Skip to Content
WorldWorld
  • Article
  • Open Access

25 May 2026

24 Pages

Disentangling the Macro-Effects of Foreign Aid: The Role of Institutional Conditions in 132 Recipient Countries

,
and
1
Lisbon School of Economics and Management (ISEG), Universidade de Lisboa, Rua do Quelhas, 6, 1200-781 Lisboa, Portugal
2
MARE-Marine and Environmental Sciences Centre, Escola Superior de Ciências Empresariais, Instituto Politécnico de Setúbal (ESCE/IPS), Campus do IPS, Estefanilha, 2914-503 Setúbal, Portugal
3
ARNET-Aquatic Research Network and BRU-IUL (Business Research Unit), Escola Superior de Ciências Empresariais, Instituto Politécnico de Setúbal (ESCE/IPS), Campus do IPS, Estefanilha, 2914-503 Setúbal, Portugal
4
Portugal and CEMAPRE/REM-Research in Economics and Mathematics, Lisbon School of Economics and Management (ISEG), Universidade de Lisboa, Rua do Quelhas, 6, 1200-781 Lisboa, Portugal

Abstract

This study revisits the debate surrounding the impact of Official Development Assistance (ODA), also known as foreign aid, on two macro-indicators: economic growth and child mortality. Unlike previous studies, which assessed the interaction of aid with composite indicators of recipient countries, this study examines the impacts of individual recipient factors, such as corruption, democracy, income, wars and exports. To overcome the issue of an inverse causal relationship potentially existing between the amount of aid received and macro-performance, a model of donor aid allocation is specified within an instrumental variables framework. The results show that ODA is more likely to be positively associated with economic growth in countries with lower levels of corruption. This positive association is evident when the level of corruption is at least one standard deviation lower than the recipient’s average. The interaction of ODA with recipients’ levels of democracy, income, wars or exports does not show a significant association with growth. The association between ODA and child mortality appears to be generally more significant, with a positive sign, than that obtained for economic growth, although the magnitude is relatively modest.

1. Introduction

In order to contribute to achieving the Sustainable Development Goals (SDGs), developed countries have committed to promoting the mobilisation of additional financial resources for developing countries, from a variety of sources, to address their development needs. However, recent economic shocks have posed challenges to achieving these goals, particularly with regard to leveraging private resources. In this context, the potential contribution of Official Development Assistance (ODA) to achieving these goals has been highlighted.
Therefore, it is particularly important to assess the impact of aid flows in recipient countries, especially at a macro level. This is particularly important for donors who allocate aid based on the performance of previous aid. Failure to properly assess the impact of such aid can lead to distortions in the allocation of future aid, which could have negative consequences for its effectiveness.
This article aims to assess the potential macro effects of ODA flows from developed to developing countries on two indicators of recipients: economic growth and child mortality. The study is based on the hypothesis that increased aid inflows and better institutional and social conditions in recipient countries should be associated with improved outcomes in terms of both economic growth and child mortality, as explained in detail in the following section.
The empirical literature assessing the macroeconomic impacts of aid on the social and economic indicators of recipient countries has reached markedly different conclusions. Some papers conclude that aid has a positive impact in all circumstances, or at least in certain policy environments or when provided by specific donors. In contrast, another set of papers concludes that, regardless of the policy environment, donors or categories of aid, aid does not seem to have a positive impact in most developing countries.
Examples of papers concluding that aid consistently promotes growth, at least under certain conditions, include Burnside & Dollar [1,2], Moreira [3], Karras [4], Bobba [5], Lin Moe [6], Arndt et al. [7], Kilby & Dreher [8], Minoiu & Reddy [9], Salisu & Ogwumike [10], Byron [11], De Matteis [12], Minasyan et al. [13], Lee et al. [14], and Yiheyis & Woldemariam [15].
Papers that conclude that aid does not seem to have a positive impact on growth in most circumstances include Easterly [16], Dalgaard et al. [17], Rajan & Subramanian [18,19], Stockemer et al. [20], Clemens et al. [21], Jin & Oh [22], Benmamoun & Lehnert [23], Driffield & Jones [24], Lessmann & Markwardt [25], Wako [26], Nwude et al. [27], and Avom et al. [28].
Moreover, studies undertaken at the micro level, such as evaluations of specific projects, suggest that aid has a consistently positive effect on recipients (Dalgaard & Hansen [29]), which accordingly raises the possibility of a micro–macro paradox (first named as such by Mosley [30]). These mixed results may be explained by the way in which ODA interacts with certain social, political and economic conditions in recipient countries.
