1 Introduction
When allocating foreign aid, donor countries face a problem of incentivizing recipient countries to invest in bureaucratic quality in an effort to increase the effectiveness of aid spending. To address this problem, a new wave of aid conditionality—political conditionality—has emerged with the intention of incentivizing needed investments and reforms (see Molenaers et al.
2015). However, political conditionality suffers from the same difficulty with credible implementation that has led many observers to claim that aid conditionality has failed (see Collier
1997, Alesina and Weder
2002, Dreher
2009 and Öhler et al.
2012). In particular, conditional aid faces a problem of non-contractibility—measures of good governance are partially subjective, and donor countries often face an ex post incentive to circumvent conditionality. The problem of non-contractibility results in a “Samaritan’s problem” that occurs when the recipient country, knowing it will receive assistance in any case, has no incentive to implement costly reforms (see Mosley et al.
1995 and Pedersen
1996 for a discussion of problems of time-inconsistency in aid spending).
1
Given the difficulty of implementing conditionality in bilateral aid, several papers consider the delegation of the allocation decision to a third party as a solution to the Samaritan’s problem (see Svensson
2000, Hagen
2006 and Annen and Knack
2018; see also Schneider and Slantchev
2013 for an argument that delegation can solve commitment problems in international cooperation). Delegation is a well-known solution to hold-up problems in general (see Aghion and Tirole
1997), and may facilitate commitment to aid conditionality in the case of the Samaritan’s problem. However, this solution depends on the existence of an independent party with verifiable preferences that diverge from the donor country’s in the precise direction that facilitates recipient-country investment. Therefore, delegation to international aid agencies may not always mitigate the Samaritan’s problem—as argued by Easterly (
2003) and Hagen (
2006), aid agencies often focus on ease of disbursement or recipient-country need rather than aid effectiveness.
In practice, donor countries most commonly retain decision rights on aid allocation, even after committing funds to a multilateral organization. A typical example is the European Development Fund, where country representatives from each of the 27 EU-member states jointly decide on the allocation of aid to recipient countries. As illustrated in the following quote, since the donor countries have divergent preferences over where to allocate aid, discussion and bargaining is required to reach a collective decision:
...the European Union (EU) has to decide on the allocation of aid to individual countries....as can be expected, this process is plagued with difficulties as each stakeholder argues in favour of its own criteria for allocation and country-specific interests. (Negre
2013, p.1)
Our paper studies a model of bargaining over aid allocation in multilateral organizations, and illustrates how this bargaining process can lead to a more efficient allocation of aid spending relative to bilateral aid allocation, and subsequently incentivizes recipient countries to invest in measures that increase the effectiveness of aid. Additionally, we analyze how the decision-making rule in the multilateral institution impacts the incentive of recipient countries to invest. We find that multilateral institutions that take decisions via majority rule provide recipient countries with a higher incentive to invest, which is consistent with our novel empirical finding that multilateral organizations that take decisions via majority are more responsive to changes in their recipient countries’ governance policies.
While the first part of our paper takes a positive approach and compares investment decisions under different institutions, we also consider a normative approach and detail which decision rule is optimal from the donor countries’ perspective. Here, we find that since the higher incentive to invest under a majority rule comes at a cost of limiting the total number of projects that are funded, it is not always optimal for multilateral organizations to adopt a majority rule. Instead, our analysis suggests that a majority rule should only be adopted when there are high positive spillovers of recipient countries’ investment on other areas, and when investment has an intermediate level of productivity.
Our model relies on two key assumptions. First, we assume donor countries have individual biases over which country receives aid—to make the Samaritan’s problem as stark as possible, each donor country only values the product of aid spending in one of the recipient countries.
2 There is a large empirical literature assessing country biases in both bilateral and multilateral aid spending (for a survey of this literature see Hoeffler and Outram
2011, and Fuchs et al.
