8204
Cash on Delivery Aid:
Incentive Issues in a Multi-Model Aid System
Jessica Brown
Core Essay 3
20 November, 2008
Abstract:
This essay examines the recent proposal of a new aid modality called Cash on Delivery Aid (CODA), specifically asking whether CODA can improve upon the incentive structure of other aid modalities to deliver aid more effectively. I draw upon a body of theory that addresses incentive structures relevant to donor-recipient relationships to take lessons from empirical evidence of other aid modalities. I conclude that while CODA is very promising, its success will depend heavily on the success of other aid modalities, especially general budget support (GBS). I suggest a model combining capped GBS and unlimited CODA.
Word Count: 4,998
INTRODUCTION
The Center for Global Development (CGD) recently proposed Cash on Delivery Aid (CODA), a modalitywhereby donors pay developing countries for the development outcomes they achieve. This is thought to incentivise recipient governments to supply greater effort towards development. In this paper I explore the nature of this incentive. I contrast it with incentives of other aid modalities and analyse it in the broader context of a donor-recipient relationship where each party is also subject toincentives from other relationships of which they are a part. I seek to answer the question whether CODAcan improve upon the incentive structure of other aid modalities to deliver aid more effectively. I present a body of theory that addresses incentive structures relevant to donor-recipient relationships, namely game theory, moral hazard, and principal-agent theory. I draw upon these theories to take lessons from empirical evidence of other aid modalities that provide insight for CODA. I examine different forms ofex ante conditionality, where donors deliver aid with the condition that recipients subsequently adhere to conditions, and ex post conditionality where donors disperse funds after recipients meet conditions. I conclude thatCODA is very promising, butits success will depend heavily on the success of other aid modalities, especially general budget support (GBS). Moreover, CODA cannot simply rely on GBS being delivered separately; Success will depend on a coordinated aid package incorporating the two modalities. I suggest a package combining capped GBS and unlimited CODA.
CASH ON DELIVERY AID
Birdsall and Barter, and Birdsall, Savedoff and Vyborny summarize the CODA model. In a CODAcontract donors and recipients agree on a specific measure of progress such as number of children completing primary school. At a later set date, donors pay recipients an agreed-upon sum for every primary school completer over the number of completers established in a baseline study. CODAis “hands-off”; payments are not tied to any prescribed policies or inputs. The only conditionality is ex postsince countries are required to have achieved specific outcomes. Mutually agreed-upon independent auditors verify results. The CGD specifies that CODAshould complement notreplace current aid modalities, offering a waytoincrease the amount and effectiveness of aid, which donors committed to do in the Paris Declaration on Aid Effectiveness. The CGD foresees several distinct benefits ofCODA. First, focusing on specific outcomes and the transparency of payment conditions promotes accountability of donors and recipients to each other and to their citizens. Donors do not have a valid reason to withhold payments if they can point to specific results achieved from their taxpayers’ funds. Recipientsare accountable to donors because they cannot receive funds until after they achieve results. They are more accountable to their citizens who witness their government’s viable opportunity to earn no-strings-attached funding and commitment to achieve specific, measurable development objectives. Second, being “hands-off”ensures that recipients assume responsibility which strengthens rather than underminesdeveloping country institutions. Third,CODAnecessitates monitoring and evaluating outcomes which strengthens local institutions’ capacity to learn from results. Fourth, CODAcomplements many of the Paris Declaration goalsincluding increased harmonization between donors, reduced administrative burden on recipients, and increasing aid levels.
THEORIES
Several theories examine incentives in the context of collective action problemsand are applicable to aid delivery. I applythree to CODA.
Game Theory
Gibson et al apply game theory to aid effectiveness, starting with the assumption that the donor (“Samaritan”) acts out of concern for the benefit of others (38-39; 89). The authorsuse atwo-party gamemodel with ordinal payoffs, wherein both parties benefit if the Samaritan chooses to help and the recipient chooses to exert high effort, but the recipient benefits even more if they receive the help but extend only low effort:
RecipientHigh Effort / Low Effort
Samaritan / No Help / 2, 2 / 1, 2
Help / 4, 3 / 3, 4
Here donors “are better-off helping no matter what the recipient does” (39). Game theory warns that if a Samaritan-donor’s threat to withdraw aid is not credible, recipients have noincentives to exert high effort. This leads us to question whether the threat to withdraw aid is credible in the CODA model.For CODA to succeed recipients mustbelieve they will receive no payments if they fail to progress, which forces us to consider what happens when recipients fail (a consideration of Renzio and Woods).
