The misty governance eﬀect of foreign aid
There is a recent but fast-growing body of the aid literature that documents the eﬀect of foreign aid on diﬀerent aspects of governance since the end of the nineties. Theoretically, foreign aid can both help and hinder governance. Empirically, there is yet no consistent evidence or at least few agreements on how foreign aid aﬀects the quality of governance. Critics argue that aid has been counterproductive in that it has supported governments that were hostile to economic growth and poverty reduction. Pros sustain that aid has fomented or at least accelerated the building of an improved governance oriented to growth and social development. In practice, diﬀerent forces may condition the governance eﬀect of foreign assistance. First, donors may aﬀect the quality of governance in both a positive and a negative direction depending on their own behaviors. Second, even if donors have good designs, aid – or certain types of aid – can possibly undermine the long-term development of governance depending on recipients’ characteristics.
Some scholars, supported by Knack and Rahman (2007) and Busse and Groning¨ (2009), have upheld that foreign aid is adverse to good governance. Knack (2001), for example, found that foreign aid undermines the rule of law and the quality of bureaucracy, both measured by International Country Risk Guide (ICRG) indicators. Controlling for aid en-dogeneity – in the sense that well-governed countries tend also to attract more aid – Knack (2004) showed that the quality of institutions decreases in countries receiving high aid in-flows. Brautigam¨ and Knack (2004) used an aggregated measure of governance provided by the ICRG data set to confirm that African aid dependent countries have a poor quality of governance. They expounded that aid dependent countries rely more on foreign assistance than on their own citizen’s taxation, which lowers pressure for accountability.2 Foreign aid may also attract greed over aid funds and postpone necessary reforms by making it easier to bear the cost of non-reforming. Rajan and Subramanian (2007) supported these results. They assumed that the manufacturing sector is dependent on good governance — as contracts enforcement and investment protection. Since they found that foreign aid is associated with a decrease in the share of manufacturing in GDP, they concluded that aid lowers the quality of governance (and the need to improve it). Djankov et al. (2008) cor-roborated the “curse” of aid in recipient countries, no matter how governance is measured. They used a model based on a wide panel of 108 countries between 1960 and 1999 to show that the adverse eﬀect of aid is even stronger than is the relationship between governance and oil.
On another hand, there are several studies arguing that foreign aid is beneficial for governance. An argument for a positive channel is aid conditionality, which requires that institutional reforms are undertaken by the current recipient country in order to be eligible again as an aid recipient country. Chauvet and Guillaumont (2003) denoted that foreign assistance, which enlarges public projects expenditures, is able to improve the quality of governance if aid allocations are conditioned on reforms commitments. Another argument is the matter of the new aid architecture. The Cold War era has not encouraged the development of good governance in Africa owing to donors’ political and strategic interests (Claessens et al., 2009). Once Cold War ended however, aid became more targeted towards governance issues (Charron, 2011). Focusing on African countries, Goldsmith (2001) claimed that political institutions, measured by Freedom House indicators, rely on foreign assistance to keep operating public services and reforms. Dunning (2004) confirmed the benefits of aid on good governance and showed that this connection strengthens in the post-Cold War period. Finally, Tavares (2003) found that foreign aid also decreases corruption thanks to higher public salaries and transfers of knowledge.
The aid-governance literature does not point out a clear agreement on the eﬀects of aid on the quality of governance. This essay is interlinked to this literature. It provides a new contribution that may explain why foreign aid does or does not improve the quality of governance. The central contribution of this essay is the hypotheses that both the depen-dence of a country on natural resources and the type of aid donors matter in determining the relationship between foreign aid and governance.
The role of natural resources
Our first hypothesis is that the country’s dependence on natural resources conditions the influence of foreign aid on the quality of its governance. Foreign aid is probably prone to maintain a low quality of governance in resources dependent countries.
Hypothesis 1.1 The dependence of a recipient government on rents generated from natural resources extraction harms the positive eﬀects of aid on the quality of governance.
