Proposition 3 Optimal pricing for a monopsonistic Principal entails (assuming ep issufficiently high):
• When d/W > 1 or when V is very large, p∗ = pnc∗(non-cooperative pricing)
14
• When d/W is close to 0, p∗ = pc∗ < pnc∗(cooperative pricing)
• When d/W is in between, p∗follows this sequence (possibly left-truncated when ep islarge) according to V : pnc∗, p, pc∗, p, pnc∗ (discontinuous pricing)
• pc∗ and pnc∗ decrease with V and increase with epProof. See appendix
V
p=1/Y
p=(1-c/2)/Y
p p
cp
*ncp
*cp
cp
Π nc
Π c
cp *cp *ncp cp p
Π Π c*
Π n c*
V increases
Figure 5: Optimal pricing rule when d/W is intermediary
Hence, according to parameters related to farmers and group’s characteristics, optimal
pricing can change and will induce different effort equilibria in the credit repayment
game. Note that the monopsony can achieve optimal profits along the [Πnc(pnc∗);Πc(pc∗)]
interval. In addition, an increase in ep does not necessarily ensure the monopsony toachieve higher profits since it will shift the two profit curves to the right (figure 5 above)
and cooperative effort might be less likely to hold in equilibrium. Last, note that a change
15
in d/W from 0 to 1 will make decrease the range of V s for which p∗ = pc∗ between p and
p. The assumption of perfect information now needs to be relaxed.
3 Imperfect information and welfare analysis
A fully-informed Principal has the ability to achieve optimal profits according (i) to the
group’s capacity to peer monitor and enforce cooperative equilibrium, and (ii) to the
level of farmers’ effort as a function of prices. Complete information about V , d/W ,
and c enables the Principal to know which pricing would be associated to which effort
equilibrium and its level in the credit repayment game. Let us first relax the assumption
that the Principal has perfect information about d/W , which means that the monopsony
cannot determine which equilibrium will be enforced by the group but knows both levels
of cooperative and non-cooperative effort equilibria.
We start by assuming that there is a probability µ that the Principal contracts with
a group endowed with d/W < 1, and we consider the case where it is optimal for the
Principal to induce a cooperative effort equilibrium by pricing pc∗ (for intermediary V ).
With probability 1 − µ, d/W > 1 and the group will play non-cooperatively for sure.
Assuming participation constraint is not binding at optimum, we only need to ensure the
contract to be incentive-compatible for the most efficient group (Laffont and Martimort,
2001). It can be written:
Eπ(pc, ec(pc))− dγc(pc) ≥ Eπ(pnc, ec(pnc))− dγc(pnc) (17)
= Eπ(pnc, enc(pnc)) +Ψ(δe(pnc), pnc)− dγc(pnc)
where Ψ(δe(pnc), pnc) = (δe(pnc))2[(c/2 + pncY − 1 + V )] = Wγcc(pnc), such as defined
in (A.4). Ψ(δe(pnc), pnc) − dγc(pnc) is the information rent that the Principal must let
to the efficient group (with a fixed transfer for instance). Since we assume non-binding
participation constraints, the maximization program of the Principal can be written:
maxpc,pnc
µΠc(pc) + (1− µ)Πnc(pnc) such that Eπ(pc, ec(pc))− dγc(pc) ≥Wγcc(pnc)− dγc(pnc)
(18)
Denoting λ as the Lagrange multiplier associated to the incentive-compatibility constraint
of the efficient group, we therefore obtain the following first-order conditions:
Πc0(pc) = 0 => pc = pc∗ (19)
Πnc0(pnc) =λµ
(1− µ)[∂Ψ(δe(pnc), pnc)
∂pnc− dγc0(pnc)] (20)
16
We obtain that second-best solutions rely on the sign of the expression in brackets. It can
be shown (see appendix) that the expression may be positive for low V s and is always
negative for high V s. The information rent induces an inefficiency for the Principal on
non-cooperative pricing, accounting for strategic complementariness when V is low and
strategic effort substitutability when V is high. According to the increasing value of
V , it turns out that non-cooperative pricing may be first distorted downward and then
(unambiguously) upward. Adjusting the non-cooperative price enables the Principal to
minimize the informational loss on profits. Conditions are summarized in the below
proposition. Note that, according to parameters, the pooling contract with pnc = pc = pc∗
is possible under specific values of the parameters, yielding no informational rent to the
efficient group, but a high distortion on non-cooperative pricing.
We now assume that the Principal has perfect information on d/W but cannot deter-
mine the effort reactions of farmers because of imperfect information on c for instance.
Consider the Principal faces an efficient group with a low cost of effort and probability υ
and an inefficient one with probability 1− υ. Then it can be shown by the same reason-
ing that the Principal has to let an information rent to the efficient group and to adjust
pricing for the cost-inefficient group in the same direction and for both cooperative and
non-cooperative prices.
Proposition 4 Under imperfect information on d/W , pc = pc∗ and pnc > pnc∗ if d/W
and V are not too low. In the reverse case, pc = pc∗ ≤ pnc < pnc∗. Imperfect information
on the individual cost of effort also entails the same kind of adjustments for cost-inefficient
groups on both cooperative and non-cooperative pricing.
Proof. See appendix
This makes sense with our former results. When d/W and V are very low, cooperative
effort is an equilibrium and its likelihood increases with price under the lowest values of V
because of strategic complementarity of efforts. So, it is better to decrease non-cooperative
price in this specific case in order to make the contract incentive-compatible since it will
reduce the opportunity for efficient farmers to choose the contract for the inefficient ones
by lower profits and a less likely cooperative equilibrium.
We now want to look at the effect of imperfect information on welfare. We first define
an overall objective function for the whole industry (agribusiness Principal and producers).
