Article

A Financial Contracting Approach to the Role of Supermarkets in Farmers' Credit Access

American Journal of Agricultural Economics (Impact Factor: 1.33). 02/2008; 92(4). DOI: 10.2139/ssrn.1102579
Source: RePEc
ABSTRACT
Traditional moneylenders monitor farmers to ensure that their investment is not diverted. Modern farming contracts offered by supermarkets in developing countries often entail a loan component, and monitoring arises as well. However, unlike moneylenders, supermarkets do care about the attributes of the product. Whether such attributes are obtained is influenced largely by the advice and the extension services received by farmers. We build a financial contracting model where we show that supermarkets optimally undertake both the monitoring and the advisory missions. This contract is shown to potentially enhance credit access for small farmers but sometimes also involves excessive monitoring.

Full-text

Available from: Philippe Marcoul, Jun 26, 2015
A Financial C on tracting A pproac h to the R ole of Superm ark ets in
Farm ers’ C redit A ccess
Philippe Marcou l
and L uc Veyssiere
September 7, 2007
Abstract
Over the last decade, supermarkets have spread to developing countries and have becom e major actors
in the fo od ma rketing system in these countries. Their involvement in farmer’s production is wellkown.
In this paper , we constru c t a simp l e nancial contracting mo del to analyze this activity and especially
its impact on farmers’ credit access.
In our framework, the sup ermarket is modeled as a procurement organization, not only able to
advise contracting farmers on their p roduction practices, but also to partly nance their activity; i.e. to
substitute for convention al lend ers. We show that by bun dling advising and moneylend ing activities the
sup erm arket reduces the agency cost incurred to insure prope r incentives in th e proc ureme nt process.
This reduction in agency costs can extend nancing to smaller farmers who would otherwise remain
credit constrained. We also point out to reasons as to why sup ermarket sometimes prefer small farmers.
Finally, supermarkets usually set higher fo od standards. By examining the impact of the introduction
of higher fo od standard s, the paper sh ows that the co existence o f the m arket for d om e stic retailers and
sup erm arket products may broade n the credit acce ss of farmers.
Keywords: Financial Contracting, D evelopment, Financial Intermediation, Fo od Standards, Organiza-
tion of Pro duction, Sup ermarket
JEL classication: O17, O 33, O50, Q1 2, Q13
Corresponding author: Philippe Marcoul, Departm ent of E conom ics, Iowa State University, Ames, IA 50011-1070. Phone:
(515) 294-6311, Fax: (515) 294-0221; E-mail: marcoul@iastate.edu.
Veyssiere Luc, D epartm ent of Econom ics, Iowa State Un iversity, Am es, IA 5001 1-1070. Phone: (515) 294-461 1, Fax : (515)
294-0221; E-m ail: luc@iastate.edu.
1
Page 1
1Introduction
In the last two decades, we have witnessed an impressive development of supermarket c hains in developing
countries. Saturation and intense competition in retail markets of developed countries, together with sub-
stan tial margins oered by investing in developing markets, ha ve largely contributed to the emergence of
supermark et chains.
1
In countries where a substantial portion of the population lives in rural areas, the rise
of supermarkets, that arguably aect the livelihood of farmers, is a sensitive issue. Although they represent
a source of investment in local economies, their real welfare impacts are hard to assess and remain contro-
v ersial. On the one hand, many empirical studies have found that supermarkets tend to leave behind or
exploit small growers, preferring to concentrate their procurement of fresh agricultural products on larger
scale operations (Dolan and Humphrey 2000; Dolan, Humphrey and Harris-P ascal 2001, Trail 2006).
2
On
the other hand, although many gro wers successfully work with supermarkets, it is not clear whether growers
who fail to enter a business relationship with them are worse o relative to the period preceding their entry.
In addition, other recent case studies have somewhat challenged the view that supermarkets have only a
negative impact on small growers. In particular, these studies show that in niche markets small growers per-
form remarkably well and remain an attractive supply source for supermarket chains (Boselie, Henson and
Weatherspoon 2003, Henson, Masakure and Boselie 2005 and Minten, Randrianarison and Swinnen 2007).
However, while arguments on both sides are compelling, it is somewhat dicult, in light of these (rather)
contradictory observations, to forge a clear understanding of the impact of supermarkets on grower activit y.
The objective of this paper is to contribute to this debate by providing a theoretical framework to analyze
the impact that supermarkets have on growers’ credit access.
There exists an important descriptive literature on supermarkets in developing countries. This literature
describes and discusses what these retail chains are trying to accomplish and how they achieve their goals.
First, it must be noted that besides the growing (local) demand for fresh food products that they try to meet,
supermarkets or their aliated grocers demand a substantially higher quality in the products they procure.
Thus, supermarkets not only need to sell more in local markets, but they need to oer safer and higher quality
products, as well. Therefore, the natural response of supermarkets has been to develop their own standards in
coun tries where public food quality standards are often inadequate and lack proper enforcement. However,
the quest for higher quality and safer food products cannot be achieved without innovative procurement
practices. These practices revolve around the creation of ve rtical relationships with growers through the
1
Fo r instance , Carrefou r, a French-based supermarket cha in, earned on averag e three times highe r margin s in its Arg entine
op erations than in those lo cated France (Reardon et al., 2003 ).
2
While the lo cal demand for foo d is globally increasing, superm arket chains established in developing countries also exp ort
a substantial p ortion of their production to develop ed countries (Dolan and Humphrey 2000). Thus, supermarket production
will only exclude a p ortion of the growers that remain uninvolved with the sup ermarket.
2
Page 2
establishment of tighter procurement contracts. Although the specic form of the contractual relationship
between the gro wer and the supermarket can vary greatly depending on the context, there is arguably a
common denominator.
Typically, supermarkets require their growers to make a substantial up-front investment into their oper-
ations. This investm ent ranges from new equipment purchases to the establishment of quality control and
coordination systems. The literature analyzing supermark et procurement practices also reports that super-
markets are playing new roles in the production process. These roles essentially consist of a combination of
intense production monitoring and advising, sometimes using the support of public partners (Boselie, Henson
and Weatherspoon 2003). In practice, the advising is performed on the spot, when supermarket emplo yees
visit producers and discuss with them problems encountered during the growing cycle. The typical advice
ranges from the proper way to apply fertilizers to the safe handling of pesticides. In addition, supermarkets
also take on a monitoring role that essentially protects their investment in the growers’ operations. Indeed,
the relationship between farmers and supermarkets features a strong moral hazard component. For instance,
to certify that product standards are met, but also that procured quantities are sucient, supermarkets must
make sure growers follow specic procedures and do not cheat or misrepresent their eorts and/or actions.
3
Finally, although supermarkets rarely provide cash credit to farmers, they extend loans in the form of
input advances that are reimbursed later when the crop is sold.
4
These input loans, which range from seeds
to fertilizers and pesticides, cover most of the necessary inputs and their amount can be substantial relative
to expected crop payments.
5
Supermarkets also attempt to absorb some of the growers’ risks related to
market conditions. This is usually achieved by committing to input and output prices prior to planting.
Such commitments arguably result in lower liquidity needs for growers and are, in that sense, equivalent
to additional loans. Overall, supermarkets’ objectives seem to ensure that the nancial and production
risks faced by their grower base are sustainable and compatible with a long-term dedication to safe and
high-qualit y products (Henson, Masakure and Boselie 2005).
The organization of production by supermarkets, nevertheless, raises several questions. For instance, it
is not clear from a theoretical standpoint why supermarkets should provide such a bundle of services. It is
conceivable that advising services could be provided independently of input loans. Farmers could nance,
possibly using moneylending services, the purchase of the inputs necessary to carry over the production
process.
6
Supermarkets would then purchase the crop, pro v ided that it met a certain quality threshold.
3
The most comm on form of cheating faced by supermarkets is one in which farmers sell part of their crop (for a higher
price) to other gro cers or lo cal markets an d, therefore, do not deliver the quantity that was agreed up on (G ow and Swinnen
2001 and Minten, Randrianarison, and Swinnen 2007).
4
Cash advances are, in fact, widespread in transition countries (Gow and Swinnen 2001).
5
For instance, Boselie, Henson and Weathersp o on (2003) repo rts that it takes a numb er of plantings for pro ducers to achieve
anetoverallprot.