We base our assumption on the following reasons. First, revenues from natural resources, at least some of them, tend to increase rent-seeking problems, to weaken the quality of governance and to create political instability (Ades and Di Tella, 1999; Jensen and Wantchekon, 2004; Collier and Hoeﬄer, 2005; Collier, 2006b; Dalgaard and Olsson, 2008; Vicente, 2010; Bhattacharyya and Hodler, 2010). Papyrakis and Gerlagh (2004) confirmed that the corruption eﬀect of natural resources neutralizes more than 40 % of the beneficial eﬀect of natural resources on economic growth. Producing high rents, natural resources activities are a honey pot, which increases patronage politics, corruption and high inequal-ities between those who hold these rents (namely the oligarchy) and the others (Collier and Hoeﬄer, 2009; Morrison, 2007). Oligarchies are able to avoid taxation and to resist the adoption of institutional reforms that would limit their choices and force them to be more responsible (see, for example, Djankov et al. (2008)). In turn, we assume that aid allocated to dependent countries would be less prompt to foment institutional reforms. Second, in resource-rich countries, investment in extractive industries is immediately more profitable than in productive industries. Natural resources enable the domestic country to derive large rents from their extraction, without any consequent investment of time and money (Leite and Weidmann, 1999). Rents on natural resources in turn may finance and support the existing government. The government, which otherwise would have collapsed, hoards the benefits of resources flows to stay in place and finance its own activity. Besides, the immediate economic benefit derived from the abundance of natural resources is partly oﬀset by the adverse eﬀects of high commodity prices on the domestic exchange rate, which impedes the development of exports of local manufactures. Yet, as shown by Rajan and Subramanian (2011), extractive industries do not require as sound institutions as manufactures do, which may also hamper the development of good governance. As a consequence, aid in hands of a government that holds rents from natural resources is expected to be diverted toward extractive industries instead of productive activities.3 Aid is presumably less preferable in a country that derives substantial rents from its natural resources because the recipient government would have no incentives enough to allocate aid funds towards institutional reforms.
The role of bilateral versus multilateral donors
Our second hypothesis (portrayed in Appendix C) is that aggregating diﬀerent types of aid may hide intrinsic variations derived from donors’ aid motives. The mechanism to successfully improve governance is assumed to have something to do with the way in which aid is allocated.
Hypothesis 1.2 Multilateral aid, more opted for the development of good governance than bilateral aid, bears the beneficial eﬀect of aid on the quality of governance.
The recent literature agrees on the necessity to consider that the eﬀect of foreign aid is diﬀerent before and after Cold War, partly because of geopolitical interests (see Dunning (2004)). Though containing the communist expansion during Cold War was not the only and main preoccupation for all the donor community, the end of Cold War has presumably changed some of donors’ views and strategies. But even in the post-Cold War period, empirical studies did not agree on the eﬀect of aid on governance. Perhaps because an aggregated measure of aid blurs the picture. The growing debate on the fact that diﬀerent types of donors may behave diﬀerently has led to the conclusion that bilateral and multilateral donors have diﬀerent motives when allocating foreign assistance (Neumayer, 2003b; Dollar and Levin, 2006).
Our prior is that bilateral aid may aﬀect diﬀerently governance than multilateral aid because of the motives underlying aid allocations. A related extended literature has en-hanced the diﬀerences in these types of donors’ behaviors. According to Acharya et al. (2006), foreign aid would be more eﬀective if allocated by multilateral agencies due to less donors proliferation.4 The success of the Marshall Plan (1947) is often attributed to the fact that the United States were the only donor responsible for the program (Knack and Rahman, 2007). Besides, multilateral agencies appear generally to have a greater devel-opmental focus than bilateral donors do (Burnside and Dollar, 2000; Alesina and Dollar, 2000; Neumayer, 2003c). Though multilateral institutions are not totally preserved from political influence (Frey and Schneider, 1986), bilateral donors are less likely to pressure on multilateral funds than on their own allocations. Multilateral aid is less tied to politi-cal interests because individual donors’ interests are diluted. Finally, there is a consensus among multilateral agencies to be more explicitly attentive to the concern of governance since the end of the nineties, in particular thanks to conditions over aid allocations (Dol-lar and Levin, 2006).5 Allocating aid to countries that commit on political reforms – aid conditionality – means that recipient countries either fulfill minimal reforms to increase their governance quality or receive lower aid funds. Alesina and Dollar (2000) found that bilateral donors do not only target poor countries but also countries with whom they have close commercial, political and historical ties. These connections may aﬀect the aid eﬀec-tiveness because close ties between donors and recipients give to recipient countries the possibility to resist institutional reforms asked by donors (Ram, 2003; Headey, 2008).
