When is foreign aid policy credible? Aid dependence and conditionality

Journal of Development Economics Vol. 61 Ž2000. 61–84 www.elsevier.comrlocatereconbase

When is foreign aid policy credible? Aid dependence and conditionality Jakob Svensson

)

DeÕelopment Research Group, DECRG-MG, The World Bank, 1818 H Street N.W., Washington, DC 20433, USA Received 1 May 1998; accepted 1 March 1999

Abstract Disbursements of foreign aid are guided Žin part. by the needs of the poor. Anticipating this, recipients have little incentive to improve the welfare of the poor. In principle, conditionality could partly solve the problem, but this requires a strong commitment ability by the donor. Without such a commitment technology, aid will be allocated Žpartly. to those in most need, and the recipient governments will exert low effort in alleviating poverty. Contrary to conventional wisdom in the aid literature, we show that tied project aid and delegation of part of the aid budget to an Žinternational. agency with less aversion to poverty improve welfare of the poor in the recipient countries. q 2000 Elsevier Science B.V. All rights reserved. JEL classification: E61; F35 Keywords: Aid policy; Credibility; Policy design

1. Introduction

Over the past few years Kenya has performed a curious mating ritual with its aid donors. The steps are: one, Kenya wins its yearly pledges of foreign aid. Two, the government begins to misbehave, backtracking on economic reform )

Tel.: q1-202-473-9109; fax: q1-202-522-3518; e-mail: [email protected]

0304-3878r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 3 8 7 8 Ž 9 9 . 0 0 0 6 1 - 9

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

62

and behaving in an authoritarian manner. Three, a new meeting of donor countries looms with exasperated foreign governments preparing their sharp rebukes. Four, Kenya pulls a placatory rabbit out of the hat. Five, the donors are mollified and the aid is pledged. The whole dance then starts again. ŽThe Economist, 1995. When is foreign aid policy credible? The quotation above on the relationship between the Kenyan government and the donor community suggests that this is indeed a relevant question. Still, the voluminous literature on foreign aid and development has not dealt with incentive problems in the donor–recipient relationship in general, and credibility and institutional design issues in foreign aid policy in particular. 1 This paper attempts to close this gap. In many developing countries, foreign assistance is an important source of revenue. As an example, for the 50 most aid-dependent countries the mean value of aid as share of central government expenditures for the period 1975–1995 was 53.8% ŽWorld Bank, 1998.. What has this vast amount of foreign aid achieved? To the extent that aid is justified by the benefits to the recipient, rather than to the donor, it might be reasonably judged on two criteria: growth and poverty-alleviating public expenditures. Taking this as a starting point, Boone Ž1995; 1996. concludes that aid primarily goes to consumption and that there is no relationship between aid and growth, nor does it benefit the poor as measured by improvements in human development indicators. The poor macroeconomic impact of aid raises questions of the efficiency of foreign aid and aid policy. Even though the state of the art is somewhat unclear, some general conclusions have emerged in the literature. First, it is argued that the weak macro performance of aid in many developing countries is largely due to unsound domestic policies in the recipient countries. Second, conditionality is a way to deal with such macroeconomic mismanagement Žsee, e.g., Summers and Pritchett, 1993 and various World Bank publications.. Third, the most efficient way to give aid is through untied program support Žsee, e.g., Cassen et al., 1986; Krueger et al., 1989.. Hence, tying aid to a specific source in the donor country is bad for the poor in the recipient country, and is viewed only as a method to increase the commercial impact of the aid program. An additional argument commonly voiced by parts of the NGO community is that the donors, and in particular the World Bank and the IMF, pay too little 1

Exceptions are Pedersen Ž1993; 1996. and Mosley et al. Ž1995.. Mosley et al. Ž1995. study conditionality as a bargaining game between the recipient and the donor. Pedersen Ž1996. analyzes a two-period game where the donor gives aid in period 1 to boost investment. He shows that the effect of aid on investment critically depends on the timing of the game Ži.e., if the donor acts as a Stackelberg leader or follower.. Drazen Ž1999. provides an excellent survey of the political economy literature on foreign aid.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

63

attention to poverty-reduction Žsee, e.g., Havnevik, 1987 and the discussion in Summers and Pritchett, 1993.. According to the critics, this explains the low poverty-reducing impact of aid. In this paper, we argue that once we start looking at the incentive effects of aid some of these conclusions need to be rethought. Our model suggests that one reason for the poor aggregate record of past aid disbursements is a moral hazard problem adversely affecting the aid recipients’ incentives to undertake structural reforms. In principle, conditionality could partly solve the problem, but this requires a strong commitment ability by the donor. Without such a commitment technology aid disbursements are Žpartly. guided by the needs of the poor, resulting in low effort on the part of the recipient governments to alleviate poverty. Thus, conditionality is not necessarily the method to deal with macroeconomic mismanagement. Lacking a commitment technology, it is shown that tied project aid as well as delegation of part of the aid budget to an Žinternational. agency with less aversion to poverty, will improve welfare of the poor. Thus, given that time-inconsistency problems are important, well-executed tied aid may in fact be an efficient way to give aid. Moreover, letting a donor with less aversion against poverty be in charged of the disbursement decision may be the best way to assist the poor. This paper is related to the literature on Samaritan’s dilemma and soft-budget constraints. 2 By studying the problem within a simple principal–agent model, our analysis should be regarded as complementing this work. One of the contributions of the paper is to integrate these conceptual frameworks with the policy discussion on foreign aid. This provides a framework to formalize the idea of conditionality. Moreover, by analyzing a setup with several recipients, a coordination problem in foreign aid policy is identified: strategic manipulations by each recipient in order to increase the share of the aid disbursed lead to inefficiently low exertion of effort. The second contribution is normative. We describe two institutional arrangements that mitigate the time-inconsistency problem in foreign aid policy. The first one; delegation of part of the aid budget to a donor with less aversion to poverty, bears resemblance to a well-known result in the monetary policy literature, namely, that delegation can relax binding incentive constraints in situations of strategic interaction. 3 We show that a similar argument applies to foreign aid policy. The second institutional arrangement; tied project aid, is novel. By introducing a third party, i.e., private firms, into the game between the donor and the recipients, a conflict of interest between the beneficiaries of aid is created. This

2

See Kornai Ž1980a; b. for discussions and Dewatripoint and Maskin Ž1991. and Qian and Roland Ž1994. for models of the soft-budget constraint. The Samaritan’s dilemma is formalized by, among others, Lindbeck and Weibull Ž1988.. See, for example, Coate Ž1995. for references. 3 See, e.g., Persson and Tabellini Ž1997. for references.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

64

in turn constrains the donor’s ex post incentives, thereby providing the necessary incentives for the recipient governments to induce effort. We assume that the donors care about the welfare of the poor. In reality, of course, the donor’s preferences are likely to be a complex function of both altruistic and self-interest motives. In fact, the empirical literature on the determinants of aid suggests that aid is driven by both these motives. 4 We also assume that the recipients care about the welfare of the poor, but realistically, that the recipient also pursues other goals. The normative arguments we lay out in the paper continue to be valid as long as the recipient government at least partly cares about the needs of the poor. We believe that in most developing countries this presumption is correct. The model’s basic prediction is a two-way relation: foreign aid is Žpartly. disbursed according to the needs of the poor, and the anticipation of this adversely affects the recipients’ incentives to carry out policies that would reduce poverty. This result is consistent with empirical evidence. A recent World Bank study finds that policies Žand welfare. have worsened in most aid-increasing countries, and Burnside and Dollar Ž1997. show that there seems to be no statistical relationship between policy change and aid flows. The remainder of this paper is organized as follows. In Section 2, the model is presented, and, in Section 3, the optimal contracts under commitment are described. In Section 4, we study the equilibrium outcome under discretion. Alternative aid institutions are presented in Section 5. Throughout the paper, the most important results are summarized in propositions, while most of the formal analysis is relegated to Appendix A.