6
In develop ing c ou ntries, credit lo an s ex ten ded by traditiona l m on ey lend ers use growers’ crops as collate ral. To m ake su re
3
Page 3
The mere fact that such organization of production does not prevail in practice suggests that substantial
benets exist in bundling these tasks. In particular, to the extent that supermarkets are keen to have a large
grower base, it is possible that this organization of production will allow more farmers to access credit. More
generally, we wonder how the emergence of supermarkets will modify credit access for small growers.
In this paper, we analyze the market for growers’ loans using a simple model of nancial intermediation.
In our framework, growers need to make a nancial investment before they can produce for the supermarket.
An organization in which supermarkets advise, extend a loan and monitor gro wers is preferred by the
supermarket. In other w ords, bundling these tasks in the nancial contract results in an organization in
which motivation costs or agency rents are reduced. A llocating the two tasks to the supermarket implies
that, as a monitor, the value of a high quality crop is increased when the probability of success increases as
well; thus the supermarket also has an incentive to advise diligently. We show that rent contraction results
in more poor growers obtaining loans.
Our denition of the supermarket procurement process is very much similar to that of con tract farming.
Production nance by contract farming usually involve technical advising and monitoring. As described
b y Conning (2000), contract farming, apart from the advising part, is not dierent from traditional money
lending. In particular, it possesses all the informal aspects of moneylending. However, this type of lending
has become prevalent in many developing countries. For instance, Conning (2000) reports that, during the
last 20 years, that production nance has become dominant in Chile. Our multitask approach to this type of
contract can explain their relative superiority with respect to banking nance or traditional moneylending.
We also analyze the implications of our model for the grower’s eort. We show that the organization
of the production by the supermarket has some motivational consequences for growers. In particular, in
our framework, poorer farmers tend to exert higher levels of eort and consequen tly produce higher quality
products. Many recent empirical contributions tend to echoe our nding that smaller farmers maybe prefered
b y supermarkets (Boselie, Henson and Weatherspoon 2003, Henson, Masakure and Boselie 2005 and Minten,
Randrianarison and Swinnen 2007).
Finally, we explore the implication of higher standards set by supermarkets on the nal market. In this
variant of our model, loan access is endogenized and ultimately determined by the competitive outcome on
the nal market. We sho w that stronger standards can benet growers, as they obtain loans more often.
In what follows, we briey presen t the existing literature on lending in developing countries that is relevant
to our work. We also relate our paper to the corporate nance literature on advising and venture capital.
that the grower repays his loan, the moneylenders closely m onitor him during the crop cycle to make sure that he do es not
secretly side-sell and then default on th eir loan by pretending to have a bad harvest (See Aleem 1990 and H o a n d S t ig litz
1998). U nlike the advising part, the monitoring exerted by the sup ermarket is very similar to that of traditional m oneylending
(See Conning 2000 and Minten, R andrianarison and Swinnen 2007 for the case of sup ermarket m onitoring.)
4
Page 4
The model developed in the main section formalizes the basic idea of bundling advising and monitoring
tasks in the same nancial contract. We then study growers’ incentives in this setup. Finally, we study how
standards aect competition in the nal market, and thereb y inuence growers’ access to loans.
2 Re lation to the litera tu re
The literature on moneylending in developing countries starts from the premise that borrowers in developing
coun tries usually have weak balance sheets, and therefore have diculty accessing nancing. Most of the
contributions in this eld describe mechanisms by which borrowers are able to commit to repay their loans.
7
One of the main mechanisms to facilitate access to nancing in the absence of adequate collateral is group
lending. In Beasley and Coate (1995), a lender extends a loan to a group of persons jointly responsible for
its repayment. Each borro wer can be diligent and decide to repay his loan. When he or she is tempted
to default, the rest of the group will subject him to intense social pressure, so that shirking incentives are
weakened. Thus, this mechanism essentially makes use of the ability of the agents to monitor each other’s
actions (Barnerjee, Besley, and Guinnane 1994).
Group lending also uses the fact that members of the group are well-informed agents. Ghatak (1999)
shows that a group that is jointly liable can act as a screening device when agents have superior information
about each other’s project protability.
8
As investigated by Aleem (1990) and Ho and Stiglitz (1998),
the informal lending activity in developing countries is usually performed by local agents who can easily
monitor borrowers. Ho and Stiglitz (1998), especially emphasize the fact that the moneylending activity is
an informationally intensive activity characterized by monopolistic competition.
Similarly, our work assumes that supermarkets are especially well-informed local agents as, in practice,
they employ well-informed local agents to perform the monitoring activity.
9
In addition, the monitoring
activity of supermarket employees is very close in nature to that of moneylenders.
10
7
A compr e h e n sive r e v ie w of this litera tu r e is o u t of th e sco pe of this p a per an d we o n ly allu de to key c o ntrib u t i ons. For a
go od re view of this literature, we refer the reader to Arm endáriz de A ghion an d M orduch (2005).
8
Interest in g ly, w ithou t ap pe a lin g to advers e sele c tio n , Armen d a r iz de Ag h io n and Go llie r (2000) show tha t gro u p joint
liabi lity als o has th e prope rty o f mak ing cr o ss pled g in g po ss ib le and t h e re fo re can enha n c e lendin g activ ity. Th e idea i s to
mutu a lize risks in such a way tha t a p e rso n ’s lo a n succ es s c a n s erve t o re p ay an o t he r pe rs o n ’s lo a n fa ilure.
9
For instance, M inten, R andrianarison and Swinnen (2007) describ e the organization of the pro curem ent activity by retail
chains in M adagascar. They write (p. 11):
Every extension agent, the chef de culture, is responsible for abou t thirty farmers. To sup ervise these, (s)he
coordinates ve or six ex ten sio n assistants (assistant de culture) th at live in th e villag e itse lf. Th e chef de cultu re
has a p erm anent salary paid by the rm .
10
Minten, Ran drianarison and Sw innen (2007) also describes the frequency and the purpose of the m onitoring (p 12):
During the cultiva tion pe riod of the vegetables under contract, th e contractor is visited o n average more than
on ce (1.3 t imes ) a week. Th is intensive mon it o r in g is to en s u r e c o r re c t pro duc tio n ma n age ment a s we ll a s to avoid
‘sid e-s e llin g ’
5
Page 5
Our work also shares common features with the literature on venture capital. Casamatta (2003) studies
under which conditions an entrepreneur (in fact, a borrower) should hire an advisor. This article provides a
rationale for the existence of venture capitalists by showing that these advisors have to provide funds as well.
In Casamatta (2003), when the venture capitalist advices diligently, the probability of a successful project
increases.
Although the investment scale of the project is quite dieren t, the supermarket plays a role that is,
arguably, qualitatively identical to that played by venture capitalists. Indeed, supermarket emplo yees do not
their activit y to growers’ monitoring ; they continuously advise them on best production practices. Moreover,
it is well documented that supermarket agents have substantial knowledge in horticulture and, in that sense,
are valuable advisors.
11
Overall, allocating these two tasks to a single agent (i.e., the supermarket) helps to reduce the agency
rents that would have to be distributed to several agents had the supermarket contracted at arm’s length
with the growers. In the spirit of the literature on micro-nance described earlier, our main result is to show
that, by combining these two tasks, the supermarket allows poorly collateralized growers to obtain a loan; a
loan that they would otherwise not obtain. This result is, to the best of our knowledge, novel.
From a purely theoretical standpoint, our contribution also relates to recent work on the design of con-
tracts involving multitasking agen ts. Laux (2001) shows how, in a limited-liability contracting environment,
wage cost can be reduced by assigning several independent projects to a single agent rather than sev eral
agents. By paying the agent only when all projects succeed, the principal can relax the agent’s limited
liability constraint by punishing the agent for a given project by taking awa y payment on another.
More recently, Hueth and Marcoul (2007) model producer cooperatives by assuming that members provide
not only work (as input providers) but also monitoring of managerial activity (as directors). The resulting
multitasking structure is shown to strictly lower motivation costs.
3 Superm ark et procurement organization
In developing countries, the supermarket not only behaves as an external consultant (that provides pro-
duction advice), but also endorses the role of conventional moneylenders. The literature on micro-credit in
developing countries has emphasized the role of moneylenders as important actors in farming areas. Tra-
11
Again, M inten, Rand rianarison and Swinnen (2007) write:
The second constraint is hum an capital and long duration required for training of the a ssistan ts de culture which
organize and sup ervise the contracting farmers in the eld. It is estim ated that it takes on average two or three
years u ntil the r m will b e ab le to give him/he r full respo n s ib ility in the eld. This slows down growth and
expansion.
