II. FINANCIAL INSTITUTIONS AND VILLAGE AGENTS

The idea of using village agents grew from the realisation that a portion of the blame for the poor performance of the traditional credit programmes landed at the feet of the various institutions charged with channelling the increased amounts of credit to the rural areas (such as co-operatives, government owned agricultural and development banks, and urban and rural private banks).(3) To be fair, formal financial institutions operating in rural financial markets face a number of problems that do not appear in other segments of a country's financial system. Among the problems that financial institutions face are: (1) the synchronic financial demands placed on the institution due to the seasonal nature of agricultural production, (2) potential clients that are widely dispersed over a geographical area, (3) the covariance of risk among borrowers, and (4) the large number of small transactions that are made by borrowers. In addition, potential clients are in an industry that faces unanticipated movements in prices, incomes, and yields [Adams and Vogel, 1986; Braverman and Guasch, 1989].

Information and incentive problems are two of the fundamental problems that hinder the performance of formal financial institutions in allocating credit. The information problem arises because lenders find it costly: (1) to gather information on the likelihood of a borrower defaulting on a loan, and (2) monitor the actions of the borrower once a loan is given. The incentive problem arises from the difficulties that lenders face in ensuring repayment of a loan. These problems for a lender, which are endemic to credit markets, are even more severe in the rural areas of low-income countries.(4)

In addition to information and incentive problems many formal financial institutions are also hampered by inappropriate operating procedures, which have often been transferred directly from their urban operations. The result is that the typical information gathering systems employed by financial institutions (for instance, lengthy paper requirements and borrower interviews), and the large staff required to process the information lead to large administrative and processing costs.(5) The costs per dollar lent fall as the size of the loan increases, and therefore, loans to small farmers are relatively more costly to the financial institution. The combination of information and incentive problems, and large administrative and processing costs makes granting loans to small farmers unattractive to formal financial institutions. The end result is that financial institutions base their loan decisions on observable wealth, and therefore, the small farmer is either rationed or screened out.

Not surprisingly alternative credit delivery mechanisms have been proposed that attempt to do two things: (1) mitigate the information and incentive problems that hamper formal financial institutions, and (2) lower their administrative and processing costs.(6) One such proposal calls for the use of a member of the village community (like a village head, clan leader, or village elder) to act as an agent in screening potential borrowers and collecting repayment. The village agent possesses an information advantage over the financial institution in regard to the potential borrower's risk characteristics due, in large part, to his proximity to the applicant. For instance, he would know the applicant farmer's farm management skills, his current asset holdings and, more than likely, the small farmer's repayment performance on previous loans. In the event that further information was needed by the financial institution, the village agent would more than likely have a cost advantage in obtaining the additional information. The agent's proximity may also give him a cost advantage in monitoring and pursuing loan repayment.

Aside from the agent's proximity, the village agent may also have an advantage in loan collection due to his access to village-level contract enforcement mechanisms. These mechanisms, which are usually unavailable to formal financial institutions, allow the agent to credibly threaten with a social sanction a member of the community for non-compliance with contract terms. For instance, an agent may be able to impose a social cost on a member of the village by denying him access to religious ceremonies or other village-wide activities.(7) Therefore, depending on who is recruited to serve as an agent, the financial institution may be able, in varying degrees, to harness the use of these village-level enforcement mechanisms to aid it in loan collection.

What is often ignored by proponents of this credit delivery mechanism is that in an attempt to solve the information and incentive problems of the financial institution one is possibly creating additional information and incentive problems. Use of village agents means adding another layer to the rural financial market hierarchy. Problems will arise if the agent's objectives are different from those of the financial institution.

There are a number of incentive and information problems that a financial institution could encounter when contracting with a village agent. There exists the possibility of collusion between the agent and the farmer in the initial screening process. Because of the spatial nature of the agricultural production, the financial institution is unable to observe the effort undertaken by the agent in the monitoring of loan terms, or in loan collection. Similar problems arise if the agent is also required to provide some form of extension service (for instance, delivery of inputs) to the borrowing farmer. In either case the financial institution will find it difficult and costly to verify if the agent revealed his private information truthfully, or exerted the necessary effort in his tasks.

III. THE BASIC MODEL

In this section I construct a model that captures some of the information and incentive problems faced by a financial institution that contracts with a village agent. To simplify matters I assume that there is one financial institution that is contracting with one village agent. In addition, the village agent is dealing with only one prospective borrower at a time. The contract that is offered to the village agent requires that he announce his private information concerning the creditworthiness of the applicant small farmer and, in the event that a loan is granted, to collect the repayment from the farmer. The agent is compensated for performing these services. The compensation that he receives is a function of his announcement and whether the loan is repaid or not repaid.

I assume that there are only two types of farmers. A 'good' farmer, g, is a farmer that possesses characteristics that make him a low credit risk to a financial institution. On the other hand, a 'bad' farmer, b, is a farmer that possesses characteristics that make him a high credit risk. Since the village agent is recruited from the same community as the farmers I assume that the agent knows the farmer's type.

If a loan is granted by the financial institution the agent is required to collect the repayment from the farmer. While the decision to grant a loan to a farmer is ultimately the decision of the financial institution, I simplify matters by assuming that the financial institution will always grant a loan to those farmers that are identified as creditworthy. When collecting the repayment the agent can exert either a high effort level [e.sub.H], or a low effort level, [e.sub.L] [less than] [e.sub.H]. Effort is costly to the agent and unobservable to the bank.

The good and the bad farmers have a probability of repayment [p.sub.i](e), i = g, b. The probability of repayment is also a function of e, the effort level undertaken by the agent. To simplify matters I assume that the bad farmer never repays a loan, thus [p.sub.b](e) = 0, [for every] e. Therefore, for the remainder of this article p(e) will always refer to the probability of repayment of the good type. I assume that the repayment of the loan by the farmer is all or nothing (hence, I do not allow for partial repayment).

Now consider a small farmer that approaches a village agent for a loan. The only possible announcements that the village agent can make to the financial institution concerning the farmer's type are: [a.sub.g] for a good farmer and [a.sub.b] for a bad farmer.