We then derive the social optimum pricing rule and compare it to the solutions derived
earlier. The optimal effort (assuming that it yields positive profits for the agribusiness
and the farmers’ group) of the supply chain solves:
17
maxeepEY − pY + e(2− e)(pY + V )− ce2/2 (21)
which yield the following fist-order condition:
ePO =ep∆Y + 2pY + 2V
c+ 2pY + 2V(22)
Note that the social Pareto-optimal level of effort cannot be a cooperative equilibrium.
Indeed, equalizing this expression with (5) yields: ePO = 1 + (ep− p)∆Y/2 > 1 to ensure
that the agribusiness will get positive profits. The only feasible Paretian effort is thus
non-cooperative. By the same reasoning, we find that ePO = enc(enc) = (p−p)∆Y+2+VpY+1+V
,
which cannot be an interior solution either. Because of different objectives and vertical
differentiation, there does not exist any pricing rule that can yield a Pareto-optimal effort
level in the credit repayment game11. The Pareto-optimal effort is thus not achievable.
We need to look at the Pareto-optimal pricing rule, given the fact that e is a solution of
the credit repayment game. As earlier, denoting x = C,NC, we look at the solution of:
maxp
epEY x − pY + ex(2− ex)(pY + V )− cex2
/2 (23)
yielding:∂ex
∂p=
Y (1− ex(2− ex))ep∆Y + 2(1− ex)(pY + V )− cex(24)
*cp *ncp *cp *ncp ncpop c
pop p Note : Imperfect information on the group type makes non-cooperative pricing move following the arrows. The dash lines represent two social and Principal optimal rules
cep
∂∂
*ep
∂∂
poep
∂∂
ncep
∂∂
Figure 6: Comparison of two social optimum and Principal rules (ep increases)11It can be shown that the social optimum level of effort can be induced under individual liability.
18
The Pareto-optimal rule can be compared to the one of the monopsony. It can be repre-
sented in the above figure. The optimal pricing rules under Principal-agent contracting
are below the social Pareto-optimal ones. The largest inefficiency applies to cooperative
pricing because the optimal pricing rule for the Principal always entail that pc∗ < pnc∗.
Imperfect information is then Pareto-improving for non-cooperative pricing when V is not
too low, because of Proposition 4. Note that increases in ep are Pareto-improving sincePrincipal optimal and social optimal rules become closer.
4 Robustness
In this section, we discuss the validity of our results when we generalize the group design
by allowing larger or heterogeneous groups of farmers. We particularly study how size
and heterogeneity impact on effort and peer-monitoring incentives. We restrict ourselves
to the case of cooperative ec and non-cooperative enc(enc) equilibria. We also discuss the
issue of repeated interactions between the Principal and the farmers’ group, and how it
would affect our results.
4.1 Group size
Increase in the size of the group would change incentives for effort as the outcome would
largely depends on the behavior of others but not in the cooperative case as everybody
endogenize the same behavior. There are more people to share the deficit of defaulting
farmers with but it is possible to have more defaulting farmers. One’s own contribu-
tion to the likelihood of accessing future credit is reduced. Size changes the probability
distribution of ex ante expected profits.
We first look at the effect of size on cooperative and non-cooperative effort levels, then
we discuss the effect on peer monitoring. We assume groups are formed by an even number
of members. With n symmetric risk-neutral farmers, the cooperative effort (joint-profit
maximizing) is solved for:
maxe
k=n/2Xk=0
µn− 1k
¶[en−k(1− e)k][pY − 1 + nV − k(1− pY )
n− k]− ce2/2 (25)
An intuitive way to identify the channel whereby size influences effort incentive is to
look at the cumulative probability distribution of profits according to e, holding it fixed.
Then, effort incentives would be higher if the cumulative distribution exhibits first-order
stochastic dominance. The below graph shows that the answer is not obvious and that
there is no first-order stochastic dominance. However, according to our parameters, there
19
are distributions (for intermediary sizes) that may second-order stochastically dominate
the others and provide the optimal incentives.
When farmers play non-cooperatively, larger groups not only decrease incentives to get
V but also individual profits since the cumulative distribution of profits is less affected
by own actions. Strategic effort substitutability and complementariness are of smaller
scopes. As a consequence, the non-cooperative effort level decreases with group size.
V ( ) 2p Y Y V+ − + 1pY V− + Note : the cumulative distribution in bold represents profits when n=2 and the one in dash is when n is large
Prob 1 1-e² (1-e) (1-e)²
Figure 7: Cumulative probability distribution of profits when x = C
These discussions allow us to conjecture what changes in terms of peer monitoring
incentives. It exists a range of group size where cooperative effort incentives are higher,
meaning that γnc and γc take on higher values. Because of the curvature of the peer-
monitoring threshold with respect to V , then the frontier curve of parameters for which
cooperative effort can be supported as an equilibrium would shift upward, thus increasing
the occurrence to reach the cooperative equilibrium. For these specific parameter values,
then it exists an optimal size where effort and farmers’ profits are the largest. However,
when the group has a low ability to enforce the cooperative equilibrium, smallest groups
are desirable.
20
4.2 Group heterogeneity
Consider one homogeneous credit group in which the two farmers have the same bV . To ac-count for heterogeneity in our model, we study the two-player game with two asymmetric
farmers, endowed with two different V such that:
bV = V + V
2and ∆V = V − V (26)
We aim to compare effort and peer-monitoring incentives in the homogeneous group with
those provided by the heterogenous one. The cooperative effort remains the same because
the credit group maximizes the joint profit of farmers, i.e. ec(bV ).When farmers maximizetheir own profit, then each farmer will exert enc(V ) and enc(V ). It can be obtained:
enc(V ) =(p(Y + Y )− 2 + V )(1− pY − V ) + c(p(Y + Y )− 2 + V )
c2 − (1− pY − V )(1− pY − V )> enc(bV )
enc(V ) =(p(Y + Y )− 2 + V )(1− pY − V ) + c(p(Y + Y )− 2 + V )
c2 − (1− pY − V )(1− pY − V )< enc(bV ) (27)
The average non-cooperative effort in an heterogenous group is below the level of an
homogenous one if and only if enc(V )+enc(V )2
< enc(bV ), which is equivalent (after computa-tions) to
(1− pY − bV )22
[(∆V )2 − 4bV ] + (∆V )2
4((p(Y + Y )− 2 + 2V − 2c) > 0 (28)
This is true whenever heterogeneity is high for a given level of bV . The ratio (∆V )2
Vmust
be sufficiently high to ensure this condition.