This possibility has previously been explored (e.g., Burnside & Dollar [1], Dalgaard et al. [17], Salisu & Ogwumike [10]), but the interactive factors analysed in these papers were limited to the following indicators: (i) policy quality indices (which involve subjective judgements about economic policies); (ii) the proportion of land in the tropics; and (iii) composite institutional indices (which average various recipient characteristics, thus masking the effects of individual components). The influence of certain fundamental recipient characteristics, which may also be relevant to explaining aid effectiveness, remains to be investigated.
The main novelty of the present article lies precisely in assessing the possible effects of ODA in a distinct way from the studies mentioned in the previous paragraph, specifically by examining ODA interaction with certain social, political and economic indicators of recipient countries that previous studies have not considered (or at least not individually, thus obscuring its particular effects); namely, the levels of corruption, democracy, income, wars and exports.
This research also contributes to the literature in two other ways: (i) it proposes a new model of aid allocation by donors, within an instrumental variables approach, to overcome the problem of the inverse causal relationship between the amount of aid received and macroeconomic performance, based on the assumption that the worst performers may well receive more aid; and (ii) it considers a larger number of donors and recipients than those analysed in previously published papers.
The remainder of the paper is organised as follows: Section 2 presents a theoretical framework for the macroeconomic role of aid; Section 3 describes the methodology of the empirical assessment; Section 4 examines the results of the econometric analysis; and finally, Section 5 presents the main conclusions.

2. Theoretical Framework

This section outlines the theoretical framework of the macroeconomic role of aid, as well as its relationship with the variables to be included in the empirical model. It considers current frameworks, as well as the long-standing dual-gap model. Despite the criticism it has received, this model is still generally considered to be the most effective for integrating aid flows into the macroeconomic framework.
According to Keynesian-inspired approaches, developing countries were seen as consisting of two highly distinct sectors: a capitalist one (which generates savings, investment, and growth) and a traditional one (which lacks the capacity to generate savings and focuses only on subsistence). In this context, growth would only be achieved if the capitalist sector absorbed the surplus labour of the traditional sector, thereby increasing its weight in the economy as a whole. For these approaches, the main problem to be solved was to break the cycle of underdevelopment—if growth could be initiated, it would at some point become self-sustaining. The first stage consisted of the so-called “big push”, where a significant amount of investment is launched to stimulate the simultaneous development of various industrial sectors.
According to Riddell [31], conventional aid theory has its origins in Keynesian economics applied to industrialised countries. Foreign aid could thus be seen as a means of overcoming the insufficiency of domestic savings in low-income countries, thereby contributing to greater convergence in income levels. In particular, aid should provide the capital resources necessary for recipients to achieve a level of savings and investment sufficient to generate a path of self-sustaining growth. It was recognised that it was virtually impossible for low-income households to accumulate the desired level of savings, and that developing regions lacked the physical and human capital to attract private investment, and therefore there seemed to be no alternative to foreign aid. Other factors that made aid a relevant element of growth strategies were the belief in medium- and long-term government macro-planning and the simplicity of the Harrod–Domar model for assessing the amount of aid needed to achieve a given growth target (Thorbecke [32]).
Since economic growth is highly dependent on investment, which requires corresponding savings, developing countries face the problem of stimulating both investment and the savings needed to finance it. The assessment of the savings/investment gap was developed by Chenery & Strout [33] through the so-called dual-gap model, which analyses the reasons that justify the need for foreign capital. This framework became the standard theoretical justification for the provision of development aid (Riddell [31]).
In particular, economic growth requires investment in capital goods, which can either be imported, requiring foreign exchange, or produced domestically, requiring domestic savings, although it should be noted that certain categories are only available abroad or through foreign direct investment. In a context of scarce foreign exchange reserves, countries tend to rely on the latter. The usual understanding of foreign capital has been that it is complementary to domestic capital, with the aim of closing the savings/investment gap and inducing faster growth. However, the dual-gap approach argues that foreign capital can also be seen as a complement to foreign exchange, and that it should therefore cover the larger of the two gaps (savings/investment or exports/imports) if a given growth target is to be achieved. If domestic and external resources are not substitutable, then growth will be constrained by the most limiting factor—either domestic saving or foreign exchange.