6
Page 6
ditionally, growers have relied on moneylenders, as the latter have an informational advantage and excel in
curbing farmers’ incentives not to reimburse loans (see, for instance, Armendáriz de Aghion and Morduch,
2005). The purpose of this section is to understand the economic rationale beyond the supermark et procure-
men t system and its implication on credit access by farmers. In particular we seek to understand why the
supermarket does not conne its activity to advising farmers on production, but rather, also gets involved
in some lending activity.
The contractual relationship between farmers and the supermarket is modeled via a framework similar
in spirit to Holmstrom and Tirole (1997).
12
However, unlike the conventional nancial intermediary of
Holmstrom and Tirole (1997), we characterize a contractual framework in which the supermarket not only
monitors borrowers, but also provides advice that enhances the value of their projects.
3.1 Presen tation of the model
Consider a project that returns R, but requires a xed investment I. Farmers would like to implement this
project, but unfortunately are nancially constrained; farmers have nance A with I>A.Weassumethat
there is a moral hazard component in the farmer’s behavior; he can either choose to be diligent, in that case,
he is successful with probabilit y p
H
(and fail with probability 1 p
H
), or he might shirk and, in that case,
will always fail.
13
In addition, when unsupervised by a monitor, we assume that the farmer enjoys a private
benet B. Based on a survey of Ivory Coast agricultural producers, Biais, Azam, Dia and Maurel (2001)
estimate that this opportunity cost of eort is important. Specically, they report a value for B as large as
40 percent of the investment.
Farmers are assumed to be protected by limited liability, i.e. investors can at most seize the realized
outcome. This assumption is in line with Innes (1990) and all the following literature on nancial contracting.
Finally, shirking is assumed not to be socially optimal (i.e. B<I), while the net value of the project under
diligence by farmers is strictly positive (i.e., p
H
R I>0).
The external nance necessary to the development of the project can be provided by three distinct
in vestors.
The local bank can provide I A to farmers at a gross rate normalized to 1. The bank is a passive
but rational investor; it extends a loan as long as it can recoup it in expectation. Banks are passive
in vestors in the sense that they do not supervise borrowing farmers. As a result, banks rely primarily
on collateral-based enforcement of their loans.
12
For applications of this fram ework to developing countries, see, for instance, Straub (2005) that explores informal credit
markets and Conn ing (1999), C onning and Kevane (200 3) and C onning a nd Ud ry (2006) on m icro nance.
13
In prac tic e , shirk in g farme rs m ay b e suc c es sf ul, alth o u g h at a lower rat e. T h i s assu mp tio n simplify th e an a ly tic a l result s
obtained, it does not aect their qualitative natu re.
7
Page 7
Moneylenders play are nancial intermediary. Their function is to monitor farmers and thereby alleviate
the moral hazard problem. In fact, by monitoring, moneylenders reduce the farmers’ opportunity cost
of being diligent by reducing the benetofshirkingtob,withB>b. Finally, they lend capital to
farmers at a gross rate γ>1.
The supermark et, like the moneylenders, plays the role of nancial intermediary. Unlike the "conven-
tional" moneylenders, it will not only monitor farmers, but also advise them on production. Advising
raises the probability of success of the project to p
H
+p
A
with p
H
+p
A
< 1.
14
Note that a supermarket
whose advice is worthless (i.e. p
A
=0)isequivalenttoamoneylender.
The proper implementation of the project requires that all agents be provided with adequate incen tives.
In particular, the contract design problem consists in optimally sharing the project return, R,amongthe
contracting parties. The optimal sharing rule is such that it guarantees the participation of all agents without
destroying incentives for diligent behavior.
To understand the rationale behind the bundling of monitoring and advising by the supermarket, t wo
diering organizations of production are considered: an organization where the supermarket acts as an
external consultant, restricting its action to advising farmers, and an organization where the supermarket
not only advises, but also acts as a moneylender.
It is important to understand that whether the supermarket takes on a monitoring role or not, farmers
always have the choice between two sources of nancing: direct nancing by the bank and indirect nancing
via a nancial in termediary (either the supermarket or a moneylender).
3.2 Monitoring and advising as separate tasks
We consider a rst organization of production, where the supermarket limits its role to advising and where
nancing remains in the hands of the moneylender and the bank. In this case, there are four parties involved
in the nancial contract: the farmer, the moneylender, the bank and the supermarket. Th us, the project
return R is divided up, so that
R = R
f
+ R
m
+ R
l
+ R
s
, (1)
where R
f
, R
m
, R
l
and R
s
denote the success-contingent stakes of the project obtained by the farmer,
the moneylender, the bank and the supermarket, respectively.
15
14
This additive sp ecication im plies that eort by the farmers and advising by the sup erm arket are not com plementary. T heir
joint realiza tion is not required to im plement th e project. Instead , each contributes separa tely to imp rove the pro tab ility of
the project. This may seem a very strong assumption, but it is d one on purp o se as assuming synergies would simply reinforce
our main results.
15
In c a se of f a ilu re , the projec t yield s no r et u rn. This imp lie s, tog e th e r with limited liab ility, th at each contrac ting party
receives no payment.
8
Page 8
3.2.1 Direct nancing
In the absence of moneylenders, the project return is only divided between the bank, the farmer and the
supermark e t (i.e. R
m
=0). As previously mentioned, in this nancial contract, the share received by each
part y should be such that it does not destroy each agent’s incentive.
Farmers. For a farmer to be diligent, he should receive at least
(p
H
+ p
A
) R
f
A p
A
R
f
+ B A,
or equivalently,
R
f
B
p
H
. (2)
This incentive compatibility constraint requires that the farmer earns at least as much from being diligent
as from shirking. Note that as a direct consequence of the substitutability of eort exerted by the supermark et
and the farmer, this constraint holds whether the supermarket is properly advising or not.
The supermarket. Advising by the supermarket is also subject to moral hazard. The opportunity
cost of advising for the supermarket is c>0. To guarantee proper incentives, the supermarket should
receive at least as much when being diligent as while shirking. The incentive compatibility constraint of the
supermarket is thus written as
(p
A
+ p
H
) R
s
c p
H
R
s
,
or equivalently,
R
s
c
p
A
. (3)
Bank. The banking sector is assumed perfectly competitive, and in order to accept a loan application
the bank should at least break-even. The break-even condition is expressed as
(p
A
+ p
H
) R
l
I A. (4)
The left-hand side of (4) is referred to as the expected pledgeable income, while the right-hand side is the
market value of the fund supplied by the local bank. The pledgeable income is the maxim um amount that
can be promised to investors without destroy ing the incentives of the agents in volved in the nancial contract
(here, the farmer and the supermarket). Giv en that R
m
=0and according to the sharing rule (equation
9
Page 9
(1)), it is straightforward that the maximum expected pledgeable income is (p
A
+ p
H
)(R R
s
R
f
).This
implies that
(p
A
+ p
H
)(R R
s
R
f
) I A. (5)
Furthermore, to make the analysis non-trivial, the following assumption is made regarding advising.
Assumption 1
p
A
(R R
f
R
s
) p
H
R
s
This assumption guarantees that the maximum pledgeable income for farmers is increased with advising
by the supermarket. This assumption implies that p
A
R
p
H
+p
A
p
A
c, which means that the rise in the net
present value of the project is superior to the minimum stake necessary to insure proper incentives from
the supermarket. In other words, the net present value of the project is increased by advising from the
supermarket. This assumption is necessary in order for farmers to be willing to participate.
3.2.2 Indirect nancing
Let us now consider the case where farmers borrow from an intermediate moneylender (i.e., R
m
0). The
moneylender is an active investor, in the sense that she visits and monitors farms to guaran tee that farmers
exert sucient eort.
The moneylender. As in Holmstrom and Tirole (1997), the monitoring activity is subject to moral
hazard. The opportunity cost of monitoring for the moneylender is m, and the incentive compatibility
constrain t of the moneylender requires that
(p
H
+ p
A
) R
m
m p
A
R
m
,
or equivalently,
R
m
m
p
H
. (6)
Farmers. By monitoring farmers, the moneylender reduces the farmers’ benets of shirking to b,with
B>b. Hence, assuming proper monitoring by the moneylender, the incentive compatibility constraint of
the farmer can be rewritten as
R
f
b
p
H
. (7)
10
Page 10
Bank. In the presence of intermediation via a moneylender, the break-even condition can be expressed
as
(p
A
+ p
H
)(R R
s
R
f
R
m
) I I
m
A. (8)
Here, I
m
denotes the amount of capital invested by the moneylender in the farm project that she monitors.