Since farmers have heterogeneous preferences, their motivations to peer monitor each
other are different. The rule of the game is that the chosen peer-monitoring level is the
lowest among the desires of farmers. Meanwhile, the needed level of peer-monitoring to
enforce cooperative equilibrium are different among farmers. The first consequence is
that the cooperative equilibrium will be less likely to occur for low V s since it would
require more costly peer monitoring. Second, the cooperative equilibrium would be less
likely to hold for higher V s. Third, the two farmers might be located in different areas
of Figure 3, and this eventuality is more likely when heterogeneity increases. If they
are in the same area, nothing changes: each play either cooperative or non-cooperative
with the (higher) level of peer monitoring required to support the equilibrium. If not,
the equilibrium level of peer monitoring is the lowest of farmers’ desires. It can induce
both farmers to play non-cooperatively for low V s instead of asymmetrically (one wants
cooperative, and the other one do not). This is inefficient when it induces the high-V
21
type to play a lower effort level (non-cooperative instead of cooperative). This is more
likely when heterogeneity increases.
Overall, some restricted level of heterogeneity can be efficient for the group to increase
the average incentive for non-cooperative effort but over a certain threshold, the effect on
the most motivated farmers is worse than the positive effect on the less-motivated one.
In addition, cooperative equilibrium is less likely to hold.
It is possible to derive several implications, knowing that size and heterogeneity are
positively correlated. When the group has a high d/W then small groups are desirable.
When d/W is lower, intermediary groups are desirable since cooperative equilibrium can
be enforced under higher effort levels and low heterogeneity is consistent with cooperative
equilibrium. However, the optimal size as if the group was purely homogeneous may entail
inability to enforce higher cooperative effort equilibrium since heterogeneity might be too
high. Hence, the optimal size may be consistent with some level of heterogeneity which
allow farmers to play cooperatively with higher level of cooperative equilibrium (according
to the probability distribution of profits in the group). That is why larger groups exhibit
more free-riding and less cooperative behavior.
If group formation is under a free-adhesion principle, then farmers will accept new
members up to the point that it will no longer contribute to increase their individual
profits. It is not likely that free association of members will lead to the Pareto-optimal
group size and heterogeneity. Matching by affinities induces the most performing and
motivated farmers to go together while groups of less-performing ones would like to be
more heterogeneous to provide better incentives and expected profits to their members.
As a result, groups can be too small or too large. Starting from the optimal group size
and composition, the structure might not be coalition-proof with the willingness of the
most-performing farmers to establish their own group. The final matching outcome will
depend on the initial distribution of individual types. It is then likely that the optimal
group design is just a framework that is not sustainable as an equilibrium of a matching
process. However, any regulation of cooperative formation could be worse because it
will define membership through rigid exogenous parameters (such as origin, geographical
location, village residence...). Many contract farming schemes with smallholders took
place with village groups where low incentives for production were observed. Allowing
free association of members could clearly be a channel of welfare improvement.
4.3 Repeated interactions
If the credit-repayment game is repeated n times, then it is possible to endogenize the
value of V. Then, further implications can be derived. The probability to get play the
22
nth time is [e(2− e)]n. Let us just state that the value of contracting and accessing credit
is the value of expected profit minus an opportunity cost Z, which is the value of the
alternative option, assumed as less valuable than contracting. Hence, if the conditions
(parameters) of the game do not change over time
V =nX
k=1
[e(2− e)]k{e2[pY − 1] + e(1− e)[p(Y + Y )− 2]− ce2/2− Z} (29)
when e is the cooperative effort. Otherwise:
V =nX
k=1
[e+ e0(1− e)]k{ee0[pY − 1] + e(1− e0)[p(Y + Y )− 2]− ce2/2− Z} (30)
when e is the non-cooperative effort reaction as a function of e0, the other farmer’s effort.
All our conclusions from the above sections about the impact of V on contracting can then
be reinterpreted in terms of the parameters of the model. From the comparative statics
derived in the appendix and the above equations, it can be established that V reinforces
the positive effects of price p and ∆Y and the negative one of the effort cost c on effort
incentives. This also amplifies the effects of price and cost on the probability of effort
equilibrium. As a consequence, since price p is a decreasing function of V in the optimal
decision of the Principal, then there exists a unique crossing point where p∗ is such that
it is an optimal price for the given V (p∗). Optimal pricing is lower when c increases since
the price elasticity of supply is higher for both effort types. The combination of lower
pricing rules and lower V means that the effect of cost on optimal pricing is ambiguous
(increase effort incentives on one hand, but higher margin on the other hand).
With repeated interactions, the issue of imperfect information becomes less relevant
since the Principal could assess the parameters of the group as time goes by. Hence, if V
is large, repeated interactions would not be Pareto-improving from a social point of view.
5 Conclusion
This paper explored the design of contract farming with groups of smallholders, addressing
the problem of ex ante moral hazard on agricultural production through joint liability and
peer monitoring. The scope of ex ante moral hazard problem can be overcome by the use of
internal peer monitoring. However, we showed that effort incentives and peer-monitoring
ones go in opposite directions according to individual agents and group’s parameters.