2. A model of aid dependence and conditionality To focus on the time-inconsistency problem in foreign aid policy and ways to mitigate it we will throughout the analysis treat the donor and the recipient governments as single decision units. Hence, problems in aid policy stemming from divided governments are assumed away. 5 Moreover, to simplify the exposition, we consider only two informational settings; one in which reform effort is perfectly observable; the other when only the aggregate outcome is observable. These two setups are not considered to be the best characterizations of reality, but provide a simple yet rigorous structure in which to analyze normative issues in foreign aid policy. 4

See, e.g., Trumball and Wall Ž1994. and Burnside and Dollar Ž1997.. See Cassella and Eichengreen Ž1994. and Svensson Ž1999. for models of decentralized decision making and the effects of foreign assistance. 5

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

65

2.1. The model Consider the following model consisting of one bilateral donor and two recipients. The two recipients are assumed to be identical in all relevant respects ex ante. We assume that the donor and each recipient have agreed, ex ante, on a vector of reform measures to be undertaken. The actual degree of implementation of the reform program, or simply reform effort, however, is one of the recipients’ choice variables. Denote the degree of reform implementation by recipient j; that is, reform effort, by i j , where i j g w0,ı x. The outcome of the reform is uncertain. Specifically, income and consequently public funds, yj , as a result of the reform is yj s

½

g with probability b with probability

q Ž1yq.

where g ) b ) 0. Exerting more effort makes the good state more likely. Hence, we postulate that the probability q is an increasing and concave function q Ž i j ., with q Ž0. G 0, q Ž ı . - 1. Each recipient also chooses a consumption pair Ž c1 j ,d j . g Rq= Rq, where c1 j denotes non-development spending by recipient j, and d j denotes development spending; that is, poverty-alleviating public expenditures. The government runs a balanced budget. Hence, c1 j q d j F yj . The donor has altruistic motives for giving foreign aid. Specifically we assume that aid is used to produce a good or service, denoted by h j , benefiting the poor in country j. h is assumed to be an increasing and concave function hŽ a j ., where a j represents the Žnon-contingent. amount of aid disbursed to country j. 6 The poor derives utility from consuming the good, which is either produced by the donor’s resources according to hŽP., or provided by the recipient government, d. Alternatively, one could think of the government producing the good with a linear technology. Thus, total Žpublic. consumption of the poor in country j is c 2 j s d j q hŽ a j .. The j-th recipient’s additively separable utility function is 2

wj s Ý u Ž c n j . y w i j

Ž 1.

ns1

where w is the marginal cost of reform effort Žor implementation.. To reduce the notational complexity of the problem, we assume that government consumption; 6

The consideration of aid as a factor of production has a long tradition, dating back at least to the study by Chenery and Strout Ž1966.. In this setup, a j can be interpreted either as project aid, or as program aid such as import-support. An alternative interpretation is to think of aid as pure cash transfers, but that due to bureaucratic or institutional factors the recipient has limited absorption capacity. Thus, aid will have a falling marginal product Žcf. Cassen et al., 1986.. See Burnside and Dollar Ž1997. for an empirical justification of this assumption.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

66

that is, non-development spending, is evaluated exactly the same way as consumption by the poor, through the differentiable, increasing and concave function uŽP.. 7 The donor agency is risk-averse with preferences for consumption of the poor in the recipient countries. Its preference function, w D , is the sum of the utility of the poor in the two recipient countries. 2

w D s Ý u Ž c2 j .

Ž 2.

js1

The donor is endowed with an exogenous aid budget A. The amount of resources earmarked for development aid by donor countries is often determined internally in the budget process. Therefore, it is in this context reasonable to assume it to be exogenous. Note, however, that the total amount of aid disbursed to each individual country will be endogenous in the model. 8 The budget constraints for the donor is then simply 2

Ý a j F A.

Ž 3.

js1

We assume that the uncertainties facing the two recipients Žor more exactly the underlying shocks. in the two countries are independently distributed. Thus, there are four possible aggregate states

°Ž b , b . with probability ~Ž b ,g . with probability

ss

Ž g , b . with probability

¢Ž g ,g . with probability

Ž 1 y q Ž i1 . .Ž 1 y q Ž i 2 . . Ž 1 y q Ž i1 . . q Ž i 2 . q Ž i1 . Ž 1 y q Ž i 2 . . q Ž i1 . q Ž i 2 .

where Žg , b . denotes the aggregate state when recipient 1 is in a good state and recipient 2 is in a bad state, and symmetrically for the other three aggregate states. We let S denote the set of all possible aggregate states. It is also convenient to define the subset of symmetric states, Ss , and the subset of asymmetric states, Sa , given by Ss s Žg ,g .,Ž b , b .4 and Sa s Žg , b .,Ž b ,g .4 . We make the following assumption. Assumption 1. a j G 0 for j s 1,2. Thus, aid flows must be non-negative, implying that the donor cannot make wealth transfers between the two countries. 7

Z

For technical reason, we also assume that u ) 0. Implicit in this setup is the assumption that the donor always will disburse all aid. As discussed in Svensson Ž1998., the incentives to always disburse are strong, partly because future allocations depend on the donor’s ability to disburse current funds. 8

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

67

The model defines a two-stage game among the recipients and the donor. The game starts with the recipients non-cooperatively choosing degree of implementation; that is, reform effort, i g w0,ı x. Reform effort by the two recipients influences the aggregate state s. In the second stage of the game, the two recipients and the donor determine the allocation of public funds and foreign aid, respectively. We consider two different policy environments. In the first case, the donor has access to a commitment technology implying that even though foreign aid is disbursed after the aggregate state has been realized the allocation of aid Žconditional on reform effortroutcome. is determined prior to the realization of s. In this setup, the donor and the recipients play a non-cooperative stage game with the donor moving first. In the second policy environment, the timing is changed so that the choice and disbursement of aid are made simultaneously after the state s has been realized. Thus, aid disbursements are made under discretion.

