Bank-Moneylender Credit Linkages:
Theory and Practice∗
The Bush School of Government and Public Service,Texas A&M University
Bush School Working Paper # 415
No part f the Bush School transmission may be copied, downloaded, stored, further transmitted, transferred, distributed,
altered, or otherwise used, in an form or by an means, except: (1) one stored copy for personal use, non-commercial use,
or (2) prior written consent. No alterations of the transmission or removal of copyright notices is permitted.
Bank-Moneylender Credit Linkages: Theory and
Bush School of Government and Public Service
Texas A & M University
College Station, TX 77843-4220
Phone: (979) 458-8015
Fax: (979) 845-4155
I would like to thank Jonathan Conning and especially Mark Schreiner for helpful comments.
All errors remain my own.
Bank-Moneylender Credit Linkages: Theory and Practice
Abstract — This paper critically reviews proposals for banks and moneylen-
ders to link together in disbursing credit to rural areas of developing countries.
The linkages suggest that banks should compensate moneylenders according to
the moneylenders’ opportunity costs and information contribution. These mech-
anisms’ appeal lie in their self-equilibrating and self-sustaining character. With
these attractive features, bank-moneylender linkages can emerge as a serious al-
ternative to group lending-based microﬁnance. The paper also provides evidence
primarily from Indonesia on incentives similar to those suggested by the theoreti-
cal models. It concludes that with the appropriate regulation of informal lenders
and with incentives provided to commercial banks, linkages provide an unexplored
In meeting the credit demand of farmers, governments in developing countries
have sponsored formal institutions in environments where private banks reluc-
tantly enter on their own. These institutions have performed poorly.1 Many of
their failures can be traced to the diﬃculties associated with disbursing and col-
lecting credit in risky agricultural environments. Informal private lenders such as
moneylenders, friends, relatives, and landlords can overcome some of the lending
constraints. By residing close to villagers and free of the bureaucratic layers of
formal creditors, these private lenders can meet demand in a quick and ﬂexible
manner. Private lenders are still limited by the size of the market. Formal lenders
can rely on nationwide funds from savings mobilization and re-ﬁnancing support
With comparative advantages in diﬀerent areas, linking the formal and infor-
mal sectors can improve the disbursal of credit to farmers. Linkages fall under
two categories: explicit or implicit. Under the former, formal lenders actually hire
other lenders. Under the latter, formal lenders recognize that borrowers resort
to toher lenders as well and incorporate that information in their lending deci-
For example see the evidence in Adams, et al. (1984).
sions. In creating linkages, formal lenders must structure incentives for informal
lenders to cooperate and not collude with borrowers. Thus, lenders face a mech-
anism design problem and the recent development of principal-agent models aid
in formalizing these incentives. In this respect, the presented models diﬀer from
previous literature in that banks act as active proﬁt maximizing participants.2
Linkages would exploit the advantages of each sector. For example, banks could
issue large production loans and request moneylenders to monitor and enforce that
loan. In monitoring the loan, moneylenders adapt their own ﬂexible practices to
the “bank” loan. In this manner, banks access borrowers to whom they would
otherwise not lend and borrowers access loans that would otherwise be beyond
Surprisingly, the rural credit markets literature does not systematically address
the linkages potential. Morduch addresses this lacuna in a review of alternative
mechanisms to microﬁnance and states that, “Unfortunately, for now policymakers
have little to go on beyond a handful of small-scale case studies and ... theoretical
Hoﬀ and Stiglitz (1998) provide a model with passive formal lenders. Their model and
similar types will not be covered here. They do not address the issue of interaction because
they do not explicitly model the formal lender’s behavior but treat it as ﬁxed. By isolating
the economic proﬁtability of bank lending, the approach in this paper is also more relevant to
ﬁnancial liberalization eﬀorts (for example in India, see RBI (1998)). Currently, banks in many
developing countries face a soft constraint in that they rely on reﬁnancing their deposits from
governments but non-proﬁtable banks do not usually obtain continuing funding
examples and counter-examples.”3 In all fairness, the literature mentions linkages
but in disparate areas and has not discussed its potential rigorously.4 For example,
consider the following suggestion: “Better linkages would enable banks to beneﬁt
from the outreach and local knowledge of informal lenders, ... improving the
overall eﬃciency of the ﬁnancial system.”5
This suggestion begs the following questions: how should policymakers con-
struct linkages ? Under which circumstances will banks willingly participate in
linkages ? What are some constraints that prevent these linkages from being con-
structed ? This paper will provide theoretical and empirical support to answer
these questions. The search for eﬀective linkages forms part of a wider program in
the microeconomics of development. This new line of research recognizes that in-
stitutions arise to exploit their relative advantages and respond to the constraints
around them. Linkages can help bridge the persistent dualism in many developing
countries by providing a step in the development process.
Consider a formal lender in rural credit markets. As outlined by others, for-
mal lenders face the following problems : at loan disbursal, they face diﬃculty in
Morduch (1999), p. 1575.
Mahabel (1954) is an early qualitative discussion of linkages. Linkages are mentioned only
in passing in Hulme and Mosley’s (1996ab) two volume study of rural ﬁnance for the poor.
Linkages do appear in a separate subsection in Ray’s 1998 textbook, Development Economics.
Steel, Aryeety, Hettige, and Nissanke (1997), p. 827.
diﬀerentiating between good and bad borrowers (adverse selection).6 While bor-
rowers use the loan, lenders cannot verify borrowers’ dedication to their projects
since they may divert the production funds (moral hazard). Ex-post, lenders face
diﬃculties observing and verifying output to a third party (costly state veriﬁca-
tion) and extracting repayments (enforcement). A comparative analysis of four
proposed linkages in the literature reveals that a well structured incentive system
can potentially overcome these four problems. The linkages suggest that formal
lenders should compensate informal lenders according to the informal lender’s
opportunity costs and information contribution.