Hence, the optimal contract for a monopsony entails specific pricing rules which can be
23
discontinuous as a function of these parameters. Imperfect information on individual and
group’s parameters may be welfare-enhancing.
More importantly, this paper provides several policy implications. First, any price-
regulation policy should be made accordingly to the available information on both indi-
vidual and groups (cooperatives) of smallholders. Second, group design will affect both
individual and group incentives and regulation of cooperative formation should then en-
able farmers to match by affinities through free association and co-option.
Our results are valid in the case where enforcement problems are only linked to ex
ante moral hazard issues and incomplete markets. However, peer monitoring is also a
useful device to reduce the scope of ex post moral hazard through strategic defaulting
on credit or side-selling the output. This enforcement problem occurs when there exist
alternative options for farmers, leading to less incentives for the agribusiness to provide
inputs. As shown by Delpierre (2008), competition would lead to a decrease of input or
credit provision, thereby justifying the observed competition-coordination trade-off by the
empirical literature (Poulton et al., 2004). Several solutions exist from a social welfare
point of view. First is to regulate the market structure so as to minimize side-selling
while promoting a certain degree of competition. Second is to promote capacity-building
for farmers’ associations so as to constitute a significant bargaining power when facing
a monopsonistic or duopsonistic downstream industrial structure. These two solutions
could be examined in future research.
Last, we should also consider policies that aim to overcome market incompleteness and
smallholders’ liquidity constraints by reducing transaction costs and information asymme-
tries. In this case, contract farming is less relevant when input use is provided to groups.
Second, the partial slackening of the liquidity constraint has ambiguous effects on input
use and income (Delpierre, 2008). Hence, these two sets of policies-industrial regulation
and more accessible rural markets-should be carefully designed and related to one another
during their respective implementation.
Our theoretical results might be translated in terms of testable hypotheses if empirical
studies for contract farming with smallholders are designed to capture all individual and
group-relevant information. This is left for future research.
References
[1] Akiyama, T., Baffes, J., Larson, D. F., and P. Varangis. 2001. Commodity Market
Reforms: Lessons of two Decades.World Bank Publications (eds.): Washington, DC.
24
[2] Armendariz de Aghion, B., 1999. On the Design of a Credit Agreement with Peer
Monitoring, Journal of Development Economics 60:79-104.
[3] Bardhan, P.K., 1980. Interlocking Factor markets and Agrarian Development: A
review of issues. Oxford economic Papers, New Series 32(1):82-98.
[4] Basu, K., 1983. The Emergence of Isolation and Interlinkage in Rural Markets.Oxford
economic Papers, New Series, 35(2):262-80.
[5] Bell, C., 1989. Credit Market and Interlinked Transactions [in] Handbook of Devel-
opment economics, North Holland (ed.), Vol 1.
[6] Besley, T., Coate, S., 1995. Group Lending, Repayment Incentives and Social Col-
lateral. Journal of Development Economics 46(1):1-18.
[7] Bijman, J., 2008. Contract Farming in Developing Countries: an Overview. Working
Paper, Wageningen University: Wageningen, the Netherlands.
[8] Braverman, A., J.E. Stiglitz, 1982. Sharecropping and the Interlinking of Agrarian
Markets. American economic review 72(4):695-715.
[9] Delpierre, M., 2008. Farm Input Use in a Context of Liquidity Constraints and Un-
enforceability. Journal of African Economics 18(3):496-528.
[10] Ghatak, M., T., Guinnane, 1999. The Economics of Lending with Joint Liability:
Theory and Practice. Journal of Development Economics 60(1):195-229.
[11] Glover, 1984. Contract Farming and Smallholder Outgrower Schemes in Less-
Developed Countries. World Development 12:1143-57.
[12] Key, N., D., Runsten, 1999. Contract Farming, Smallholder, and Rural Develop-
ment in Latin America: the Organization of Agroprocessing Firms and the Scale of
Outgrower Production. World Development 27(2):381-401.
[13] Laffont, J.J., D. Martimort, 2001. The Theory of Incentives: The principal-agent
model. Princeton University Press: Princeton, US.
[14] Little, P.D., and M.J. Watts, 1994. Living under Contract: Contract Farming and
Agrarian Transformation in Sub-Saharan Africa. Madison eds., University of Wis-
consin Press.
[15] Mitra, P., 1983. A Theory of Interlinked Rural Transactions. Journal of Public eco-
nomics, 20(2):167-91.
25
[16] Morduch, J., 1999. The Microfinance Promise. Journal of Economic Literature
37(4):1569-614.
[17] Poulton, C., Gibbon, P., Hanyani-Mlambo, B., Kydd, J., Maro, W., Larsen, M.N.,
Osorio, A., Tschirley, D., and B. Zulu 2004. Competition and Coordination in the
Liberalized Cotton Sectors of Southern and Eastern Africa. World Development
32(3):519-36.
[18] Winter-Nelson, A., A. Temu, 2002. Institutional Adjustment and Transaction Costs:
Product and Inputs Markets in the Tanzanian Coffee System. World Development
30(4):561-574.
26
6 Appendix
V
γ
γ c
1-pY
γ nc A
B
D C
Figure A.1: Areas of effort equilibrium in the (γ, V ) space
6.1 Proof of Lemma 1
Proof. We have to analyze the shape of the two functions γc(V ) and γnc(V ). We find
that∂γc
∂V=
∆e
W
c(c− p∆Y )
(c+ 2pY − 2 + 2V )2 > 0 (A.1)
because of the assumption made in (6). The second derivative is:
∂2γc
∂V 2=
c(c− p∆Y )2
W (c+ 2pY − 2 + 2V )4 −4c∆e(c− p∆Y )
W (c+ 2pY − 2 + 2V )3 (A.2)
which is negative when V is sufficiently high. This is notably the case when V > 1− pY .