All these studies have shown that bilateral aid and multilateral aid have diﬀerent motives. However, this concern has been largely omitted from the academic discussion of the eﬀect of aid on governance. We enter into the debate by considering the distinction between both types of donors to analyze the eﬀects of bilateral and multilateral aid on the quality of governance. Alesina and Weder (2002) opened this branch by investigating the eﬀect of aid on corruption in an OLS estimation. However, they found no significant diﬀerence between bilateral and multilateral donors in reducing corruption between 1975 and 1995. Charron (2011) nuanced this result showing that the diﬀerence between both types of donors becomes significant only after the end of Cold War. Specifically, after 1997 and international commitments on a governance focus, multilateral aid succeeds in reducing corruption while bilateral does not, no matter the time period. Charron (2011) applied the “diﬀerence” General Method of Moments (GMM) estimator on dynamic panel data covering 82 recipient countries to avoid a possible simultaneity bias between aid and governance, a specific econometric issues that will be discussed in Section 1.4. The potential for a reverse causation, not taken into account in Alesina and Weder (2002), may also explain the diﬀerence recorded between the two studies findings.
The data and the variables
In order to investigate whether rents on natural resources aﬀect both bilateral and multi-lateral aid eﬀects on governance, we use annual available data for 52 African aid recipient countries, from 1997 to 2008 (see Table 1.8 for the list of countries). Our panel data is unbalanced (data are not available each year for all countries). Following Busse and Gron¨-ing (2009), we average the data over three years to flatten out cyclical fluctuations.6 The sources and definitions of the variables are reported in Table 1.11. Descriptive statistics for the variables are provided in Table 1.1 and depicted in details in Appendix 1.A.3.
Variables of interest
Our dependent variable is a proxy for the quality of governance. There are many sources that produce ratings on the quality of governance.7 The most frequent measure in aca-demic research (as Knack (2001) and Brautigam¨ and Knack (2004)) is that compiled from the International Country Risk Guide (ICRG), a commercial service providing informa-tion on governance for investors and lenders. The ICRG quality of governance is the mean value of the ICRG measures of corruption, law and order, and bureaucracy quality (source: the Quality of Governance Institute). Corruption stands for the eﬃciency of government (whether positions are assumed through nepotism or ability) and its stability. Law and order stands for the impartiality of the legal system and the enforcement of law. Bureau-cracy quality stands for the quality in public services. The ICRG indicator is scaled from 0 to 1. Higher scores indicate higher quality of governance. The lowest value of the quality of governance within the sample is 0.083 for Somalia in 2008 and the highest value is 0.875 for Namibia in 1997.
To account for foreign aid we use the Net Oﬃcial Development Assistance (ODA), which refers to the disbursement of aid granted and to loans with a grant proportion of at least 25 percent. Among aid measures used in the empirical analysis, aid intensity (or aid dependence) scales ODA by the recipient’s GNI. This measure accounts for the de-pendence of a country on foreign aid (Brautigam¨ and Knack, 2004). Multilateral ODA is ODA allocated by an international agency, institution, or organization to an aid recipient country. Bilateral ODA is ODA allocated directly by one donor to one aid recipient coun-try. Annual data of bilateral ODA and of multilateral ODA are available from the World Development Indicators (WDI) and from the Organization for Economic Cooperation and Development (OECD) databases. The average recipient country of our sample receives 12.7 percent of total ODA in GNI (among which 40 percent is allocated by multilateral agencies). The highest allocation (144 percent of total aid in GNI) was directed to Liberia in 2008.