3. Conditional aid when the donor can commit 3.1. Optimal contract when reform effort is obserÕable: first-best Suppose that the donor can commit to a policy rule ex ante and, as a benchmark, that the degree of implementation is fully observable. Thus aid is conditional on both s and i. 9 The optimal contract Žand equilibrium of the game. can be solved by working backwards. Let a j Ž s,i j . denote the contingent amount of aid disbursed to country j. That is, a j s a j Ž s,i j ., j s 1,2, ;s g S, is the policy rule, or contract, that specifies the amount of aid disbursed as a function of aggregate state and reform effort. In the last stage of the game, the donor disperses according to the specified contract, and the two recipients allocate public funds between c1 and d. Solving the problem we can derive two consumption functions; C1 j Ž s,i j . and C2 j Ž s,i j ., where C1 j Ž s,i j .w C2 j Ž s,i j .x denotes government consumption Žtotal consumption by the poor. in country j in aggregate state s, given the aid inflow a j Ž s,i j .. Since government consumption is assumed to be evaluated exactly the same way as consumption by the poor, it follows that Cn j Ž s,i j . ' C j Ž s,i j . s 1r2w yj q hŽ a j Ž s,i j ..x for n s 1,2, j s 1,2. 9 Note that even though the donor’s preferences are defined over the welfare of the poor, rather than reform effort, the contract is conditional on i not the composition of spending. We believe this is both a realistic characterization of conditional aid and provide insightful results on the relationship between reform implementation, conditionality and fungible aid flows. First, the typical reformraid package conditions aid on ‘efficiency-enhancing’ reforms which will indirectly benefit the poor Žsee Branson and Jayarajah, 1995.. Second, this setup allows us to separate the discussion of fungibility and conditionality, a discussion which in practice has been rather confused.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

68

The optimal contract can now be determined by solving the program of maximizing the donor’s expected utility subject to the budget constraint Ž3., the recipients’ individual rationality ŽIR. constraints, and the consumption functions C1 j Ž s,i j . and C2 j Ž s,i j .. Thus, given i, the optimal composition of aid is defined by the program, 2

max

Ý Ý Q Ž s,i . u Ž C j Ž s,i . . a Ž s,i . ,a Ž s,i .4 1

2

Ž 4.

js1sgS

subject to 2

Ý a j Ž s,i . F A,

;s g S

Ž 5.

js1

and 2 Ý Q Ž s,i . u Ž C j Ž s,i . . y w i j G Ewj0 ,

j s 1,2

Ž 6.

sgS

where we have exploited the properties of the symmetric Nash equilibrium in the last stage of the game, and where QŽ s . denotes the probability of aggregate state s g S, a i ŽP. is a vector of state-contingent aid disbursements to country j for all s g S, and Ewj0 is recipient j: s expected utility without aid. Note that Eq. Ž6. is the IR constraint for recipient j. Let superscript fb denote the first-best equilibrium. The important properties of this subgame-perfect equilibrium are summarized below. Proposition 1. The optimal contract when i is contractible (first-best) is characterized by three conditions: (a) the donor fully bears the risks, implying full consumption smoothing across countries in all states; (b) the optimal amount of reform effort is higher than in the equilibrium without aid; (c) the IR constraints bind. Proof. See Section A.1.

B

Since the degree of implementation is contractible it is always optimal to give aid to those in most need, resulting in an equalization of the marginal utilities of aid across countries. Moreover, since the donor does not derive any utility from government consumption, the IR constraints must bind. Thus, the poor skim off the entire surplus from the recipient government, and the recipient government is no better off with aid than without. By giving conditional aid in an environment where the donor can commit and where i is verifiable, the donor in practise buys a certain amount of reform effort in exchange for the aid it disburses. Thus, even though part of the dispersed aid is spent on government consumption which does not benefit the poor and even though aid is fungible, conditional aid is remarkably

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

69

effective by inducing the government to adopt poverty-reducing polices. This result is at the heart of the conditionality argument. 3.2. Optimal contract when adjustment effort is not Õerifiable: second-best In reality, the degree of implementation Žreform effort. is seldom observable in all of its dimension. Below, we consider the case where the recipients’ policies are not verifiable at all. This rather extreme setup provides us with a benchmark that will be useful in the second part of the analysis. When i is not verifiable, the optimal contract can only be made conditional on the observable state s. Thus, a j s a j Ž s ., j s 1,2, ;s g S. We make one additional assumption. Assumption 2. 2 qX Ži. D u y Ž QŽ g,b . qY Ži. DuXrŽ qX Ži.Ž1 y q Ži... - 0. Where D u s uŽŽ1r2.g q Ž1r4. A. y uŽŽ1r2. b q Ž1r4. A. is the utility difference between the two states when aid is split equally across states and recipients; that Ž wrDu. is the degree of reform effort is, a j s Ar2 for all s and j, and i ' qy1 i each recipient would exert if aid followed the simple rule a j s Ar2 for all s and j. Exploiting the symmetry of the model, assumption 2 is the first-order condition of the donor’s maximization program with respect to i j Žsee Section A.2. evaluated at a j s Ar2. Assumption 2 implies that it is never optimal for the donor to write a contract such that the country in a good state receives more aid than the country in a bad state. We believe this is a realistic assumption, and it will hold provided that the utility function uŽP. is sufficiently concave. As in Section 3.1, the equilibrium of the game is found by working backwards. The optimal composition of public funds is identical to that described in Section 3.1, with a j Ž s,i j . replaced by a j Ž s .. The optimal contract is also derived as in Section 3.1, except that there are now two additional constrains, namely the IC constraint, 10 q X Ž i 1 . w V 1 y L1 x s w

Ž 7.

where:

V 1 ' 2 q Ž i 2 . u Ž C1 Ž g ,g . . q 2 Ž 1 y q Ž i 2 . . u Ž C1 Ž g , b . .

Ž 8.

L1 ' 2 q Ž i 2 . u Ž C1 Ž b ,g . . q 2 Ž 1 y q Ž i 2 . . u Ž C1 Ž b , b . .

Ž 9.

and corresponding constraint for recipient 2. The IC constraint Ž7. has an intuitive interpretation. The left-hand side of Eq. Ž7. captures the expected gain of higher reform effort, treating the other recipient’s choice of i as given, while the 10

We have replaced the infinite set of relative incentive constraints with a single ‘‘first-order constraint’’ Žsee Section A.2..

70

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

right-hand side is the marginal cost, w . The expected benefit is the product of the marginal increase in the probability of a good state, times the relative change in utility of such an increase in q. We denote the second-best equilibrium with superscript sb. Solving the maximization program yields Proposition 2. Proposition 2. The optimal contract when i is not contractible (second-best) is characterized by three conditions; (a) risks will be shared between the donor and the recipients, implying less than full consumption smoothing across countries; sb Ž sb Ž fb Ž sb Ž Ž . . . . . asb j s s Ar2 for all s g Ss , and a1 b ,g s a 2 g , b - a1 b ,g , and a1 g , b fb sb Ž s a 2 b ,g . ) a1 Žg , b . ; (b) the optimal amount of reform effort is lower than in the first-best, i sb - i fb ; (c) the recipients will be strictly better off and the donor (and the poor) will be strictly worse off compared to the first-best. Proof. See Section A.2.