Linkages provide an alternative to the more popular solution to credit dis-
bursal, joint-liability lending or group lending (hereafter JLL). Most microﬁnance
organizations adopt JLL as opposed to individual lending policies. The survey by
Ghatak and Guinnane (hereafter G-G) denote this practice as a primary reason
for their success. This paper follows a similar structure to G-G but focuses on
aspects of lending other than JLL, which has been exhaustively covered in the
two major surveys by both G-G and Morduch. As Conning and Fuentes (2000)
note, even JLL institutions require a staﬀ member to monitor and oversee the
group. This paper then also serves in evaluating contracts where a microﬁnance
For example, see Besley (1994), Ghatak-Guinnane (1999), and Ray (1990).
institution hires oﬃcers that have greater information and enforcement powers.
Thus, this paper both complements previous work and provides an alternative to
Some limits of the scope of the paper follow. As the title indicates, the paper
will limit itself to banks and moneylenders. Cooperatives, the other major formal
ﬁnancial institution in many developing countries, are generally poorly run with
their economic motives weak. Moneylenders are individuals who lend at an interest
either full or part time. This paper will also not focus on linkages through savings
such as ROSCAS.7 Furthermore, on the credit side it will discuss the pure lending
aspects of trader-lenders and not their inter-linked aspects. It does not include
lenders such as friends and relatives who help out in times of need. Friends and
relatives may serve as linking agents but since they do not participate as proﬁt
maximizing agents, the economic incentives are diﬃcult to discern.
This paper ﬁrst identiﬁes the relative advantages of banks and moneylenders.
It then incorporates these advantages in reviewing proposed theoretical models
of linkages. The paper then ﬁnds several successful cases in Indonesia which
incorporate some of the incentives suggested by the theory. In contrast to JLL,
it ﬁnds that linkages provide additional attractive features. It concludes that for
Nagarajan-Meyer (1996) provide an example of this type of linkage.
viable linkages, banks need additional “carrots” to enter while informal lenders
need “sticks” in regulation.
2. Bank-Moneylender Linkages: Theory
In this section, we ﬁrst outline the models and address how moneylenders can help
overcome information and enforcement constraints. We explore the models in a
unifying manner through simple equations. Throughout, we will use the following
structure: borrower’s output Y has two values: high (Yh ) and low (Yl ) where
Yh > Yl = 0. The probability of high (low) output is P (1 − P ). Each borrower
requires a loan amount L and needs to repay amounts Rh and Rl for high and
low output, respectively. Assume that all projects are socially proﬁtable, i.e. that
P Yh +(1−P )Yl = Y > L. Assume that borrowers face limited liability constraints
and lenders can extract only what borrowers declare. Then, it follows that Rl = 0.
Normalize the borrowers’ alternative from borrowing to zero. Denote the bank’s
and moneylender’s cost of funds respectively as γ and ρ, where following the
literature, γ ≤ ρ.8
In order to focus on the pure problems of information, assume that banks and
Moneylenders must rely on their own funds. Commercial banks,with nationwide branches
and re-ﬁnancing access from governments, face a lower cost of funds.
moneylenders are risk-neutral. All the linkages share similar assumptions on infor-
mation available to lenders. Banks cannot observe the borrowers’ actions (moral
hazard), types (adverse selection), and/or verify incomes (enforcement and costly
state veriﬁcation). In evaluating when banks would link, we ﬁrst note that banks
can lend on their own. We focus on the incentive constraint that banks employ
to induce borrowers to “truth tell.” To simplify the technical details, we will
assume that these constraints bind which can be formally proven in a more com-
plete model. For sustainable linkages, not only moneylenders but also banks must
willingly participate. We then add the moneylender as a linkage agent. In con-
trast to banks, moneylenders have superior information (or enforcement powers).
We then compare the bank’s proﬁts on their own (denoted π B ) to those with the
linkage (with moneylenders) (denoted π L ) in evaluating when banks would link.
The linkages can be divided into two broad categories: explicit and implicit. In
the explicit linkages, banks hire moneylenders. In the implicit linkages, banks
alter their own loan contract, aware of the presence of moneylenders.
2.1. Moral Hazard (MH): Moneylenders Monitor Borrowers
With moral hazard, the bank cannot explicitly observe how the borrower runs her
project because it ﬁnds it too costly to observe the borrower’s actions (Conning
(1999, 2002)). Borrowers can choose a good or bad action (diligence or non-
diligence), where the probability of a high output is greater for a good action
than a bad action, i.e. Pg > Pb .9 If the borrower is non-diligent, then she receives
a private beneﬁt which is an increasing function of the loan size, B(L). As in
standard moral hazard models, an asymmetry rises. In case of non-diligence,
borrowers share their lower expected returns with banks but can capture the full
value of the private beneﬁt.
The borrower will choose to undertake the good action as long as the returns
are greater than the bad,
Pg (Yh − Rh ) ≥ Pb (Yh − Rh ) + B(L)
which, assuming that it binds, simpliﬁes to the following:10
Rh = Yh − (2.1)
where 4P = Pg − Pb . The bank does not extract the full amount in case of high
income and leaves a surplus (referred to as the enforcement rent, see Conning
We can equivalently model the moral hazard as high and low eﬀort with a disutility in
choosing high eﬀort.
Following the tie-breaker rule, assume that the borrower will choose the good action.
(1999)) which depends upon the amount of the borrower’s private beneﬁt and
the sensitivity of 4P . If 4P were large, then diligence probability is high and
the bank requires lower repayments. After substituting the binding condition
(in which the borrower chooses the good action), the bank proﬁts simplify to the
π B = (Pg )(Yh − ) − γL (2.2)
From above, in order to obtain positive proﬁts, the bank would set a bound on
the loan size, which would in turn limit the private beneﬁt.11 The bank could
increase its proﬁts by hiring a moneylender.