2014), and this literature provides evidence that idiosyncratic political motivations influence the allocation of aid spending, particularly in bilateral aid. Since different donor countries exhibit different political biases, e.g., due to ties to former colonial countries, this is consistent with the notion that donor countries have individual biases. What is more, the scholarly literature that compares bilateral and multilateral aid typically argues that donor-country political interests are less prevalent for multilateral aid, and takes the relative absence of political motives as a reason why multilateral aid is more effective for promoting development (Headey
2008, Milner and Tingley
2013).
3 Our model provides a structured explanation for why multilateral aid is more effective: By bargaining over the allocation of aid spending, donor countries commit to a mechanism for allocating aid that mitigates individual donor-country biases.
Second, we assume that donor countries can commit to allocate aid spending via a bargaining process. In practice, donor countries most often commit funds to a multilateral organization prior to bargaining over the precise allocation of the aid spending to individual projects. For example, the European Development Fund (EDF) is funded on a voluntary basis according to a pre-set formula. The allocation of the EDF budget to projects in individual recipient countries, however, is determined after funds have already been committed to the general budget. Importantly, after the donor-countries’ aid budgets have been allocated to the EDF, the member countries have effectively committed to disburse this aid budget through the EDF—disagreement over where to allocate the joint aid budget does not result in an automatic reimbursement of the EDF’s budget and a reversion to bilateral aid.
4
Our model considers a setting of three donor and three recipient countries. The donor countries each have an aid budget, and commit to implement aid bilaterally or via a multilateral fund with a decision rule of either unanimity or majority.
5 The recipient countries then choose an observable, but non-contractible, level of investment that increases the expected effectiveness of aid spending, after which the donor countries determine the allocation of aid. In this setting, the donor countries would like to incentivize the recipient countries to invest ex ante to increase the productivity of aid spending. However, knowing that the donor countries will allocate all spending to their preferred recipient country ex post under bilateral aid, the recipient countries have no incentive to invest ex ante. Alternatively, if donor countries commit to allocate via a multilateral aid fund, then the donor countries bargain over allocation outcomes ex post à la Nash. In contrast to bilateral aid, Nash Bargaining results in an allocation outcome that reflects aggregate efficiency (that is, efficiency according to the donor countries’ preferences).
Therefore, under multilateral aid, a recipient country with a higher effectiveness will receive a higher level of aid, which in turn induces competition over ex ante investments by recipient countries and mitigates the Samaritan’s problem.
6 That is, it is precisely the process of preference aggregation in the multilateral fund that enables multilateral organizations to better implement aid conditionality—when donor countries with heterogenous preferences pool their resources, competition among recipient countries over aid intensifies since bargaining functions as a commitment to reward recipient countries for higher investments.
Additionally, by explicitly modeling the bargaining process of the multilateral fund, we are able to link the investment outcomes to the decision rule used to allocate funding. That is, our analysis demonstrates that the recipient countries’ incentive to invest in reform is also a function of the decision rule used within the fund. Specifically, we show that majority rule further increases the incentive for the recipient countries to invest, since higher investment increases the probability that a recipient’s project will be selected by the endogenous majority coalition. However, the higher incentive to invest comes at a cost of limiting the total number of projects that are funded, which implies that majority will only outperform unanimity when the productivity of donor-country investment is at an intermediate level.
Empirically, the predictions of our model are in line with the finding that aid allocated through multilateral agencies is more selective than bilateral aid (Eichenauer and Knack
2018), which suggests that multilateral allocation rewards recipient countries that invest in good governance. Our model also provides a clear prediction regarding the impact of decision rules on aid allocation: Majority results in a more selective budget allocation, and hence provides a greater incentive for recipient countries to invest in good governance. While there exists theoretical literature on optimal decision-rules in international organizations (see Harstad
2005, Maggi and Morelli
2006), empirical evidence regarding the relationship between decision rules and outcomes at the international level is lacking. Accordingly, we provide novel empirical evidence on the relationship between voting rules in multilateral organizations and the selectivity of aid allocation. We find that the qualitative prediction of our model is in line with the empirical evidence—multilateral organizations that take decisions via majority adjust aid allocation more in response to changes in their recipient countries’ governance policies, suggesting that recipient countries have a higher incentive to invest in good governance when multilateral funds allocate aid via majority.