B

Again this result is intuitive. The second-best contract is a compromise between giving aid to those in most need and providing optimal incentives. In order to induce the recipient to exert higher effort, aid flows in bad states must be lowered and aid flows in good states raised. Note, that there are two forces that drive the equilibrium away from the first-best. First, there is a ‘moral hazard’ in that full consumption smoothing lowers the incentive to invest or exert effort ex ante. This distortion is reinforced by an adverse competition for aid across recipients. Since the two recipients act non-cooperatively, they do not internalize the effect of domestic policy choices on the equilibrium aid flows. Given that the donor’s resources are limited, the choice of reform effort in country 1 will affect the expected welfare of country 2, and vice versa. The higher effort exerted by recipient 1, the less likely it ends up in the bad state and the less likely it receives as much foreign assistance, implying that expected aid to country 2 rises. Hence, there exists a positive externality between expected aid disbursement to country 2 and reform effort in country 1. The recipients will not internalize this externality when choosing i, resulting in too low reform effort in both countries.

4. Discretion: third-best Contracts of the form described in Sections 3.1 and 3.2 have been suggested to solve the moral hazard problem present in many donor–recipient relations. However, enforcing such contracts are difficult. Ex post, once the recipients’ choices of reform effort are determined and the shock realized, the donor agency has incentives to increase disbursements to the country in most need. The anticipation that this will happen will in turn affect the incentive to carry out politically costly reform policies ex ante.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

71

The main difference from Section 3 is that we relax the assumption that the donor can commit to a policy rule in advance. Or put differently, the cost of deviating from a pre-announced policy is small. Consequently, in each period, the donor and the recipients simultaneously and non-cooperatively choose the allocation of aid and public funds, respectively. We denote the equilibrium adjustment level in the discretionary equilibrium by i d . The important properties of this subgame-perfect equilibrium are stated in the following proposition. Proposition 3. The discretionary equilibrium entails full consumption smoothing d sb Ž . adj Ž s . s a fb j s , for j s 1,2 and s g S, but low reform effort, i - i . Proof. See Section A.3.

B

Once the recipients’ choices of i are made and the shock realized, the donor agency has incentives to deviate from the pre-announced policy rule, or contract, and increase disbursements to the country in most need. The optimal allocation of aid ex post is identical to the first-best allocation, and implies an equalization of the marginal utilities of aid across countries. The anticipation that this will happen will in turn affect the incentive to carry out politically costly reform policies ex ante. In other words, the donor’s incentive to push the outcome towards the first-best will drive the equilibrium towards the third-best. The inability to commit results in the first-best allocation of aid across countries, but low reform effort. The welfare implications could be summarized as follows. The donor and the poor are strictly better off in the second-best equilibrium than in the discretionary environment, while the changes in the recipients’ expected welfare are ambiguous. The reason for this is that the individual government does not internalize the positive externality between i j and expected aid disbursement to the other recipient. As shown in Section A.4, i d is lower than what would be the case if the two recipients cooperated and maximized joint welfare. If the gain of increased aid flows in bad relative good states is outweighed by the loss of too low reform effort, the recipients’ are also better of Žin expected terms. in the second-best equilibrium. Note that each recipient will choose a strictly positive effort level even though lower effort will be Žpartly. compensated by increased aid flows. The reason for this is that due to a fixed aid budget, only Ar2 is disbursed to each individual recipient in the worse state of nature Ž b , b .. In summary, conditionality could partly solve the incentive problem Žinevitably. present in many situations in which the donor and recipients interact, but this requires a strong commitment ability by the donor. Without such a commitment technology, aid will be allocated Žpartly. to those in most need, and the recipient governments will exert low effort in alleviating poverty. In Section 5 we describe two arrangements under which the incentives for ex post recontracting are eliminated, and which are able to sustain second-best outcomes.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

72

5. Alternative aid institutions The question we ask in this section is whether it is possible to design institutions so as to push the discretionary equilibrium closer to the second-best equilibrium. Two different scenarios are considered. In the first case, the set of policy instruments available for the donor is expanded by allowing for tied project aid. In the second scenario, we allow the donor to delegate responsibility to another agency with different preferences over the allocation of foreign assistance. 5.1. Tied aid In this section, we describe an alternative method to deal with the incentive problems in foreign assistance, namely, tied project aid. We define tied aid as project aid contracted by source to private firms in the donor country. Project aid in general, and tied project aid in particular, have received a great deal of attention and criticism in the aid literature. A general conclusion that has emerged from this research is that if aid is highly fungible, targeting assistance to specific projects is essentially a futile exercise. 11 Furthermore, tying aid resources by source and end-use to firms within the donor country is seen only as a way to increase the commercial impact of the aid program. As seen below, these conclusions may in fact be reversed once we take into account the time-inconsistency problem in foreign assistance. The main reason for this result is that tied aid is contractible. That is, contrary to many international agreements where there are no third party or institution that can enforce contracts, tied project aid is contractible within the donor country. Furthermore, such a contract is credible not only because of the use of legal institutions within the donor country, but because the third party involved, i.e., private firms within the donor country, is likely to enforce the contract for profit-maximizing reasons. Hence, by exploiting domestic institutions the donor achieves some commitment power in the international policy game. Or put differently, by introducing a third party into the game between the donor and the recipients, a conflict of interest between the beneficiaries of aid is created. This in turn constrains the donor’s ex post incentives, thereby providing the necessary incentives for the recipient governments to induce effort. To make the analysis more realistic we assume that tied aid is less efficient than non-tied aid. In other words, contractibility acts as a constraint that reduces efficiency. However, it turns out that tied aid can serve a useful role even though it is only an imperfect substitute for non-tied aid.

11

Feyzioglu et al. Ž1996. show that aid in general is highly fungible.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

73

5.1.1. A modified model Let the amount of tied project aid to country j be denoted by t j . A project realized through tied aid takes time to implement. We capture this by assuming that at the start of the game, the donor writes Žbinding. contracts with two domestic firms. Each firm then implements the project; firm 1 in country 1 and firm 2 in country 2. We assume that the contracts cannot be conditional on aggregate state, which seems like a realistic setup. The projects are finalized after the shocks are realized. Non-tied aid, denoted a ntj , is chosen ex post; that is, once the aggregate outcome is known. Tied aid is likely to be less efficient than non-tied aid. This is captured by assuming that a fraction m of resources used for tied project aid is wasted, with m ) 0. When m ) 0, the second-best outcome can of course never be implemented. In this case, the donor faces an additional trade-off; wasting aid resources vs. creating incentives to induce the recipients to choose higher effort. Intuitively, under discretion, effort is too low whereas the allocation of aid is set optimally. At the margin, it is therefore optimal to accept some waste of aid resources in exchange for higher effort ex ante. This intuitive argument is formalized in Proposition 4 and Corollary 1. Let the indirect utility function of the donor be WD Ž m ,m., where m is a vector of exogenous parameters, and denote the donor’s expected welfare under discretion as EWD3 . Proposition 4. If m s 0 the donor can implement the second-best equilibrium by using a combination of tied and non-tied aid. Specifically, t j s tU s a1sb Žg , b ., and a ntj Ž s . s Ž A y 2 tU .r2 for s g Ss , a1nt Ž b ,g . s a2nt Žg , b . s A y 2 tU , a1nt Žg , b . s a 2nt Ž b ,g . s 0, for j s 1.2, results in the second-best leÕel of effort. Proof. If t j s a1sb Žg , b ., j s 1,2, the first-order condition of the donor’s maximization program ex post with respect to a1 , subject to the budget constraint Ý j a j q 2 t j F A, in aggregate state Žg , b . is, uX Ž g q d Ž a1nt Ž g , b . q a1sb Ž g , b . . . hX1 y uX Ž b q h Ž a nt 2 Žg ,b . qa1sb Ž g , b . . . hX2 - 0

