CREDIT AND TRAINING PROVISION TO THE POOR BY VERTICALLY CONNECTED NGO S AND COMMERCIAL BANKS

Transcription

1 CREDIT AND TRAINING PROVISION TO THE POOR BY VERTICALLY CONNECTED NGO S AND COMMERCIAL BANKS Gherardo Gino Giuseppe Girardi Econoics, Finance and International Business London Metropolitan University g.girardi@londoneac.uk Work in progress, do not quote Third draft: July 007 Abstract: An NGO trains potential entrepreneurs and lends the oney in period. In period, those who are successful in their projects are eligible to apply for bank loan Epowerent is defined as the aount of training which the NGO provides to each trainee, and outreach as the nuber of successful trainees at the end of the first period. We ask the following question: does the presence of a socially responsible agent (the NGO) ensure that the arket outcoe coincides with the social optiu? We find that the answer depends on the type of social objective: if the policy aker seeks to axiize the welfare of the poor, then the arket correctly provides epowerent but underprovides outreach. If, on the other hand, the social planner seeks to axiize the social surplus (the su of welfare of the poor and bank profits), then the arket correctly provides both epowerent and outreach. We then assess, in the context of the odel, the validity of the policy of capping icrofinance interest rates that is practised in soe countrie Finally, we allow the NGO and a bank to erge, and copare the resulting arket outcoe with the social optiu; we then apply the results to the present situation in Latin Aerica, where banks are entering the very poor segent of the arket in which NGO s operate. Keywords: NGO, bank, outreach, epowerent Introduction How do NGO s and banks interact in the provision of credit to the poor? There is soe discussion of this issue in icrofinance anuals, newspapers and the interne However, there is little in the for of rigorous analysis, both epirical and theoretical. McIntosh and Wydick (005) odel copetition between icrofinance institutions, and Conning (999) studies an NGO s proble of axiizing outreach to the poor whilst reaining financially viable and ensuring project sustainability. Iportantly for our work, it appears that no theoretical work has been done on vertical linkages between NGO s and banks, in particular on what coentators tend to refer as graduation. This is the idea that, having started and grown on subsidized NGO credit, icro enterprises graduate when they have grown sufficiently and are large and financially viable enough to switch to non-subsidized, coercial bank credi This paper is, to the best of our knowledge, a first attept at the theory. We odel a coon situation where an NGO trains (potential) entrepreneurs and lends the oney. Soe of these trainees succeed in that their projects do well. In this case, they have sufficient oney to use as collateral and can therefore apply for a bank loan. It is not clear at all fro the literature whether NGO s care about graduation, and, consequently, very little is known about what strategies they adopt, if any, to enable their clients to graduate. An exaple of an NGO which we know cares about graduation is the Indian NGO BYST (Business and Youth Starting Together). BYST trains prospective entrepreneurs, ainly by providing the with a business entor, and lends the oney at low interest rates and without collateral requireen It then organizes eetings between successful prograe participants and banks See for exaple Ledgerwood (999).