In this context, as noted by Ghatak [34], if the growth rate generated by domestic saving is lower than that generated by foreign exchange, then a certain proportion of the latter would be irrelevant to growth if the constraint were to persist. On the other hand, if the growth rate generated by domestic saving is higher than the growth rate allowed by foreign exchange, then a share of the former would be irrelevant.
Bearing in mind that most developing countries are in a situation where one of the above constraints prevails, leading to a waste of resources, it should be emphasised that in cases where foreign exchange scarcity is the main constraint, unemployed domestic resources should be used to increase foreign exchange earnings or to increase the productivity of imports. On the contrary, in cases where domestic saving is the insufficient factor, foreign exchange should be used to increase domestic saving and the productivity of domestic resources.
The question of the absorptive capacity of recipient countries is also raised; i.e., their ability to use capital productively and achieve satisfactory rates of return. This capacity depends on the level of income, the growth rate, the availability of skills and also the savings rate, as well as the quality of the economic and educational infrastructure.
Degnbol-Martinussen & Engberg-Pedersen [35] note that the dual-gap model involves not only the provision of grants but also soft loans, as negotiating foreign loans on market terms to increase capital formation has not been an attractive alternative to aid for many developing countries because they lack the creditworthiness to obtain acceptable interest rates. However, the strong emphasis on the volume of capital as a trigger for growth is contested by those who consider that its qualitative aspects are more important, while others have stressed the relevance of the recipient countries’ conditions and policies, factors linked to the concept of structural adjustment. Technical cooperation to improve knowledge and skills is therefore seen as having a very positive impact on the development process; namely, in the framework of theories that consider human capital as the main driver of development.
Some authors, such as Easterly [16], question the key assumption that aid actually finances investment rather than consumption. This assumption only holds if investment is constrained by liquidity shortages and incentives to invest are encouraging. If investment is low due to poor incentives, then aid will not encourage investment. Indeed, aid could even reduce incentives to invest if recipients believe that future poverty will attract future aid. In either case, aid would be financing consumption rather than investment.
The most frequently cited criticism of the dual-gap model is the assumption that foreign and domestic resources are not substitutable (Bird [36]). In fact, if a more flexible stance is adopted, allowing for a high degree of substitutability, and if there is a significant chance of import substitution or increased exports, then the demand and supply of foreign exchange can be altered without adversely affecting growth, and thus its scarcity cannot be considered a relevant constraint, leaving only one gap. Furthermore, if foreign resources change the structure of output in such a way as to reduce the capital–output ratio, then the assumption of fixed ratios in the dual-gap model is challenged. However, since the assumptions about the way in which resources are used are only really feasible in the long run but not in the short run, the principle of total substitutability may not be entirely realistic, and the dual-gap model remains a reasonable alternative.
Within this framework and under other approaches, foreign aid should therefore contribute significantly to the economic growth of recipient countries. Accordingly, the independent variable representing the statistical expression of foreign aid received (ODA, as defined in Section 3), included in the empirical models, is expected to exhibit a positive and significant coefficient.
It should be noted that certain aspects of the political economy of aid could lead to a different outcome. On the donor side, geopolitical, commercial and bureaucratic interests may take precedence over the original goal of promoting socio-economic development. On the recipient side, aid can be used to reinforce patronage networks and internal elites may influence how aid is allocated. Ultimately, the domestic political framework determines whether aid reinforces or undermines development perspectives.
The issue of the sectoral allocation of aid and its strategic rationale may also be of paramount importance. Theoretically, the debate balances growth-oriented allocation, which involves investing in productive capacity and infrastructure to yield high economic returns and self-sustaining development, against direct social-sector spending, such as on health or education, which targets immediate poverty reduction but may not address underlying structural constraints. Modern frameworks also highlight the issue of fragmentation, whereby donors may disperse their funding across a proliferation of projects in their own preferred sectors without any external coordination. The lack of alignment between sectoral choices and partner-country strategies, coupled with the absence of donor specialisation in areas where they have a comparative advantage, may contribute to outcomes that are poorer than expected.
Moreover, according to the Keynesian-inspired conventional approach to aid, by alleviating the problem of lack of domestic savings in low-income countries, aid should contribute to increasing convergence of recipient income levels. This is consistent with the assumption that, given the right conditions (i.e., a sufficient endowment of usable resources and the absence of major impediments to industrialisation), less developed countries should grow faster than more developed countries, with the former increasingly converging with the latter. This hypothesis is based on the idea that poorer countries have advantages that the early industrialisers did not, in that they can now draw on capital, skills and technologies that have already been developed and deemed more productive by advanced countries, without having to bear the costs of research. As low-income countries adopt the more productive technologies developed by the leading countries, a convergence towards the technological frontier can be expected.