Ž 10 .

which is strictly negative. Hence, it is optimal, ex post, to allocate all available aid to country 2. By construction, then, the total level of aid disbursed to country 2 is Ž . Ž . t j q A y 2 t j s asb 2 g , b . By symmetry, the opposite result holds in state b ,g . Ž . As shown in Section 3.2, aid allocations asb j s for s g S, j s 1,2, result in i j s i sb . B

74

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

Corollary 1. There exists a threshold Õalue m ˆ ) 0 such that for any m g w0, mˆ . the donor is strictly better off (compared to the third-best) using a combination of tied and non-tied aid. Proof. Both the budget constraint and the donor’s utility function are continuous. Hence, by Berge’s Maximum Theorem the value function W is continuous in all arguments. As shown in Section A.2, EWD2 ) EWD3 , i.e., there is a discrete jump in welfare between the second- and third-best outcomes. From Proposition 4, we know that W D Ž0,m. s EWD2 . Corollary 1 is then immediate from the continuous property of W. B When m ) 0, the optimal composition between non-tied and tied aid can be found in two steps. 12 First, let the aggregate net level of aid disbursed be denoted by An , where An ' Ž A y 2 m t .. Second, let the equilibrium pair of net aid flows that implements a given reform effort level ˆı ) i d as a function of m , be denoted by  a1Ž s; m ., a 2 Ž s; m .4 for all s g S, where Ý 2js 1 a j Ž s; m . F A n . Since  a1Ž s; m .,a2 Ž s; m .4 is the optimal aid-pair given An , it must be the case that  a1Ž s; m .,a2 Ž s; m .4 is the solution to the second-best problem Ždescribed in Section 3.2. of maximizing expected utility Ž4., subject to the budget constraint Ž5., the IR constraint Ž6., and the IC constraint Ž7., with a j Ž s . replaced by a j Ž s; m ., j s 1,2, s g S, and with A replaced by the net-budget An . Third, we know that in order for the donor to induce the recipients to choose i j s ˆı ) i d , the allocation of aid must be such that in equilibrium the marginal utilities of aid in the asymmetric states differ. To implement such an aid program, a sufficient share of the aid budget must be ‘tied’ to guarantee that ex post, the donor cannot equalize the marginal utilities of aid. This implies that, given t j , ex post all available funds will be disbursed to the country where the marginal utility of aid is the highest. Thus, if tU denotes the optimal level of tied aid, the optimal composition of non-tied is a1nt Ž g , b . s a2nt Ž b ,g . s 0 a1nt Ž b ,g . s a2nt Ž g , b . s A y 2 tU a1nt Ž s . s

1 2

Ž 11 .

Ž A y 2 tU . , ;s g Ss .

Fourth, since all aid disbursed to recipient 1 in aggregate state Žg , b . is in the form of tied aid, and since the implementation of ˆı requires a net flow of aid to 12 An equivalent way to solve the problem is through backward induction. However, due to nested implicit functions, the evaluation of how the endogenous variables change with m is simplified by exploiting the results of previous sections.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

75

recipient 1 of a1ŽŽg , b .; m . in state Žg , b . Žand symmetrically for recipient 2 in state Ž b ,g .., it follows that the equilibrium level of tied aid must be t Ž m . s a1 Ž Ž g , b . ; m . r Ž 1 y m . .

Ž 12 .

For a given m , Eq. Ž12. gives a mapping from the space of possible t into itself: a given t implies a total net level of aid disbursed from the following budget constraint; Ž13.

and vectors of net aid flows wa 1Ž s; m .,a 2 Ž s; m .x from the second-best maximization program, which in turn implies a new level of tied aid from Eq. Ž12.. The fixed point of this mapping is tU . The following proposition states the important result of this section. Proposition 5. If the amount of tied aid is giÕen by tU and the amount of non-tied aid is giÕen by Eq. (11), the donor minimizes the waste of aid resources while at the same time creating incentiÕes to induce the recipients to choose a giÕen effort leÕel ˆı ) i d . Proof. Note first that ex post t is fixed, i.e., neither firm wants to reduce their project size. Second, since it is always optimal ex post to allocate all available aid funds to the country in most need, the total net aid allocated to country 2 w1x in state Žg , b . wŽ b ,g .x is

Ž 1 y m . tU q A y 2 tU s a2 Ž Ž g , b . ; m . .

Ž 14 .

Hence, the net allocation will be given by wa 1Ž s; m .,a 2 Ž s; m .x and the IC constraint Ž7., with a j Ž s . replaced by a j Ž s; m . binds. Suppose that t - tU . This would increase the amount of resources available ex post Ž A y 2 t ., and would reduce the expected difference in the marginal utilities of aid Žsince more resources would be channelled to the country in most need.. As a result, the recipients would exert too low reform effort ex ante. Thus, t - tU cannot be optimal. Suppose instead that t ) tU . This would result in increased waste Ž m t ., and too little aid would be given to the country in most need. Both effects unambiguously reduce the donor’s welfare, so this cannot be optimal either. B By credibly tying up part of the aid budget, the donor ties its own hands. As a result, it is no longer possible to equalize the marginal utilities of aid across countries ex post. Therefore, the necessary incentives are created to induce the recipients to exert higher effort. Note that it is not optimal to tie up all aid resources, tU - Ar2. Thus, there exists a trade-off between flexibility and credibility. Due to the uncertain environ-

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

76

ment, it is optimal to provide more aid to countries in bad states. These resources would not be available if all aid was tied. Corollary 2. An increase in m leads to a reduction in tU and a reduction in net aid disbursed. Proof. See Section A.5.