2 in an effort to ensure that these participants receive credit to continue with their business enterprises. What we do know is that policy akers are concerned about NGO clients oving on to bank loan For exaple, India s National Bank for Agriculture and Rural Developent is dedicating a lot of effort and resources to create linkages between self-help groups and banks (NABARD (007)). NABARD goes as far as seeing icrofinance as an instruent for graduation. The process of graduation, according to NABARD, is ade up of two stages: in the first stage a poor person and potential entrepreneur oves fro subsistence living to owning a icroenterprise; in the second stage, his or her business grows to becoe a sall/ediu enterprise (SME). NABARD clais that icrofinance operators are unable to eet the credit needs of borrowers whose business has grown significantly, so that they are left in the lurch. NGO s are seen as training ground for future SME entrepreneur In particular, NGO s are valued for their ability to nurture self-help groups (a very coon for of icrofinance in poor countries) and to entor the in their relationship with bank An iportant advantage of the self-helpgroup/bank linage odel is that the NGO akes available to the bank the client s credit and entrepreneurial history, enabling banks to select 3 those candidates ost likely to success as SME anager As of today, there are about 800,000 self-help groups in India [with between 0 and 0 individuals in each group] who are benefiting fro linkages with bank The ain issue that we wish to explore are the following: (a) The interaction between altruistic agents (NGO s) and opportunistic ones (coercial banks); (b) The trade-off between the nuber N of poor people that an NGO helps ( outreach ) and the level of training ( epowerent, e) that it iparts to each person; (c) The role of collateral as required by banks in explaining the trade-off between N and e; (d) The role of institutions in the trade-off between N and e, particularly the effectiveness of the legal syste in punishing those who voluntarily default 4 ; (e) Most iportantly, whether the NGO trains too any or too few poor people relative to the social optiu (over or under provision of N), and whether it trains each one of the too uch or too little (over or under provision of e). (f) How the under/over provision of N and e is affected by the objectives of the NGO and of society, for exaple whether they try to axiize the nuber of people above the poverty line (absolute poverty approach) or whether they seek the greatest happiness for the greatest nuber (utilitarian approach); (g) How the under/over provision of N and e is affected by whether the NGO has a short tie horizon, so it does not take into consideration that its training helps its clients to obtain coercial loans in the future, or a long tie horizon. The connection between this paper and the literature on corporate social responsibility is that the NGO is a socially responsible agent, whilst coercial banks are not: their objective being profit axiization, they are opportunistic agent We therefore have a odel of an agent, the NGO, which adopts the stakeholder approach, whilst coercial banks adopt the stakeholder approach (see De Wit and Meyer(004)). The question we ask is, does the presence of a socially responsible agent in a vertical relation with an opportunistic agent result in the arket outcoe coinciding with the social optiu; we find that, if the policy akers shares the sae objective as the socially responsible agent, the answer is no. The paper proceeds as follow In part we explain the odel and solve the banks and the NGO s proble. In part 3 we consider the social optiu, and how it copares with the arket optiu. In part 4, we allow the NGO and a bank to erge, and copare the new See 3 We don t odel this kind of selection, leaving it to a future paper on asyetric inforation between NGO s and bank 4 We would not want to punish, though, those who voluntarily default because they have to pay for necessities like a relative s operation resulting fro an unforeseen edical coplication. This is particularly true when private insurance is too expensive and when the public health syste is too thin.

3 arket outcoe with the social optiu; we apply the result to the present situation in Latin Aerica. Part 5 concludes and offers soe avenues of further research. The odel We assue that there are periods, with lending and training by the NGO in period, and lending by the bank in period. The gae tree looks as follows: Figure : net return to clients of NGO/banks q R -(+r )L p R -(+r )L -q 0 R -cl -p 0 Tie (period) In period, an NGO trains N prospective entrepreneurs and lends the oney. There are high skill and low skill trainees, whose ability is unknown to everyone (including to the trainees) at the start of the training prograe. The trainees do not know their skill types as they have never before participated in a training prograe of the type iparted by the NGO. The different skill types are present in the population in proportions p and (-p). At the end of the training prograe (i.e. at the end of period ), the skill type becoes known to all 5. In the second period, the fraction p of trainees who are successful approach a bank and apply for a loan. The bank can distinguish the successful trainees fro the unsuccessful ones as the success or failure of their project is observable. At this point a further separation takes place. The bank faces high and low skill borrowers, whose ability is unknown to everyone (including to the borrowers) at the start of the second period. The borrowers do not know their skill type as 5 In practise, there is an iportant coplication. Many people who successfully pass the first stage never wish to proceed any further. This ay be because they have obtained what they wanted, naely an increase in the flow of incoe for their failies, or because they are risk averse. The class of people who proceed to the second stage is altogether different: they are entrepreneurs with a keen abition to succeed and quite possibly a risk-loving attitude. A odel accounting for these different types ay yield rich insights for researchers and policy aker 3