Our paper’s main contribution is to the literature on aid conditionality (see Dreher
2009 and Molenaers et al.
2015 for an overview), showing that delegation to a multilateral fund functions as a commitment to implement conditionality due to the endogenous objective that arises when donor countries bargain over aid allocation. In particular, this result shows that the Samaritan’s problem can be mitigated through delegation even in the absence of an independent third party with appropriate preferences. Most closely related to our work is Annen and Knack (
2018), who consider a model of delegation to an institution that exogenously maximizes the joint welfare of member nations, and focus on characterizing the decision of whether to join a fund as a function of the level of donor-country bias. Our model differs from theirs in the sense that we explicitly model decision-making within multilateral funds rather than assuming that the multilateral fund maximizes a certain objective, and are therefore able to characterize recipient-country investment as a function of the decision rules of the multilateral fund.
7
Additionally, our work contributes to the literature on optimal decision rules in international organizations (see Harstad
2005, Maggi and Morelli
2006, and Barbera and Jackson
2006).
8 In particular, our results regarding the optimal voting rule in multilateral funds are related to Harstad (
2005), who shows how a majority rule can mitigate a holdup problem in a setting where investments by members of a club (or countries in a union) are expropriated ex post. A key difference in our findings, however, is that majority rule is not always optimal for donor countries, even though the costs of investment are fully borne by the recipient countries. Instead of simply choosing the decision rule that maximizes investments, unanimity may be optimal for multilateral aid organizations because it ensures that all recipient-country projects receive funding, whereas majority increases investment precisely by providing more funding to the most effective projects.
More broadly, our paper also relates to the literature on aid allocation and selectivity. A number of recent contributions investigates what determines the allocation of aid. Schneider and Tobin (
2016) show that governments allocate more resources to multilateral organizations that are similar to the donor in terms of the countries they support.
9 Annen and Knack (
2019) show that aid receipts increase with the quality of recipient-country governance.
10Dietrich (
2013) shows that donors bypass national governments when institutional quality is low; Dietrich and Murdie (
2017) report that shaming by International Non-Governmental Organizations has the same effect. According to Knack (
2013), donors rely on recipient-country aid management systems to a larger degree when such systems are of higher quality, pointing to an allocation of aid that builds administrative capacity in recipient countries. Our analysis contributes to this literature by demonstrating how the design of decision-making institutions within multilateral funds can impact the investment choices of the recipient countries.
A number of recent papers also have investigated the conditions under which donors prefer bilateral over multilateral aid. According to the standard view, multilateral aid allows different donors to share the burden of aid-giving, at the cost of losing control over how exactly the aid is spent (Milner and Tingley
2013, Reinsberg et al.
2017). As holds true for multilateral cooperation at large, multilateral aid can realize efficiency gains, pool risks, materialize economies of scale, and encourage wide cost sharing (Abbott and Snidal
1998). However, it is important to note that bilateral aid also has access to benefits of scale in implementation since bilateral aid is often channeled through international institutions (see Eichenauer and Hug
2018 or Reinsberg et al.
2017 for analyses of donors’ decisions to participate in trust funds and the implementation of earmarked aid). Additionally, by avoiding the bargaining process inherent in multilateral aid, donors have more direct control and freedom to use bilateral aid as a tool to promote their own political interests. Our paper shows that, despite the ability to achieve economies of scale with bilateral aid, donor countries benefit from delegating aid to a bargaining process as a commitment to reward recipient countries for investment in administrative capacity and bureaucratic quality. Related, Dreher et al. (
2020) show that powerful governments refer to multilateral organizations when they aim to hide contentious foreign policies from domestic audiences. They allocate aid bilaterally when extending favors to allies, while they channel funds via multilateral organizations to non-allied recipients, where more visible bilateral aid is more contentious.
Our paper proceeds as follows: Section
2 provides illustrating empirical evidence to motivate the formal analysis. Section
3 presents the theoretical framework. Section
4 contains the analysis of aid allocation (Section
4.1) and investment decisions (Section
4.2); Section
4.3 contains our normative analysis and characterizes the optimal decision rule. Section
5 concludes. We present formal proofs for all results in the
Appendix.