Assume that the moneylender has access to a linear monitoring technology
c which determines whether the borrower chooses a good or bad action. Now
monitoring decreases the private beneﬁt the borrower obtains and results in a
beneﬁt function B(c, L). The binding incentive compatibility constraint (Equation
2.1) from before now alters to the following:
Rh = Yh −
Here the bank would want to optimally set L = 0 but that would aﬀect Yh in a fully speciﬁed
Now, in contrast to lending on its own (Equation (2.1)), the bank can extract
a higher repayment amount through the increased monitoring and leave a lower
enforcement rent. However the bank needs to hire the moneylender and pays
wages wh and wl , respectively, for high and low output.12 For the moneylender
the participation constraint follows:
Pg wh + (1 − Pg )wl − c ≥ Pb wh + (1 − Pb )wl
The above, assuming that it binds, simpliﬁes to the following:
4w = (2.3)
where 4w = wh − wl . The left hand side (4w) indicates the compensation
diﬀerential the bank would pay the moneylender. The compensation diﬀerential
(4w) directly relates to the monitoring costs and inversely relates to the diligence
probabilities (4P ). When the diﬀerence (4P ) is large, the bank need not have
to compensate the moneylender as much to ensure that the borrower chose dili-
gence. This linkage provides a convenient advantage: the bank pays lower wages
when moneylenders do not provide a valuable monitoring role and consequently,
For now, allow for the possibility of wl 6= 0.
banks need not employ moneylenders. It calls for a ﬂexible credit policy across
regions depending on moneylenders’ opportunity costs and borrowers’ diligence
After substituting the repayments and the moneylender’s wages, the bank’s
proﬁts then yield (here, WLOG set wl = 0):
(B(c, L) + c)
πL = (Pg )(Yh − ) − γL (2.4)
Comparing π B to πM , banks will hire moneylenders as long as B(L)−B(c, L) >
c. In other words, if the incremental lowered diversion through monitoring is
greater than the monitoring costs, banks can increase their proﬁts by linking with
2.2. Enforcement (E): Moneylenders Enforce Repayments
In the most commonly observed and suggested linkage, assume that banks on their
own cannot enforce repayment and need to hire moneylenders (Fuentes (1996)).14
This linkage shares many features with the (MH) linkage, sometimes denoted as
Note that surprisingly this decision does not depend upon 4P , the diﬀerential gain between
good and bad actions since the moneylender’s wages absorbs the gains
In this static case and no collateral, banks will not lend on their own and trivially, πB = 0.
With dynamics, as will be seen later, banks can exclude borrowers from future credit access
ex-post moral hazard. Similar to the (MH) linkage, the moneylender may choose
diligence (eg ) or non-diligence (eb ), where eg > eb represents the moneylender’s
eﬀort in recovering a bank loan.15 This linkage diﬀers from (MH) in that the
probability (P ) is now of borrower repaying rather than action choice.
The above discussion indicates that the wages paid to the moneylender will
have a similar structure to Equation (2.3) where now 4e = (eg − eb ) substitutes
for the monitoring costs c in the previous equation. The 4P now corresponds to
the incremental increased probability the borrower repays if the moneylender puts
in high eﬀort. Again focussing on the binding condition, we obtain in a similar
4w = (2.5)
The smaller the diﬀerential probabilities (i.e. 4P is small), then a lower re-
sponsiveness of moneylender’s high wage to the repayment probabilities. In this
case, the moneylender’s value added is small. The bank induces the moneylender
to work harder by increasing his wages and may choose not to hire the money-
lender. The linkage also reveals a self-equilibrating character: banks will not
hire moneylenders when their value is less. In contrast to explicitly hiring money-
Usually, these eﬀort levels are referred to as “high” and “low.” But to be consistent with
the previous notation, we refer to them as “good” and “bad,” from the bank’s perspective.
lenders, banks “free ride” from the information of moneylenders. The following
linkages address this option.
2.3. Adverse Selection (AS): Moneylenders Screen Borrowers
The bank cannot diﬀerentiate between good and bad borrowers (Jain (1999)).
Moneylenders, with better information, lend only to the good types at the lenders’
cost of funds ρ. Good borrowers (g) have a higher probability of high output than
bad (b): Pg > Pb . Denote as λ (1 − λ) the proportion of good (bad) borrowers.
The bank can ﬁrst oﬀer a pooling contract in which both types obtain the
same loan. Here, we allow borrowers the option to borrow from moneylenders.
Thus, in order to attract the good borrowers it must provide the same terms as
moneylenders, where R denotes the pooling repayments:
Pg (Yh − R) ≥ Pg Yh − ρL
Simplifying and assuming that the above binds,
Notice that the banks’ repayments now take into account the presence of money-
lenders by tieing their repayments to the moneylender’s cost of funds. The proﬁts
under the pooling contract yield:
π B = λρL + (1 − λ) − γL (2.6)
Banks can extract the moneylender’s information in the following manner. Banks
can distinguish between good and bad borrowers by using the additional informa-
tion that the good borrowers have access to moneylenders and that all borrowers
rely on a critical minimum amount. The bank separates by deliberately under-
ﬁnancing the good, using the implicit knowledge that the good will resort to
moneylenders for the rest of the funds.16 The bad will not obtain any funds from
the banks but will not mimic the good’s contract since they cannot obtain the
remaining funds from moneylenders.17 In other words, the bank oﬀers two con-
tracts: one with higher repayments and no ﬁnancing (the bad will choose), the
other with lower repayments and underﬁnancing (the good will choose).
Since the bank cannot observe the riskiness of the borrower, it oﬀers loans
Note that this implicit linkage can incorporate the (MH) case as well where now moneylen-
ders would only lend to borrowers who choose good projects. The bank, again knowing that
moneylenders engage in this type of lending, would then coﬁnance these borrowers.
For certain parameter values, Jain ﬁnds that the bad will obtain full ﬁnancing from the
bank as well. In this knife-edge result of either obtaining full or no funding, for the purposes
of the paper we will ignore the uninteresting result of full funding.
contingent upon what it can observe: repayments.18 The bank must still provide
the good with the alternative of borrowing directly from the moneylender so that
(where M now represents the partial loan from the moneylender):
Pg (Yh − Rg ) − ρM = Pg Yh − ρL
Solving for Rg , we obtain:
ρ(L − M)
Thus the bank lowers the good’s required repayments by the loan amount it
obtains from moneylenders. The bank’s proﬁts now from good types only since
the bad now do not have access to bank loans follow:
π L = λ(ρ − γ)(L − M) (2.7)
Comparing π B to π L , the higher the proportion of bad borrowers (1 − λ), the
higher 4P , and the lower the cost diﬀerence (ρ − γ), the more likely the bank
For the bad borrower, the incentive compatibility constraint yields the following: Pb (Yh −
Rb ) ≥ Pb (Yh − Rg ) − ρM . Assuming that the above binds, the constraint simpliﬁes to the
following: Pb (Rb − Rg ) = ρM . The diﬀerential gain from repaying for a bad borrower is exactly
oﬀset by access to moneylenders (ρM ).