4 they have never before anaged a large, foral business, with all its requireents, and with credit fro a coercial bank. The different skill types are present in the population in proportions q and (-q). At the end of period, the skill type becoes known to all. Let us begin our analysis with the second period. Banks are evenly distributed around a Salop circle, with bank clients (forerly NGO trainees) uniforly distributed along i Figure : the Salop odel Bank i+ Borrower j n x j x j Bank i The distance between each bank is /n. Let α be the borrower s cost of travelling unit of distance to the bank (the higher α, the ore differentiated the banks are fro the borrower s perspective). Let the bank i's interest rate by r i and its collateral rate by c i. The arginal borrower, who is indifferent between bank i and bank i+, is characterized by x j =x and thus q ( R ( + r ) L ) + ( q)( R c L) αx = q( R ( + r ) L ) + ( q)( R c L) x i i i+ i+ α As there are custoers on both sides of the bank s location, the nuber of borrowers is given by pqnx, with equation giving x. We assue liquidation costs of collateral δc i L. Bank profits per custoer j, Π j, are given by () n j [ q r + ) + ( q) c δc ] Π = L ( () i i i Bank i will axiize profits subject to the condition that there is no voluntary defaul By setting the collateral rate is to prevent voluntary defaul After receiving the oney fro the bank (i.e. "ex post"), the borrower does not default as long as the loss fro defaulting (weakly) exceeds the gain: 4

5 ( r ) L c L P (3) + i i + where P is the exogenous (expected value of the) penalty which a successful entrepreneur would be charged by the court if he defaulted. P easures the degree of effectiveness of a country s legal syste: if the likelihood of the law catching and punishing voluntary defaulters rises, then P rise It could also easure the cost of not being able to borrow in the future as a result of having defaulted. Each bank axiizes total profits, i.e. the nuber of borrowers pqnx (where x is given by equation ) ultiplied by profits per borrower (equation ), subject to (4): = pqn n α (3) { q[ ( R ( + r ) L ) ( R ( + r ) L )] + ( q)( [ R c L) ( R c L) ]} i i i+ i i+ * L [ q( r + ) + ( q) c δc ] i i If δ=0, the result is a linear cobination of c i and r i, not specific values of these variable However, if δ>0, then each bank would want to iniize c i (and axiize r i ); in this case, (4) is binding. In equilibriu, r i = r i+ = r i- = r and c i = c i+ = c i- = c. The solutions are i * (4) δ ( ) α P P + q qp L L α r = +, c = + L (5) Ln δ Ln δ Note that the interest rate which banks charge increases with (-q) because this iplies a greater risk of default, which ust be copensated by a higher interest rate for banks to break even. What is the NGO's objective? We assue the NGO adopts an absolute poverty approach whereby it wants as any people as possible to get across the poverty line λ. In other words, the NGO wishes to axiize pn, where pn is the nuber of successful trainees in the first period. For siplicity, we assue that, if the bank financed project does not succeed, a successful trainee in period will obtain a flow of incoe fro his or her period investent equal to his flow of incoe in period ; in other words, a achine bought in period thanks to a positive project outcoe will yield the sae return across periods 6. To ake the link explicit with the terinology of icrofinance and also with the capabilities literature, We call the aount of training per person e "epowerent", and the nuber of successful trainees at the end of the first period pn outreach. Epowerent is useful in increasing the proportion of successful trainees, i.e. in increasing outreach. Consider now what happens in period. The NGO receives an aount of donated oney 7. We suppose that the NGO can borrow fro the banking sector 8. Then it can borrow at 6 We also do not allow NGO clients to spread interest payents over both period 7 Many, if not ost, NGO s rely priarily on funds fro donors based in rich countries as a source of incoe. 5

6 the start of period on the strength of the inflow of cash in the for of interest payents at the end of the first period. For siplicity and to avoid introducing eleents that are not central to our analysis, we assue that the NGO does not need to deposit collateral with the banks, and that it pays a fixed rate of interest r 0 to the banks of zero. Then the budget constraint of the NGO is + pnl + r ) = N( L + ) (6) ( e The right hand side of the equation shows the expenditure per trainee, which is the su of the loan given to each trainee and the cost of training, ties the nuber of trainee The left hand side shows total incoe. We assue that project success in period is defined as achieving a return on investent which puts the entrepreneur above the poverty line λ, i.e. R >λ. How uch oney does the NGO wish to leave its successful clients with, via its choice of interest rate? The answer depends on the objective of the NGO. If the NGO's tie horizon is short and so it does not care about aking it possible for its clients to obtain coercial loans in the second period, then it siply leaves its clients with λ, the poverty line. Thus, R ( + r ) L = λ. If clients apply for bank loan and their project fails, then, as a result of losing the collateral, they will fall below the poverty line. This goes against the NGO s objective. However, this need not bother the NGO: the NGO ay reason that if entrepreneurs choose to run the risk of falling below the poverty line, they indicate that running such a risk is ore attractive for the than to stay with a constant level of incoe given by the poverty line. If, on the other hand, the NGO s tie horizon is long and hence it is concerned with aking sure that no client falls below the poverty line at the end of the second period, then it will ake sure that the client with the greatest transport costs in period will not fall below the poverty line if the bank financed project turns out unfavourably. It will leave the with just enough to ake it possible for the successful clients to pay the banks' collateral and incur α transport costs and still be at the poverty line. In this case, R ( + r ) L = λ + cl + n. We can suarize the attitude of the NGO as follow α R ( + r ) L = λ + ϕ cl +, where ϕ [,0] (7) n In words, if the NGO s tie horizon is long, then φ=; if short, φ=0. Suppose that training increases the probability that the NGO project succeeds: p = β e, < and p (8) 8 This is not a critical assuption: in a odel with any periods, the NGO could siply use interest received fro trainees trained in a previous period to train trainees in the current period. Which ever way we look at it, because the interest is received with a delay of one period, we need to discount the interest received. 6