11
2 Empirical analysis of decision rules and multilateral aid
Before we proceed with the model, we provide descriptive evidence regarding the relationship between voting rules in multilateral organizations and the sensitivity of aid allocation to changes in their recipient countries’ governance policies. To address this question, we draw on the Creditor Reporting System (CRS) of the OECD’s Development Assistance Committee (DAC), which provides recipient- and year-specific data on aid given by a broad range of international organizations. We consulted the statutory documents of these international organizations, as well as secondary sources, in order to code the organizations’ decision-making rules for allocating funds across members.
12
Out of the 46 international organizations included in the CRS, only five allocate funds via a strict unanimity rule.
13 This is too small a sample to provide sufficient empirical evidence on the difference between unanimity and majority. However, an additional 15 organizations take decisions via a consensus rule. While consensus rules have commonly been coded as unanimity in the literature on international organizations (Blake and Lockwood-Payton
2015), they differ from an explicit unanimity rule in the sense that country representatives are strongly encouraged to seek unanimity, but may fall back on a majority decision rule if unanimity cannot be attained (see Gould
2017 for a detailed discussion on the implementation of consensus rules).
14
Given the small number of organizations in our sample that use a traditional unanimity rule, we separate organizations that allocate funds according to a (weighted or unweighted) majority rule from those that explicitly encourage decisions via consensus in our empirical analysis. Therefore, our assumption is that the bargaining power of countries in the minority is higher under a consensus rule relative to a traditional majority rule—arguably, organizations with the strong norm of decision-making via consensus are closer to unanimity, in terms of our model, than organizations deciding based on majority rule alone.
Table
A1 in the Online
Appendix lists the 46 international organizations included in our data. For each organization, we list the main decision-making rule for allocating funds, in concert with information about whether we code decision-making to conform with the norm of consensus as well as sources on which we base our coding. As can be seen, 41 organizations decide with majority and five require unanimity. 20 organizations included in our data use either consensus or unanimity. In the empirical application,
Consensus Rule is thus a binary indicator that is one for these 20 organizations, and zero for the others.
One way to test the impact of decision rules on aid allocation is to compare the weighted average governance quality of the aid portfolio across multilateral organizations. However, the multilateral organizations in our sample vary widely in their scope and regional focus, and a large degree of the variance in the allocation of aid spending may be unrelated to their ability to overcome the Samaritan’s problem. A test of the sensitivity of aid allocation to recipient country investment in good governance thus has to consider how the aid allocation of a given institution changes in response to
changes in the recipient countries’ governance policies. Accordingly, we test the relationship between decision rules and aid allocation using the following regression equation:
$$ \begin{array}{@{}rcl@{}} Aid_{ijt}=\beta_{1} Bureaucratic \ Quality_{jt}+\beta_{2} Consensus \ Rule_{i}\\ +\beta_{3} Quality_{jt}*Consenus_{i}+\beta_{{4}} \mathbf{X_{jt}}+\eta_{i} +\eta_{j} +\tau_{t} + \epsilon_{ijt}, \end{array} $$
(1)
where
Aidijt is the amount of funds committed by international organization
i to a specific member
j in a year
t in millions constant 2010 US$,
Consensus Rule is a measure of consensual decision-making,
Bureaucratic Quality an indicator of recipient country institutional quality, and
X constitutes a set of control variables. One difficulty, in terms of translating the model to the empirical setting, is identifying the set of relevant recipient countries for each international organization. For example, most countries of the world are member of the International Bank for Reconstruction and Development (IBRD), but some have never received funding and are thus unlikely to adjust their policies in reaction to the IBRD’s decision rules. Our regressions therefore only include countries that have received funding from a specific organization in at least one of the years of our sample.