The inclusion of the initial GDP per capita variable (Ln_IniGDPpc) in the empirical model thus reflects the assumption that countries in earlier stages of economic development will grow faster than those in more advanced stages, especially if they receive foreign aid to overcome the lack of domestic savings. A negative sign for the respective coefficient is therefore to be expected. Furthermore, including the interactive term between aid and initial GDP per capita (ODA.Ln_IniGDPpc) reflects the possibility that aid’s effects might differ according to recipients’ income levels, with greater effects expected in low-income countries and smaller ones in upper middle-income countries.
The inclusion of the initial life expectancy variable (IniLE), alongside the convergence hypothesis, expresses the assumption that initial development is reflected not only in GDP. In fact, some countries have a particularly high GDP per capita—namely, certain commodity producers that benefit from favourable international price trends—but these countries have modest levels of human development. In such cases, the life expectancy indicator is expected to capture constraints on development that are not reflected in GDP. The same reasoning applies to the inclusion of the initial child mortality rate variable (IniCM), while the consideration of the interactive term between aid and initial child mortality (ODA.Ln_IniCM) expresses the hypothesis that the effects of aid might differ according to the recipients’ health situation, with greater effects expected in countries in a more precarious situation.
The inclusion of the variable expressing the weight of exports in the recipient’s economy (Expo) in the empirical model is intended to evaluate whether countries with a greater focus on foreign trade grow or develop at a different rate to those with a greater focus on domestic markets. As previously stated, the dual-gap approach suggests that in order to close the external gap, it is necessary to increase recipients’ foreign exchange earnings by promoting the export of goods and services. However, it is important to note that the positive outcomes of this variable may be hindered by the occurrence of what is termed ‘Dutch disease’, which postulates that large and sustained inflows of foreign exchange (whether by aid or by exports) can appreciate the real exchange rate, thereby reducing the competitiveness of a country’s tradable sectors, such as agriculture or manufacturing. It is posited that expenditure financed by aid on non-tradable goods and services, such as construction or local services, has the potential to elevate the prices of such goods and services. This, in turn, has the effect of diverting resources away from the export sector.
In theoretical terms, even in scenarios where total output experiences a temporary increase, the loss of export dynamism has the potential to adversely impact long-term productivity and structural transformation. Consequently, aid-effectiveness frameworks emphasise meticulous macroeconomic management, delivering aid in a manner that mitigates inflationary pressures. The incorporation of the interactive term between aid and exports (ODA.Expo) within the empirical models aims to underscore the potential for aid to function in a manner contingent on the degree of openness exhibited by the recipient nations, as well as on the availability of foreign exchange. The potential repercussions of Dutch disease may be amplified, or otherwise mitigated, by the inflows of aid.
The incorporation of variables reflecting the absence of corruption (Corr) and the level of democracy (Dem) in the recipient countries into the empirical models serves to underscore the significance of the institutional framework in facilitating growth and development, given its potential to engender an orderly system and to mitigate the level of uncertainty in trade, thereby reducing transaction costs through the diffusion of information. Furthermore, the inclusion of the interactive terms of these variables with aid (ODA.Corr and ODA.Dem) aims to express the possibility that the effectiveness of such aid might depend on the recipient’s institutional context.
In effect, the governance theory posits that the quality of institutions, reflected in factors such as the rule of law, accountability, and control of corruption, exerts a decisive influence on the impact of aid. The capacity of aid to generate growth is contingent upon the robustness of the policies and institutions in place; in the context of deficient governance, there is a risk that aid may be appropriated by elites or become a vehicle for the distortion of incentives. This line of thought underscores selectivity, giving more aid to countries with good governance, conditionality, requiring governance reforms, and a shift toward using aid to improve governance itself; for example, through budget support, dialogue or technical assistance on public finance. Furthermore, the absorptive capacity of beneficiaries and the effective utilisation of capital is a salient concern, given the inherent limit to the productive deployment of aid by a recipient nation prior to the collapse of the marginal return. The concept is rooted in the existence of constraints such as limited administrative skills, weak public financial management, infrastructure bottlenecks, and a shortage of skilled workers. This means that aid beyond a certain threshold can result in diminishing returns, inflation, or waste. In the context of effectiveness frameworks, absorptive capacity is a concept that underlies the gradual scaling up of aid, with the volumes of assistance being aligned with the country-owned systems as opposed to the imposition of a fixed allocation.