7 R and R are fixed (for exaple at 0% of L and L ). The NGO's proble is then Max pn N, e, r (5),(6),(7),(8) (9) Let us consider the solutions to this proble. r is used by the NGO to ensure that its clients achieve the poverty line λ if the NGO has a short tie horizon (φ=0), or λ+cl if the NGO has a long tie horizon (φ=). This iplies that r is lower if the NGO has a long tie horizon: α R λ + ϕ cl + n r =, ϕ [0,] (0) L Fro the NGO s budget constraint (6), N = L + e p( + r L () ) The optial epowerent level fro the NGO's point of view is therefore L e = () Intuitively, epowerent e (training per head) rises with L, to achieve an expected return in excess of the cost of the investent L. Epowerent also rises with, reflecting a greater return on investent (reduced diinishing returns to investent). Interestingly, epowerent is not affected by the interest rate; however, a higher interest rate iplies a greater outreach (see). That the level of epowerent is not affected by the interest rate can seen by siplifying the NGO s proble as follows: βe Max, r L + e βe t (7) e + [( r ) L ] (3) Siplifying further by dividing the top and botto of (3) by e, we have 7

8 Max β [ r ) L ] e, r Le + e β ( + (7) (4) Where r is given by (0). It is clear that axiizing (4) involves iniizing its denoinator, so that only L and affect e, contrary to what one ight expect 9. Proposition. Consider an increase in the level of effectiveness of the legal syste in punishing (unjustifiable) voluntary default (higher P). Then, in the case of an NGO with a long tie horizon, the nuber N of poor people trained increases, but the level of epowerent e provided by the NGO reains unaffected. In the case of an NGO with a short tie horizon, N and e are both unaffected. Proof. That de/dp = 0 is clear fro (). As for N, note fro (5) that a higher P reduces c; fro (0) with φ=, a lower c causes r to rise, which, using (), akes N fall. Intuitively, if the legal syste becoes ore effective, as reflected in a higher penalty charge P for those who voluntarily default, the banks can afford to lower the collateral they deand fro borrower This eans that the NGO is able to lower the level of wealth it leaves its clients with after they successfully coplete the training prograe; this lower required level of wealth enables the NGO to raise the interest rate it charges to its clients, generating ore resources for the NGO with which to train ore people. 3 The social optiu How do the arket levels of e and N copare with the socially optial levels? We assue that the social planner's objective is the sae as the NGO's, to facilitate the coparison and avoid obtaining results which depend on discrepancies in preferences 0. This eans that both the NGO and the policy aker take an absolute approach to poverty, i.e. they want to get as any people across the poverty line as possible. The social planner s objective differs fro the NGO s only in that the social planner always has a long tie horizon, whilst the NGO s tie horizon ay be short or long. What is the source of inefficiency in this odel, which iplies that the arket outcoe and the social planner's outcoe ay not coincide? It is that the NGO is unable to extract the full returns fro its training. There are two sources of return which the NGO does not capture. The first is the second period surplus to the entrepreneur in the good state, after repaying the loan and paying interest to the bank. The second is profits of the bank We allow the policy aker to appropriate both surpluses via lup su taxation. In the case of a utilitarian approach, we allow the social planner to appropriate bank profit We assue that the social planner allows the arket to deterine the values of r and c according to equation (5), and will deterine r. However, these variable are ere transfer 9 This results is obtained also if the NGO axiizes the expected utility of trainee However, in this case a 0 payoff in the period bad state is required. If the payoff was positive (negative), epowerent would rise (fall) with the interest rate. 0 Discrepancies in preferences ay be very iportant, but we wish to focus on issues of econoic efficiency. 8