About a third of African countries are rich in natural resources, particularly in gold, diamonds, platinum (namely minerals), oil and gas (see Table 1.B.1 in Appendix 1.A.3). To assess the influence of natural resources on African countries’ governance, we use three distinct measures, denoted Oil, Gas, and Min, appraised in percentage of GDP, and an aggregate, denoted Nat, where Nat = Oil + Gas + Min (source: WDI). These measures provide the share of oil, gas, and minerals in the GDP of the recipient country. In other words, they capture a country’s dependence on natural resources.
Following the existing literature, control variables are used to capture the determinants of the quality of governance and recipients’ characteristics. Indeed, although African coun-tries are (in average) major aid recipient countries, these countries still have (in average) poor institutional quality. However, it cannot be claimed that aid is directly responsible for the entire shape of governance in Africa. We need to control for all the determinants of governance in order to measure the net eﬀect of aid on governance only. We give an exam-ple. In time of conflict, a country may attract more humanitarian aid and assistance to help for reconstruction. More aid may be associated to lower governance just because countries tend to have a lower quality of governance during the time of conflict. Do not control for the determinants of governance when estimating the aid-governance nexus may produce a false or biased correlation between high levels of aid and the worsening of the quality of governance in recipient countries. The literature on the determinants of governance usu-ally imposes economic growth, social development, conflicts, ethnic heterogeneity, natural resources, history, and geographical location as determinants of governance.
We follow Busse and Groning¨ (2009) by using the economic growth rate (source: WDI) to capture the extent of the influence of economic growth on governance, and the share of rural population (source: WDI) to proxy for social development. Busse and Groning¨ (2009) argued that greater revenues support institutional reforms. Gundlach and Paldam (2009) found that income explains the long term quality of institutions because economic growth can lead citizens to ask for institutional changes suitable for investments. Rural countries have been shown to leave aside the available human capital (Lucas, 2004) and the development of manufacture, which requires strong institutional rules (Rajan and Subramanian, 2011).
We use the ethno-linguistic fractionalization index, which measures the probability that two citizens in a country belong to the same ethnic or linguistic group (source: Alesina et al., 2003) and the number of deaths occurred in an internal or external conflict8 (source: WDI) to control for conflict and ethnic heterogeneity (see La Porta et al. (1999) and Collier (2001)). The degree of fractionalization, say the degree of heterogeneity among citizens, reflects the number of groups in competition. In heterogeneous countries, public resources tend to be diverted towards military, non-productive or rent-seeking sectors (Aghion et al., 2004), and governance presumably weakens (Alesina et al., 1999). Similarly, because conflicts need more public resources dedicated to the military sector, conflicts presumably decrease the quality of governance (Addison et al., 2001; Busse and Groning¨, 2009).
We use a dummy that equals unity for tropical countries (source: CIA Factbook) to point out that tropical weather has hampered the development of sound institutions (Sachs and Warner, 1997; La Porta et al., 1999; Easterly and Levine, 2003; Rodrik et al., 2004). According to Acemoglu et al. (2001), a potential explanation is the inheritance of colonial history. Settlers were not able to build metropolitan institutions where they could not permanently settle. Instead, in areas where they had to face tropical diseases and mortality, they have built extractive institutions, which persist even after colonial independence.
Historical and religion characteristics are also common determinants of the current shape of governance (La Porta et al., 1999; Goldsmith, 2001; Treisman, 2000; Alesina and Dollar, 2000). Aside our tropical dummy, we consider three other variables designed to capture (i) the legal system legacy, (ii) the religious legacy and (iii) the institutional legacy. To proxy for these variables, we use a dummy that takes one for English common law countries (source: La Porta et al., 1999); the shares of Catholic and Muslim populations in countries in 2007 (source: CIA Factbook) and the degree of political freedom (source: Freedom House). First, cultural beliefs and religious traditions shape the citizens incentives to ask for changes in terms of institutions. On the report of La Porta et al. (1999) and Treisman (2000), citizens from Catholic and Muslim countries are less likely to confront the existing government because of the hierarchic social construction. Second, the English common law – inherited from the 17th century with the Parliament to control the political power and protect individual rights – is associated to a lower weight of the government over the society, which decreases opportunities for corruption (La Porta et al., 1999). The recent and rapid creation of African states since the end of the colonial period has often be built over the colonial institutions inheritance (Bloom et al., 1998; Engerman and Sokoloﬀ, 2002). African countries have inherited from their former institutions that might be more or less similar to the European institutions depending on the colonial environment and endowment. During the colonial period, extractive institutions have been developed in resources rich countries while in non-tropical countries, settlers may have exported their institutional outlines and experiences as well as their language and others specific ties. These schemes persist, at least in part, to the present (Acemoglu et al., 2001).