2.4. Costly State Veriﬁcation (CSV): Moneylenders Verify Output
The bank now cannot observe if the incomes of borrowers are high or low (Bolton-
Scharfstein (1990),Varghese (2004)). Since the bank cannot observe the two states,
borrowers would always claim they suﬀered bad times and the bank will never lend
since 0 < L. However, with an additional period, the dynamics of the lending
sustain a solution.
Banks can separate good (high income) and bad (low income) borrowers again
based on what they can observe, i.e. repayments. As in the (AS) linkage, repay-
ments are not suﬃcient but banks can now employ an additional instrument, the
threat not to lend anew (denote β i the probability of obtaining a loan conditional
on output i = h, l). Only if borrowers repay, they will obtain more loans (thus,
we set β h = 1).19 The threat of termination provides good borrowers an incentive
to repay. The incentive compatibility constraint for the good borrowers follows:
Yh − Rh + P Yh ≥ Yh − Rl + β l (P Yh ) (2.8)
Since Rl = 0, then β l = 0 since that would relax the above constraint. Assuming
This result can be proven in a full model,see Varghese (2004).
that the constraint binds, then the repayments for high income yield:
Rh = P Yh
The bank thus requests repayments that will cover next period’s expected income.
Bad borrowers cannot mimic good borrowers since repayments for good borrowers
(Rh ) are too high (P Yh ). In other words, the bank oﬀers two contracts: Rh > Rl
but with the additional stipulation that the good will receive loans and the bad
will not. The bank’s proﬁts under this separating contract would yield:
π B = P (P Yh − γ 2 L) − γL (2.9)
The separation comes at a cost since Y = P Yh > L for all borrowers in the
next period and the bank does not seize the socially eﬃcient opportunity.
Moneylenders recognize this opportunity. Moneylenders with their superior
information serve as linking agents by providing loans to excluded borrowers who
can then repay banks and enjoy continued access. An advantage of this linkage is
that banks need not rely on the threat of termination to separate the good from
the bad since the bad can now borrow from moneylenders. In a reversal of the
(AS) linkage, the bad and not the good borrowers are active in both markets.
Now the bank needs to induce the moneylender to participate, where R refers to
the pooling repayments, RM refers to the moneylender’s repayments:
−R + ≥0
Assuming that the moneylender will extract the full amount in the good state
and competition among moneylenders forces the above constraint to bind, we
The bank’s repayments now relate inversely to the moneylender’s cost of capi-
tal. As in the previous linkages, this equation reveals a self-equilibrating character.
With a higher cost to induce the moneylender to participate, banks require lower
repayments from the borrowers. Now the bank proﬁts yield:
πL = − γ 2 L − γL
Comparing π B to π L , the bank opts to link when P and ρ are low. One can
also show without normalizing Yl = 0, that when the dispersion between incomes
(Yh − Yl ) is high, the bank prefers linking with moneylenders. In these situations,
banks ﬁnd it more diﬃcult to diﬀerentiate between income types and would rely
on the moneylender.
2.5. Summary and Discussion
We have reviewed four models of linkages. The theoretical linkages precisely
outline the mechanisms of bank-moneylender linkages. The models indicate that
banks and moneylenders complement each other, increasing the available lending
opportunities.20 While the (AS) and (MH) linkages focus on the ex ante screening
aspects, the (E) and (CSV) are concerned with loan recovery.
Banks need not always link with moneylenders. A theoretical analysis of
the linkages reveals that banks will link when the information value added (∆P )
is high, the monitoring costs or eﬀort of the moneylender (c) are low and the
diﬀerential cost of funds (ρ − γ) is low. With complete data, an empirical exercise
could map these values onto observable variables. The information value added
(∆P ) can be captured with proxy variables which measure banks’ knowledge of
borrowers. These include trustworthiness, access to collateral, access to credit
With overwhelming qualitative evidence,the ADB study essentially comes to the same con-
clusion and as eloquently stated by Jain, the interaction between the two sectors creates a
information, legal recourse, or when idiosyncratic shocks form a major component
of the output. The monitoring costs c would be higher when lenders engage in
higher marginal activities. Finally, the diﬀerential cost of funds (ρ − γ) is related
to the costs c above with the larger spread for less well developed and integrated
the ﬁnancial markets.
The theory also reveals that wages should be contingent on repayments, which
are observable. The linkages are also self-equilibrating in that payments to money-
lenders adjust according to their contribution and costs. The above linkages are
not purely theoretical, policymakers have implemented these in a number of de-
veloping countries. The practical execution of the linkages must overcome some
issues which are absent in the theoretical models.
3. Formal-Informal Linkages: Practice
Many of the attempted linkages draw from Indonesia, “the world’s laboratory of
rural ﬁnancial markets.”21 Two general surveys on rural credit by Hulme and
Mosley (1996ab) and Ghate, (1992, sponsored by the Asian Development Bank,
hereafter referred to as ADB) delineate the adopted practices. We will ﬁrst
explore the pay structures of the bank oﬃcers in the (MH) and (E) linkages. In
Gonzalez-Vega and Chavez, quoted in Hulme and Mosley (1996b), p.32.
these linkages, banks explicitly hire oﬃcers to monitor repayments and are the
most prevalent. The pay structure reﬂects lenders’ eﬀorts at overcoming incentive
In Indonesia, regional development banks established KURKs, village units
which disburse loans at weekly mobile bank oﬃces. In order to monitor at this
level, the KURKs actually hire ex-moneylenders as commission agents (Hulme and
Mosley (1996b)). The lenders receive four percent of collected loan installments.