If the village agent exerts a high level of effort in loan collection, the probability of repayment is higher than if the agent exerts a low effort level, thus p([e.sub.H]) [greater than] p([e.sub.L]). I assume that p([e.sub.H]) [less than] 1 (failure to repay is not proof of low effort level) and both probabilities are positive and common knowledge.

The financial institution is assumed to be only able to observe whether the loan is repaid or not repaid. Therefore, the compensation that is offered to the village agent specifies a set of payments that are contingent on the announcement and whether the loan is repaid. Denote the contract that is offered to the village agent as w(.), and the two possible values that it can take as [w.sub.r] and [w.sub.n]: [w.sub.r] is the compensation that is given to the village agent when he announces [a.sub.g] and the loan is repaid by the farmer; [w.sub.n] is the compensation when the announcement is [a.sub.g] and the loan is not repaid.

The utility function of the village agent is assumed to be additively separable in wages and effort: V(w(.)) - e. Let V be unbounded, strictly increasing, and concave. By assumption the inverse function [V.sup.-1](.) exists. That is for all v [element of] R there exists a w such that V(w) = v. The village agent's reservation utility is normalised to be 0. The financial institution is assumed to be risk-neutral in income.

I can summarise the sequence of events as follows. The village agent is endowed with private information, i, on farmer type, where i [element of] {b, g}. The contract, w(.), is offered by the financial institution to the village agent. If the agent accepts the contract he announces the farmer's type, where [a.sub.i] [element of] {[a.sub.b], [a.sub.g]}. A loan is given by the financial institution to the good small farmer. The village agent then chooses effort level e, where e [element of] {[e.sub.L], [e.sub.H]}. The loan is either repaid or not repaid by the farmer. A payment is made to the agent by the financial institution.

IV. COSTLESS INFORMATION REVELATION AND UNOBSERVABLE EFFORT

I assume that it pays for the financial institution to hire the village agent only if he exerts a high effort level in obtaining repayment. The financial institution will then seek the least cost compensation scheme that will induce the agent to reveal his private information truthfully and to exert a high level of effort in loan collection. In minimising expected compensation costs the financial institution faces three constraints:

(IR-1) p([e.sub.H])V([w.sub.r]) + (1-p([e.sub.H]))V([w.sub.n]) - [e.sub.H] [greater than or equal to] 0

(IC-2) V([w.sub.n]) - [e.sub.L] [less than or equal to] 0

(IC-3) p([e.sub.H])V([w.sub.r]) + (1-p([e.sub.H]))V([w.sub.n]) - [e.sub.H] [greater than or equal to] p([e.sub.L])V([w.sub.r]) + (1-p([e.sub.L]))V([w.sub.n]) - [e.sub.L]

The first constraint guarantees that the agent can do no worse, in expectation, than his reservation utility if he accepts the contract. Constraints 2 and 3 are referred to as the incentive compatibility constraints. Constraint 2 ensures that the village agent is better off revealing that a bad a farmer is bad (making no announcement) and receiving no compensation, rather than lying (announcing the farmer is good) and exerting a low effort level in loan collection.(8) Finally, constraint 3 ensures that the agent finds it in his best interest to exert a high effort level in loan collection.

The financial institution's problem can then be written as choosing [w.sub.r], [w.sub.n], to:

min[p([e.sub.H])[w.sub.r] + (1-p([e.sub.H]))[w.sub.n]]

subject to constraints (1), (2), and (3). Following Grossman and Hart [1983] the financial institution will choose levels of utility that correspond to a set of payments that align the agent's objectives to its own. The optimal compensation scheme can be summarised as follows:

Proposition 4.1

The set of payments which the agent receives as compensation are:

V([w.sub.r]) = [e.sub.H] + {([e.sub.H] - [e.sub.L])/[Delta]p} x (1-p([e.sub.H]))

V([w.sub.n]) = [e.sub.H] - {([e.sub.H] - [e.sub.L])/[Delta]p} x p([e.sub.H])

where [Delta]p [equivalent to] (p([c.sub.H]) - p([c.sub.L])).

Proof: see Appendix.

It is easier to follow the basic intuition of the result if the compensation is discussed in terms of the utility derived by the agent [Kreps, 1990]. The proposition simply states that in the event that the loan is repaid, the village agent will receive as compensation V([w.sub.r]), which is greater than the agent's disutility from effort. Since the agent exerts a high effort level in equilibrium, the village agent is rewarded by an amount V([w.sub.r]) - [e.sub.H] [greater than] 0. In the case that the loan is not repaid the village agent receives a utility level V([w.sub.n]) as compensation, which is less than his costs from exerting a low effort level. Once again, since the agent exerts a high effort level in equilibrium, this payment constitutes a penalty equal to V([w.sub.n]) - [e.sub.H] [less than] 0.

The optimal compensation scheme can be interpreted in the following manner. In order for the financial institution to guarantee that it will face the higher probability of repayment when lending to small farmers it must induce the agent to exert a high level of effort in loan collection. Since the financial institution cannot observe the agent's actions, in order to give the agent the proper incentives the financial institution must share the risk it faces when extending loans. Because the agent is risk-averse, the financial institution must compensate the agent for this increased exposure to risk. Thus, when the loan is repaid the agent's utility is greater than his reservation utility (normalised to zero) plus any costs of effort. On the other hand, if the financial institution does not receive a repayment, the agent is correspondingly penalised with a level of utility that is less than his cost of effort.

This result can be seen graphically [Varian, 1992]. In state-utility space the agent's indifference curves are just straight lines, negatively sloped [ILLUSTRATION FOR FIGURE 1 OMITTED].(9) For the compensation scheme to be incentive compatible, constraint 3 must be binding at equilibrium. Solving for V([w.sub.r]), the following relationship can be obtained.

V([w.sub.r]) = V([w.sub.n]) + ([e.sub.H] - [e.sub.L])/[Delta]p (4)

Equation (4) says that the agent will be willing to exert high effort only if his utility in the state where the borrower repays exceeds that when he does not repay by the amount ([e.sub.H] - [e.sub.L])/[Delta]p. It also follows that the village agent's indifference curves, corresponding to high and low effort, respectively, intersect above the 45 degree line.

The financial institution's preferences are given by the concave isocost curves. Costs are decreasing as the curves move toward the origin. The slope of the isocost curves is given by:

[dv.sub.r]/[dv.sub.n] = -(1-p([e.sub.H]))[Z.sub.n]/p([e.sub.H])[Z.sub.r]

where [Z.sub.i] [equivalent to] d[V.sup.-1]/[dw.sub.i] for i = n, r.