We explore the causal relationship between aid and governance in aid recipient countries using dynamic panel data and accounting for the persistent nature of domestic institutions. We estimate the following benchmark equation:
govit = αi + ρgovit−1 + β1 maidit + β2 baidit + γ1 natit + γ2 maidit × natit + γ3 baidit × natit + φ′ Xit + λt + εit
where govit indicates the measure of the quality of governance for the country i at time t; αi indicates the fixed individual eﬀects on each country; govit−1 is the lagged value of the dependent variable; maidit and baidit are respectively multilateral and bilateral aid flows divided by GNI; natit is the share of natural resources rents in GDP; maidit × natit and baidit × natit are interaction terms; Xit is a vector of control variables; λt indicates temporal dummies, and εit is the error term.9 Econometric problems may arise when estimating equation (1.1) with the Ordinary Least Squares estimator. First, the causality between foreign aid and governance may run in both directions, making foreign aid and the error term not independent. Second, the lagged value of the governance term in the right-hand side causes a problem of auto-correlation. Third, fixed-country eﬀects (as the size of the country or its location) may be correlated with regressors while they are part of the error term. While a fixed eﬀects instrumental variables estimation may cope with these issues, good instruments (highly correlated with the endogenous regressor but uncorrelated with the dependent variable) are diﬃcult to find.10 Another way to cope with these issues is to draw instruments from within the panel dataset itself. To estimate equation (1.1), we therefore use the Blundell and Bond (1998) estimator (hereafter the “system” GMM estimator), designed for dynamic panel data.11 It estimates simultaneously equation (1.1) written in levels and equation (1.1) written in first diﬀerences.
The system GMM estimator performs better than the “diﬀerence” GMM estimator proposed by Arellano and Bond (1991) (and used in Charron (2011)) as it uses additional moment conditions.12 Precisely, estimations are much more eﬃcient in small samples in time when applying the system GMM estimator. While we have to keep the limitations of using GMM estimators in mind – regarding the choice and quantity of instruments in particular –, the system GMM estimator is able to provide consistent results for such models. Besides, we run robustness checks using alternative estimators.
The treatment of endogeneity
We now comment on the issue of endogeneity took up in Subsection 1.4.1. First, using dynamics to capture the eﬀect of lagged Gov on current Gov makes the lagged dependent variable inherently correlated with the error term. Second, aid donors’ allocations may be conditioned on the recipient’s quality of governance (Burnside and Dollar, 2000; Alesina and Dollar, 2000; Svensson, 2000; McGillivray, 2005; Younas, 2008).13 Aid is potentially endogenous to governance and correlated with the residuals. Third, the quality of gov-ernance may explain parts of economic growth (Knack and Keefer, 1995). Mauro (1995) showed that corruption decreases economic growth, either directly or through political instability. Finally, according to Le Billon (2003), a change in corruption or political liberalization aﬀects significantly the probability and duration of conflicts.
The two-step GMM estimator proposed by Blundell and Bond (1998) provides asymp-totically eﬃcient, robust and reliable results for such models when facing endogeneity, dynamic issues and heteroscedasticity (Windmeijer, 2005). The lags of endogenous vari-ables are used as instruments for equation (1.1) written in first diﬀerences and the lagged diﬀerences of the endogenous variables are used as instruments for equation (1.1) written in level. We do not include additional (external) instruments. Specifically, the estimated aid coeﬃcient is not biased by reverse causality and only measures the direct eﬀect of aid on governance.