If aid delivery is analogous to a repeated game then there is a risk that recipients will become aid-dependent. For example, if a developing country experiences a food shortage, donors are often quick to provide emergency food aid. Recipients know that Samaritan-donors will not stand for large numbers of people starving. Or they may know that self-interested donors have incentives to subsidize their own agricultural sector due to domestic pressures, and food aid is one way to do so. With this relative certainty that donors will bail them out in a crisis, recipients have little incentive to allocate their scarce resources to increasing agricultural productivity. They then allocate them elsewhere, potentially to a non-productive use. Here aid dependence arises due the fungibility of aid in the context of a repeated game. Moreover, dependency breeds further dependency since the lower incentives to invest in increasing productivity means that future opportunities for the recipient to earn revenue from a growing tax base may be lost (Gibson;52, 89).
Presuming donors’ threats to withdraw aid under CODA are credible, incentives to produce certain resultswould likely encourage recipients to invest productively if this is a necessary step towards achieving the results that will reward them with aid. However, the problem of how to provide aid based on need remains; I doubt the United States is interested in paying for results in Finland.
Moral Hazard
Moral hazard is the idea that people (or governments) behavemore riskily when they have insurance. Moral hazard considers situations where two parties make an agreement, yet one party has no incentives to stick to their portion of the agreement. Klein and Harford frame the dependency problem in terms of moral hazard (60). Taking the food aid example, recipient countries engage in riskier activities because they have insurance in the form of emergency assistance from developed countries. Moral hazard is a risk when an insurer has imperfect information about the party they are insuring. Azam and Laffont note “that governments of the South have a level of drive for fighting poverty which is their private information” (27). This can make it difficult for donors to provide recipients insurance because donors are unsure if recipients will behave according to the donors’ interests.The main warning from moral hazard is that when aid is provided with ex ante conditionality, recipients have little incentive to adhere to the conditions.
While some authors, notably those of the UN Millennium Project, advocate aid levels based on the financing gap needed to achieve development, Easterlyshowsthe moral hazard problem of giving aid in this manner. If recipients can expect to receive more aid with a larger gap, they can simply lower taxes, demonstrate greater need, and receive more aid. With imperfect information about recipients, donors cannot be sure that recipients will not take more risks by lowering taxes, knowing that with the insurance provided through foreign aid they will not have to bear the consequences. By paying only after recipients achieve results, CODA renders the information problem obsolete. But again, it fails to address the issue of how to target aid where it is most needed.
Principal-Agent Theory
Inprincipal-agent theoryone party, a “principal”, tries to incentivise another party, or “agent”, to do what the principal wants. Any insurance the principal provides the agent corresponds to diminished incentives for the agent to supply effort and/or avoid risk. In the case of aid delivery we may think of donors as principals and recipients as agents. Principals must adequately compensate agentsto motivate themto participate in a contract. Agents have alternative uses for their resources, so principals must offer agents incentives greater than or equal to those offered via alternative opportunities. Principal-agent theory usually considers agents to be inherently risk-averse. Agents prefer a certain return of X to a gamble where the expected return is X (but could be higher or lower than X).
Principal-agent theory raises several concerns regarding incentives in aid contracts. First:recipients’ opportunity costs. In an environment with other aid agencies (alternative principals that could offer contracts to recipients), the incentive effects of one aid agency may be negated or diluted. Gibson et al provide a case in point, where one recipient claimed that if SIDA were to act on its threat to withdraw funding, they could simply approach the Japanese who had softer standards (75). With modalities using ex post conditionality, different standards on the level of performance rewarded create a “race to the bottom” (119). Thus, principal-agent theory highlights the scope for multilateral/coordinated aid delivery and raises concerns about the implications for CODA’s success should coordination fail.This is a key concern of de Renzio and Woods who argue that, CODA“would most likely be an additional modality that recipient governments have to ‘accommodate’, with potential additional transaction costs, and in competition with new sources of development assistance such as Chinese in Africa” (3).
Second, the trade-off between incentives and insurance is crucial in considering CODA, given that recipients are risk-averse. If CODAis to be implemented alongside existing aid modalities then we cannot assessits potential in isolation. We may hypothesise that combining CODAand other modalities such as general budget support (GBS) may result in the incentives of CODA offsetting the moral hazard problems of providing insurance through GBS. Recipients would still have an incentive to raise taxes, for example, if it would help them to achieve the outcomes that CODA rewards. However, the greater the proportion of aid provided through GBS, the lower the incentive effect of CODA would be. Achieving the appropriate balance of incentives and insurance (the essence of contract theory) is no small feat.