It is possible than the function is convex for the lowest values of V according to the values
of parameters c, p, ∆Y . For the other peer monitoring level, we find:
∂γnc
∂V=
cδe
W{ 2(c− p∆Y )
(c+ 2pY − 2 + 2V )2 −(1− enc)
c+ pY − 1 + V} (A.3)
which is always negative when V > 1 − pY . However, it can be positive for the lowest
values of V if c is high. Recall that γnc > γc when V < 1 − pY and γnc = γc when
V = 1− pY . For high values of V,the curve is convex. Note that the slope of γnc does not
really matter for low values of V since γc determines the frontier of equilibrium occurrence
according to peer-monitoring levels. Then the curves displayed in Figure A.1 are one
possibility. In any cases, γc is an increasing function and γnc decreases for high values of
V , driving the results stated in Lemma 1. For in-between levels of peer-monitoring, there
is no Pure-strategy Nash equilibria, only mixed ones exist. A mixed-strategy equilibrium
implies that a player is indifferent between the two strategies with the probability β to
27
play cooperatively and (1− β) to play non-cooperatively. At the equilibrium, the below
condition should be satisfied:
βEπ(ec, ec) + (1− β)Eπ(ec, enc(ec)) = βEπ(enc(ec), ec) + (1− β)Eπ(enc(enc), enc(enc))− γW
β =γnc,c − γ
γnc,c − γc
yielding the result of Lemma 1.
6.2 Proof of proposition 2
Proof. Note that mixed-strategy equilibria is only relevant in the regions [γnc,c, γc] and
[γc, γnc,c] because of Lemma 1. For large V , it means that mixed-strategy equilibria exist
in the region of cooperative equilibrium, which is Pareto-dominating, and supported by
lower peer-monitoring costs. Hence, mixed-strategy equilibria are only relevant for low
V . Denoting: γcc = Eπ(ec,ec)−Eπ(enc(enc),enc(enc))W
= min(γnc[1+2(pY −1+V )/c], 1), farmers
choose between:
• γ∗ = 0 if Eπ(enc(enc), enc(enc)) > Eπ(ec, ec) − dγc since peer-monitoring invest-
ment acts as a sunk cost, players choose the minimal amount to ensure a Pareto
improvement by a shift in effort equilibria.
• γ∗ = γc if Eπ(ec, ec)− dγc > Eπ(enc(enc), enc(enc))
• γ∗ = γ when Eπ(β)−dγ > Eπ(enc(enc), enc(enc)) and Eπ(β)−dγ > Eπ(ec, ec)−dγc.We obtain the indifference curves that delimit the parameters ranges within which
effort and peer-monitoring equilibria occur. Indifference between cooperative and
non-cooperative behavior is defined by the indifference curve:
d
W=
γcc
γc(A.4)
Let us check whether and when mixed strategies can Pareto-dominate the pure ones.
We know that it only occurs for low V. Indifference between mixed strategy and the
cooperative one is defined by the curve:
d
W=
γcc + γnc,c
γc − γnc,c(A.5)
This curve is below the one described in (A.4) for the lowest values of V and there-
fore imply that the mixed strategy will be preferred to the cooperative one up to
the point where γnc,c > 0. Doing the same for the indifference between mixed strat-
egy and non-cooperative one implies that the mixed strategy is preferred to the
28
non-cooperative one when the latter is Pareto-dominated and the former Pareto-
dominates the cooperative one, such that:
d
W=
β(γcc + γc)
γ− 1 < γcc + γnc,c
γc − γnc,c<
γcc
γc(A.6)
Then, only (A.5) is relevant to determine the region of Pareto-dominating mixed-
strategy equilibria, such as depicted by Figure 2. This curve is strictly increasing for
low V and crosses the curve of (A.4) when V < 1− pY . Then, the mixed strategy
is an equilibrium for the lowest values of V and intermediary levels of dW
< 1.
6.3 Comparative statics
Parameters of the model
Endogenous variables p c ∆Y V d/W
ec + - + + 0
enc(enc) + - + + 0
enc(ec) + - + + 0
δe +/- - - +/- 0
∆e + +/- - + 0
Pr(ec) +/- -/+ +/- +/- -
Pr(enc) -/+ +/- -/+ -/+ +
6.4 Proof of proposition 3
Proof. According to feasible equilibria of the credit-repayment game, the two first bullets
are trivial and are depicted in Figure 4. They are just an application of the equations
regarding the existence of indifferent prices for the farmers’ effort reaction. In the in-
termediary case, farmers play cooperatively only if price is between p and p. So, it is
optimal for the monopsony to propose pc∗ when the price that maximizes profit under
cooperative equilibrium is lying on [p, p]. But the monopoly can afford higher profits that
the non-cooperative one when p < pc by proposing p and when p > pc by proposing p.
Otherwise, nothing else better than pnc∗ can be done. An increase in ep induces a transla-tion of optimal pc∗and pnc∗ as well as for pc (larger effect) and pc (smaller effect) so that
the sequence of optimal prices in the discontinuous case can be left-truncated with respect
to V. Because both effort equilibria increase with V and their price elasticity decrease,
then optimal pricing rules can be lower for the monopsony, meaning a higher marketing
margin on the output.
29
6.5 Proof of proposition 4
Proof. We need to study the sign of ∂Ψ(δe(pnc),pnc)∂pnc
−dγc0(pnc). After several computations,we find that
∂Ψ(δe(pnc), pnc)
∂pnc=
(Y + Y )(2− pncY )− 2V Y − 2Yc+ 2pncY − 2 + 2V (A.7)
− (pncY − 1 + V )
(c+ pncY − 1 + V )2{(Y + Y )c+ V Y −∆Y }
dγc0(pnc) =cd
W∆e
∂∆e
∂pnc< 0 (A.8)
Expression (A.7) has an ambiguous sign, but is undoubtedly positive when V is low and
negative when V is high. Note that (A.7) is already negative when V = 1−pncY . It meansthat the distortion on non-cooperative pricing involves a move along the non-cooperative
profit curve of the monopsony along its decreasing slope when V is not too low. That
means that the difference between cooperative and non-cooperative pricing will increase
with V (as well as the information rent). For the very low values of V, the distortion may
be of opposite sign provided dWis low.