It can now be clearly seen that if the financial institution is able to observe and enforce an effort level of [e.sub.H], then V([w.sub.r]) = V([w.sub.n]), and the slope of the iso-cost curve [C.sub.1], is equal to the slope of the agent's indifference curve through [e.sub.H] (see pt.A). Point A is equivalent to the first-best outcome. However, since the financial institution cannot observe the agent's effort level, the best that it can do is move to point E. At point E, the financial institution is on a higher iso-cost curve [C.sub.2] [ILLUSTRATION FOR FIGURE 1 OMITTED].

The amount of risk that the village agent is exposed to (that is, how wide is the gap between the repayment and non-repayment compensation utilities) will depend on the effect of the agent's actions on the probability of repayment, and on the agent's ability to bear risk. If the agent is risk-neutral and his effort level is unobservable, inducing the agent to exert a high level of effort is straightforward. Because transferring risk to the agent is costless, the financial institution can impose very large rewards and penalties on the agent so as to influence his actions. In effect the financial institution can transfer all the risk to the agent by demanding a fixed payment from the agent regardless of whether the loan is repaid. The village agent becomes the residual claimant.

In order to determine how much risk a risk-averse agent is exposed to, recall the relationship from equation (4). One can see that the larger the effect of a high effort level on the probability of repayment, the smaller will be the difference in the utility levels that are given to the agent as compensation. Put another way, as the probability of repayment that the financial institution faces becomes more sensitive to the effort level chosen by the agent, the less is the exposure to risk that is needed for the compensation scheme to be incentive compatible.

If [Delta]p (or p([e.sub.H]) - p([e.sub.L])) is large, then even a small gap in V([w.sub.r]) - V([w.sub.n]) has a large effect on the village agent's payoffs and hence on his incentive to exert a high level of effort. In other words, exposure to small amounts of risk are enough to induce a high level of effort from the agent. On the other hand, if [Delta]p is small, then a large gap in V([w.sub.r]) - V([w.sub.n]) is needed in order to induce a high level of effort. Since exposing the agent to risk is costly, the financial institution may find it altogether unprofitable to hire an agent to collect the repayment when [Delta]p is small.

The structure of the compensation scheme provided to the agent implies that the financial institution will have a preference for agents that have a large effect on the probability of repayment (where [Delta]p is large). If the agent's efforts are interpreted as the cost or pressure that he is able to impose on the borrower when collecting the loan, the financial institution will then prefer an agent that can impose large amounts of these costs on the borrower in order to influence repayment. Thus, a village chief, or another member of the community with similar power will be preferred by the financial institution as an agent, precisely because of their ability to harness the use of social sanctions and other village-level enforcement mechanisms to influence repayment.

In some rural areas borrowers may not have developed the discipline necessary to consistently repay formal financial institutions because in the past their debts may have been forgiven for political purposes, or they may divert their funds for other purposes (for instance, paying for a religious ceremony, or paying a debt to a local moneylender).(10) In these circumstances the presence of a collecting agent, someone who can diligently visit the farm and the borrower, may have a large effect on the probability of repayment.(11)

Another advantage of using a village head or a clan leader as a village agent is due to his greater ability to bear or to spread risk compared to other members of the rural community. This allows the financial institution to transfer the same amount of risk to the agent at a lower cost, and thus reduces the costs of the incentive scheme.

V. UNOBSERVABLE SOCIAL COSTS AND EFFORT

One implication of the model above is that inducing the agent to reveal his private information is costless to the financial institution (it does not cause an increase in the financial institution's costs). The assumption that drives this result is that the village agent receives no benefit nor incurs a cost, either pecuniary or non-pecuniary, from revealing his private information to the financial institution. Thus, the agent has no incentive to misrepresent his information. Although there exist few empirical studies that allow us to fully evaluate the validity of this assumption, it is certainly true that in many agrarian settings social forces play an important role (such as kinship ties) in everyday interactions, including economic ones [Clough, 1981; 1986].

In the event that the agent does receive some benefit, or incurs a cost from revealing his private information, the financial institution may find it more difficult and costly to derive an incentive compatible compensation scheme. In this scenario the financial institution would not only have to induce the agent to exert effort, but it must also induce him to truthfully reveal his private information. A simple modification to the basic model can illustrate this point. Suppose that it is common knowledge that the agent incurs some cost s, from other members of the village if he reveals damaging information about a farmer to the financial institution. I further assume that s can take only two values. Let [s.sub.b] be the cost incurred by the agent when he announces [a.sub.b] and the farmer is a bad type, while [s.sub.g] is the costs incurred by the agent when he announces [a.sub.b] and the farmer is a good type. I assume that [s.sub.g] [greater than] [s.sub.b] [greater than] 0. This assumption implies that the agent will incur positive social costs from revealing to the financial institution that a bad farmer is bad; however, the social costs will be higher if the agent identifies that a good farmer is 'bad'. While this is a strong assumption, it does allow me to clearly show the possible role that these types of costs can play in designing an incentive compatible scheme. The design problem for the financial institution is now clear. In addition to inducing a high effort level from the agent it must now also induce the agent to truthfully identify the bad types.

In addition to [w.sub.r] and [w.sub.n], the compensation scheme offered to the agent must now include [w.sub.b]. Let [w.sub.b] be the payment given to the agent when he announces [a.sub.b]. Finally, I assume that the probability of encountering a good farmer in the population is [Pi], where [Pi] [less than] 1.

For expositional purposes I will first isolate the effects of costly information revelation. Suppose that the financial institution is able to observe and enforce a high level of effort from the agent, but is unable to observe whether the agent truthfully revealed the true type of the potential borrower. The problem for the financial institution reduces to ensuring that the agent truthfully reveals his private information. In order to make the analysis more tractable, I simplify the basic model by assuming that p([e.sub.L]) = 0. That is, if the agent shirks in loan collection, the borrower will not repay. I continue to assume that 0 [less than] p([e.sub.H]) [less than] 1.