The whole system builds a web of incentives, with other participants such as the
village headman (who provides some of the (AS) linkage advantage) screening
borrowers and receiving one and a half percent of pre-tax proﬁts. One of the
KURKs’ successes was to minimize the guaranteed element in a bank worker’s
Another Indonesian bank (BUPB) oﬀers ﬁeld oﬃcers minimum guarantees plus
two percent of fully repaid loans seven and a half percent of savings (which thus
includes a link through savings) (Fuentes (1996)). The evidence stretches beyond
Indonesia to Sri Lanka where banks use informal lenders termed PNNs. These
14,000 PNNs lend bank loans to borrowers with no documentation but have to
follow bank regulated interest rates and loan amounts (ADB).
In eight other ﬁnancial intermediaries in Indonesia, village agents (but not
necessarily moneylenders) screen and collect loans (Chaves and Gonzalez-Vega
(1996)). The agents’ wages depend on observable variables such as collected re-
payments, loan installments, and primarily adjusted proﬁts. Proﬁts are adjusted
since some events go beyond the control of lenders. This ﬂexible system varies in
its implementation across villages in that wages are village speciﬁc and incorporate
the variables outlined in the theoretical section.
The term “on-lending” refers to an implicit version of the above when lenders
typically work as traders, i.e. inter-linked credit. In this case, banks aware of the
moneylenders’ presence deliberately increase credit so that moneylenders may lend
the increased loans without following bank regulations. Moneylenders would then
lend on their own. The Philippines has a long history of deliberately increasing
In the NAP Program of 1984, end-users and input suppliers received cheap
credit if they extended credit to farmers. In particular, a senior oﬃcial of one
of the largest commercial banks claimed that “some of the informal lenders are,
in eﬀect, conduits of bank funds.”23 Credit-layering, an extreme version of on-
Conning (2002) provides more direct evidence of the (MH) linkage. He observes that in
Chile contract farming ﬁrms establish contracts with farmers by signing letters of credit. These
notes are technically legally binding, but diﬃcult to enforce. The ﬁrm then shows the letters
to the bank and requests the bank to coﬁnance these projects. The bank agrees as long as the
ﬁrm invests a certain fraction of the money itself.
Quoted in Floro-Ray, p.40.
lending, is a cascading series of transactions where banks lend to informal lenders
who lend to others and so on. On-lending is widespread even when banks do not
deliberately increase credit. Frequently, borrowers from banks re-lend at higher
interest rates: with examples from Thailand (Coleman (1999)), the Grameen Bank
(Rahman (1999)), Malaysia and Pakistan (ADB). The above sources indicate that
the percentage of loans that lenders on-lend range from twenty to upwards of
With regard to the implementation, linkages (MH) and (E) have strong promise
but a number of countries have not fully implemented them. This possibility does
not seem to arise from the reluctance of moneylenders. As Karmakar (1999) ex-
plains, in informal talks with moneylenders in India, many have oﬀered to act as
agents as long as they can lend with an agency commission to meet their operating
costs. Policymakers’ attitude towards informal lenders vary in a number of coun-
tries. As mentioned in the Philippines, the government has actively intervened to
incorporate the informal sector into the overall strategy of agricultural develop-
ment (Floro-Ray (1997)). In sharp contrast, in India historically the government
has actively excluded the informal sector. In the 1980s though, the Indian gov-
ernment as in many other countries has reversed its philosophy. By launching
a program where commercial banks participated with informal lenders, linkages
have become more viable in India. (reference: RBI)
In the (AS) linkage, banks screen borrowers with the complicity of moneylen-
ders. Though not the same implicit structure as the (AS) linkage, the aforemen-
tioned Indonesian banks engage in explicit screening mechanisms. In particular,
within the discussed KURKs, village heads screen borrowers. Robinson (?) re-
ports on a similar scheme also in Indonesia (PSP-Kupedes) with traders recom-
mending borrowers. Customers with good banking records recommend members
from their business networks as borrowers. The head of the network has his name
and preferential treatment at stake which explains the low amount of defaults.
For now, the (AS) linkage may be a case where theory is ahead of practice.
Practically, banks can implement the (AS) linkage in the following manner.
Suppose a bank operates in an area with active informal lenders. If banks know
the required loan size of the project, they can deliberately underﬁnance borrowers
knowing that low risk borrowers can always resort to moneylenders. The bank’s
information requirement is high in this linkage. The bank must not only know the
critical minimum amount required by the borrower but also their other ﬁnancing
sources with information such as borrowers’ liquid wealth and access to other
Varghese (2002) provides evidence on the (CSV) linkage, where moneylenders
provide loans to borrowers who can then repay banks. Using ICRISAT data from
Indian villages, he ﬁnds that in general repayments to banks ﬂuctuate with income.
However for households that obtain loans from moneylenders, banks’ repayments
do not ﬂuctuate with income. Thus, borrowers can use moneylenders to smooth
cash ﬂows so as to meet bank obligations better. In another interpretation of this
linkage, banks provide production loans and moneylenders provide bridge loans
in order to ensure access. In a similar but slightly diﬀerent set-up in Bangladesh,
Sen reports that “recovery agents” help borrowers roll over bank loans for a fee
(ADB). Borrowers then obtain a new formal loan and found that even after
paying the fee, found the loans worthwhile. Karmakar provides another twist on
this linkage from moneylenders in India who provide bridge loans for borrowers
who are waiting to receive sanctioned bank loans.
In microﬁnance the above is known as informal “bicycling” which occurs when
borrowers who repay a microﬁnance institution on time obtain immediate access to
another larger loan. Informal lenders, aware of this situation, provide bridge loans
to borrowers who are short of cash to pay the microﬁnance lender. A number of
studies focus on another aspect uncovered by the (CSV) linkage: the consequences
of denying future loan access by formal lenders. For example, as the Masagna-
99 formal credit programs expanded in the Philippines, repayments deteriorated
under the formal sector, borrowers were excluded and the lending shifted back to
informal lenders. This same phenomenon occurred in rural Thailand when the
BAAC attempted to expand credit (ADB). From these case studies, policymakers
can follow a more inclusive approach to informal lenders in the launching of new
credit programs. Informal lenders can serve an invaluable role in the incipient
stages by enabling borrowers to enjoy continuing access to formal sector loans.24
As seen above, the evidence mainly from the largely successful Indonesian
experiments serve as lessons for other countries. The question remains on the
replicability of the Indonesian experiments. For example, Hulme and Mosley ar-
gue that the Indonesian system works because of a clear system of control (the
village head) which may not work in all places. Similarly, compensation to money-
lenders reﬂects the opportunities foregone and these vary with respect to Indone-
sia. Finally, in some regions moneylenders are not prevalent so that linkages must
be attempted with other agents (see Steel, et al. on the continent of Africa and
These studies and others do not provide evaluation on neither the eﬀectiveness
of particular linkages nor whether formal lenders actually hired moneylenders
An important caveat is the anecdote told by Coleman inThailand. He found that the village
bank recorded a 100 % repayment rate to the NGO but that 67 % of borrowers borrowed from
moneylenders to repay the village bank. Later, borrowers repaid the moneylenders with the
loan from the village bank, creating a “vicious debt cycle.”