30
PREVIOUS DISCUSSION PAPERS 1.01 Yoav Kislev - Water Markets (Hebrew). 2.01 Or Goldfarb and Yoav Kislev - Incorporating Uncertainty in Water
Management (Hebrew).
3.01 Zvi Lerman, Yoav Kislev, Alon Kriss and David Biton - Agricultural Output and Productivity in the Former Soviet Republics. 4.01 Jonathan Lipow & Yakir Plessner - The Identification of Enemy Intentions through Observation of Long Lead-Time Military Preparations. 5.01 Csaba Csaki & Zvi Lerman - Land Reform and Farm Restructuring in Moldova: A Real Breakthrough? 6.01 Zvi Lerman - Perspectives on Future Research in Central and Eastern
European Transition Agriculture. 7.01 Zvi Lerman - A Decade of Land Reform and Farm Restructuring: What Russia Can Learn from the World Experience. 8.01 Zvi Lerman - Institutions and Technologies for Subsistence Agriculture: How to Increase Commercialization. 9.01 Yoav Kislev & Evgeniya Vaksin - The Water Economy of Israel--An
Illustrated Review. (Hebrew). 10.01 Csaba Csaki & Zvi Lerman - Land and Farm Structure in Poland. 11.01 Yoav Kislev - The Water Economy of Israel. 12.01 Or Goldfarb and Yoav Kislev - Water Management in Israel: Rules vs. Discretion. 1.02 Or Goldfarb and Yoav Kislev - A Sustainable Salt Regime in the Coastal
Aquifer (Hebrew).
2.02 Aliza Fleischer and Yacov Tsur - Measuring the Recreational Value of Open Spaces. 3.02 Yair Mundlak, Donald F. Larson and Rita Butzer - Determinants of
Agricultural Growth in Thailand, Indonesia and The Philippines. 4.02 Yacov Tsur and Amos Zemel - Growth, Scarcity and R&D. 5.02 Ayal Kimhi - Socio-Economic Determinants of Health and Physical Fitness in Southern Ethiopia. 6.02 Yoav Kislev - Urban Water in Israel. 7.02 Yoav Kislev - A Lecture: Prices of Water in the Time of Desalination.
(Hebrew).
8.02 Yacov Tsur and Amos Zemel - On Knowledge-Based Economic Growth. 9.02 Yacov Tsur and Amos Zemel - Endangered aquifers: Groundwater
management under threats of catastrophic events. 10.02 Uri Shani, Yacov Tsur and Amos Zemel - Optimal Dynamic Irrigation
Schemes. 1.03 Yoav Kislev - The Reform in the Prices of Water for Agriculture (Hebrew). 2.03 Yair Mundlak - Economic growth: Lessons from two centuries of American Agriculture. 3.03 Yoav Kislev - Sub-Optimal Allocation of Fresh Water. (Hebrew). 4.03 Dirk J. Bezemer & Zvi Lerman - Rural Livelihoods in Armenia. 5.03 Catherine Benjamin and Ayal Kimhi - Farm Work, Off-Farm Work, and Hired Farm Labor: Estimating a Discrete-Choice Model of French Farm Couples' Labor Decisions. 6.03 Eli Feinerman, Israel Finkelshtain and Iddo Kan - On a Political Solution to the Nimby Conflict. 7.03 Arthur Fishman and Avi Simhon - Can Income Equality Increase
Competitiveness? 8.03 Zvika Neeman, Daniele Paserman and Avi Simhon - Corruption and
Openness. 9.03 Eric D. Gould, Omer Moav and Avi Simhon - The Mystery of Monogamy. 10.03 Ayal Kimhi - Plot Size and Maize Productivity in Zambia: The Inverse Relationship Re-examined. 11.03 Zvi Lerman and Ivan Stanchin - New Contract Arrangements in Turkmen Agriculture: Impacts on Productivity and Rural Incomes. 12.03 Yoav Kislev and Evgeniya Vaksin - Statistical Atlas of Agriculture in Israel - 2003-Update (Hebrew). 1.04 Sanjaya DeSilva, Robert E. Evenson, Ayal Kimhi - Labor Supervision and Transaction Costs: Evidence from Bicol Rice Farms. 2.04 Ayal Kimhi - Economic Well-Being in Rural Communities in Israel. 3.04 Ayal Kimhi - The Role of Agriculture in Rural Well-Being in Israel. 4.04 Ayal Kimhi - Gender Differences in Health and Nutrition in Southern Ethiopia. 5.04 Aliza Fleischer and Yacov Tsur - The Amenity Value of Agricultural Landscape and Rural-Urban Land Allocation.
6.04 Yacov Tsur and Amos Zemel – Resource Exploitation, Biodiversity and Ecological Events.
7.04 Yacov Tsur and Amos Zemel – Knowledge Spillover, Learning Incentives
And Economic Growth. 8.04 Ayal Kimhi – Growth, Inequality and Labor Markets in LDCs: A Survey. 9.04 Ayal Kimhi – Gender and Intrahousehold Food Allocation in Southern
Ethiopia 10.04 Yael Kachel, Yoav Kislev & Israel Finkelshtain – Equilibrium Contracts in
The Israeli Citrus Industry.