It can be hypothesised that nations exhibiting reduced corruption and politically induced economic distortions are likely to possess superior economic and social governance infrastructures. Consequently, a positive sign in the empirical model would have been anticipated for the variables corruption (Corr) and democracy (Dem). This assumption is supported by the recent empirical literature, which suggests that economic growth is inversely associated with the level of corruption (e.g., Gründler & Potrafke [37]) and is positively related to the level of democracy (e.g., Kabir & Alam [38]).
Conversely, events such as conflicts can negatively impact the functioning of institutions and economies. Therefore, it is advisable to include a variable such as the number of battle-related deaths (BRD) in the empirical models. The additional consideration of the interactive term with aid (ODA.BRD) reflects the possibility that the impact of aid might vary depending on the security situation of the recipients.
With regard to theoretical frameworks that specifically consider the relationship between foreign aid and child mortality, the main focus is on the fact that the effectiveness of aid depends on the interaction of various factors, such as whether aid is specifically allocated to the health sector, how recipients adjust their public health spending in response, and the capacity of local institutions to deliver services.
The relationship between aid and domestic public health spending highlights the issue of the political economy of aid, specifically how governments respond to external funding. The problem of fungibility arises as recipients may use aid to replace their own planned health expenditure, thereby redirecting domestic funds to other, often less productive, areas of the budget. This is a key reason why health sector aid is often perceived as failing to deliver positive health outcomes.
The effectiveness of health aid is also considered to be influenced by governance, particularly poor institutional quality and the prevalence of socio-political instability. It has also been suggested that fungible aid is more likely to be used efficiently in democracies. Therefore, health aid may be effective when a country’s political and institutional systems provide the appropriate incentives for public funds to be spent effectively on health services. However, even when health aid is targeted and governments are willing to spend, results depend on absorptive capacity—that is, the maximum amount of aid that a country’s health system can productively use. This capacity is largely determined by the availability of basic infrastructure and administrative capabilities. This concept may explain why massive inflows of aid do not automatically translate into better health, but instead may lead to waste and minimal impact. Improving child mortality may be seen as primarily dependent on a country’s institutions’ ability to deliver services.
The findings of empirical research concerning the effectiveness of aid in reducing child mortality are not entirely straightforward. For instance, Mishra & Newhouse [39] conclude that, although the estimated effect is negligible, health aid nevertheless has a beneficial and statistically significant effect on infant mortality. This is evidenced by the fact that a doubling of per capita health aid is associated with a 2 per cent reduction in the infant mortality rate. However, studies such as Wilson [40] demonstrate that aid for health has no effect on general mortality, as well as other types of aid, such as water development.

3. Methodology

This section provides a quantitative analysis of the impact of aid, based on an evaluation of the theoretical and empirical literature identified in previous sections.
The dependent variables considered are the average GDP per capita growth rate (GrGDP) and the average annual child mortality rate decline (DCM). Another interesting variable to consider would be the variation in the Human Development Index as defined by the United Nations; however, the absence of data for many recipients during the period covered by this study makes this possibility unfeasible.
In the baseline models (corresponding to models 1 and 2 in Table 1 and Table 2, respectively), the main independent variable of interest is the amount of aid received by each beneficiary (ODA). The remaining independent control variables, proposed according to the above-detailed theoretical framework, are the following: initial GDP per capita, log transformed (Ln_IniGDPpc), initial life expectancy (IniLE), initial rate of child mortality (IniCM), number of battle-related deaths per million inhabitants (BRD), absence of corruption perceptions index (Corr), democracy index (Dem), and the weighting of exports in the economy as a share of GDP (Expo).
Table 1. Assessment of the macro effects of aid on growth (GrGDP).
Table 2. Assessment of the macro effects of aid on child mortality (DCM).
The following models are therefore considered, where i is the recipient country and t is the period under consideration (sub-periods of the range 1990–2015):
GrGDPit = β1 ODAit + β2 Ln IniGDPpcit + β3 IniLEit + β4 BRDit + β5 Corrit + β6 
Demit + β7 Expoit + β8 GrGDPi,t−1 + εit
DCMit = β1 ODAit + β2 Ln IniGDPpcit + β3 IniLEit + β4 IniCMit + β5 BRDit + β6 
Corrit + β7 Demit + β8 Expoit + β9 DCMi,t−1 + εit
Furthermore, as stated in Section 1, it is also pertinent to evaluate the potential impact of ODA by examining its interaction with specific social, political and economic indicators of recipient countries that have not been taken into consideration in previous studies.