Finally, we make the following assumption.
Assumption 2
B b m
This assumption is necessary in order for intermediation by a moneylender to be a viable option. It
simply states that the reduction in the private benet of the farmers, B b, is greater than the cost of
monitoring, m. It is intuitive that under these conditions monitoring is socially desirable, since it increases
the number of nanced farmers. If this assumption does not hold, then moneylending will not be protable.
Proposition 1 In the money lending case, there exists two thresholds of liquid assets given by A
a
= I
(p
A
+ p
H
)
h
R
c
p
A
B
p
H
i
and A
am
= I I
m
(p
A
+ p
H
)
h
R
c
p
A
b+m
p
H
i
. When the supermarket behaves
as an external consultant (i.e. limits its activity to advising), the optimal contract between the farmer, the
bank, the moneylender and the supermarket has the following fe a tures:
when A A
a
, farmers borrow solely from the bank.
when A
a
A A
am
, farmers borrow from banks and the moneylender.
when A<A
am
, farmers do not have access to credit.
Moreover, as an external consultant, the advising rent of the supermarket is given by
Φ
a
=
p
H
p
A
c.
Proof. Farmers can either be directly nanced b y the bank or indirectly via a moneylender.
Direct nancing:
Let us rst consider the case where farmers are directly nanced. Substituting back (2) and (3) into (5)
implies that
A A
a
= I (p
A
+ p
H
)
R
c
p
A
B
p
H
¸
.
11
Page 11
In other words, to obtain a loan with direct nancing, farmers should justify a minimum lev el of assets,
at least equal to A
a
, to the bank.
Indirect nancing:
Farmers not directly nanced by the bank (i.e., such that A A
a
) can turn to a moneylender to obtain
a loan. By the same reasoning, substituting back (7), (6) and (3) into (8) implies that
A A
am
= I I
m
(p
A
+ p
H
)
R
c
p
A
b + m
p
H
¸
. (9)
Verifying that A
a
A
am
for any I
m
0, it follows that farmers with nance A, such that A
a
A A
am
,
are nanced via a moneylender. Finally, farmers with nance such as A A
am
do not oer a sucient
guaran tee to receive a loan either from the bank or the moneylender. Furthermore, note that to guarantee
the participation of the moneylender, it is necessary that (p
A
+ p
H
)
m
p
H
m+ I
m
; i.e. its expected rent from
moneylending is at least as high as its cost. Because a positiveshareoftheprojecthastobeforfeitedtoa
moneylender, when given the choice, farmers will seek to avoid borrowing from a moneylender.
Supermarket advising rent:
TherentofthesupermarketisgivenbyΦ
a
=(p
A
+ p
H
) R
s
c =
p
H
p
A
c.
It is important to understand that the lower the level of collateral necessary in order to be nanced (i.e.,
the lower A
am
), the greater the number of farmers nanced. For the sak e of comparison, the nancial contract
in absence of a supermarket has been reported in the appendix. It is straightforward that given Assumption
2, more farmers get nancing when advised by the supermark et (i.e. the level of nance necessary to access
credit is reduced when advising is part of the contract).
3.3 Bundling monitoring and advising
We now explore the possibility that the supermarket decides to take on two missions; namely, moneylending
and advising farmers. In practice, the nancing part often takes the form of an input advance on seeds,
pesticides or fertilizers.
16
Unlike the previous case, the multitasking nature of the supermarket now generates
several incentive constraints. First, the supermarket must be given reward R
s
, such that it does not want
to shirk on the advising task alone:
(p
A
+ p
H
) R
s
m
s
c p
H
R
s
m
s
,
16
Thefactthatthesupermarketoers inputs rather than cash has several rationales. First, there a re economies of scale in
procurem ent; sup erm arkets or grossists often serve several thousands of sm all growers. S econd, there is arguably less scope for
diversio n of physic a l in pu t s, a lt ho u g h it is still po s sib le t h a t fa rmer s may try to resell t h em in a seco n d a ry ma rket.
12
Page 12
or
R
s
c
p
A
. (10)
Note that we assume that the supermarket is as ecient as the moneylenders in the monitoring activity, and
we let m
s
= m. The supermarket must also monitor the farmer. Such monitoring includes, among other
things, making sure that the farmer does not divert the inputs, and following the farmer closely at harvest
time to make sure he does not resell his harv est to other retailers.
17
The incentive constraint is written as
(p
A
+ p
H
) R
s
m c p
A
R
s
c
or
R
s
m
p
H
. (11)
Finally, the supermarket can decide to shirk on both tasks, in which case the incentive constraint is
written as
(p
A
+ p
H
) R
s
m c 0
or
R
s
m + c
p
A
+ p
H
. (12)
Overall, the supermarket will be diligent in both tasks if constraints (10), (11), and (12) hold true. Thus,
the minim um stake consistent with supermarket diligence is
R
s
max
½
c
p
A
,
m
p
H
,
m + c
p
A
+ p
H
¾
. (13)
If we denote by I
s
the amount of capital invested by the supermarket in every farm that it monitors, we can
state the following result:
Proposition 2 When the supermarket not only advises, but also acts as a moneylender, there exists a
threshold of farmers’ asset, dened by A
S
am
= I (p
A
+ p
H
)
h
R
b
p
H
max
n
c
p
A
,
m
p
H
,
m+c
p
A
+p
H
oi
I
s
,such
that the optimal contract passed between the farmer, the bank and the supermarket has the following regimes:
when A A
a
, farmers borrow solely from the bank.
when A
a
A A
S
am
, farmers borrow from banks and the supermarket.
when A<A
S
am
, farmers do not get funded.
17
Mo re gen erally, the cost of m on itoring th e bo rrower in d eveloping co untries can rep resent up to 39% o f the a m ount lend ed.
See Aleem (1990) for a description of m onitoring practices.
13
Page 13
Proof. Here, the nancial contract involves three parties: the farmer, the bank and the supermarket.
Therefore, the return R is shared among them:
R = R
f
+ R
s
+ R
l
.
As in proposition 1, the farmer will not require a loan from the supermarket if A A
a
and prefers
direct nancing with the bank. However, when the farmer has insucient wealth (i.e., when A<A
a
), the
supermarket substitutes for the moneylender and oers credit and advising. The project is then funded if
(p
A
+ p
H
)[R R
s
R
f
] I A I
s
or
A A
S
am
= I (p
A
+ p
H
)
R
b
p
H
max
½
c
p
A
,
m
p
H
,
m + c
p
A
+ p
H
¾¸
I
s
. (14)
Yet, to understand the behavior of the supermarket, it is critical to derive the rent associated with
moneylending for both the moneylender and the supermarket.
3.4 Financial In termed iation
This section examines nancial intermediation by the moneylender and the supermarket. In particular, it
determines the amount of capital invested, as well as the gross rate c harged by these two nancial interme-
diaries.
Moneylender. By denition,
(p
H
+ p
A
) R
m
= γI
m
. (15)
Here, γ denotes the gross rate charged by the moneylender. Together with equation (6), this implies that
the amount of capital borrowed via intermediation by the moneylender should be at least
I
m
=
p
H
+ p
A
γp
H
m. (16)
In fact, all farms monitored will demand precisely this minimum level of capital. More would be exces-
sive ly costly and less would be inconsistent with proper incen tives from the moneylender. Furthermore, the
moneylender rent can be expressed as
14
Page 14
Γ
ml
=
(γ 1) (p
A
+ p
H
)
γp
H
1
¸
m. (17)
As a result, in its most general form, the gross rate charged by the moneylender can be expressed as
γ =
p
A
+ p
H
p
A
m p
H
Γ
ml
m. (18)
This gross rate captures dieren t forms of the moneylending market structure. For instance, Ho and
Stiglitz (1998) argue that there is monopolistic competition in moneylending. This case is captured b y
Γ
ml
=0and γ =
p
A
+p
H
p
A
; price competition between moneylenders is such that the gross rate is set to co ver
exactly the cost of moneylending (i.e., m + I
m
). How ever, it can also be argued that due to the scarcity
of moneylenders and the high transaction cost that plagues developing countries, moneylenders ma y enjoy
a niche market in nancial intermediation. Because it may be too costly for borrowers to seek another
moneylender, moneylenders enjoy a positive ren t Γ
ml
> 0. In that case, the gross rate charged is given by
equation (18).