9 payents across agents, and do not affect the total return, which the policy aker appropriates by taxing successful entrepreneurs and banks; there is one exception, however: liquidation costs depend on the collateral and these are not transfer payents, since liquidation costs are erely lost without any agent appropriating the. Taking into account the average transport cost of pnα period borrowers and the liquidation cost of collateral δcl, the budget constraint of 4n the policy aker is α + pn ( R + L ) + qn( R + L ) + ( q) ( R δ c ) = N( L + pl + e) 4n (5) (5) shows incoe on the left hand side and expenditure on the left hand side. Re-arranging, we have N = L + e p q) R + L + qr δc ) + ( q p) L ( α ( q) δc 4n (6) The social planner s proble is Max pn N, e, r (5),(7),(8),(5) (7) The proble siplifies thus: βe Max e, r L + e βe ( q)( R ) + L + qr (7) + ( q p) L α ( q) δc 4n (8) r is given by (0), again, does not depend on e, and is the sae as the NGOs, since we are assuing that the NGO and the social planner share the sae objective The ter in square brackets in () contains but r, but is larger to include taxation receipt Following our discussion of Proposition, the ter in square brackets in the denoinator of () is irrelevant in deterining e. It follows that the socially optial e is the sae as the arket s, i.e. (). N will be bigger, however, reflecting the extra resources which the social planner has at its disposal via taxation. Proposition. The NGO correctly provides epowerent e (training per head), but it underprovides outreach (i.e. it trains too few people, N) relative to the first best social optiu. 9

10 Proof. For e, by solving (8) and coparing with (); for N, by coparing (6) with () for given e. Since the policy aker s r is the sae as the arket s, a policy of capping r or subsidizing it below its arket level would be unwise. This is a policy soeties ipleented in poor countries, for exaple in India under the Integrated Rural Developent Prograe and in the Philippines up to 98 (Arendariz de Aghion and Murdoch, 005). Proposition 3. The interest rate of a icrofinance NGO should not be capped or subsidized. Proof. () applies to both the NGO and the social planner, and the values of R, c and λ are the sae. In assessing the generality of Proposition 3, it is worth reebering our assuption that the NGO s and the policy aker coincide in wanting to axiize the nuber of poor people who (just) cross the poverty line. Up to now, we have not allowed epowerent e to affect any of the second period variable However, q could well increase with epowerent, and α should fall as people becoe better able to copare the services of banks, increasing copetition aong bank Suppose that q stays constant and that α falls with e, so that α = =. Then (8) is p βe odified as follows: βe Max e, r L + + e βe [( q)( R ) + L + qr + ( q p) L ( q) δc] (9) 4n (7) The socially optial value of e is now L + = 4n e (0) Copared to e of the NGO (see (), the social planner s e is greater. This is because the social planner tries to achieve a return to training per head in excess not only of the cost of the loan, but of borrowers transport costs in the banking sector as well. The NGO has no incentive to do so because it cannot appropriate via taxation the extra returns to the banks and banks clien Note that an increase in the nuber of banks reduces transport costs and hence the degree of underprovision of epoweren Finally, consider a policy aker who, instead of being concerned exclusively with the welfare of the poor, is concerned with axiizing the social surplus, that is, the su of welfare of the poor (i.e. the nuber of entrepreneurs who cross the poverty line) and bank profit Bank sector profits nπ are given by inserting the solution values of r and c(see (5)) into the expression for bank profits (4), the forula being 0