Therefore, in addition to the variables already included in Models 1 and 2 above, the specifications below (1a and 2a) include interactive terms between aid and other independent variables (represented as ODA·X); namely, the levels of initial income (ODA·Ln_IniGDPpc), initial child mortality (ODA·IniCM), battle-related deaths (ODA·BRD), absence of corruption (ODA·Corr), democracy (ODA·Dem), and weight of exports (ODA·Expo). The contribution of each of these terms is analysed separately in models 1a.1 to 1a.5 in Table 1, and 2a.1 to 2a.5 in Table 2.
GrGDPit = β1 ODAit + β2 Ln IniGDPpcit + β3 IniLEit + β4 BRDit + β5 Corrit + β6 
Demit + β7 Expoit + β8 GrGDPi,t−1 + β9 (ODAit·Xit) + εit
DCMit = β1 ODAit + β2 Ln IniGDPpcit + β3 IniLEit + β4 IniCMit + β5 BRDit + β6 
Corrit + β7 Demit + β8 Expoit + β9 DCMi,t−1 + β10 (ODAit·Xit) + εit
All of the above models employ a GMM instrumental variables approach (see Arellano & Bond [41]; Blundell & Bond [42]) to address the issue of ODA endogeneity, which can be attributed to an inverse causal relationship; that is to say, countries in a difficult economic or social situation are more likely to benefit from increased aid, which in turn generates an unanticipated negative or neutral relationship between aid and growth or social progress. In addition, the GMM enables efficient estimation in the presence of heteroskedasticity. This method uses assumptions (moment conditions) on specific moments of the variables rather than assumptions on the entire distribution. The number of moment conditions is greater than the number of parameters, providing a direct way to evaluate the specified models.
The estimates presented are based on a GMM dynamic panel data analysis, using averages for each of the five sub-periods considered (1990–1994, 1995–1999, 2000–2004, 2005–2009 and 2010–2015). Correction has also been applied for recipient country (cross-sectional) fixed effects (Arellano–Bond type), using the first-difference transformation method. This methodology is particularly well-suited to panel data such as that used in this study, which contains a large number of cross-sections (132 recipient countries) and a small number of time periods. However, it should be noted that panel data with many short sub-periods should be avoided, as these are prone to high short-term volatility in both the dependent variable and the measurement of ODA. Additionally, the chosen specification employs the lag of the dependent variable (GrGDPi,t−1 and DCMi,t−1) as regressors. The instruments include lags that vary from observation to observation, meaning their number increases dynamically over time as more predetermined variables become available.
The analysis is based on data from 1990 to 2015, reflecting the start and end dates of the Millennium Development Goals (formerly the SDGs). This period is particularly relevant for the issue of aid effectiveness, as the United Nations defined the pursuit of certain quantified goals by 2015 based on 1990 data.
However, of the 132 recipients examined (see Appendix A), there are some cases where it is not possible to consider the entire analysis period, mainly because some countries only existed as statistically independent entities for part of the period or were added to the list of recipients later. Data for the following countries start in the respective years: Armenia, Azerbaijan, Georgia, and Micronesia (1991); Kyrgyzstan, Tajikistan, Turkmenistan, and Uzbekistan (1992); Eritrea, North Macedonia, and South Africa (1993); Serbia (1994); Moldova (1997); Timor-Leste (1999); Montenegro (2003); Belarus and Ukraine (2005). The descriptive statistics and data sources for the variables included in the models are presented in detail in Appendix B. The correlation matrix between regressors can be found in Appendix C.
The dependent variable concerning the average growth rate of GDP per capita (GrGDP) is measured at constant prices. The other dependent variable (DCM, or average annual decline in child mortality rate, per 1000) is defined as the probability of dying between the ages of one and four.
The main independent variable of interest (the amount of ODA received by each beneficiary) is measured by its ratio to the recipient’s GDP. The category considered is net ODA (disbursements less repayments and recoveries). According to the OECD [43], the concept of ODA in force during the period of this study was as follows: “Official development assistance is defined as those flows to countries and territories on the DAC List of ODA recipients and to multilateral development institutions which are: (i) provided by official agencies, including state and local governments, or by their executive agencies; and, (ii) each transaction of which: (a) is administered with the promotion of the economic development and welfare of developing countries as its main objective; (b) is concessional in character and conveys a grant element of at least 25 per cent (calculated at a rate of discount of 10 per cent)”.