B

When m increases, more aid will be waisted. A smaller net aid budget leads to an increased difference in utility in the symmetric states; that is, uŽ C j Žg ,g .. y uŽ C j Ž b , b .., and consequently higher incentives to reform. The donor can therefore reduce the amount of tied aid and still provide necessary incentives for the recipient to exert reform effort ˆı. 5.2. Delegation A well-known result from the political economy literature on monetary and fiscal policy is that delegation to an agent with different objectives may help relax binding incentive constraints. In this section, we show that a similar result applies to the time-inconsistency problem in foreign aid policy. In real life, there are many donors interacting on the aid-scene. Of these the World Bank and the IMF have come to play an important role. The policies of these institutions are often criticized for being to too conservative and inflexible, pursuing policies that may increase efficiency but at the cost of increased poverty, i.e., cuts in poverty-reducing public spending. Many critics argue that bilateral donors should distance themselves Žand withdraw financial support. from the policies of these institutions, and instead emphasize poverty alleviation more strongly Žsee, e.g., Havnevik, 1987 and the discussion in Summers and Pritchett, 1993.. We show below, however, that this claim should in fact be reversed. A donor with less aversion to poverty, or stronger emphasis on aggregate efficiency, will increase welfare of the poor. To simplify, we assume that the donor’s relative poverty aversion, denoted by h , is constant. Hence,

hs

yuY Ž c n j . c n j uX Ž c n j .

.

Ž 15 .

Given this assumption, we can exploit the first-order condition Ž32. in Appendix A to determine the relative poverty aversion for the donor considered above as

hˆ s

hX Ž a1fb Ž g , b . . uX Ž C2fb Ž g , b . . hX Ž afb 2 Žg ,b . .

.

Ž 16 .

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

77

Notice that the donor’s welfare function becomes more Rawlsian as h increases. In other words, a higher h implies a higher aversion to relative poverty or less emphasis on aggregate consumption or efficiency Žsee Behrman and Sah, 1984.. Evaluating Eq. Ž16., it is obvious that a lower h leads to a shift in aid flows away from the recipient country with the lowest income Žrevenue.. We can now state the main result of this section. Proposition 6. The second-best outcome can be implemented by delegating responsibility to a donor agency with less relatiÕe aÕersion to poÕerty. Proof. It is immediate from Eq. Ž16. that a donor agency with relative aversion to poverty given by hU , where

hU s

hX Ž a1sb Ž g , b . . uX Ž C2sb Ž g , b . . hX Ž asb 2 Žg ,b . .

Ž 17 .

will implement the second-best. Since the numerator is smaller and the denominator is larger in Eq. Ž17. than Eq. Ž16., hU - hˆ . B

Due to the donor’s inability to commit, it is optimal for a bilateral donor to delegate responsibility to a donor agency with less relative aversion to poverty. The second-best entails giving less aid for social spending to those countries in most need in order to induce more effort ex ante. A donor with less aversion to poverty will do just that, and the recipients will react by exerting higher reform effort. In equilibrium, the poor will be strictly better off Žin expected terms. than if the donor would disburse aid according to the discretionary equilibrium described in Section 4. Thus, delegation to a donor agency agent with less relative aversion to poverty results in higher expected consumption of the poor. Note that in the case considered above, the execution of the whole aid program is delegated to an agency with h s hU . However, the second-best could also be implemented by delegating part of the aid budget to an agency with h - hU and disburse the remaining aid according to the first-order condition Ž32.. 5.3. Discussion Since the mid-1980s, the IFS institutions ŽWorld Bank, IMF. have come to play a much more important role in foreign aid policy. A larger share of bilateral donor aid is now channelled directly Žor indirectly. through these organizations. In this paper we have provided a rationale for this form of delegation. Many aid advocates have questioned this shift in policies and called for a more poverty oriented approach, i.e., for donors to be more responsive to the short-term needs of

78

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

the poor. However, if the policy game described above is important in reality, a stronger emphasis on aggregate efficiency relative poverty alleviation will on average increase welfare of the poor. It is important to note that this result is not conditional on the World Bank Žor other agencies. having access to a commitment technology, which does not seem to be the case ŽMosley et al., 1995; Svensson, 1998., but that the agency’s discretionary disbursement of aid ‘favors’ relatively better performers. We have also identified a potentially positive role of tied aid. The result in Section 5.1 is conditional on the efficiency loss of tied aid Ž m . not being too large. In most donor countries the composition and nature of tied aid is strongly influenced by special interests ŽKrueger et al., 1989., which results in high efficiency losses. If the donor agencies instead could exploit the inherit commitment technology in tied aid, the loss in efficiency could be reduced substantially, making tied aid a potential mechanism to overcome time-inconsistency problems in foreign aid policy. 6. Conclusion The present model has abstracted from a number of issues influencing the game between the donor and the recipient. The analysis may therefore be biased and it would be inappropriate to draw definite conclusions, let alone to make final policy recommendations. Nevertheless, some important insights emerge from the analysis. First, it is shown that one reason for the poor aggregate record of foreign aid may be a moral hazard problem that adversely affects the aid recipients’ incentives to undertake structural reforms. In principle, conditionality could partly solve the problem, but this requires a strong commitment ability by the altruistic donor. Contrary to conventional wisdom in the aid literature, we show that without such a commitment technology, delegation of part of the aid budget to an international agency with less aversion to poverty as well as tied project aid may improve welfare for all parties. Acknowledgements I am grateful for useful comments and suggestions by Patrick Kehoe, Peter Norman, Lars Persson, Torsten Persson, Gerard Roland, Peter Svedberg, and ´ participants in seminars at IIES, Stockholm University, Development Economics Conference in Bergen, ECARE, University of Rochester, University of Pennsylvania, Universitat Pompeu Fabra, and The World Bank. I also thank two anonymous referees for constructive comments. The findings, interpretations and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries that they represent.

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

79

Appendix A A.1. Optimal contract when adjustment effort is Õerifiable The first-order conditions with respect to a1Ž s,i . are Q Ž s,i . uX Ž C1 Ž s,i . . hX Ž a1 Ž s,i . . yuX Ž C2 Ž s,i . . hX Ž a2 Ž s,i . . q 2 Q Ž s,i . l1 uX Ž C1 Ž s,i . . hX Ž a1 Ž s,i . . yl2 uX Ž C2 Ž s,i . . hX Ž a2 Ž s,i . . s 0,

for all s g S

Ž 18 .

where l j is the Lagrange multiplier associated with the IR constraint. By symmetry, l j s l in equilibrium. By concavity of uŽP. it is the only solution. Thus, the first-best equilibrium entails full consumption smoothing across countries, independent of q and w . Denote the equilibrium without aid with superscript 0. Since the donor does not derive any utility from government consumption, the IR constraints bind in equilibrium. Hence, the optimal degree of implementation Žreform effort. is implicitly defined by the IR constraint Ž6.. That is q Ž i 1 . V 1fb q Ž 1 y q Ž i 1 . . L1fb y 2 q Ž i 0 . u

1

g q Ž1 y q Ž i 0 . . u

ž / 2

s w Ž i1 y i 0 .

1

ž / 2

b

Ž 19 .

where

V 1f b ' 2 q Ž i 2 . u Ž C1f b Ž g ,g . . q 2 Ž 1 y q Ž i 2 . . u Ž C1f b Ž g , b . .

Ž 20 .