The problem for the financial institution is to choose [w.sub.r], [w.sub.n], and [w.sub.b] to:

min {[Pi][p([e.sub.H])[w.sub.r] + (1-p([e.sub.H]))[w.sub.n]] + (1 - [Pi])[w.sub.b]}

subject to:

(IR-g) p([e.sub.H])V([w.sub.r]) + (1-p([e.sub.H]))V([w.sub.n]) - [e.sub.H] [greater than or equal to] 0

(IR-b) V([w.sub.b]) - [s.sub.b] [greater than or equal to] 0

(IC-g) p([e.sub.H])V([w.sub.r]) + (1-p([e.sub.H]))V([w.sub.n]) - [e.sub.H] [greater than or equal to] V([w.sub.b]) - [s.sub.g]

(IC-b) V([w.sub.b]) - [s.sub.b] [greater than or equal to] V([w.sub.n]) - [e.sub.H]

Equations IR-g and IR-b are the agent's participation constraints for the good and the bad farmers, respectively, while equations IC-g and IC-b are the corresponding incentive compatibility constraints. Put differently, these constraints ensure that the agent finds it in his best interest to report a good farmer as good (IC-g), and a bad a farmer as bad (IC-b). Since it is assumed that the financial institution can observe e, no constraint is needed to ensure a high level of effort.

Proposition 5.1

If the financial institution is able to observe the agent's effort level, but unable to observe social costs [s.sub.g] and [s.sub.b], then the optimal compensation scheme for the village agent is: V([w.sub.r]) = V([w.sub.n]) = [e.sub.H], and V([w.sub.b]) = [s.sub.b].

Since the agent's choice of effort level is observable and enforceable, the financial institution compensates the agent for his disutility of effort, regardless of whether the loan has been repaid or not repaid (no incentives are needed to induce effort). Thus, IR-g must be binding and V([w.sub.r]) = V([w.sub.n]) = [e.sub.H]. A binding IR-g constraint implies that IC-b and IR-b are equivalent. In order to induce the agent to truthfully reveal bad farmers at minimum cost, IC-b must be binding, thus implying that V([w.sub.b]) = [s.sub.b]. Since it is assumed that [s.sub.g] [greater than] [s.sub.b], the optimal solution implies that IC-g is not binding.

The proposition states that in spite of the village agent having private information, the financial institution is still able to achieve the first-best outcome. This outcome implies that the agent is unable to use his private information strategically. To see this, suppose that an agent identifies a good farmer as 'bad'. Given the norms of the village he will incur social costs, [s.sub.g] [greater than] 0. However, the agent is compensated by an amount [s.sub.b]. Since [s.sub.g] [greater than] [s.sub.b], the agent will always have the incentive to report a good farmer as good. Not surprisingly, constraint IC-g is not binding at equilibrium. The problem that remains for the financial institution is to induce the agent to truthfully identify bad farmers. This can be done by compensating the agent for his social cost [s.sub.b].

While the agent does have private information concerning the true type of the potential borrower, the nature of these social costs prevents the agent from acting in a strategic manner. In effect, the social rules and norms that prevail in the village help to discipline the agent's announcements. Put differently, the financial institution is able to harness the social norms of the village to aid it in inducing truthful revelation.

Now suppose that the financial institution is unable to observe either the agent's level of effort, or the cost the agent incurs from revealing his private information. The financial must now design a compensation scheme that induces a high level of effort and truthful reporting [Myerson, 1982].

The problem for the financial institution is once again to choose [w.sub.r], [w.sub.n], and [w.sub.b] to:

min{[Pi][p([e.sub.H])[w.sub.r] + (1-p([e.sub.H]))[w.sub.n]] + (1-[Pi])[w.sub.b]}

subject to: IR-g, IR-b, IC-g, IC-b, and

(IC-e) p([e.sub.H])V([w.sub.r]) + (1-p([e.sub.H]))V([w.sub.n]) - [e.sub.H] [greater than or equal to] V([w.sub.b]) - [e.sub.L]

The design problem for the financial institution is as before except for the addition of IC-e. This constraint ensures that the agent find it in his best interest to exert a high level of effort in loan collection. The solution to this problem leads to the following proposition:

Proposition 5.2

If [s.sub.g] [greater than] [s.sub.b] and [s.sub.g], [s.sub.b], and e are unobservable to the financial institution, the optimal compensation scheme for the village agent is: V([w.sub.r]) = [e.sub.L] + ([e.sub.H] - [e.sub.L])/p([e.sub.H]), V([w.sub.n]) = [e.sub.L], and V([w.sub.b]) = [s.sub.b].

In order to reduce the costs of the incentive scheme, the financial institution would like to have as many constraints binding as possible. Note that if the constraints IR-b and IC-b are binding, then constraints IR-g and IC-e are equivalent. In order to ensure that the agent participates in the mechanism and tells the truth at the lowest possible cost, the financial institution would like both IR-b and IC-b to be binding. Thus, V([w.sub.b]) = [s.sub.b] and V([w.sub.n]) = [e.sub.L]. In order for the agent to exert a high effort level at the lowest possible cost, the constraint IC-e must also be binding. This implies that V([w.sub.r]) = [e.sub.L] + ([e.sub.H] - [e.sub.L])/p([e.sub.H]). Since [s.sub.g] [greater than] [s.sub.b], it can be easily shown that constraint IC-g is not violated, while IR-g is redundant.

The structure of the compensation scheme is as follows. If the loan is repaid the village agent will receive as compensation a wage-utility that is greater than his costs. That is, in equilibrium V([w.sub.r]) [greater than] [e.sub.H]. On the other hand, if the loan is not repaid the agent will receive as compensation a wage-utility that is less than his costs. Since in equilibrium the agent will exert [e.sub.H], the payment V([w.sub.n]), which is less than [e.sub.H], constitutes a loss in utility. Finally, the financial institution must now also compensate the agent for identifying a bad farmer. Thus, the compensation package now includes a payment, V([w.sub.b]), that is just equal to the disutility the agent incurs from reporting a bad farmer as bad.

As was seen earlier, the fact that the village agent must bear some risk in order to have the appropriate incentives increases the costs to the financial institution. However, once again, the agent is unable to use his private information strategically. The nature of the social costs within the village impose a constraint on his behavior that the formal financial institution is able to harness. Thus, while the financial institution's costs increases due to costly revelation (adding a payment V([w.sub.b]) = [s.sub.b]), the agent does not earn a rent from his private information. In fact, the increase in costs (relative to the first-best outcome) is due solely to the inability of the financial institution to observe the agent's effort level.