as agents. As mentioned before, in not all the cases the hired agents were ex-
moneylenders. One would still expect moneylenders with the learned practice of
lending to be the most adept at continuing this tradition. Alternatively, banks
lend to NGOs, where NGOs guarantee and assist in loan recovery as in Sri Lanka or
Bangladesh (ADB and Fuentes). No study provides independent evaluation on the
eﬀectiveness of moneylenders with respect to non-moneylenders. Hulme-Mosley
state that training costs of staﬀ represent a signiﬁcant portion of microﬁnance
institutions’ costs. In hiring moneylenders, banks need not expend resources on
training since moneylenders have already incurred these costs. Furthermore,
the implicit linkages have the added advantage of banks not needing to directly
interact with moneylenders.
The evidence squares with the theory with its emphasis on bank worker in-
centives and its ﬂexibility. Not surprisingly, the Indonesian experiment was built
on the work of foreign consultants who were adept at creating incentive based
systems. The review of the actual practices also points towards other dimensions
not covered in the theory which should be considered for the success of linkages.
One research point we can begin to address are the advantages of linkages over
other mechanisms such as group lending.
4. Linkages vs Joint Liability Lending
In this section, we compare and contrast linkages to joint liability lending. Group
lending or JLL circumvents the problems banks face by issuing joint liability
contracts: members of a group are liable for one another. Formally, joint liability
consists of the following for a two person group: if a borrower will repay her
loan but her partner will not repay the loan, then the borrower must repay an
additional c to the bank. This incentive constraint replaces the moneylender’s
participation constraint in the linkages. Previous literature has not evaluated the
advantages of joint liability lending (JLL or group lending) over linkages or any
other alternative credit delivery mechanisms. For example, in Ghatak-Guinnane’s
(hereafter G-G) survey of JLL, the alternative to group lending is individual bank
We will formally only analyze the eﬀects of group lending on the (MH) contract
(Conning (1999)). Now we need to account for four possible outcomes (with
four corresponding repayments): Yhh , Yhl , Ylh , Yll where the ﬁrst subscript denotes
the borrower’s outcome and the second the partner’s outcome. In the optimal
contract, the bank rewards the successful borrower but penalizes the unsuccessful
borrower.25 The only relevant constraint is for a successful borrower to truth-tell
and monitor the partner in Nash Equilibrium:
Pg Pg (Yhh − Rhh ) − c ≥ Pg Pb (Yhh − Rhh ) + B(c, L) − c
Thus, one constraint captures the production and monitoring problem. The
above, assuming that it binds, simpliﬁes to the following:
Rhh = Yhh −
Again substituting the above in bank proﬁts yields the following (denote this
π G ):
π G = (Pg Pg )(Yhh − ) − γL
Comparing the above to Equation (2.4), assuming that Yhh = Yh ,note that
π G > πB if c > 4P (1 − Pg )Yh . For large monitoring costs and high probability
of a good outcome, group lending dominates linkages. The advantage of group
lending is that it absorbs the monitoring costs c within the group but requires a
See Conning (1999) for a more thorough discussion of a model with collateral.
higher likelihood of a good outcome for success.
For brevity, for the other problems we will limit ourselves to the G-G informal
explanations (please refer to their article for formal derivations). In the (AS)
version. the safe will associate with the safe and the risky are left with the risky
resulting in positive assortative matching. Banks oﬀer two contracts: one with
high interest rates and low joint liability (the risky will choose) and one with
low interest and high joint liability (the safe will choose). In the (E) case, if one
member will not pay back but the other pays both her own and her partner’s JLL
dominates individual liability. Also, if the community imposes social sanctions on
a member who does not pay her partner’s then JLL becomes more attractive. For
the (CSV) case, now the bank has to induce the borrower to report the truth for
high borrower’s returns and low partner’s returns. G-G argue that the partner has
an incentive to audit the borrower due to partial liability. So now the bank need
only audit when the whole group announces its inability to repay (which occurs
with a lower probability). Group lending still introduces further constraints as
borrowers are prone to collude with each other. To avoid this possibility, we need
to introduce a non-collusion constraint. With this constraint, for large monitoring
costs borrowers will not reveal the truth about each other. Due to this limitation,
many borrowers would graduate from group loans to linkages or individual lending
to reach larger scales.
To address greater problems that arise in group lending we limit our analysis
to an informal analysis. The lending solution proposed by JLL creates its own
problems since the solution relies on interdependence among borrowers. The
incentive constraint imposed in JLL is not as innocuous as the participation con-
straint imposed in linkages since it introduces interdependence. Practically, in
close knit village communities, borrowers reluctantly sanction delinquent borrow-
ers (see G-G on Ireland and Burkina Faso). Wydick also ﬁnds that with Accion in
Guatemala friends do not make reliable group members since members are often
softer on friends. Sometimes social ties among possible borrowers are too weak
to support feelings of group solidarity as demonstrated in the failed transplant
in Arkansas (Ghatak-Guinnane). On a theoretical level, interdependence creates
the following: bad borrowers create negative externalities on good ones. The
group-lending structure may be less ﬂexible than individual lending for borrowers
in growing businesses and those that outstrip the pace of their peers (Morduch).
Due to the interdependence, the JLL enjoys more success in areas of high
population density and steady and frequent income streams (Conning-Fuentes).