11.04 Zvi Lerman, Csaba Csaki & Gershon Feder – Evolving Farm Structures and Land Use Patterns in Former Socialist Countries. 12.04 Margarita Grazhdaninova and Zvi Lerman – Allocative and Technical Efficiency of Corporate Farms. 13.04 Ruerd Ruben and Zvi Lerman – Why Nicaraguan Peasants Stay in
Agricultural Production Cooperatives.
14.04 William M. Liefert, Zvi Lerman, Bruce Gardner and Eugenia Serova - Agricultural Labor in Russia: Efficiency and Profitability. 1.05 Yacov Tsur and Amos Zemel – Resource Exploitation, Biodiversity Loss
and Ecological Events. 2.05 Zvi Lerman and Natalya Shagaida – Land Reform and Development of
Agricultural Land Markets in Russia.
3.05 Ziv Bar-Shira, Israel Finkelshtain and Avi Simhon – Regulating Irrigation via Block-Rate Pricing: An Econometric Analysis.
4.05 Yacov Tsur and Amos Zemel – Welfare Measurement under Threats of
Environmental Catastrophes. 5.05 Avner Ahituv and Ayal Kimhi – The Joint Dynamics of Off-Farm
Employment and the Level of Farm Activity. 6.05 Aliza Fleischer and Marcelo Sternberg – The Economic Impact of Global
Climate Change on Mediterranean Rangeland Ecosystems: A Space-for-Time Approach.
7.05 Yael Kachel and Israel Finkelshtain – Antitrust in the Agricultural Sector:
A Comparative Review of Legislation in Israel, the United States and the European Union.
8.05 Zvi Lerman – Farm Fragmentation and Productivity Evidence from Georgia. 9.05 Zvi Lerman – The Impact of Land Reform on Rural Household Incomes in
Transcaucasia and Central Asia.
10.05 Zvi Lerman and Dragos Cimpoies – Land Consolidation as a Factor for Successful Development of Agriculture in Moldova. 11.05 Rimma Glukhikh, Zvi Lerman and Moshe Schwartz – Vulnerability and Risk
Management among Turkmen Leaseholders. 12.05 R.Glukhikh, M. Schwartz, and Z. Lerman – Turkmenistan’s New Private
Farmers: The Effect of Human Capital on Performance. 13.05 Ayal Kimhi and Hila Rekah – The Simultaneous Evolution of Farm Size and
Specialization: Dynamic Panel Data Evidence from Israeli Farm Communities.
14.05 Jonathan Lipow and Yakir Plessner - Death (Machines) and Taxes. 1.06 Yacov Tsur and Amos Zemel – Regulating Environmental Threats. 2.06 Yacov Tsur and Amos Zemel - Endogenous Recombinant Growth. 3.06 Yuval Dolev and Ayal Kimhi – Survival and Growth of Family Farms in
Israel: 1971-1995. 4.06 Saul Lach, Yaacov Ritov and Avi Simhon – Longevity across Generations. 5.06 Anat Tchetchik, Aliza Fleischer and Israel Finkelshtain – Differentiation &
Synergies in Rural Tourism: Evidence from Israel.
6.06 Israel Finkelshtain and Yael Kachel – The Organization of Agricultural Exports: Lessons from Reforms in Israel.
7.06 Zvi Lerman, David Sedik, Nikolai Pugachev and Aleksandr Goncharuk –
Ukraine after 2000: A Fundamental Change in Land and Farm Policy?
8.06 Zvi Lerman and William R. Sutton – Productivity and Efficiency of Small and Large Farms in Moldova.
9.06 Bruce Gardner and Zvi Lerman – Agricultural Cooperative Enterprise in
the Transition from Socialist Collective Farming. 10.06 Zvi Lerman and Dragos Cimpoies - Duality of Farm Structure in
Transition Agriculture: The Case of Moldova. 11.06 Yael Kachel and Israel Finkelshtain – Economic Analysis of Cooperation
In Fish Marketing. (Hebrew) 12.06 Anat Tchetchik, Aliza Fleischer and Israel Finkelshtain – Rural Tourism:
Developmelnt, Public Intervention and Lessons from the Israeli Experience.
13.06 Gregory Brock, Margarita Grazhdaninova, Zvi Lerman, and Vasilii Uzun - Technical Efficiency in Russian Agriculture.
14.06 Amir Heiman and Oded Lowengart - Ostrich or a Leopard – Communication
Response Strategies to Post-Exposure of Negative Information about Health Hazards in Foods
15.06 Ayal Kimhi and Ofir D. Rubin – Assessing the Response of Farm Households to Dairy Policy Reform in Israel. 16.06 Iddo Kan, Ayal Kimhi and Zvi Lerman – Farm Output, Non-Farm Income, and
Commercialization in Rural Georgia. 17.06 Aliza Fleishcer and Judith Rivlin – Quality, Quantity and Time Issues in
Demand for Vacations. 1.07 Joseph Gogodze, Iddo Kan and Ayal Kimhi – Land Reform and Rural Well
Being in the Republic of Georgia: 1996-2003. 2.07 Uri Shani, Yacov Tsur, Amos Zemel & David Zilberman – Irrigation Production
Functions with Water-Capital Substitution. 3.07 Masahiko Gemma and Yacov Tsur – The Stabilization Value of Groundwater
and Conjunctive Water Management under Uncertainty. 4.07 Ayal Kimhi – Does Land Reform in Transition Countries Increase Child
Labor? Evidence from the Republic of Georgia. 5.07 Larry Karp and Yacov Tsur – Climate Policy When the Distant Future Matters: Catastrophic Events with Hyperbolic Discounting. 6.07 Gilad Axelrad and Eli Feinerman – Regional Planning of Wastewater Reuse
for Irrigation and River Rehabilitation. 7.07 Zvi Lerman – Land Reform, Farm Structure, and Agricultural Performance in
CIS Countries. 8.07 Ivan Stanchin and Zvi Lerman – Water in Turkmenistan. 9.07 Larry Karp and Yacov Tsur – Discounting and Climate Change Policy. 10.07 Xinshen Diao, Ariel Dinar, Terry Roe and Yacov Tsur – A General Equilibrium
Analysis of Conjunctive Ground and Surface Water Use with an Application To Morocco.