Similar implications arise from this compensation scheme with regard to the choice of an appropriate agent. Since V([w.sub.r]) is decreasing in p([e.sub.H]), the financial institution has the incentive to choose as agents those members of the rural community who have a large effect on the probability of repayment. Those members of the rural community likely to have such an influence on repayment rates are the village chief, the village headman, a clan leader, or anyone else who may be able to use their authority and status within the village to influence repayment. Another potential benefit of these types of agents is their ability to offset or divert any social sanctions directed their way (a lower [s.sub.b]), therefore mitigating their effects. Some caution is required in the interpretation of these social costs, however, since their effects are still not well understood and their magnitudes unknown.

VI. DISCUSSION

In two articles, McGregor [1988; 1989] describes how development credit programs in Bangladesh have used a broker model in linking the banking system and the rural poor. These 'brokers' or intermediaries may be NGOs or designated officials. Potential borrowers are screened by the project staff, or by the other nominated intermediaries. In addition to screening potential borrowers, the project (often through a staff member) may also make decisions about who will be eligible for loans, the writing of loan applications, the submission of applications to the bank, and the collection of loan repayments. McGregor emphasises that there is little or no direct contact between the target group (the rural poor) and the banking system.

As an example of the broker model, McGregor cites the work of the Swanirvar (Self Reliance) development programme. Over time its credit programme has evolved to be almost identical to the Grameen Bank, with one important difference. The banking functions are not carried out by the programme itself, rather it is done through a national agreement with the nationalised commercial banks. McGregor suggests that a key role is played by the Swanirvar Union-level worker who acts much like a loan assistant. His position is described as being 'semi-voluntary', and his income is calculated on the basis of the amount of loans disbursed and recovered. However, some confusion exists as to the true nature of the compensation scheme since Maloney and Ahmad [1988] report a small fixed fee per week per borrower as the compensation to the loan facilitator. No further information is provided by Maloney and Ahmad as to whether this is the only compensation given to the loan assistant.

Maloney and Ahmad also report and critique various rural credit projects and organisations (mostly NGO's) that are linked to banks and use committees to nominate borrowers. In their recommendations they suggest that expectations should be built into the system so that those who nominate a borrower for a loan have the moral responsibility to encourage the borrower to repay and, in addition, place verbal pressure on him in case of nonrepayment [Maloney and Ahmad, 1988; 189]. Maloney and Ahmad also suggest that commissions be given for loan recovery and incentives to recover loans [ibid.: 204].

Mone [1994] provides a description of the credit delivery services of Bank Umum Purba Danarta (BUPD), a non-profit bank in Indonesia. The bank is organised into three layers. At the top is the Staff which decides upon bank policy and manages the bank's operations. Below the Staff are the Station-Managers (STM). The station managers, in turn, supervise a maximum of twenty full-time field-officers, Pembina Purna Waktu (PPW). The bank decentralises its services by allowing the field officers to approve all loans up to a limit (Rp 150,000). For a loan amount above this limit the field officer must get approval from his station manager, and in some cases the Staff.

The bank uses the field officer to screen and monitor borrowers, and collect loan repayments. After a few years of experience, some field officers are able to specialise in one sector or industry, such as the fishery or machine-tool industry. Relationships are first established by a field officer providing a prospective borrower with a savings account. It is these personal relationships that the bank wishes to establish before providing credit services. As Mone points out, after establishing a personal relationship with a field officer, a borrower feels 'socially obliged' to repay a loan. In addition, this relationship allows the field officer to be better informed about the risk characteristics of the borrower prior to the loan being granted, and to effectively monitor the borrower after the loan has been made.

In order to give the field officer the proper incentives, the bank designed a compensation scheme that is based solely on his performance. The field officer's incentive scheme is contingent on full loan repayments and the amounts of deposits generated. The field officer receives seven and a half per cent annually of every savings account he takes care of plus an additional two per cent of every fully repaid loan. While the bank does not provide any base pay it has set a minimum standard which should ensure the field officer a 'sufficient' level of income. No detail is given by the author, however, as to how this minimum standard is enforced or to its overall efficiency.

Chares and Gonzalez-Vega [1996] provide further evidence from Indonesia on a similar type of delivery mechanism. In their article the authors present a description and analysis of rural financial institutions that utilise village chiefs, clan leaders, or local residents from the area as agents in screening, monitoring, and in loan collection. In explaining the apparent success of Indonesian rural financial institutions, Chaves and Gonzalez-Vega point to the ability of these institutions to incorporate the wealth of information on potential borrowers that is available at the village level. Instead of utilising traditional screening methods, a financial institution is now able to use character references provided by the agent, or simply have the agent sign the application form of the borrower. In fact, the authors report that all the rural financial institutions in the regions covered by their investigation required the loan applicant to provide a character reference from the clan leader. Furthermore, the authors note that this announcement of borrower type by the agent often served as a collateral substitute.

In addition to providing screening services, the agents were also used in monitoring loan terms and in loan collection. Thus, in addition to incorporating the agent's superior information, financial institutions also attempted to access the village-level enforcement mechanisms that are available to the agent because of his position in the village. Having access to these local enforcement mechanisms served to increase the probability of repayment that the financial institutions faced in the rural areas, since there were doubts as to their ability to enforce contracts through the legal system.

In most of the Indonesian delivery mechanisms the village agent was also delegated the authority to grant loans and to set loan amounts. There also appeared to be no effective supervision by the government of the effective interest rate charged, and of the use of the loan funds by the agent. Although most of these village units operated within a general framework of financial policies set by the government, actual implementation varied greatly across village units and banks. Chaves and Gonzalez-Vega argue that this allowed for greater flexibility at the village level and allowed the agents to adapt to the specific environment.

In efforts to mitigate the agency problems that arise in this form of delivery mechanism (such as collusion in the screening process and low effort level in loan collection), the financial institution offered the village agent a compensation scheme that consisted of a fixed wage plus a bonus, or an incentive payment. Similar to the optimal compensation scheme derived above, the Indonesian compensation scheme was also a function of some observable variables, such as loan instalments collected, or profits. Furthermore, the authors suggest that the total remuneration of the managers (the village agents) consisted of an 'efficiency wage'. By an efficiency wage the authors mean a level of compensation that is higher than the opportunity cost of the worker. The authors argue that this raises the costs to the agent of being fired, and hence, provides him with the incentive to work hard in his assigned tasks.