Supermarket gross rate. Similarly,
(p
H
+ p
A
) R
s
= δI
s
. (19)
Here, δ denotes the gross rate charged by the supermarket. Using logic similar to the moneylender case,
the incentive constrain t of the supermarket implies that
I
s
=
(p
H
+ p
A
)
δ
max
½
c
p
A
,
m
p
H
,
m + c
p
A
+ p
H
¾
. (20)
Therefore, the rent of the supermark e t when advising and lending money can be expressed as
Φ
s
=(p
H
+ p
A
)
δ 1
δ
max
½
c
p
A
,
m
p
H
,
m + c
p
A
+ p
H
¾
(m + c) . (21)
The next Lemma, which will be needed later, claries when the supermarket behaves as a multitask
agency.
Lemma 3 The supermarket provides both advising and moneylending services if
max
½
c
p
A
,
m
p
H
¾
m + c
p
A
+ p
H
.
Proof. The participation of the supermark et is warranted if Φ
s
0. Assuming that max
n
c
p
A
,
m
p
H
,
m+c
p
A
+p
H
o
=
15
Page 15
m+c
p
A
+p
H
,thenwehave
Φ
s
=
µ
δ 1
δ
1
(m + c) < 0.
AccordingtoLemma3,when
m+c
p
A
+p
H
max
n
c
p
A
,
m
p
H
o
the supermarket will limit its activity to advising.
On the contrary, when
m+c
p
A
+p
H
< max
n
c
p
A
,
m
p
H
o
it may engage in moneylending activity. In this case, the
gross rate charged by the supermarket is expressed as
δ =
(p
H
+p
A
)c
(p
H
+p
A
)cp
A
[Φ
s
+(m+c)]
(p
H
+p
A
)m
(p
H
+p
A
)mp
H
[Φ
s
+(m+c)]
if
c
p
A
m
p
H
,
if
c
p
A
m
p
H
.
(22)
3.5 Bundling tasks to extend supply base
Let us assume that the objective of the supermarket is to extend its supply base. Tw o arguments can
justify the importance to the supermarket of having a large supply base. First, with a large base of farmers
the supermarket can achieve economies of scale. Another argument is that a supply base of numerous
farmers who are geographically dispersed acts as an eective mechanism to reduce the risk of widespread
crop failures due to disease and (to a lesser extent) weather, thus safeguarding the ability to fulll customer
orders (Henson, Masakure and Boselie 2005).
In this section, we show that by bundling advising and monitoring, supermarkets can extend credit access
to a larger number of farmers, and thereby increase its supply base. The next Proposition is one of the main
resultsofthispaper.
Proposition 4 (Monitoring and Advising) Assuming that the supermarket and the moneylender charge
the same gross rate (i.e., δ = γ), and that max
n
c
p
A
,
m
p
H
o
m+c
p
A
+p
H
, then a pro c urement organization in which
the supermarket bundles advising and monitoring strictly increases the number of farmers who get funded.
Thus, a larger number of farmers can supply the supermarket.
If max
n
c
p
A
,
m
p
H
o
m+c
p
A
+p
H
, then the supermarket only acts as an external consultant and moneylenders
provide nancial services to farmers.
Proof. According to Lemma 3, it is straightforward that to guarantee the participation of the supermarke t
it should be that max
n
c
p
A
,
m
p
H
o
m+c
p
A
+p
H
. Under this assumption, the number of farmers funded will strictly
increase if and only if A
S
am
<A
am
. First, substituting back (16) into (9) implies that
A
am
= I (p
A
+ p
H
)
R
c
p
A
b
p
H
γ 1
γ
m
p
H
¸
.
16
Page 16
Assuming that δ = γ, then according to (20), I
s
is given b y
I
s
=
(p
H
+ p
A
)
γ
max
½
c
p
A
,
m
p
H
¾
.
Bythesamereasoning,weobtain
A
S
am
= I (p
A
+ p
H
)
R
b
p
H
γ 1
γ
max
½
c
p
A
,
m
p
H
¾¸
.
Let us dene = A
am
A
S
am
,orequivalently
=(p
A
+ p
H
)
c
p
A
+
γ 1
γ
µ
m
p
H
max
½
c
p
A
,
m
p
H
¾¶¸
.
By denition, A
S
am
<A
am
is equivalen t to > 0. Assuming that
c
p
A
m
p
H
,then
=
p
A
+ p
H
γ
(γ 1) m
p
H
+
c
p
A
¸
> 0.
Similarly, assuming that
c
p
A
m
p
H
,then
=(p
A
+ p
H
)
c
p
A
> 0.
Therefore, we conclude that A
S
am
<A
am
.
Heuristically, when both tasks are exerted by the same agent, diligence in monitoring also favors diligence
in advising and the reverse also holds true.
Formally, bundling both monitoring and advising together results in a contract that reduces the minimum
rent necessary to insure proper incentives. To see this, note that the minim um rent necessary to insure proper
incen tives (i.e., with Φ
s
= Γ
ml
=0) when both advising and monitoring are exerted by separate agents is
MS =
p
H
+ p
A
p
A
c + m.
When both tasks are performed by the same agent, this rent reduces to
MB = c + m,
which implies that
17
Page 17
MS MB =
p
H
p
A
c. (23)
3.6 Bundling tasks to raise prots
The previous section made several strong assumptions about the beha vior of the supermarket and the
moneylender. First, the supermarket and the moneylender were assumed to charge the same gross rate. In
reality, this need not be the case, given the obvious discrepancies between a moneylender and a supermarket.
Second, in the last section, the goal of the supermarket was to increase its supply base. Although the
existing literature makes a strong case for the existence of signicant gain associated with scale economy in
procurement, one can legitimately argue that keeping a large supply base of producers ma y be extremely
costly for the supermarket. In fact, as is explicit in Lemma 3, the supermarket may sustain losses when
max
n
c
p
A
,
m
p
H
o
m+c
p
A
+p
H
. In addition, as shown in Proposition 1 with a procurement organization only
advising farmers, a supermark et will earn a positive rent Φ
a
=
p
H
p
A
c. Hence, it might be more protable for
the supermark et to delegate monitoring to a moneylender and to restrict its role to advising.
In this section, we generalize our results to the case where the supermarket seeks to maximize its prot
and where the gross rate charged by the supermarket and moneylender does not necessary match.
Proposition 5 (Supermarket Pa rticipation) Assuming that max
n
c
p
A
,
m
p
H
o
m+c
p
A
+p
H
,therealwaysex-
ists a range of gross rates that can be charged by the supermarket, under which all farmers seek to borrow
from the supermarket and the supermarket prefers a procurement organization that advises and monitors, to
a procurement organization that only advises.
Proof. Farmers’ borrowing preferences
To understand their preferences, we rst need to determine farmers’ expected rent from borrowing either
from a moneylender or the supermarke t. When farmers borrow from a moneylender, the supermarket
still advises them on their production practices. In that case, the expected rent perceived by a farmer
corresponds to the expected net return of the project, less not only the share perceived by the moneylender
and supermarket to properly monitor and advise, but also their respective opportunity costs. Mathematically,
this corresponds to
U
f|ml
=(p
A
+ p
H
) R I
Γ
ml
+ m +(p
A
+ p
H
)
c
p
A
¸
, (24)
where, U
f|ml
denotes the farmer’s expected rent when borrowing from a moneylender. Similarly, when a
farmer borro ws from the supermarket, his rent has to be forfeited by the supermarket’s expected net return,
18
Page 18
as well as its opportunity costs of monitoring and advising, i.e.,
U
f|s
=(p
A
+ p
H
) R I [Φ
s
+ m + c] , (25)
where U
f|s
denotes the farmer’s expected rent when borrowing from the supermarket. Therefore, a farmer
will seek to borrow from a supermarket when U
f|s
>U
f|ml
. According to (24) and (25), this will be the case
as long as
Φ
s
Γ
ml
+
p
H
p
A
c. (26)
Supermarket preference
To conclude the proof, we need to show that there exists a range of gross rates charged b y the supermarket
under which equation (26) is veried and the supermarket prefers a procurement organization that advices
and monitor.