15
Our indicator of
Bureaucratic Quality is from the International Country Risk Guide (PRS Group
2017), and measures whether bureaucracies are “somewhat autonomous from political pressure and ... have an established mechanism for recruitment and training,” on a zero to four scale. We control for the recipient country’s (log) GDP per capita (in constant 2010 US$), (log) population, and its share of exports relative to GDP, which are control variables included in most aid allocation regressions.
16
We estimate the aid allocation models with Tobit, which is well-suited for modeling the allocation of aid if we assume whether or not countries receive aid and, if they do, how much they receive are determined by the same process.
17 Aid allocation models are also often estimated as two step models, where a first set of regressions focuses on the decision whether or not to provide funding to a recipient in a year and the second stage explains the amount of funding received, given a country was selected to receive funding (e.g., Fleck and Kilby
2010). We provide such analyses for completeness. As is typical in the related literature, we focus on logged values of aid (adding a value of one to keep zero observations in the sample). For comparison, we also report results using an inverse hyperbolic sine transformation and a regression that does not transform aid (i.e., that focuses on absolute dollar values).
18
We estimate the model with a set of fixed effects.
19 Our preferred specification includes dummies for recipient countries (
ηj) and years (
τt), given that the decision-making rule is constant over time within donor organizations and thus collinear with donor fixed effects. We however also report a specification including dummies for donor organizations (
ηi), where the coefficient of
Bureaucratic Quality is captured by the fixed effects. The coefficient of the interaction is then exclusively identified by changes in
Bureaucratic Quality within recipient-countries over time. We cluster standard errors at the international organization level.
Table
1 shows the results. Column 1 reports the basic Tobit regression, excluding fixed effects for donors, with log aid commitments as dependent variable.
20 Focusing on the coefficient of
Qualityjt ∗
Consensusi, our results show that consensus rule results in a less selective budget allocation compared to majority rule.
21 This result holds when we use an inverse hyperbolic sine transformation (in column 2) or focus on the level of aid commitments (in column 3) rather than taking logs, include fixed effects for donors (in column 4) or exclude control variables (in column 5).
Table 1
Consensual Decision-making and Bureaucratic Quality, 1985-2016
(log) GDP per capita | 0.135*** | 0.156*** | 6.723* | 0.158*** | | –0.149 | 0.018 | 0.137*** |
| (0.03) | (0.04) | (3.48) | (0.00) | | (0.16) | (0.02) | (0.03) |
(log) Population | 0.141*** | 0.183*** | –19.440*** | 0.121*** | | 1.086** | 0.045 | 0.203*** |
| (0.02) | (0.02) | (1.58) | (0.00) | | (0.49) | (0.06) | (0.01) |
Exports/GNI | –0.103 | –0.110 | –18.780 | –0.106*** | | –0.379*** | –0.020 | –0.025 |
| (0.54) | (0.62) | (51.14) | (0.03) | | (0.13) | (0.31) | (0.45) |
Bureaucratic quality | 0.010 | 0.007 | 7.304 | –0.020* | 0.019 | 0.034 | –0.002 | –0.496*** |
| (0.12) | (0.14) | (12.56) | (0.01) | (0.10) | (0.07) | (0.01) | (0.10) |
Consensual decision-making, dummy= 1 | –1.247*** | –1.420*** | –65.613*** | | –1.387*** | –1.116** | –0.047 | –1.988*** |
| (0.20) | (0.24) | (19.07) | | (0.19) | (0.43) | (0.08) | (0.19) |
Consensual decision–making * Bureaucratic quality | –0.249** | –0.275** | –51.170*** | –0.127*** | –0.179* | –0.351* | –0.005 | 0.421*** |
| (0.10) | (0.12) | (10.18) | (0.01) | (0.10) | (0.20) | (0.02) | (0.10) |
Triple interaction | | | | | | | | –0.0004*** |
| | | | | | | | (0.00) |
Method | Tobit | Tobit | Tobit | Tobit | Tobit | OLS | OLS | Tobit |
Aid is in | log | ihs | levels | log | log | log | binary | log |
IO fixed effects | no | no | no | yes | no | no | no | no |
Recipient fixed effects | yes | yes | yes | yes | yes | yes | yes | yes |
Year fixed effects | yes | yes | yes | yes | yes | yes | yes | yes |
Number of observations | 63,986 | 63,986 | 63,986 | 63,986 | 70,314 | 18,143 | 63,986 | 63,986 |
Number of IOs | 45 | 45 | 45 | 45 | 45 | 42 | 45 | 45 |
Number of recipients | 94 | 94 | 94 | 94 | 100 | 88 | 94 | 94 |
Results are very similar when we focus on the second stage of the aid allocation process, using the sample with positive aid commitments only. Both in terms of statistical significance and in quantitative terms, results from this OLS regression (shown in column 6) are very similar to those of column 1. To the contrary, results from a linear probability model of the binary selection stage shows no significant interaction (column 7).