11 pqnα( δ ) nπ = () n Thus, the policy aker s proble is qα( δ ) Max pn + N, e, r n (5),(6),(7),(8) () The proble siplifies to: βe Max r L + e [ + qα( δ )] βe [( r ) L ] e, + (7) (3) This proble has exactly the sae structure as the original one of the NGO, (see (3)). Hence N and e will be the sae as the NGO s, leading to the following proposition. Proposition 4. If the social planner ais to axiize the social surplus, eaning the su of welfare of the poor and of bank profits, the NGO correctly provides both epowerent e and outreach N. Proof. Fro recognizing that (3) has the sae structure as (3), so that the solution N and e are the sae. 4 NGO-bank ergers In Latin Aerica there is a trend at present for coercial banks to expand their operations by offering credit to very poor people, this being the niche arket of icrofinance NGO s (see Chowdri (004)). This ove can take one of two fors: (a) Either a bank relies on the assistance of foreign advisors like ACCION to obtain knowledge about the new arket they wish to enter, or it acquires a financial organization serving the poorest segent (not necessarily an NGO). A bank s otivation is usually profits; however, considering that the ove would be towards a arket segent that is not obviously profitable (Arendariz de Aghion and Murdoch (005)), there ay be other otivating factors at play of a social justice nature. Our odel allows us to consider the second for of expansion into the poorer segent of the arket, that is, by acquisition of the NGO. For this purpose, we consider a erge between one bank in the Salop circle and the NGO. The new group ay either of two objectives indeed in practise it is likely to have a hybrid of these two objective The first is to axiize Such ergers/acquisitions are often otivated by issues of inforational asyetry: the bank obtains highly trained staff failiar with the bank history of clients, whilst NGO s obtain cheaper access to credi

12 profits; this is likely to be the doinant objective in practise. The second is to axiize the welfare of the poor as the NGO originally intended; this is less likely in practise but of theoretical interes What if the group seeks to axiize profits? Considering the forula for bank sector profits (), the proble the group faces is qα( δ ) Max pn,, N e r n (5),(6),(7),(8) (4) The reader will be now be failiar with the structure of the proble, which has the following solutions: r as in (0), e is in () and N is given by N = L (5) ( + r ) L + e p ) Hence the group s N is the sae as the N before the erger (copare (5) with ()). In the case where the erged entity wants to axiize the welfare of the poor, the group can use profits of the one bank to add to donation The new budget constraint is pnqα( δ ) + pnl ( + r ) + = N( L + n e ) (6) The group s proble is thus Max pn N, e, r (5),(6),(7),(8) (7) The solutions are as follows: r is given by (0), e is given by () and N is given by N = L + e p ( + r ) L qα( δ ) + n (8) Hence, N is greater than before the erger (copare (8) with ()), as there are ore funds available, but is less than the socially optial N (given by (6)), since the group has less funds available than the policy aker.

13 These results lead us the following proposition: Proposition 5. If the erged entity seeks to axiize the welfare of the poor, it succeeds in reducing the underprovision of outreach; if it seeks to axiize profits, the underprovision of outreach is unaffected. Epowerent is always optially provided. Proof. By coparing (), (5) and (7) in the case of outreach; by recognizing that the group s proble has the sae structure as that of the original NGO (see (4)) in the case of epoweren The greater provision of outreach when the group s objective is to axiize the welfare of the poor is due to the fact the group sacrifices profits in order to increase outreach. Corollary. The governent should allow the erger to take place, as it cannot result in a worse arket outcoe, and can result in a better one. Fro the point of view of corporate social responsibility, it is interesting that the negative effect on society of a takeover of a socially responsible corporation (the NGO) by an opportunistic one (a bank) is exactly offset by the positive effect of the externality being internalized. In other words, although the shift in objectives is socially bad, the bank now wants to help the poor in order to raise profits; furtherore, the greater nuber of assisted poor raises profits of all other banks, too. We now conclude and offer soe proising avenues of further research. 5 Conclusion and extensions We constructed a odel where an NGO and coercial banks are vertically related in providing credi To the best of our knowledge, this is unexplored territory in econoic literature. We ade two fundaental assuptions in the odel: that epowerent does not affect second period variables like the probability of project success and borrowers cost of transport, and the NGO and the policy aker want to axiize the nuber of people who cross the poverty line. We showed that epowerent is always optially provided by the NGO, even when the NGO erges with a bank. Outreach is, however, underprovided in alost every case. The exception is given by the case where the social planner seeks to axiize the social surplus inclusive of bank profits, rather than exclusively the well being of the poor, in which case the NGO correctly provides both epowerent and outreach. We also showed that a policy of capping NGO interest rates is unwise. This result is consistent with the conclusion found in ost odern literature. Finally, we showed that the authorities should always allow a erger between the NGO and a bank to take place, as it leaves the underprovision of outreach either unaffected (if the new group axiizes profits) or aeliorated (if the new group axiizes welfare of the poor). This result ay help to infor policy akers, especially in Latin Aerica where the downscaling of coercial banks into the poorer segent of the arket is a trend. In general, we hope that these results will help a policy aker to decide if and how to subsidize either epowerent or outreach. Specifically, and not disregarding the liitations of the odel, if a governent intervenes in icrofinance arkets with the ai of reducing the nuber of poor people as uch as possible and is not concerned with the profits of banks, epowerent should never be subsidized; outreach, on the other hand, should always be subsidized. 3