In a procurement organization that only advices, the rent of the supermarket is Φ
a
=
p
H
p
A
c.Thus,to
guaran tee the participation of the supermarket and condition (26), it is straightforward that Φ
s
should be
such that
p
H
p
A
c Φ
s
Γ
ml
+
p
H
p
A
c. (27)
It is important to understand that condition (27) is very general and holds true for any combination
of gross rate charged by either the moneylender or the supermarket. Furthermore, according to equation
(22), a monotonic and continuous relationship between δ and Φ
s
exists . Therefore, a range of gross rate
exists that can be charge d by the supermarket that veries both conditions (27) and (26). Given that Φ
s
is
positively correlated with δ, the minimum gross rate that can be c harged by the supermark et is obtained by
substituting back Φ
a
in to equation (22), i.e.,
δ
min
=
(p
H
+p
A
)c
(p
H
+p
A
)cp
A
[Φ
a
+(m+c)]
(p
H
+p
A
)m
(p
H
+p
A
)mp
H
[Φ
a
+(m+c)]
if
c
p
A
m
p
H
,
if
c
p
A
m
p
H
.
For this gross rate, the supermarket will be indierent between the two procurement organizations and
all farmers will seek to borrow from it rather than a conventional moneylender. By the same reasoning,
according to (26), the maximum rent that can be perceived by the supermark et is Φ
max
s
= Γ
ml
+
p
H
p
A
c.In
that case, farmers are made indierent between borrowing from a moneylender and a supermarket. Again,
19
Page 19
the maximum gross rate that can be charged is obtained by substituting back Φ
max
s
in to equation (22), i.e.,
δ
max
=
(p
H
+p
A
)c
(p
H
+p
A
)cp
A
[Φ
max
s
+(m+c)]
(p
H
+p
A
)m
(p
H
+p
A
)mp
H
[Φ
max
s
+(m+c)]
if
c
p
A
m
p
H
,
if
c
p
A
m
p
H
.
Note that the range of gross rates that can be charged by the supermarket will depend on the size of the
moneylender’s rent. In particular, when Γ
ml
=0, then δ
min
= δ
max
. Intuitiv ely, by bundling both tasks the
supermarket can capture a share of the moneylender rent in addition to its advising rent.
While Proposition 6 shows that supermarkets always have a nancial interest in implementing procure-
ment system that bundles advising and monitoring, the impact of such organization on farmers’ credit access
remains to be assessed.
Corollary 6 Assuming max
n
c
p
A
,
m
p
H
o
m+c
p
A
+p
H
, then a procurement organization that advises and monitors
extends the number of farmers nanced,aslongasitdoesnotchargeitsmaximumgrossrate.
Proof. First, note that according to equation (17), A
am
can be rewritten as
A
am
= I (p
A
+ p
H
)
R
c
p
A
b
p
H
¸
+ Γ
ml
+ m.
By the same reasoning, according to (21), A
S
am
can be rewritten as
A
S
am
= I (p
A
+ p
H
)
R
b
p
H
¸
+ Φ
s
+ m + c.
Clearly, in a procurement organization that bundles advising and monitoring, the number of farmers
nanced will strictly increase (i.e., A
S
am
>A
am
), as long as the supermarket rent is such that Φ
s
< Γ
ml
+
p
H
p
A
c = Φ
max
s
.
The reasoning behind these results is as follows. By bundling both the monitoring and advising, the
supermarket can squeeze out the rent of the moneylender. As a result, a lower share of the project has to
be forfeited to insure proper incentives. This agency cost reduction provides farmers with a larger share of
the project and makes the supermarket nancial service more appealing to farmers.
According to corrolary 6, the development of the supermarket procurement system is conduciv e to the
extension of credit to smaller farmers (i.e., with lower levels of nance),aslongasthesupermarketdoesnot
capture the entire benet from the agency cost reduction. Such scenario implicitly puts all the bargaining
power in the hands of the supermarket. In reality, farmers always have the option to turn to conventional
20
Page 20
moneylenders, making such scenario unlikely. In fact, the observed long waiting list to enter into the
supermarket procurement system corroborates strong farmer preference to ward supermarket procurement
system (on this issue, see Henson, Masakure and Boselie, 2005).
4Smallfarmersworkharder
So far, our paper has analyzed the supermarket procurement system and its implications, notably on access
to credit by farmers. It has been shown that one important feature of the supermarket procuremen t system
is that by bundling monitoring and advising it may broaden access to credit for smaller producers. In
particular, smaller farmers (i.e. with lower nancial capac ity) can, with the supermarket, receive credit in
the form of agricultural inputs (such as fertilizer applications).
Although the early empirical literature has found that supermarkets seem to prefer to procure their
products from large producers (see, for instance, Dolan, Humphrey and Harris-Pascal 2001; Fernando et al.
2003; Reardon et al. 2003 and Brown and Sander 2007). More recent case studies have contrasted this view
and argue instead that, absen t substantial economies of scale, supermarkets actually like to procure from
small farmers (Boselie, Henson and Weatherspoon (2003)). For instance, Henson, Masakure and Boselie
(2005) suggest in a case study that small farmers are indeed competitive. They write, “small producers
supplying ‘Hortico’ have managed to achieve lower rejection rates for certain nontraditional vegetables than
large scale farmers.” Moreover, they emphasize how fast producers of dierent scales learn. The authors
write that “small producers consistently perform as well as large producers, if not better.” Therefore, unless
there exists substantial economies of scale in production, small producers seem to put more care and eort
into production.
While a compelling explanation for the use of small farmers by supermark ets may be related to their
growing concerns for a socially responsible image, in this section, we demonstrate that the reasons for
supermarkets to extend market access to smaller farmers can be motivated by the greater willingness of
smaller farmers to exert additional eort.
This section develops a variant of our model that focuses on farmers’ eorts. Unlike in the previous
section, farmers can now c hoose among three possible eort levels. In addition, we also assume that, besides
liquid assets A
l
, farmers hold illiquid assets A
i
. The latter assets are t ypically made up of land, farm buildings
or personal homes. For our purposes, it is important to recognize that a social loss is often incurred whenever
these assets change hands.
18
For instance, a farmer may have a sentimen t al attac hment to the family farm,
but the housing market will not reect this specic valuation; or he may have developed specichuman
18
Kranton and Swamy (1999) made a similar assumption to a nalyze m oneylending in India. Furthermore, the imp ortance
of farm-sp ecic human capital has b een em pirically shown by Rosenzweig and Wolpin (1985).
21
Page 21
capital to operate machines and this knowledge is lost whenever the item is seized. This assumption is
made, for instance, by Kranton and Swarmy (1995). Formally, we let βA
i
denote the market value for the
moneylender of an illiquid asset that a farmer values A
i
. As such, whenever illiquid assets change hands, a
social loss of (1 β) A
i
occurs.
Farmers can now exert three levels of eort: high, medium and low. The high level of eort requires that
the farmer to privately incur an additional cost of K>0. The net monetary surplus of the project is, thus,
as follows:
U
f
=
U
H
= p
H
R (I + K)
U
M
= p
M
R I
U
L
= B I
if high eort is exerted,
if medium eort is exerted,
if low eort is exerted.
Here, p
H
and p
M
denote the probability of success for high and medium levels of eort, respectively. To
make the problem interesting, the following assumption is made.
Assumption
K
p
H
p
M
>R.
This assumption implies that it is socially optimal to exert the medium level of eort or, in other words,
all farmers prefer a medium level of eort.
Yet farmers are nancially constrained and must obtain external funding. External funds are borrowed
from a lender that can, without loss of generality, be referred to as a bank, a moneylender or a supermarket.
However, when farmers do not have enough cash to fund the project, the lender will provide nancing only if
the farmers collateralize their assets. In the presence of collaterization, the farmers’ utility can be expressed
as
U
f
=
U
H
= p
H
R (I + K) (1 p
H
)(1 β) A
i
U
M
= p
M
R I (1 p
M
)(1 β) A
i
if high eort is exerted,
if medium eort is exerted.
(28)
Lemma 7 Under asset collaterization, a farmer prefers to exert medium eort if and only if
A
i
A
i
=
1
(1 β)
µ
K
p
H
p
M
R
. (29)
Proof. The proof is by comparing the two expressions of equation (28).
Therefore, farmers who have to collateralize a few illiquid assets (i.e., when A
i
< A
i
), will choose to exert
medium eort, while highly collateralized farmers will choose to exert a high level of eort that decreases
the likelihood that their asset will be seized. The next result characterizes the optimal nancial contract.
19
19
Ou r optimal contract, which fea tures continge nt c o lla te riz a tio n is, in fac t, domin a te d by stochas tic nancial contracts.
22
Page 22
Proposition 8 (“Small farmers work harder”) The optimal nancial contract between the farmer and
the lender has the following features.
When the farmer has important liquid assets, i.e.. when A
l
A
l
with A
l
= I p
M
³
R
B
p
M
p
L
´
,he
obtains nancing and exerts medium eort.