Thus, less successful in agricultural environments where the nature of produc-
tion requires balloon repayments (Hulme-Mosley). Furthermore, the obsession
with the repayments record leads to some undesirable consequences such as vio-
lence against women (Rahman (1999) on the Grameen Bank). ADEMI in the
Dominican Republic switched to individual lending because it felt that credit ad-
visors relied too much on peer pressure for loan repayments and did not develop a
signiﬁcant relationship with clients (Conning-Fuentes). Finally, some ﬁner points
include whether group lending leads to excessive monitoring, pressure to undertake
“safe” projects, and the costs of weekly meetings and staﬀ training (Morduch).
This latter point cannot be overemphasized as Hulme and Mosley indicate that
training costs form a large component of JLL programs. For example, with the
Grameen bank, salary and personnel costs accounted for half of Grameen’s total
costs. Over half of female trainees and a third of male trainees dropped out before
taking ﬁrst positions at Grameen (Morduch).
Until recently, the proﬁtability of JLL justiﬁed some of the negative conse-
quences. Morduch’s re-evaluation (1999) ﬁnds that most programs are subsidized
with soft loans from donors, with $26-30 m for the Grameen bank. Morduch also
reports that some donors believe only ﬁve percent of all programs today will be
ﬁnancially sustainable ever. These diﬃculties of microlending leads one in search
of a self-sustaining solution.
This search leads one back to the solution proposed here: linkages, a not fully
exploited solution with potential. The relative advantages over JLL are many:
self-sustainability, no excessive bureaucratic layer, no pressure on neighbors, and
the use of the speciﬁc capital of moneylenders. Linkages build on villagers who
would not engender the suspicion of neighbors. Individuals who traditionally
have been lenders, i.e. moneylenders, would continue as lenders. Linkages could
potentially achieve Hulme and Mosley’s three main conditions for successful credit
programs: intensive loan collection, incentive to repay, and provision for voluntary
saving. Linkages would bring the bank worker (moneylender) to the customer,
within a self-interested and decentralized decision making process.
This paper has reviewed potential linkages and provided available evidence of
linkages in rural credit markets. Theoretically and empirically, linkages are at an
inchoate stage. In exploring why linkages are not prevalent in developing countries,
one can uncover some stumbling blocks that remain. An immediate response is
that ideas such as mechanism design, institutions, and incentives are relatively
new topics. Furthermore, historically policymakers have viewed informal lenders
such as moneylenders as exploitative. The ADB study reviews many cases and
concludes that these views may be out-dated. enforcement, political economy
This paper provides an alternative role for moneylenders. For linkages to be
eﬀective, banks are needed alongside moneylenders. In certain cases, banks may
not ﬁnd it worthwhile to enter both on theoretical and practical grounds. The
theory is based on knowledge spillovers since one bank would provide another bank
with borrower training and management skills gratis. Second, banks may not enter
for the same information and enforcement issues raised throughout this paper.
Due to these reasons, commercial banks would still need additional “carrots” to
enter rural areas.
Since the linkages program is at an incipient stage, a myriad of topics remain
to be explored. Future research can investigate savings linkages, the advantages
of moneylenders over others as linking agents, the relative advantages of one link-
age over another, or ﬁnally combining linkages, as linkages within linkages. In
the future, in light of the new revisionist views of microcredit, policymakers can
explore linkages more fully as an alternative credit delivery mechanism.
We now extend the discussed linkages to incorporate the following: relax-
ing limited liability constraints on borrowers (i.e. allowing for collateral) and
moneylenders, introducing risk aversion for moneylenders, and allowing bank-
moneylender competition. We will only brieﬂy discuss the extensions.
Extension I: Collateral for borrowers
First, suppose in the (MH) linkage banks can employ collateralized loans in
which monitoring is unnecessary. Borrowers with high levels of collateral will then
choose to borrow these cheaper loans from banks For medium levels of collateral,
borrowers choose the linkage (Conning (2002)). For the (AS) model, as Besanko-
Thakor (1987) and others show, collateral allows banks to screen good and bad
borrowers, where the good borrowers now opt for the contract with collateral. The
collateral allows banks to overcome the problems in the (CSV) and (E) models
if the collateral amount would cover bank proﬁts. In sum, the discussion above
indicates that the introduction of collateral allows for less reliance on linkages.
However, this assumes away one of the main problems in developing countries
where limited wealth and limited legal systems do not allow the use of collateral.
In fact, researchers now directly focus on the eﬀects of collateral substitutes (for
example, Bond-Rai (2002)). As countries develop into more collateral-based
economies, the collateral constraint would lessen.
Extension II: Relax limited liability for moneylenders
Instead of relaxing the limited liability constraint on borrowers, we can relax
the moneylender’s constraint. In the moneylender’s case, a low wage of wl = 0
may not provide enough of an incentive for him to participate. This normalization
is harmless for the eﬀects of the contract. Again, in the (MH) and (E) linkages,
in the Equation (2.3), set wl = wMin , the minimum accepted wage. Then the
equation modiﬁes to the following:
wh = wMin +
The above now comprises a guaranteed component (wMin ) and a bonus element
which varies as before ( 4P ).26 For the other linkages, a similar normalization now
with respect to the bank’s repayments would yield similar results, with a greater
surplus provided to moneylenders and less to banks.
Extension III: Risk aversion for moneylenders
In introducing risk aversion with moneylenders, banks would need to part with
a greater surplus. Risk aversion is more justiﬁable for moneylenders than banks as
Conning and Fuentes (2002) further discuss this extension.
they cannot diversify income since they are subject to aggregate shocks within the
village economy. Risk-aversion in the (MH) and (E) models introduces concavity
in the wage function so that Equation (2.3) would change to the following where
U(·) refers to this function:
U(wh ) − U (wl ) =
Now since 4U > 4w for small w, an even greater diﬀerence in utility is
needed to compensate the moneylender for a small 4P .27 Consequently, with
risk-aversion we have the well-known trade-oﬀ between risk-aversion and incen-
tives, i.e. the larger the guaranteed element in wages, the smaller the incentive
for the moneylender to follow up on the repayments. Introducing risk-aversion
in the other two linkages would also induce banks to lower its repayments so
that moneylenders would participate and the linking option would become less
Extension IV: Introducing Competition
Jain-Ghasari Instead of providing linkages, competition between banks and
moneylenders could potentially transfer surplus from lenders to borrowers. In ad-
More precisely, for a concave function, U (wh ) − U (wl ) < U 0 (wl )(wh − wl ), and since U 0 > 0
and U 00 < 0, for small wl , U (wh ) − U (wl ) > (wh − wl ).
dition, competition would not entail the ineﬃcient layering required in linkages.