14 There are a nuber of possible lines of future research: Firstly, it would be interesting to allow for a group of NGO clients who, having succeeded in the training prograe, are satisfied with the wealth they have and do not wish to take out a bank loan. This iplies that we would have (at least) two groups of NGO clients with different preferences, which we feel is a realistic extension of the odel. Secondly, we could allow for the realistic possibility that the NGO ay wish keep lending to clients who, although they successfully passes the training prograe, for soe reason do not anage to get a bank loan, perhaps because banks are iperfect at assessing clients capabilities, or perhaps because banks require that clients ake higher profits than the NGO require Thirdly, we wish to study policy, in particular the provision of cheap credit by the central bank to coercial banks with a view to helping poor people via NGO's. Such a policy has been tried both in the Philippines and in Thailand (see the chapter on credit in Ray (997)), with a variant: it was aied priarily at aking credit ore accessible to inforal oneylenders, who do uch ore lending than NGO' Nevertheless, NGO's would be affected by such a policy, and it would be interesting to study the consequences of such a policy for the provision of training by the NGO, in particular for the trade-off between the nuber of poor people who receive assistance and the level of epoweren Fourthly, it would be useful to explore to what the extent our results are robust to the use of other kinds of social welfare functions, and concave utility. If the NGO (and the policy aker) wish to axiize the expected utility of the poor, rather than the nuber of poor who get across the poverty line, the assuption that unsuccessful NGO clients obtain a return of zero in the first period will becoe significan Fifthly, the solution for the level of epowerent, (see ()) suggest that the size of the loan is iportan It therefore akes sense to consider the consequences of endogenizing the loan size. With a continuu of returns at the end of the first period, a higher first period return could be associated with a higher probability of project success, so that rich bank clients would get cheaper bank loans than poor clients, as observed in practise. Finally, it would be interesting to see if an internally optial nuber of banks fro society s point of view is ever possible. This would be due to notion that, as the nuber of banks rises, there is a the trade-off between a lower bank interest rate and collateral, with increase the welfare of the poor, and lower banks profits, which reduce welfare of the poor where bank profits are taxed and used by the policy aker to finance training and loan provision for the poor. Acknowledgeents I thank Roan Matousek and participants at the seinar series for econoists at the Central Bank of Mexico for their advice. References Arendariz de Aghion, B., and Murdoch, Jonatahan, 005, The Econoics of Microfinance, MIT Press 4

15 Chowrdi, S.H., 004. Downscaling Institutions and Copetitive Microfinance Markets: Reflections and Case-Studies fro Latin Aerica. Coissioned by CALMEADOW. Conning, J, 999, Outreach, Sustainability and Leverage in Monitored and Peer-Monitored lending, Journal of Developent Econoics, Vol. 60, 5-77 De Wit, B., and Meyer, R., 004. Strategy: Process, Content, Contex Thoson (chapter ) Freixas, Xavier and Rochet, Jean-Charles, 997, Microeconoics of Banking, MIT Press Ledgerwood, Joanna, 999, Microfinance Handbook, The World Bank McIntosh, C. and Wydick, B., 005, Copetition and Microfinance, Journal of Developent Econoics, Vol. 78, 7-98 Murdoch, Jonathan, 999, The Microfinance Proise, Journal of Econoic Literature, Vol. 37, Deceber, pp Ray, Debraj, 998, Developent Econoics, Princeton University Press United Nations Capital Developent Fund, 005, Banking the "Missing Middle" in India: Syposiu Focuses on Challenges in Involving Coercial Banks in Microfinance, October, available at National Bank for Agriculture and Rural Developent, 007, Addressing the Missing Middle: Institutional Fraework for Graduating Microfinance Clients, by P.Satish, Chief General Manager, available at NABARD.ppt 5