When the farmer has intermediate liquid assets, i.e. when A
l
A
l
A
l
with A
l
= A
l
³
β
1β
+ p
M
´³
K
p
H
p
M
R
´
,
he exerts medium eort and obtains nancing by pledging 0 A
M
i
A
i
,whereA
M
i
=
A
l
A
l
p
M
+(1p
M
)β
.
The contract stipulates that illiquid assets are seized if the project fails, and are otherwise kept.
When the farmer has weak liquid assets, i.e., when A
l
<A
l
, he exerts high eort and obtains nancing
by pledging A
H
i
> A
i
where A
H
i
=
A
H
l
A
l
p
H
+(1p
H
)β
and A
H
l
= Ip
H
³
R
B+K
p
H
´
. The contract stipulates
that the assets are seized if the project fails and are kept otherwise.
Proof. With no collaterization, a farmer exerts medium eort if
p
M
R
b
p
L
R
b
+ B, (30)
or if
R
b
R
M
b
=
B
p
M
.
The break-even constraint holds when
p
M
¡
R R
M
b
¢
I A
l
,
or if
A
l
A
l
= I p
M
µ
R
B
p
M
.
When A
l
A
l
, liquid assets are no longer sucient and collaterization of A
i
transforms (30) into
p
M
R
b
(1 p
M
) A
i
p
L
R
b
(1 p
L
) A
i
+ B
or
R
b
R
M
b
=
B
p
M
A
i
. (31)
This more g ene ral class of co ntract would e ntail probab ility of a sset seizu re in e ach state of nature. We refrain from making
this more general analysis for two reasons. First, the analysis would be m ore involved without adding any further insight.
Second stochastic nancia l co ntra c ts s ee m less intuit ive, as in pra c tic e, wh oe ver w rit e s th e m must c ommit to these rando m
asset seizures.
23
Page 23
Using (31), it is seen that collaterization relaxes the lender’s constraint, as we now have
p
M
µ
R
B
p
M
+ A
i
+(1 p
M
) βA
i
I A
l
or
A
l
A
l
(A
i
)=A
l
[p
M
+(1 p
M
) β] A
i
, (32)
which is strictly lower than
A
l
for any A
i
> 0. To minimize the social loss, the amount of illiquid assets
pledged is optimally set to its minimum level. This level must verify the lender’s constraint (32) and we
obtain
A
M
i
=
A
l
A
l
p
M
+(1 p
M
) β
. (33)
As expected, according to (33), the size of the illiquid collateral is negatively correlated with the size of the
farmer’s liquid asset, A
l
. It is important to note that the relationship between liquid and illiquid assets
captured by equation (33) is only valid for farmers exerting a medium level of eort. When farmers have
a shallow pocket, i.e. when A
l
is smaller, the amount of collateralized assets reaches a level at which, by
Lemma 6, they prefer to exert a high level of eort. Using (29) and (32), we obtain
A
l
= A
l
µ
β
(1 β)
+ p
M
¶µ
K
p
H
p
M
R
.
When liquid assets fall further below A
l
, then farmers must pledge more than A
i
,andthebreak-even
constrain t is thus written as
p
H
µ
R
B + K
p
H
+ A
i
+(1 p
H
) βA
i
I A
l
,
with the relationship between liquid and illiquid assets now given by
A
H
i
=
I p
H
¡
R (B + K) /p
H
¢
A
l
p
H
+(1 p
H
) β
.
The reasoning behind this proposition is as follows. Although every farmer (rich or poor) would strictly
prefer to exert medium eort, those who are poor have to rely on illiquid assets to obtain loans. However,
reliance on these assets creates a social loss and distorts eort toward higher levels. In essence, small farmers
make up for their lack of assets by exerting a higher level of eort. Interestingly, the empirical analysis by
Foltz (2004) of Tunisian agriculture estimates that the amoun t of owned land is a signicant determinant
24
Page 24
of overall prots. In particular, although credit-rationed farmers have less owned land, it has a signicantly
greater impact on prot for credit-rationed farmers. These estimates are in line with the idea of small
producers exerting a higher level of eort.
If we speculate that supermarkets derive some unmodelled non-monetary benetfromhavingabetter
quality more often (i.e., a higher probability of success) then the fact that they tend to repeatedly employ
small farmers is not surprising. From that perspectiv e, the policy advertised by many supermarkets to
procure agricultural products only from small farmers may appear much less altruistic than it sounds.
Man y high-end fresh agricultural products require labor-intensive techniques. Proposition 8 shows that the
supermarket can exert a greater pressure on smaller farmers by threatening to seize their assets, thereby
reinforcing farmers’ incentives to work harder when compared to large scale producers. While it is true that
supermarkets stamp their authority on the supply chain b y monitoring and auditing production, as argued
b y (Brown and Sander, 2007), this is in the interest of both the farmers (who ha ve access to more nancing)
and the supermarkets.
While the latter two sections have focused on the organization of the supermark et procurement system,
the next section analyses another prominent feature of supermarkets; namely their capacity to set higher
food standards.
5 The superm ark et private standard
The spread of supermarkets to developing countries arguably aects the nancing of farmers. However,
supermarkets have an increasing inuence on developing countries, not only through their investments, but
also through the imposition of their private standards (Reardon and Farina 2002, Reardon and Berdegué
2002, Berdegué et al. 2003, Fernando et al. 2003 and Weatherspoon and Reardon 2003). To fully understand
the impact of the supermarket on farmers in developing countries, the implications of the introduction of
higher food standards cannot be ignored.
With a variant of our modeling framework, this section seeks to understand the link bet ween the intro-
duction of more stringent food standards downstream, and access to credit by upstream farmers. Formally,
this is achieved by endogenizing the return R of the project.
As before, production requires a xed investment I and farmers are nancially constrained. Their level
of nance, denoted by A, is assumed to be uniformly distributed between 0 and
¯
A,where
¯
A>0 denotes the
level of nance of the wealthiest farm. Furthermore, for tractability, farmers produce at most one unit of
agricultural product.
Unlike in the previous section, farmers now have the choice between two projects:
25
Page 25
Thedomesticretailerproject. Thedomestic retailer represents local retail shops The products sold
b y this domestic retailer follow the country’s food standards, α
d
, that are set by a public agency.
The probabilit y of success of a diligent farmer in meeting the public requirements is p
H
, while the
opportunity cost of eort is increasing with the stringency of the food standard and is given by α
d
B.
The overall return of the domestic retailer project is R
d
.
The supermarket project. The supermarket follows its own private standard, α
s
which is assumed
higher than the public food standard. Hence, the opportunity cost of eort for farmers is increased to
α
s
B, but it is assumed that the probability of success for farmers when exerting eort remains the same.
In addition, to comply with supermarket requirements, an upfron t investment C in infrastructure is
necessary. Indeed, the supermarket usually imposes substantial investme nts, such as irrigation systems,
greenhouses, trucks, cooling sheds and packing technologies, among other things. Finally, the o verall
return of the supermarket project is R
s
.
Although the supermarket ma y nance a fraction of its project, for tractability, it is assumed that both
projects are directly nanced by the bank. Again, farmers are protected by limited liability and diligence by
farmersissociallydesirable.
In the downstream market, consumers who are heterogeneous in terms of preferences make their pur-
chasing decisions while observing the price and nature of the products (i.e., public/supermarket private
standard). Consumers are modeled in the spirit of Mussa and Rosen (1978) and we denote by θ the
consumer-dierentiating attribute. For tractability, this attribute is assumed to be uniformly distributed
between [0,M] and each consumer purchases at most one unit of the good. Thus, M also denotes the
maximum market size. The utility of consumers with the dierentiat ing attribute θ is given by
U =
θα
s
R
s
θα
d
R
d
0
if purchased from the supermarket,
if purchased from the domestic retailer,
otherwise.
(34)
For tractability, the marginal cost of producing one unit of agricultural product is normalized to zero.
Thus, the project returns, R
s
and R
d
, also denote the per unit retail price charged by the supermarket and
the domestic retailer, respectively. Based on expression (34), the demand for the domestic retailer product
is
D
d
=
R
s
R
d
α
s
α
d
R
d
α
d
0
if R
s
>
α
s
α
d
R
d
,
otherwise.
(35)
26
Page 26
On the other hand, the quantit y demanded for the supermarket product can be expressed as
D
s
=
M
R
s
R
d
α
s
α
d
M
R
s
α
s
if R
s
>
α
s
α
d
R
d
,
otherwise.