L1f b ' 2 q Ž i 2 . u Ž C1f b Ž b ,g . . q 2 Ž 1 y q Ž i 2 . . u Ž C1f b Ž b , b . . .

Ž 21 .

0

V 1fb ) uŽŽ1r2.g . and L1fb ) uŽŽ1r2. b . positive. So i 1 / i 0 . Because the left-hand

Suppose that i 1 s i . Then since the left-hand side of Eq. Ž19. is strictly side of Eq. Ž19. is an increasing and concave function of i 1 , we know there, at the most, exists two effort levels, i 1 ) i 0 and i 1 - i 0 , such that Eq. Ž19. holds. Note that the donor’s welfare is strictly increasing i 1. Hence, i 1 ) i 0 . Because of symmetry and concavity of q, i 1 s i 2 s i fb . A.2. Optimal contract when adjustment effort is not Õerifiable Given iU the optimal composition of aid is found by maximizing Eq. Ž3. s.t. Eqs. Ž4. and Ž5., and the IC constraints 2 Ý QU Ž s . u Ž C j Ž s . . y w iU G 2 Ý Q Ž s . u Ž C j Ž s . . y w i j , sgS

; i j g w 0,ı x

sgS

Ž 22 .

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

80

where QU Ž s . denotes the probability of aggregate state s, given adjustment level iU . Since q ŽP. is concave and differentiable, and the total spending scheme w yj q hŽ a j .x is non-decreasing in i j , the indirect utility function Wj Ž i j . is concave and differentiable. The infinite set of relative incentive constraints for recipient j can therefore be replaced with a single ‘‘first-order constraint’’ Žsee Laffont, 1989 for the validity of the first-order approach., qX Ž iU . V j y L j s w ,

j s 1,2

Ž 23 .

where V j and L j are defined in Section 3.2. The Lagrangian for this problem is 2

max L Ž a 1 , l ,n . s Ý

Ý QU Ž s . u Ž C j Ž s . .

js1sgS

q l1 2 Ý QU Ž s . u Ž C1 Ž s . . y w iU sgS

q l2 2 Ý QU Ž s . u Ž C2 Ž s . . y w iU sgS

q n 1 q X Ž i U . Ž V 1 y L1 . q n 2 q X Ž i U . Ž V 2 y L 2 . .

Ž 24 . The constraints are given by Eqs. Ž6. and Ž23., and the inequality constraints l j G 0, n j G 0, j s 1,2. The four complementary slackness conditions are

l j 2 Ý QU Ž s . u Ž C j Ž s . . y w iU y Ewj0 s 0,

j s 1,2

Ž 25 .

sgS

n j qX Ž iU . Ž V j y L j . y w s 0,

j s 1,2.

Ž 26 .

By symmetry, l j s l, n j s n and C j Žg , b . s Ck Ž b ,g ., for j s 1,2, k s 1,2, j / k. By concavity of uŽP. it is the only solution. Exploiting the symmetry of the model to simplify, the first-order conditions of the program can be written as

Ž 1 q 2 l . QU Ž s . q 2 n pˆ uX Ž C1 Ž s . . hX Ž a1 Ž s . . y uX Ž C2 Ž s . . hX Ž a2 Ž s . . s 0,

for all s g Ss

Ž 27 .

and

Ž 1 q 2 l . QU Ž s . uX Ž C1 Ž s . . hX Ž a1 Ž s . . yuX Ž C2 Ž s . . hX Ž a 2 Ž s . . q 2 np qX Ž iU . pu ˇ X Ž C1 Ž s . . hX1 Ž a1 Ž s . . q Ž 1 y pˇ . uX Ž C1 Ž s . . hX2 Ž a2 Ž s . . s 0,

for all s g Sa .

Ž 28 .

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

81

In a symmetric equilibrium, l j s l and n j s n . The variables pˆ and pˇ are defined as pˆ s

pˇ s

½ ½

qX Ž iU . q Ž iU . yq Ž iU . Ž 1 y q Ž iU . . X

q Ž iU . U

Ž1yqŽ i . .

in state Ž g ,g . in state Ž b , b .

in state Ž b ,g . in state Ž g , b .

Ž 29 .

and p s 1 in state Žg , b . and p s y1 in state Ž b ,g .. Note that if the IC constraints do not bind, the first-order conditions result in the first-best levels of aid flows. To pin down the optimal effort level, we define a j Ži; s . as the optimal aid flow as a function of i and s, where i s w i 1;i 2 x is a vector of reform effort levels. a j Ži; s . is implicitly defined by the first-order conditions Ž27. – Ž28.. Note that, independent of i, a j Ži; s . s Ž1r2. A for all s g Ss . By differentiating the IC constraint Ž23. and invoking the donor’s budget constraint, we have d d i1

a1 Ž i; Ž g , b . . ' aX1 Ž i; Ž g , b . . s yaX2 Ž i; Ž g , b . . ) 0

Ž 30 .

and symmetrically for s s Ž b ,g .. By substituting a j Ži; s . for a i in Eq. Ž4., and using the budget constraint to substitute for a 2 Ži; s ., expected utility can be expressed as a function of i only; WD Ži.. Maximizing W D with respect to i 1 yields 2

2 qX Ž i 1 .

Ý q Ž i 2 . Ž u Ž C j Ž g ,g . . y u Ž C j Ž b ,g . . . js1

q Ž1 y q Ž i 2 . . Ž u Ž Cj Ž g , b . . y u Ž Cj Ž b , b . . . q

Ý Q Ž s . aX1Ž i; s .

uX Ž C1 Ž s . . hX Ž a1 Ž s . . y uX Ž C2 Ž s . . hX Ž a 2 Ž s . . s 0

sgS a

Ž 31 . and symmetrically for i 2 . Eq. Ž31. compares the marginal gain of increased effort, the first term, with the expected marginal cost, the last term. If the disbursement of aid follows the first-best allocation, the second term in Eq. Ž31. vanishes and the first-order condition Ž31. becomes strictly positive ; i. Hence, when i is not verifiable, it is no longer optimal for the donor to allocate aid so as to smooth public consumption of the poor. From the IC constraints we know that it is possible to implement i j ) i 0j iff a1Žg , b . ) a1Ž b ,g . wand a2 Ž b ,g . ) a 2 Žg , b .x. However, by assumption this is not optimal. Consequently, the recipient governments are strictly better off in the

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

82

second-best equilibrium than in the equilibrium without aid. Finally, i sb - i f b , since i f b ) i 0 . A.3. Proof of Proposition 4.1 The discretionary equilibrium is found by backward induction. In the last stage of the game, the donor determines the allocation of aid across the two countries, taking the composition of public funds in the second period as given. The first-order condition is uX Ž C1 Ž s . . hX Ž a1 Ž s . . y uX Ž C2 Ž s . . hX Ž a 2 Ž s . . s 0,

;s g S

Ž 32 .

where c j is replaced with the equilibrium composition of public funds C j Ž s . defined in Section 3.1. The optimal degree of implementation is found by maximizing expected utility w.r.t. i j , taking the equilibrium aid flows implicitly defined by Eq. Ž32. into account. The f.o.c. is qX Ž i j . V jfb y Lfb sw, j

j s 1,2.