11.07 Barry K. Goodwin, Ashok K. Mishra and Ayal Kimhi – Household Time
Allocation and Endogenous Farm Structure: Implications for the Design of Agricultural Policies.
12.07 Iddo Kan, Arie Leizarowitz and Yacov Tsur - Dynamic-spatial management of
coastal aquifers. 13.07 Yacov Tsur and Amos Zemel – Climate change policy in a growing economy
under catastrophic risks.
14.07 Zvi Lerman and David J. Sedik – Productivity and Efficiency of Corporate and Individual Farms in Ukraine.
15.07 Zvi Lerman and David J. Sedik – The Role of Land Markets in Improving
Rural Incomes. 16.07 Ayal Kimhi – Regression-Based Inequality Decomposition: A Critical Review
And Application to Farm-Household Income Data. 17.07 Ayal Kimhi and Hila Rekah – Are Changes in Farm Size and Labor Allocation
Structurally Related? Dynamic Panel Evidence from Israel. 18.07 Larry Karp and Yacov Tsur – Time Perspective, Discounting and Climate
Change Policy. 1.08 Yair Mundlak, Rita Butzer and Donald F. Larson – Heterogeneous
Technology and Panel Data: The Case of the Agricultural Production Function.
2.08 Zvi Lerman – Tajikistan: An Overview of Land and Farm Structure Reforms. 3.08 Dmitry Zvyagintsev, Olga Shick, Eugenia Serova and Zvi Lerman –
Diversification of Rural Incomes and Non-Farm Rural Employment: Evidence from Russia.
4.08 Dragos Cimpoies and Zvi Lerman – Land Policy and Farm Efficiency: The
Lessons of Moldova. 5.08 Ayal Kimhi – Has Debt Restructuring Facilitated Structural Transformation on
Israeli Family Farms?. 6.08 Yacov Tsur and Amos Zemel – Endogenous Discounting and Climate Policy. 7.08 Zvi Lerman – Agricultural Development in Uzbekistan: The Effect of Ongoing
Reforms. 8.08 Iddo Kan, Ofira Ayalon and Roy Federman – Economic Efficiency of Compost
Production: The Case of Israel. 9.08 Iddo Kan, David Haim, Mickey Rapoport-Rom and Mordechai Shechter –
Environmental Amenities and Optimal Agricultural Land Use: The Case of Israel.
10.08 Goetz, Linde, von Cramon-Taubadel, Stephan and Kachel, Yael - Measuring
Price Transmission in the International Fresh Fruit and Vegetable Supply Chain: The Case of Israeli Grapefruit Exports to the EU.
11.08 Yuval Dolev and Ayal Kimhi – Does Farm Size Really Converge? The Role
Of Unobserved Farm Efficiency.
12.08 Jonathan Kaminski – Changing Incentives to Sow Cotton for African Farmers: Evidence from the Burkina Faso Reform.
13.08 Jonathan Kaminski – Wealth, Living Standards and Perceptions in a Cotton
Economy: Evidence from the Cotton Reform in Burkina Faso.
14.08 Arthur Fishman, Israel Finkelshtain, Avi Simhon & Nira Yacouel – The
Economics of Collective Brands. 15.08 Zvi Lerman - Farm Debt in Transition: The Problem and Possible Solutions. 16.08 Zvi Lerman and David Sedik – The Economic Effects of Land Reform in
Central Asia: The Case of Tajikistan. 17.08 Ayal Kimhi – Male Income, Female Income, and Household Income Inequality in Israel: A Decomposition Analysis 1.09 Yacov Tsur – On the Theory and Practice of Water Regulation. 2.09 Yacov Tsur and Amos Zemel – Market Structure and the Penetration of
Alternative Energy Technologies. 3.09 Ayal Kimhi – Entrepreneurship and Income Inequality in Southern Ethiopia. 4.09 Ayal Kimhi – Revitalizing and Modernizing Smallholder Agriculture for Food
Security, Rural Development and Demobilization in a Post-War Country: The Case of the Aldeia Nova Project in Angola.
5.09 Jonathan Kaminski, Derek Headey, and Tanguy Bernard – Institutional
Reform in the Burkinabe Cotton Sector and its Impacts on Incomes and Food Security: 1996-2006.
6.09 Yuko Arayama, Jong Moo Kim, and Ayal Kimhi – Identifying Determinants of
Income Inequality in the Presence of Multiple Income Sources: The Case of Korean Farm Households.
7.09 Arie Leizarowitz and Yacov Tsur – Resource Management with Stochastic
Recharge and Environmental Threats. 8.09 Ayal Kimhi - Demand for On-Farm Permanent Hired Labor in Family
Holdings: A Comment. 9.09 Ayal Kimhi – On the Interpretation (and Misinterpretation) of Inequality
Decompositions by Income Sources. 10.09 Ayal Kimhi – Land Reform and Farm-Household Income Inequality: The Case
of Georgia. 11.09 Zvi Lerman and David Sedik – Agrarian Reform in Kyrgyzstan: Achievements
and the Unfinished Agenda. 12.09 Zvi Lerman and David Sedik – Farm Debt in Transition Countries: Lessons for Tajikistan. 13.09 Zvi Lerman and David Sedik – Sources of Agricultural Productivity Growth in
Central Asia: The Case of Tajikistan and Uzbekistan. 14.09 Zvi Lerman – Agricultural Recovery and Individual Land Tenure: Lessons
from Central Asia.