This estimation procedure assumes that there is no second-order autocorrelation in the error terms and that instrumentation is sound. Hence, for each regression, we test for autocorrelation and for the validity of the instruments (say that instruments are not correlated with residuals). The Hansen J test for overidentifying restrictions loses power when the number of instruments exceeds the cross section sample size (Roodman, 2009). When the ratio of countries to instruments is lower than one, the estimation procedure may be biased and coeﬃcients may be significant even if there is no statistical association. This is precisely the problem faced when using as a dependent variable the ICRG quality of governance. For most of our regressions, the data are available only for 34 countries. To overcome a possible bias in the significance of results, we control for the relative number of instruments so that this number is never large relative to the number of countries.14 For example, in the second regression reported in Table 1.2, 35 variables are used to instrument for endogenous variables. The ratio of countries to instruments (34/35) is lower than one so that we need to limit the number of instruments. Alternative estimating procedures are provided as robustness checks to prove that our results are not dependent on this choice.15
Based on this instrumentation strategy, Section 1.5 presents our analysis. The empirical results for equation (1.1) are reported in Table 1.2. They are designed to answer the following questions:
(a) Do multilateral and bilateral aid have a direct eﬀect on governance?
(b) Do natural resources undermine the positive eﬀect of aid on governance?
(c) Does the eﬀect of aid on governance depend on the type of natural resources?
Do multilateral and bilateral aid have a direct eﬀect on gover-nance?
To answer this question we estimate equation (1.1) without interaction terms. The param-eters of interest are β1 and β2 , the respective coeﬃcients of multilateral and bilateral aid. β1 is positive and β2 is negative, both significant at the 5% level.16 The results suggest that, all else equal, aid increases the quality of governance when allocated by multilateral agencies.
Let us look at two examples to illustrate the propitious eﬀect of multilateral aid on gov-ernance. Consider two countries, the Republic of the Congo and the Democratic Republic of the Congo (DRC). Their GNIs are comparable (a few more than 11,500 million current US dollars in 2008). The Republic of the Congo has received more than 5.67% of GNI in terms of multilateral aid and the DRC around 0.78% in 2008. The regression shows that an increase in multilateral aid from the amount received by the DRC to the amount received by the Republic of the Congo will increase the ICRG indicator (which is scaled from 0 to 1) by about 0.03 units, from 0.11 to 0.14 (∂gov/∂maid = 0 007 × (5 67 − 0 78) ≈ 0 035), say by 27%. Consider now Burundi and Eritrea that also have comparable GNIs (about 1,500 million current US dollars in 2008) but have received extremely diﬀerent multilat-eral aid amounts in 2008. Then, the regression shows that an increase in multilateral aid from the level of Eritrea (5.06% of its GNI) to the level of Burundi (21.73% of its GNI) will increase appreciably the quality of governance by 0.12 units (∂gov/∂maid = 0 007 × (21 73 − 5 06) ≈ 0 120).
We now briefly move to the other variables. Tropical location has a predicted significant adverse eﬀect on the quality of governance. The coeﬃcients of the share of rural population and the shares of Muslim and Catholic populations are positive and significant. Though natural resources, the heritage of English common law, conflicts and economic growth are not statistically significant, they have the expected sign. The estimated coeﬃcient of lagged quality of governance is positive, suggesting that current governance is positively correlated with future governance.
Do natural resources undermine the positive eﬀect of aid on gov-ernance?