(36)
Note that because the supermarket and domestic retailer products are assumed to be vertically dieren-
tiated, to insure the coexistence of both mark ets it is necessary to have R
s
>R
d
.
Before concluding this section, it is interesting to note that 1 can be interpreted as the marginal utility
of income (Tirole, 1988). As reported in the literature (see for instance, Reardon et al. 2003 and Trail
2006), with urbanization, the emergence of a wealthier social class is a major determinant of the diusion
of supermarkets in developing countries. The current framework captures the fact that consumers with
the lowest marginal utility of income (i.e., the highest θ) are more eager to consume the supermarket food
products.
5.1 Financing under coexistence
Let us consider the case where there is coexistence of both the domestic retailer and supermarket marketing
channels (i.e., R
s
>
α
s
α
d
R
d
). As before, the return of the supermarket project has to be shared, such that
R
s
= R
f
+ R
l
+ π
s
.
Here, π
s
denotes the rent of the supermarket. Again, to guarantee proper eort by the farmer, he should
be provided at least p
H
R
f
α
s
B. The bank’s break-even condition implies that
p
H
(R
s
R
f
π
s
) I + C A,
or equivalently,
A A
s
= I + C + α
s
B + p
H
π
s
p
H
R
s
. (37)
Recall that C denotes the necessary upfront investment to produce supermarket products. Likewise, to
be granted nancing for the domestic retailer project, the farmer’s level of nance has to be such that
A A
d
= I + α
d
B + p
H
π
d
p
H
R
d
. (38)
Here, π
d
denotes the rent of the domestic retailer. The next proposition summerizes our ndings.
27
Page 27
Proposition 9 (Marke t Segmentation) Assuming co existence of both marketing channels domestic re-
tailer and supermarket the nancial contract between farmers, the bank, the domestic retailer and the
supermarket is as follows:
when
¯
A A A
s
, farmers borrow from the bank and supply the supermarket procurement system,
when A
s
A A
d
, farmers borrow from the bank and supply the domestic retailer procure ment system,
when A A
d
, farmers have no access to credit and are excluded from the marketing systems.
Furthermore, the expected rent of the supermarket can be expressed as
p
H
π
s
= p
H
R
s
+
R
s
R
d
α
s
α
d
¡
I
¯
A
¢
C α
s
B M, (39)
while domestic retailer expected rent is given by
p
H
π
d
=
µ
p
H
+
1
α
d
R
d
α
d
B
¡
I
¯
A
¢
M. (40)
Proof. Given that farmers are uniformly distributed and produce at least one unit, the number of farmers
who supply the supermarket also denotes the total quan tity supplied by the supermark et. This quantity is
given by
S
s
=
¯
A A
s
,
or using (37),
S
s
= p
H
R
s
¡
I
¯
A
¢
C α
s
B p
H
π
s
. (41)
When A A
s
farmers do not receive funding to supply the supermarket marketing channel. However, they
can still receive funding to supply the domestic retailer as long as A A
d
. Thus, with the entry of the
supermarket, the total quantity supplied to the domestic retailer becomes
S
d
= A
s
A
d
,
or using (38),
S
d
= p
H
R
d
p
H
R
s
+ C +(α
s
α
d
) B + p
H
π
s
p
H
π
d
. (42)
At equilibrium, both the market for the domestic retailer and for the supermarket products clear, such
that D
d
= S
d
and D
s
= S
s
. Substituting back the demand expressions (equations (36) and (35)) into the
28
Page 28
respective supply equations (equations (41) and (42)), the expected rent of the supermarket is expressed as
p
H
π
s
= p
H
R
s
+
R
s
R
d
α
s
α
d
¡
I
¯
A
¢
C α
s
B M,
while domestic retailer expected rent is given by
p
H
π
d
=
µ
p
H
+
1
α
d
R
d
α
d
B
¡
I
¯
A
¢
M.
The empirical literature describing the emergence of supermark ets in developing countries has forcefully
argued that supermarkets tend to contract with large, wealthy farmers, while poorer farmers are left behind
(see for instance, Dolan and Humphrey 2000 and Dolan, Humphrey and Harris-Pascal 2001). Our model
underlies the simple logic behind these observations: supermark ets set up high standards that often require
substantial investment on the part of farmers. Thus, only wealthy farmers are able to nance the upfront
in vestment, while poorer ones are credit-rationed and will, as such, turn to the traditional segment.
Furthermore, this literature has also emphasized that the domestic retailing sector is often jeopardized
b y the entry of supermarkets. The supermarket private standards are in general more stringent than public
standards and erce competition from supermarkets can drive out domestic retailers. For example, following
the emergence of supermarkets 64,198 small shops went out of business in Argentina from 1984 to 1993, and
5240 small shops closed their doors in Chile from 1991 to 1995 (Reardon et al., 2003).
Lemma 10 The minimum price that can be charged by the domestic retailer is given by
¯
R =
α
d
[
α
d
B+
(
I
¯
A
)
+M
]
α
d
p
H
+1
.
For any price greater than
¯
R, the domestic retailer and the supermarket can coexist.
Proof. The proof is straightforward from equation (40).
According to Lemma 10, the survival of the domestic retailing sector is contingent on the domestic
project’s return, which ultimately depends on the equilibrium price. Therefore, to assert the impact of
supermarket entry, it is necessary to determine the eects of supermarket entry on retail prices. The next
section studies this question.
5.2 M arket structure and project returns
Assuming coexistence of both marketing channels (i.e., R
s
>
α
s
α
d
R
d
), rst note that giv en equations (35) and
(36), the supermarket project’s return can be written as
R
s
= α
s
M α
d
D
d
α
s
D
s
. (43)
29
Page 29
Similarly, for the domestic retailer it corresponds to
R
d
= α
d
M α
d
D
d
α
d
D
s
. (44)
Th us, the size of the market, as well as consumer preference for eac h product, determines the returns
for each project. Furthermore, while the supermarket is assumed to behave as a monopolist, the domestic
retailing sector is composed of symmetric retailing shops. The problem for a retail shop can be written as
max
d
d
{R
d
d
d
} ,
or equivalently,
max
d
d
{α
d
(M D
d
D
s
) d
d
} .
Here, d
d
denotes the quantity sold by a single retail shop. The rst order condition implies that
α
d
M α
d
(1 + χ) D
d
α
d
D
s
=0, (45)
where χ =
∂D
d
∂d
d
d
d
D
d
denotes the conjectural variation elasticity of demand. Assuming that there are N
domestic retailing shops, given that D
d
=
P
i=N
i=1
d
di
,
∂D
d
∂d
di
=1+
P
j6=i
∂d
dj
∂d
di
.Intheory,
∂d
dj
∂d
di
represen ts
the "conjecture" of rm i regarding how rm j will react to an increase in quantity by rm i.Itcanbe
divided into three classes. Negative values of the conjecture indicate adaptive behavior, a zero value may
refer to a Cournotian behavior and positive values denote expectation matching behavior. Assuming that
individual rms anticipate other rms’ behaviors, reasonable boundaries for conjecture can be set at 1
and 1 (Anderson 1977). For instance, under Cournot conjecture
∂d
dj
∂d
di
=0, while under Bertrand conjecture
∂d
dj
∂d
di
=
1
N1
.Giventhat
d
d
D
d
denotes the market share, the conjectural variation elasticity will take a value
between 0 and 1. The parameter χ allows an examination of various types of strategic interactions among
domestic retailing rms. If, for instance, N rms compete in quantity (i.e., Cournot-Nash competition),
d
d
D
d
=
1
N
and
∂D
d
∂d
d
=1(given that Cournot conjecture imposes that
∂d
dj
∂d
di
=0); thereby, χ =
1
N
. Similarly, if
N rms are engaged in price competition and have no capacity constraints, χ =0(given that
∂d
dj
∂d
di
=
1
N1
,
then
∂D
d
∂d
d
=0). Obviously, χ =0also captures the perfectly competitive outcome. Finally, if the rms
collude and act as a monopoly (expectation matching behavior), χ =1.
20 ,21
20
Conjecture and conjectural va riation elasticities have b een the subject of confusion in notation and interpretation. For a
goo d review and discussion, see Lavoie, 2005.
21
W h ile we h ave ass u med th at do mest ic reta ilin g rms are symm etric and hold th e same conjectu re, mo re g en erally, χ =
S
N
i=1
s
i
κ
i
,wheres
i
denotes the market share of rm i and κ
i
=
∂D
d
∂d
di
(see, for instance, Porter 1983).
30