22 It thus seems that the results from the Tobit analysis are driven by the second rather than the first stage. As we show in the robustness section of our formal model, this is consistent with a setting where donor countries have a partial bias so that recipient countries with high bureaucratic quality receive more, but not all, of the aid funds.
Quantitatively, the OLS results of column 6 imply that organizations which decide via consensus commit around 40 percent lower amounts of aid to countries that improve their bureaucratic score by one point compared to organizations that follow a strict majority rule. The corresponding decrease according to the Tobit estimates of column 1 (for countries that receive positive values in a year) is around 30 percent. In absolute numbers, the estimate of column 3 implies that organizations which decide consensually commit around US$ 50 million less to countries that improve their bureaucratic score by one point compared to organizations that follow a strict majority rule (with mean commitments of around US$ 30 million and a mean of US$ 106 million for positive commitments).
We conclude this section by testing whether the effect of
Qualityjt ∗
Consensusi depends on the size of an international organizations’ resources in a year. To this end, we add an interaction of
Qualityjt ∗
Consensusi with the yearly total of funds committed by an organization, to proxy their aid budget.
23 As one would expect, the effect turns stronger (i.e., more negative) when total commitment size increases.
24 This is in line with the interpretation that budgets have to be sufficiently large in order to work as incentive.
In summary, our conditional correlations provide some evidence that organizations with majority rule are more selective compared to those adopting decisions by consensus. This novel empirical finding can be rationalized in the context of our model: As we show below in Section
4.3, majority rule induces stronger competition between recipient countries since, relative to unanimity, the bargaining solution under majority is more responsive to changes in recipient country governance quality.
5 Conclusion
In this paper, we consider a formal model of the allocation of aid spending in an environment where donor countries face a bias over which recipient countries receive funding. Our analysis provides several important insights regarding the optimal design of multilateral aid organizations. As highlighted by Svensson (2000, 2003), multilateral aid organizations must focus on distributing aid in a manner that provides an incentive for developing nations to invest in reform. However, given competing national and special party interests, the question is how to enforce this objective. Here, we show that competition in the area of reform can arise endogenously when donor countries directly bargain over the allocation of aid funds, and that this competition is intensified under majority rule, as recipient countries invest in reform to increase the probability that their project will be selected by the endogenous majority coalition. These findings are consistent with our novel empirical finding that aid allocation by organizations deciding via majority is more responsive to changes in recipient-country governance quality.
From a policy perspective, our analysis suggests that a majority rule should only be adopted when there are high positive spillovers of recipient countries’ investment in reform on other areas, such as promoting good bureaucratic practices more widely, and when investment in reform has an intermediate level of productivity—if productivity is very low, then the utility loss of only funding a subset of projects is not worth the gain from additional investment in reform; and if productivity is very high, then both Unanimity and Majority result in high levels of investment in reform.
We emphasize that the predictions of our model only apply to multilateral aid funds that allocate aid spending via an unstructured bargaining process. In recent years, “earmarked” donations (aka multi-bi aid) have become increasingly common as donor countries seek to take advantage of the benefits of scale of international organizations, while ensuring that aid is distributed according to national priorities (Eichenauer and Hug
2018). However, as our paper shows, earmarking diminishes the incentive of recipient countries to invest in reforms, since it circumvents multilateral bargaining. Therefore, in this case, less structure can result in greater efficiency.