Third, we mustacknowledge that incentives in the donor-recipient relationship may impact incentives in other principal-agent relationships within donor and recipient countries. For example, due to limited funds, developing country governments (principals) are often forced to provide long term employment so that employees (agents) are willing to accept low pay (Gibson, 44). Because governmentscannot offer employees sufficient incentives they must compensate them with increased insurance. Recipients could use increased aid revenue to pay higher salaries, enabling them to incentivise higher performance through less job security. CODA’s hands-off approach does not prescribesuch changes in local incentive structures, but presents an opportunity.Another example:recipients might start a program as Cambodia recently has, experimenting with output-based funds to provide water and sanitation facilities through contracts with private providers (Klein, 97-103; Birdsall and Barder, 10).
Milner applies the principal-agent model to domestic politics in donor countries, showing that when taxpayers (principals) feel unsure about whether the aid agency (agent) is serving their interests indevelopment abroad, the agency can ease the problem by diverting aid through multilateral organizations (2). Voters in donor countries have only biased information about the benefits of their government’s aid agencies (who have incentives to misrepresent their impact) whereas multilateral agencies have fewer incentives to appeal to a particular country’s constituency (32). Milner’s analysis yields two points relevant to CODA. One, CODA presents another and perhaps better way of reassuring voters: literally using their taxes to buy results. Second, if CODAdoes in fact resolve this domestic principal-agent problem the incentives for donors to deliver aid multilaterally may decline. However,the incentive of lower donor costs to obtain information about recipients(Azam, 51-52) is still present.
EVIDENCE
Here I examine the experiences of different aid modalities with incentives in the donor-recipient relationship, and what these experiences suggest for CODA.
Ex Ante Conditionality under Structural Adjustment
The Economist’s oft-quoted piece on Aid to Kenyaquite aptlyportrays theproblem with recipients’ lack of incentivesunder Structural Adjustment (SA), deserving citation here:
“Over the past few years Kenya has performed a curious mating ritual with its aid donors. The steps are the following. One, Kenya wins its yearly pledges of foreign aid. Two, the government begins to misbehave, backtracking on economic reform 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”(The Economist, August 19, 1995).
Svensson’s study indicates this experience may have been common during SA. Examining approximately 200 SA programs, he and his colleagues “find no link between a country’s reform effort, or fulfillment of ‘conditionality’, and the disbursement rate” (383).Throughout the SA period, if donorsever withheld aid it was only temporarily, as occurred with Mozambique, Nicaragua, Tanzania, Vietnam, Zambia and Zimbabwe,and in these cases new loans were provided with essentially the same sets of policy conditions attached(Riddell, 237). In a repeated game where donors failed to penalise recipients, recipients had no incentives to change their behaviour.
Devarajan et al’s study of ten African countries reveals the importance of country ownership of policy reforms, particularly in their analysis of the “two sustained reformers,” Ghana and Uganda (15). Both countries were initially sceptical of market-based reforms. It was their own countries’ failed policies, not conditionality,which eventually led to local political movements for market-based policies (15, 22). In other countries the failure of ex ante conditionality meant that, “large amounts of aid to countries with bad policies sustained those poor policies” (6). And what has all this meant for the poor? Chen and Ravallion’s study reveals in the developing world excluding China the number of people living on less than $1 per day roseslightly from 1984 to 2001 (23). This failure has led Sharpe and many others to endorse the Country-led model. Thisrecent emphasis on country ownership has led to donor-recipient relationships based on Poverty Reduction Strategy Papers (PRSPs) produced by developing country governments (Riddell,239).
Ex Post Conditionality Modalities
Given the failure of ex ante conditionality explained above, many donors are moving towards various forms ofex post conditionality.This includes aid based on policies or outcomes that have already been achieved. I examine three of the most pertinent examples here.
Millennium Challenge Corporation
Over the last two decades, donors have become increasingly selective in their allocation of foreign aid, targeting “poor countries that [already] have reasonably good economic governance” (Dollar and Levin, 2004). More recently, the MCC exemplifies donors’ increasing selectivity in favour of good performers (Herrling and Radelet, 3). The MCC conditions their selection of recipients on the basis of data from third-party sources including Freedom House, the World Bank, WHO, UNESCO, IFC, and the Heritage Foundation (MCC).They rank countries on their commitment to rule justly, invest in their people, and promote economic growth using the following indicators: civil liberties, political rights, voice and accountability, government effectiveness, rule of law, control of corruption, immunization rates, public expenditure on health, girls’ primary education completion rate, public expenditure on primary education, business start-up, inflation, trade policy, regulatory quality, fiscal policy, natural resource management, and land rights/access (MCC).