The level of initial GDP per capita in each sub-period (measured in constant US dollars) is expressed through a logarithmic transformation (Ln_IniGDPpc), due to its perceived exponential relationship with the dependent variables. Initial life expectancy (IniLE) is measured in years.
The number of battle-related deaths (BRD) variable, published by the Uppsala Conflict Data Programme, aims to express the disruption to society and economic activity resulting from conflict situations and is measured per million inhabitants.
The absence of corruption variable (Corr) is indicative of the perception of experts and businesspeople in each country regarding the level of corruption in the public sector, as reported by Transparency International. Normalised figures have been employed due to the different index scales used during the analysis period.
The democracy index (Dem), as published by the Centre for Systemic Peace, comprises a scale in which a complete autocracy is classified with the minimum index of that scale (−10). This is the index that corresponds to a system that restricts the political participation of citizens, selects the executive agents within the political elite and operates with few institutional constraints. Conversely, a full democracy is classified with the maximum index of the scale (+10), which is defined as a system composed of institutions in which citizens are able to express their preferences, which contemplates limits to executive powers and guarantees of individual freedoms.
Finally, the weight of exports in the economy (Expo) is measured by the ratio of exports to the GDP of the recipient countries.
As mentioned above, the empirical model proposed in this study uses a GMM instrumental variables approach to address the issue of ODA endogeneity. The validity of these instrumental variables must therefore be demonstrated through statistical tests, as described in the ‘Robustness Checks’ subsection of Section 4, as well as through conceptual and logical analysis, as discussed in this section. In particular, it must be demonstrated that the selected instrumental variables are uncorrelated with the error term, while also being highly correlated with the endogenous independent variable (ODA), considering the impact of the other exogenous independent variables. In addition to having a direct impact on the endogenous ODA, the instrumental variables should not introduce the possibility of a direct effect on the dependent variables. The possibility of a reverse causal effect with the dependent variables should also be ruled out. To establish this requirement of non-correlation between the instrumental variables and the error term, this section discusses a theoretical argument on the determinants of aid allocation.
The choice of instrumental variables is based on a framework in which the amount of ODA received by a recipient is the result of the complex interactions between structural links connecting donors and recipients, the motivations of donors, and the specific characteristics of recipients. The chosen instrumental variables are log-transformed population (Ln_Pop), former colony status (UK, Fra, Spa, Por, Ita, and Oth), island status (Island), War on Terror marker (WT), and weighting of imports in world trade (Imp).
The inclusion of the log-transformed population variable (Ln_Pop) is based on the assumption that countries with smaller populations will receive more aid proportionally than countries with larger populations and thus have a higher ODA/GDP ratio. This assumption reflects the notion that donors can evaluate a country’s wealth and needs based, not on per capita indicators, but on global values. For instance, this position may be particularly pertinent in the context of aid to countries such as China and India—the world’s two most populous countries—which have received significantly lower levels of aid as a percentage of GDP compared to other countries with similar incomes. By imposing penalisations on the more populous beneficiaries, donors effectively project onto them a level of wealth and power derived from sheer size and perceived influence in the global economy and international relations. However, when assessed on purely ideological grounds, a citizen of a more populous country is no less deserving of aid than a citizen of a less populous country, despite the fact that collectively they belong to a more powerful entity. Furthermore, no direct relationship is demonstrated between population size and per capita economic growth or variation in child mortality during a given period. Similarly, no inverse relationship is evident between these dependent variables and average population size.
The former colony status marker (UK, Fra, Spa, Por, Ita, Oth) aims to reflect one of the most important systemic links between donors and recipients, implying six binary variables specified separately for countries with more historical expression at this level: the United Kingdom (37 former colonies), France (26), Spain (20), Portugal (7), Italy (3), and others (5). It is anticipated that the former colonies will receive a substantially greater allocation of aid from the former colonial power. However, the behaviour of other donors may be in opposition, thereby cancelling out or reversing the overall effect. Furthermore, no direct correlation can be anticipated between the status of each former colony and the evolution of the socio-economic indicators selected as dependent variables over the period under scrutiny.
The binary variable concerning the small island status of the recipient countries (Island) reflects the fact that the Millennium Development Goals addressed the special needs of small island developing states, given the higher costs that such a situation entails. While it can be hypothesised that island status may exert an influence on the level of GDP per capita or other social indicators in a given country, a direct correlation with economic growth or the variation in these social indicators over the period covered by this study is not evident.