Ž 33 .

It is now straightforward to show that i di - i sb . Suppose that i di s i sb , then d sb d sb w V jf b y Lfj b x - w V jsb y Lsb x j , implying that i i / i . Suppose instead that i i ) i . As the left-hand side of Eq. Ž33. is decreasing in i, this cannot be true either. Hence, i di is unambiguously lower than i sb . Because of symmetry, i 1d s i d2 s i d . A.4. Cooperation among recipients Consider the case where each recipient choose i j to maximize joint welfare, i.e., 2

max Ý

Ý

 i 1 ,i 2 4 js1sgS

Q Ž s . u Ž Cj Ž s . . y w i j .

Ž 34 .

The first-order condition for this problem ŽIC constraint. for recipient 1 is 2 qX Ž i 1 . w V 1 y L1 x q 2 qX Ž i 1 . u Ž C1 Ž b ,g . . y u Ž C1 Ž g , b . . s w

Ž 35 .

where we have exploited the assumption of symmetry. To prove that i j ) i d , we can rewrite Eq. Ž35. as

w y qX Ž i 1 . V 1f b y L1fb s qX Ž i 1 . F1

Ž 36 .

where F1 ) 0 is defined as

F1 ' 2 q Ž i 2 . u Ž C1 Ž g ,g . . y u Ž C1 Ž g , b . . q Ž 1 y q Ž i 2 . . u Ž C1 Ž b ,g . . y u Ž C1 Ž b , b . .

.

Ž 37 .

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

83

Suppose that i 1 s i d . Then the left-hand side of Eq. Ž36. is zero, while qX Ž i 1 .F1 ) 0. Hence, i 1 / i d . Suppose instead that i 1 - i d . As the right-hand side of Eq. Ž36. is falling in i 1 while the left-hand side is increasing, this cannot be true either. Thus, i 1 ) i d . Because of symmetry, i 1 s i 2 . A.5. Proof of Corollary 5.5 Note that Eq. Ž14. can be rewritten as D a Ž g , b ; m . ' a 2 Ž Ž g , b . , m . y a1 Ž Ž g , b . , m . s A y 2 t.

Ž 38 .

For a given i, suppose first that an increase in m leads to an increase in t. From Eq. Ž38. it follows then that D aŽg , b ; m . should fall. Furthermore, since An ' Ž A y 2 m t ., an increase in m and t must lead to a lower net level of aid, An . However, differentiating the IC constraint Ž23. with a j Ž s . replaced by a j Ž s, m ., yields Ždrd An . D aŽg , b ; m . - 0, i.e., a contradiction. Thus, an increase in m cannot lead to an increase in t. Suppose that an increase in m leads to a sufficiently large reduction in t such that An increases. This cannot be true either since from Eq. Ž38. it follows that a smaller t leads to larger D aŽg , b ; m .. However, Ždrd An . D aŽg , b ; m . - 0. Thus, an increase in m must lead to a reduction in t and An .

References Behrman, J.R., Sah, R.K., 1984. What role does equity play in the international distribution of development aid? In: Syrquin, M., Taylor, L., Westphal, L.E. ŽEds.., Economic Structure and Performance. New York, 1984. Boone, P., 1995. The impact of foreign aid on savings and growth. LSE, mimeo. Boone, P., 1996. Politics and the effectiveness of foreign aid. European Economic Review 40, 289–329. Branson, W., Jayarajah, C., 1995. Structural and sectoral adjustment: World Bank experience, 1980–92. Washington DC, OED, The World Bank. Burnside, C., Dollar, D., 1997. Aid, policies and growth. Policy Research Working Paper No. 1777, World Bank. Cassella, A., Eichengreen, B., 1994. Can foreign aid accelerate stabilization? CEPR Discussion Paper No. 961. Cassen, R., et al., 1986. Does Aid Work? Oxford Univ. Press, Oxford. Chenery, H.B., Strout, A.M., 1966. Foreign assistance and economic development. American Economic Review 56, 679–733. Coate, S., 1995. Altruism, the Samaritan’s dilemma, and government transfer policy. American Economic Review 85 Ž1., 46–57. Dewatripoint, M., Maskin, E., 1991. Credit and efficiency in centralized and decentralized economies. Mimeo, Harvard University. Drazen, A., 1999. Political Economy in Macroeconomics, Princeton, NJ: Princeton University Press Žforthcoming..

84

J. SÕenssonr Journal of DeÕelopment Economics 61 (2000) 61–84

Feyzioglu, S., Swaroop, V., Zhu, M., 1996. Foreign aid’s impact on public spending. Policy Research Working Paper No. 1610, World Bank. Havnevik, K. ŽEd.., 1987. The IMF and the World Bank in Africa: Conditionality, Impact and Alternatives. Scandinavian Institute of African Studies, Uppsala. Kornai, J., 1980. The Economics of Shortage. North-Holland, Amsterdam. Kornai, J., 1980b. ‘Hard’ and ‘soft’ budget constraint. Acta Oeconomica 25 Ž3–4., 231–245. Krueger, A.O., Michalopoulos, C., Ruttan, V.W., 1989. Aid and Development. John Hopkins Univ. Press, Baltimore, MD. Laffont, J.-J., 1989. The Economics of Uncertainty and Information. MIT Press, MA. Lindbeck, A., Weibull, J., 1988. Altruism and time consistency: the economics of fait accompli. Journal of Political Economy 96 Ž6., 1165–1182. Mosley, P., Harrigan, J., Toye, J., 1995. Aid and Power, Vol. 1, 2nd edn. Routledge, London. Pedersen, K.R., 1993. On development aid, income distribution, and poverty traps. Discussion Paper 7r93. Norwegian School of Economics and Business Administration. Pedersen, K.R., 1996. Aid, investment and incentives. Scandinavian Journal of Economics 98 Ž3., 423–438. Persson, T., Tabellini, G., 1997. Political economics and macroeconomic policy. In: Taylor, J., Woodford, M. ŽEds.., Handbook of Macroeconomics, forthcoming. Seminar Paper No. 630, IIES, Stockholm University, North-Holland, Amsterdam. Qian, Y., Roland, G., 1994. Regional decentralization and the soft budget constraint: the case of China. CEPR Discussion Paper No. 1013. Summers, L.H., Pritchett, L.H., 1993. The structural-adjustment debate. American Economic Review 83 Ž2., 383–389. Svensson, J., 1998, Reforming donor institutions: aid tournaments. Mimeo, World Bank. Svensson, J., 1999. Foreign aid and rent seeking. Journal of International Economics, forthcoming. The Economist, 19th August, 1995. Trumball, W.N., Wall, H.J., 1994. Estimating aid-allocation criteria with panel data. Economic Journal 104, 876–882. World Bank, 1998. World Development Indicators, Washington DC: The World Bank.