A quick comparison of the optimal incentive scheme discussed in this paper and the Indonesian scheme reveals a similar structure when the loan is repaid. Recall that under the optimal compensation scheme, the agent receives a reward when the financial institution is repaid, while in the Indonesian scheme the agent receives a bonus, or an incentive payment when loan instalments are collected.

However, a difference does arise in the structure of the incentive scheme when a loan is not repaid. Under the optimal incentive scheme the agent receives a penalty, while under the Indonesian scheme the agent receives a fixed fee and no bonus. Unfortunately, insufficient information is available to evaluate the utility derived by the agent from the fixed fee payment. If the fixed fee payment does not fully compensate the agent for all of his costs (effort) in loan collection then the agent suffers a penalty. Under this scenario the fixed fee payment would be consistent with the penalty payment under the optimal incentive scheme.

On the other hand, the efficiency wage portion of the Indonesian scheme seems somewhat troubling to explain. Chaves and Gonzalez-Vega's analysis suggests that both portions of the scheme, the performance based portion and the efficiency wage portion, are needed to give the agent the proper incentives. The authors' analysis, however, provides no rationale for why this is true. The answer may lie in the fact that in the Indonesian case the agent is delegated the authority to extend loans and set loan amounts. The delegation of these tasks to the agent will undoubtedly alter the incentive scheme offered by the financial institution. How it will be altered has yet to be explored. Further theoretical work is needed on the Indonesian scheme and similar schemes before their efficiency can be fully evaluated.

VII. CONCLUSION

This article has examined some of the agency problems that arise when a formal financial institution utilises a member of the village community to act as an agent in screening potential borrowers and in loan collection. Optimal incentive schemes were derived that induce the agent to truthfully reveal his private information regarding the creditworthiness of the applicant, and to provide the appropriate level of effort in loan collection. These incentive schemes require that the formal financial institution deviate from the optimal risk-sharing arrangement by sharing with the agent the risks of lending to small farmers. This deviation causes an increase in the financial institution's costs since it now must expend resources to align the agent's incentives with its own. In addition, this paper showed how the norms and rules that govern village life may aid the financial institution by constraining strategic behavior by the agent.

In the model the factor that induces farmer repayment is the 'cost' that the agent is able to impose on the farmer. Further research in this area, both empirical and theoretical, will need to address how social sanctions are used to induce contract compliance within a village and, in turn, under what conditions a formal financial institution will be able to access them.

One benefit of this research is that it has helped in identifying and separating the different roles that the agent may take on for the financial institution. For instance, when the agent is screening potential borrowers his role does not differ significantly from that of a credit rating agency in a more developed economy. On the other hand, the agent is also utilised by the financial institution as a collection agent. Each role that the agent may take on for the financial institution carries with it its own agency problems. In addition, each role also contributes to the costs of designing an incentive compatible compensation scheme. Thus, depending on the relative costs, it is possible that the financial institution may decide not to use the agent in one of these roles. For example, in certain agrarian settings a rural bank may have little difficulty in identifying creditworthy farmers but it may have considerable problems in enforcing repayment. In this situation the rural bank may need a collection agent but not a credit rating agency. Once again, further empirical and theoretical research can help in pinpointing the conditions under which the financial institution would find it optimal to hire the agent as a credit rating agency, a collection agent, or both.

One of the limitations of the incentive scheme that is derived in this paper is that it is independent of the loan size. The loan size is implicitly fixed in the model. While the model delegates to the village agent the decision of who will receive a loan, he is not delegated the authority to decide the loan size. This assumption will need to be relaxed in future work.

Finally, while important socio-political and cultural issues are ignored in this analysis, they may be very important factors in determining the effectiveness of this proposed delivery scheme [Crow, 1994; McGregor, 1989]. Serious considerations must be given as to how the existing power structure of the rural community may be affected by, or may affect the use of village agents in rural credit delivery.

NOTES

1. For a critique of the traditional agricultural programmes attempted by low-income countries and the various assumptions underlying them, see Adams and Graham [1981] Adams and Vogel [1986] and Adams, Graham and Von Pischke [1984]. See also Rashid and Townsend [1994] for a new methodology for evaluationg financial systems in low-income countries.

2. The agency problems that could arise in this credit delivery mechanism have been identified by Miracle [1973], Bell [1990] and Chaves and Gonzalez-Vega [1996].

3. For a discussion of the problems faced by formal financial institutions operating in rural areas, see Braverman and Guasch [1986; 1989].

4. The role that information and incentive problems play in rural credit markets is now becoming better understood. For a recent discussion of the issues see Hoff and Stiglitz [1990]. In addition, see the citations in Bell [1988], Bardhan [1989], and the chapters on credit in Hoff, Braverman and Stiglitz [1993].

5. See Donald's [1976: Ch.10] for a discussion of the banking practices of formal financial institutions and how they can be (re)formed so as to acquire information on the small farmer more effectively.

6. There are at least two other mechanisms that have received attention in the literature that attempt to link formal lenders with the rural poor. One method consists of a formal lender lending to a group of borrowers that have joint liability. The Grameen Bank of Bangladesh remains the classic example of this mechanism [Hossain. 1988]. See Stiglitz [1990], Besley and Coate [1995] and Conning [1996b] for an economic analysis of the incentives within this form of delivery mechanism. Seibel and Marx [1987], on the other hand, provide case studies from Africa of formal lenders linking with various forms of self-help groups that are similar in spirit to the group lending schemes cited above.

The second method links formal lenders with informal lenders (such as crop traders, input dealers, or village moneylenders). This mechanism would have formal lenders lend directly to informal lenders who would, in turn, on-lend to the small farmers. Although there have been earlier discussions of this form of credit delivery mechanism [Moore, 1953], a more recent argument for the use of informal lenders as agents is contained in Bell [1990]. Recent economic analysis on this form of delivery system can be found in Esguerra [1987], Floro and Ray [1993] and Hoff and Stiglitz [1996].