Formally, banks face a zero proﬁt condition in addition to the incentive compatibil-
ity condition. Moneylenders would not face the incentive compatibility condition
but incur a higher cost of funds. In diﬀerent models, both Mc-Intosh-Wydick
(2002) and Hoﬀ-Stiglitz (1998) show that competition in information markets
crease some unintended consequences. These problems arise because of the in-
creased indebtedness of borrowers and concomitant lack of information sharing
among lenders so that all participants may be worse oﬀ. Varghese (2002), in
a similar model to the (CSV) linkage above, shows that under competition the
surplus ﬂows to the good borrowers but poorer borrowers still would not obtain
further loans. As the above papers indicate, banks still face constraints with in-
formation or collateral and these would be alleviated only with great diﬃculty.
Introducing a guarantee for moneylenders means less incentives and practice might
provide directions into the actual mix between guarantee and incentive.
 Adams, D. W., D.H. Graham and J.D. von Pischke (1984). Undermining
Rural Development with Cheap Credit. Boulder, CO: Westview Press.
 Bell, C. (1990). “Interactions between Institutional and Informal Credit
Agencies in India.” The World Bank Economic Review 4(3), 297-327.
 Besley, T. (1994). “How do Market Failures Justify Intervention in Rural
Credit Markets ?” The World Bank Research Observer 91(1), 27-48.
 Besanko, D. and A. Thakor (1987). “Collateral and Rationing: Sorting Equi-
libria in Monopolistic and Competitive Credit Markets.” International Eco-
nomic Review 28 (3), 671-689.
 Bolton, P. and Scharfstein, D. S. (1990). “A Theory of Predation Based on
Agency Problems in Financial Contracting.” American Economic Review 80,
 Bond, P. and A. Rai (2002). “Collateral Substitutes in Microﬁnance.” Mimeo,
 Chaves, R. and Gonzalez-Vega, C. (1996). “The Design of Successful Rural
Financial Intermediaries.” World Development 24(1), 65-78.
 Coleman, B. (1999). “The Impact of Group Lending In Northeast Thailand.”
Journal of Development Economics 60, 105-141.
 Conning, J. (1999). “Outreach, Sustainability and Leverage in Monitored and
Peer-Monitored Lending.” Journal of Development Economics 60, 51-77.
 Conning, J. (2002). “‘Ventas Piratas’: Product Market Competition and the
Dept of Lending Relationships in a Rural Market in Chile.” Mimeo, Hunter
College, New York.
 Conning,J. and Fuentes, G. (2000). “Staﬀ Incentives for Microﬁnance Or-
ganizations in Low-Income Countries: A Review of Theory and Evidence.”
Mimeo, Hunter College, New York.
 Floro, M. and Ray, D. (1997). “Vertical Links between Formal and Informal
Financial Institutions.” Review of Development Economics 1(1), 34-56.
 Fuentes, G. A. (1996). “The Use of Village Agents in Rural Credit Delivery.”
Journal of Development Studies 33(2), 188-209.
 Ghatak, M. and T. Guinnane (1999). “The Economics of Lending with Joint
Liability: Theory and Practice.” Journal of Development Economics 60, 195-
 Ghate, P.B. (1992). Informal Finance: Some Findings from Asia. Hong Kong:
Oxford University Press.
 Hoﬀ, K. and J. E. Stiglitz (1990). “Introduction.” World Bank Economic
Review 4, 235-250.
 Hoﬀ, K. and J. E. Stiglitz (1998). “Moneylenders and Bankers: Price In-
creasing Subsidies in a Monopolistically Competitive Market.” Journal of
Development Economics 55, 485-518.
 Hulme, D. and P. Mosley (1996a). Finance Against Poverty, Volume 1. New
 Hulme, D. and P. Mosley (1996b).Finance Against Poverty, Volume 2. New
 Jain, S. (1999). “Symbiosis vs Crowding-out : The Interaction of Formal and
Informal Credit Markets in Developing Countries ” Journal of Development
Economics 59, 419-444.
 Karmakar, K.G. (1999). Rural Credit and Self-help Groups. Delhi: Sage Pub-
 Mahabel, S. B. (1954). “Institutionalizing the Moneylender.” Indian Journal
of Agricultural Economics 9, 175-178.
 Morduch, J. (1999). “The Microﬁnance Promise.” Journal of Economic Lit-
erature 37 (4), 1569 -1614.
 Nagarajan G.and R. Meyer (1996). “Linking Formal and Informal Financial
Arrangements through NGOs: An Experment from the Gambia.” African
Review of Money, Finance and Banking 1,121-133.
 Rahman, A. (1999). Women and Microcredit in Rural Bangladesh. Boul-
der,CO: Westview Press.
 Ray, D. (1998). Development Economics. Princeton: Princeton University
 Reserve Bank of India (1998). Narsimham Committee Report on Banking
Sector Reforms. Bombay: Reserve Bank of India.
 Robinson, M. (1992). “Rural Financial Intermediaries: Lessons from Indone-
sia,” Development Disscussion Paper. Boston: Harvard Institute of Develop-
 Steel, W., Aryeety, E., Hettige, H. and M. Nissanke (1997). “Informal Finan-
cial Markets under Liberalization in Four African Countries.” World Devel-
opment 25(5), 817-830.
 Varghese, A. (2002). “Can Moneylenders Link with Banks ? Theory and
Evidence from Indian Villages,” Mimeo. St. Louis: St.Louis University.
 Varghese, A. (2004). “Bank-Moneylender Linkage as an Alternative to Bank
Competition in Rural Credit Markets,” forthcoming, Oxford Economic Pa-
 Walker, T. and J. G. Ryan (1990). Village and Household Economies in
India’s Semi-arid Tropics. Baltimore: Johns Hopkins University Press.