We presently estimate equation (1.1) with our two interaction terms, maid × nat and baid×nat. Now, the parameters of interests are β1 , β2 , γ2 and γ3 , where γ2 is the coeﬃcient of maid × nat and γ3 the coeﬃcient of baid × nat. As aid is assumed to be endogenous to governance, interactions terms including aid are also assumed to be endogenous to governance. The inclusion of these interaction terms therefore increases the number of instruments employed in the regressions. The ratio of countries to instruments becomes slightly lower than one. The hypotheses underlying the estimation procedure may be violated and the reliability of our empirical results may be weakened. As seen in Section 1.4, we restrict the number of instruments. All parameters of interests are significant at the 1% level, and the estimates of β1 and β2 are similar to those of the previous regression. Note that β2 and γ1 are negative, and β1 and γ2 positive, both significantly. This suggests that natural resources alter the relationship between multilateral aid and governance by diminishing the propitious eﬀect of aid on governance, maybe because parts of aid can still be diverted from initial aid purposes in resources dependent countries. But surprisingly, estimation results suggest as well that the negative eﬀect of bilateral aid is reduced in resources-rich recipients. Bilateral aid tends (albeit slightly) to be less detrimental to the quality of governance in resources-rich countries.
Some bilateral donors might impose further constraints on aid in resource-rich countries implying more pressure on the recipient government that spends its rents with discretion (in this vein, see Kolstad et al. (2009) who discussed the Norwegian petroleum-related aid program designed to reduce corruption in oil rich countries). Another explanation might be that bilateral donors give less aid (in average) to resource-rich countries, which may reduce the harmful governance eﬀect of bilateral aid. In average, a resource-rich country in our sample receives the quarter of the average amount received by an African country. Some of donor countries may condition their assistance on the governments’ willingness to improve institutional reforms (as Australia and Denmark aid following Berth´elemy and Tichit (2004)), which can weight on total bilateral aid. Aid may therefore be (in average) lower in resource-rich countries owing to weaker institutions than in other developing countries.
Table of contents :
1 Do Natural Resources Condition the Aid – Governance Relationship?
1.2 Literature and hypotheses
1.2.1 The misty governance effect of foreign aid
1.2.2 The role of natural resources
1.2.3 The role of bilateral versus multilateral donors
1.3 The data and the variables
1.3.1 Variables of interest
1.3.2 Control variables
1.4 Estimation procedure
1.4.1 The model
1.4.2 The treatment of endogeneity
1.5 Benchmark regressions
1.5.1 Do multilateral and bilateral aid have a direct effect on governance?
1.5.2 Do natural resources undermine the positive effect of aid on governance?
1.5.3 Does the effect of aid on governance depend on the type of natural resource?
1.6 Robustness regressions
1.6.1 Alternative measure of governance
1.6.2 Alternative estimators
1.6.3 Sample selection
1.6.4 Alternative measure of bilateral aid
1.6.5 Time fixed effects
1.6.6 Extended sample
1.A.1 Defining and measuring governance
1.A.2 Governance, a channel from aid to growth
1.A.3 The benefits of good governance
1.B.1. Descriptive analysis of African countries
1.C.1. The direct effect of total ODA on governance
2 Understanding the Link Between Aid and Corruption: A Causality Analysis
2.2 Literature review
2.2.1 From aid to corruption
2.2.2 From corruption to aid
2.3 Method and data
2.4.1 Main estimations
2.4.2 Additional estimations
2.5 Robustness checks
2.A.11 Summary of the literature
2.B.1 Descriptive statistics for the CPI
2.B.2 Additional unit root tests for corruption
2.B.3 Additional tables of results
3 Is Aid Efficient?
3.2 Why and how focusing on efficiency?
3.2.1 A background on the aid-productivity nexus
3.2.2 A focus on efficiency
3.3 Our empirical strategy
3.3.1 The relationship between aid and efficiency
3.3.2 The data
3.4.1 The direct effect of aid on efficiency
3.4.2 The conditional effects of aid on efficiency
3.4.3 Robustness checks
3.A.1. An insight into the aid-investment nexus
3.B.1. The issue of instrumentation
4 The Aid – Migration Link: How does Unemployment Affect Donors’ Policies?
4.2 Empirical strategy
4.2.1 The panel gravity model
4.2.2 The issue of simultaneity
4.3 Empirical results
4.3.1 Aid allocation
4.3.2 Migration flows
4.3.3 Summary of our policy implications
4.4 Robustness Checks
4.A.11 Descriptive analysis
4.B.1. Additional estimation results
4.C.1. Supplemental estimation results
4.D.1. Summary of hypotheses and implications