7. See Wade [1988] for a description of the village institutions used to penalise and sanction members of the community for non-cooperative behaviour.

8. One constraint is missing but not needed. To see why, suppose the agent lies and announces that a bad farmer is good. Since a bad farmer never repays a loan the agent will receive V([w.sub.n]) as compensation. the agent must still choose a level of effort, however. Since [e.sub.L] [less than] [e.sub.H] the agent will always choose to exert a low effort level since it will provide him with a higher level of utility. Since the agent will never choose to lie and exert a high level of effort, the constraint never binds.

9. The agent's indifference curve can be represented as: p([e.sub.i])V([w.sub.r]) + (1 - p([e.sub.i]))V([w.sub.n]) = k, for i=L, H. Rewriting the equation we can see that, V([w.sub.r]) = k - [(1 - p([e.sub.i]))/p([e.sub.i])] x V([w.sub.n]). Thus the slope of the typical indifference curve is negative.

10. See Conning [1996a; Ch.6] for a discussion and analysis of 'pirate sales' in rural Chile.

11. As an example of how a programme can teach poor borrowers the discipline of repaying a loan, see Hossain [1988] for information on the Grameen Bank's methods.

REFERENCES

Adams, D.W. and D.H. Graham, 1981, 'A Critique of Traditional Agriculture Credit Projects and Policies', Journal of Development Economics, Vol.8, No.3, pp.347-66

Adams, D.W., Graham D.H., and J.D. Von Pischke (eds.), 1984, Undermining Rural Development with Cheap Credit, Boulder, CO: Westview Press.

Adams, D.W. and R.C. Vogel, 1986, 'Rural Financial Markets in Low Income Countries: Recent Controversies and Lessons', World Development, Vol. 14, No.4, pp.477-87.

Bardhan, Pranab, 1989, The Economic Theory of Agrarian Institutions, New York: Oxford University Press.

Bell, C., 1988, 'Credit Markets and Interlinked Transactions', in H.B. Chenery and T.N. Srinivasan (eds.), Handbook of Development Economics, Amsterdam: North-Holland.

Bell, C., 1990, 'Interactions Between Institutional and Informal Credit Agencies in Rural India', World Bank Economic Review, Vol.4, No.3, pp.297-327.

Besley, T. and S. Coate, 1995, 'Group Lending, Repayment Incentives and Social Collateral', Journal of Development Economics, Vol.46, No.1, pp.1-18.

Braverman, A. and J.L. Guasch, 1986, 'Rural Credit Markets and Institutions in Developing Countries: Lessons for Policy Analysis from Practice and Modem Theory', World Development, Vol.14, No.10/11, pp.1253-67

Braverman, A. and J.L. Guasch, 1989, 'Rural Credit in Developing Countries', Journal of Economic Development, Vol.14, No.1, pp.7-34.

Chaves, R. and C. Gonzalez-Vega, 1996, 'The Design of Successful Rural Financial Intermediaries: Evidence from Indonesia', World Development, Vol.24, No.1, pp.65-78.

Clough, P., 1981, 'Farmers and Traders in Hausaland', Development and Change, Vol.12, pp.273-92.

Clough, P., 1986, 'The Social Relations of Grain Marketing in Northern Nigeria', Review of African Political Economy, No.34, pp.16-34.

Conning, J., 1996a, 'Financial Contracting and Intermediary Structures in a Rural Credit Market in Chile: A Theoretical and Empirical Analysis', unpublished doctoral dissertation, Department of Economics, Yale University, New Haven, CT.

Conning, J., 1996b, 'Group Lending, Moral Hazard, and the Creation of Social Collateral', mimeo, Department of Economics, Williams College, Williamstown, MA.

Crow, B., 1994, 'Finance in Context: Exploring Diverse Exchange Conditions', in F.J.A. Bouman and Otto Hospes (eds.), Financial Landscapes Reconstructed.' The Fine Art of Mapping Development, Boulder, CO: Westview Press.

Donald, G., 1976, Credit for Small Farmers in Developing Countries, Boulder, CO: Westview Press.

Esguerra, E., 1981, 'The Redistributive Potential of the Masagana 99 Credit Subsidy', unpublished M.A. thesis, School of Economics, University of the Philippines, Diliman.

Esguerra, E., 1987, 'On the Use of Informal Lenders as Conduits for Formal Credit: The Case of the National Agricultural Productivity Programs in the Philippines', Economics and Sociology Occasional Paper No. 1351, Ohio State University, Columbus, OH.

Floro, M.S. and D. Ray, 1993, 'Direct and Indirect Linkages between Formal and Informal Institutions: An Analytical Approach', mimeo, Department of Economics, American University, Washington, DC.

Graham, D.H., 1992, 'Informal Rural Finance in Niger: Lessons for Building Formal Institutions', in D.W. Adams and Delbert A. Fitchett (eds.), Informal Finance in Lot-Income Countries, Boulder, CO: Westview Press.

Grossman, S.J. and O. Hart, 1983, 'An Analysis of the Principal-Agent Problem', Econometrica, Vol.51, No.1, pp.7-45.

Hoff, K. and J. Stiglitz, 1990, 'Introduction: Imperfect Information and Rural Credit Markets-Puzzles and Policy Perspectives', World Bank Economic Review, Vol.4, No.3, pp.235-50.

Hoff, K. and J. Stiglitz, 1996, 'Moneylenders and Bankers: Price Increasing Subsidies in a Monopolistically Competitive Market', Journal of Development Economics, forthcoming.

Hoff, K., Braverman A., and J. Stiglitz (eds.), 1993, The Economics of Rural Organization, New York: Oxford University Press.

Hossain, M., 1988, 'Credit for Alleviation of Rural Poverty: The Grameen Bank in Bangladesh', IFPRI Report No.65, International Food Policy Research Institute, Washington, DC.

Kreps, D.M., 1990, A Course in Microeconomic Theory, Princeton, NJ: Princeton University Press.

Larson, D.W. and R. Urquidi, 1987, 'Competitiveness of Informal Financial Markets in Bolivia', Department of Agricultural Economics and Rural Sociology Working Paper No. ESO 1426, Ohio State University, Columbus, OH.

Maloney, C. and A.B. Ahmad, 1988, Rural Savings and Credit in Bangladesh, Dhaka, Bangladesh: University Press Ltd.

McGregor, J.A., 1988, 'Credit and the Rural Poor: The Changing Policy Environment in Bangladesh', Public Administration and Development, Vol.8, pp.467-82.

McGregor, J.A., 1989, 'Towards a Better Understanding of Credit in Rural Development The Case of Bangladesh: The Patron State', Journal of International Development, Vol.1, No.4, pp.467-86.

Miracle, M., 1973, 'Notes on Developing Small Farmer Credit Institutions in Third World Countries', in A.I.D., Spring Review of Small Farmer Credit, edited and published by the Agency for International Development, Washington, DC.

Mone, E., 1994, 'Informalizing Formal Banking?: The Case of Bank Umum Purba Danarta, Indonesia', research paper, Department of Development Economics, Wageningen Agricultural University, Wageningen, The Netherlands.

Moore, F., 1953, 'Money-lenders and Co-Operators in India', Economic Development and Cultural Change, Vol.2, No.2, pp.139-59

Myerson, R., 1982, 'Optimal Coordination Mechanisms in Generalized Principal-Agent Problems', Journal of Mathematical Economics, Vol.10, No.1, pp.67-81.

Onchan, T., 1992, 'Informal Rural Finance in Thailand', in D.W. Adams and Delbert Fitchett (eds.), Informal Finance in Low-Income Countries, Boulder, CO: Westview Press.

Rashid, Mansoora and R.M. Townsend, 1994, 'Targeting Credit and Insurance: Efficiency, Mechanism Design and Program Evaluation', ESP Discussion Paper Series No.47, World Bank, Washington, DC.

Seibel, H.D. and M.T. Marx, 1987, Dual Financial Markets in Africa, Saarbruecken Ft. Lauderdale: Breitenbach Publishers.

Stiglitz J.E., 1990, 'Peer Monitoring and Credit Markets', World Bank Economic Review, Vol.4, No.3, pp.351-66.

Timberg, T.A. and C.V. Aiyar, 1984, 'Informal Credit Markets in India', Economic Development and Cultural Change, Vol.33, No.1, pp.43-59.

Varian, H., 1992, Microeconomic Analysis, New York: W.W. Norton & Co.

Von Pischke, J.D., 1991, Finance at the Frontier, Washington, DC: World Bank.

Wade, R., 1988, Village Republics, Cambridge, UK: Cambridge University Press.

Wells, R.J.G., 1978, 'An Input Credit Programme for Small Farmers in West Malaysia', Journal of Administration Overseas, Vol.17, pp.4-16.

Yaron, J., 1992, 'Successful Rural Finance Institutions', World Bank Discussion Paper No.150, Washington, DC.

APPENDIX

Proof of Proposition 4.1: The claim is that constraints 1 and 3 are binding. The proof proceeds by showing that the Lagrange multipliers associated with constraints 1 and 3, respectively, are positive. The set of payments that are derived from this sub-programme are then shown not to violate the remaining constraint (constraint 2) from the full programme. Let [[Lambda].sub.1] and [[Lambda].sub.2] be the Lagrange multipliers associated with the constraints 1 and 3, respectively. Let [[Gamma].sub.[Theta]] represent the derivative of the Lagrangian with respect to the payment received in the [Theta]th state. The first order conditions for cost minimisation are the following:

[[Gamma].sub.r] = p([e.sub.H]) - [[Lambda].sub.1]p([e.sub.H])[V[prime].sub.r] - [[Lambda].sub.2]p([e.sub.H])[V[prime].sub.r] + [[Lambda].sub.2]p([e.sub.L])[V[prime].sub.r] = 0

[[Gamma].sub.n] = (1 - p([e.sub.H])) - [[Lambda].sub.1](1 - p([e.sub.H]))[V[prime].sub.n] - [[Lambda].sub.2](1 - p([e.sub.H]))[V[prime].sub.n] + [[Lambda].sub.2](1 - p([e.sub.L]))[V[prime].sub.n] = 0

In addition there are the complementary slackness conditions. The following solutions for the multipliers can be found.

[[Lambda].sub.1] = [p([e.sub.L])[V[prime].sub.n][Delta]p - (1 - p([e.sub.H]))(p([e.sub.L]) - p([e.sub.H]))[V[prime].sub.r]]/([V[prime].sub.n][V[prime].sub.r][ Delta]p)

[[Lambda].sub.2] = [(1 - p([e.sub.H]))p([e.sub.H])([V[prime].sub.n] - [V[prime].sub.r])]/([V[prime].sub.n][V[prime].sub.r][Delta]p)

Recall, [Delta]p [equivalent to] (p([e.sub.H]) - p([e.sub.L])). Since p(.) is positive, and the derivatives of the agent's utility function with respect to [w.sub.r] and [w.sub.n] are greater than zero and are not equal to each other, and since [w.sub.r] [greater than] [w.sub.n], the Lagrange multipliers can be shown to be positive. The payments derived from the binding constraints can be denoted, in utility terms, by V([[w.sub.r].sup.*]) and V([[w.sub.n].sup.*]). In order to complete the proof it must be shown that these proposed equilibrium payments do not violate the remaining constraint. For constraint 2 not to be violated requires that [e.sub.L] [greater than or equal to] V([[w.sub.n].sup.*]). It can be shown by algebraic manipulation that this is true. Hence, constraint 2 is not violated and therefore V([[w.sub.r].sup.*]) and V([[w.sub.n].sup.*]) are solutions to the problem.

Assistant Professor, Department of Economics, Loyola Marymount University, Los Angeles, CA (USA). The author thanks George Akerlof, Pranab Bardhan, Alain De Janvry, Claudio Gonzalez-Vega, Karla Hoff, Tom Micelli, and a referee of this journal for helpful comments. The author especially thanks Manny Esguerra, Raul Fabella, Ben Hermalin, and Michael Kevane for extensive discussions. In addition, he thanks seminar participants at Berkeley, the Board of Governors, Connecticut, Food Research Institute, George Washington, New York Fed, Ohio State, UC-Riverside, USC, and the University of the Philippines-Diliman.

请问有PDF格式吗？可否发到我的邮箱alexcheung_eco@163.com谢谢

是这个

本来还以为不能下PDF，今天看了下发现可以下载

非常感谢