The variable highlighting participants in the War on Terror (WT) initiative is intended to reflect the reorientation of geographical and sectoral priorities in the allocation of aid as a result of the increased preoccupation with security issues and also a number of cases of state-building. The countries included in this category are Afghanistan, Egypt, Iraq, and Pakistan. It is important to note that the purpose of this variable is not to identify countries particularly affected by conflict (a situation that could have a direct negative association with the dependent variables), but rather to highlight those that have benefited from particular amounts of aid related to security issues. Indeed, the inclusion of countries exhibiting markedly divergent levels of conflict within their respective territories is a salient feature of this marker. Consequently, no direct association with the dependent variables can be anticipated.
The weighting of imports in world trade (Imp) is intended to reflect the possibility that countries with greater importance in international trade flows may benefit from additional attention from donors. It is relevant to note that this variable functions primarily as an additional size indicator, rather than a performance indicator. It is also pertinent to acknowledge that it refers to an average level ratio, as opposed to a growth factor, and that a direct link to economic growth cannot be established. Indeed, while it would be reasonable to assume a relation between economic growth and the growth in the share of world imports of a certain country during the period analysed, the same cannot be said about the relation between economic growth and the average share in the world trade during that period. It is not reasonable to expect that countries with a higher (or lower) weight of imports in world trade would necessarily grow faster (or slower) than the rest. In fact, there are examples of countries that have achieved growth and social progress with very different weights in world trade.

4. Results and Discussion

The results for the models described in the preceding sections are presented in Table 1 and Table 2. The positive but statistically non-significant sign of ODA on growth in Table 1, in most of the models considered, is particularly noteworthy, despite the use of an instrumental variables GMM methodology to overcome the endogeneity problem associated with a possible inverse causal relationship between aid and economic growth.
This non-significant result for ODA is not in line with the positive effects of foreign aid foreseen in several theoretical frameworks. This suggests that some of the criticisms raised against these approaches may be justified. It is possible that attaching more importance to the volume of capital as an inducer of growth than to its qualitative aspects, particularly during the “big push” phase, did not produce the expected results in terms of the simultaneous development of different industrial sectors. This outcome could also be explained by a lack of absorptive capacity in the recipient countries due to issues such as a lack of skills or poor infrastructure. According to approaches that consider human capital to be the main driver of development, the fact that most of the aid was provided in the form of capital transfers rather than skill-enhancing technical cooperation would make the positive impact of ODA less likely.
It is also conceivable that the necessary careful allocation of resources has not been achieved, in the sense that, depending on the recipient’s prevailing gap (external or domestic), free domestic resources should be used to increase foreign exchange earnings and the productivity of imports or, alternatively, to promote domestic savings and the productivity of domestic resources. Additionally, the assumption that aid finances investment rather than consumption may not have been realised, as ODA may have replaced public savings rather than supplemented them. Nevertheless, it should be noted that certain neoclassical growth models assume aid can be used to finance current consumption. This would help to overcome the problem of diminishing returns on aid-financed investment.
The results for the interaction terms between aid and various socio-political or economic indicators (ODA.X) also show that the respective coefficients are not significant, except for model 1a.3. Indeed, in this model, the interaction term between aid and the absence of corruption (ODA.Corr) exhibits a positive and significant relationship with growth at the 5% level, which is in marked contrast to the results obtained for the other interaction coefficients of ODA. A one standard deviation increase in this interactive term is associated with an additional 4.9 percentage points in average GDP growth, a figure that might be considered relevant. The positive correlation between economic growth and aid is only evident when the level of corruption in a recipient country is at least one standard deviation more favourable than the recipients’ average, a situation that, in the last period analysed, encompassed approximately 17% of recipients.
These results suggest that ODA has been less effective in stimulating economic growth in countries with high levels of corruption. Conversely, there is no evidence that the effectiveness of aid varies according to the recipients’ income level, the frequency of armed conflicts, the level of democracy, or the importance of exports to the economy.
Furthermore, examining the reduced form equation in which ODA allocation (endogenous) is explained by all exogenous independent and instrumental variables, reveals that corruption is clearly not significant for aid allocation (coefficient of −0.004, with a p-value of 0.994). Having concluded that corruption is non-significant for aid allocation but positively significant for growth, it can be postulated that the lack of evidence of aid’s positive impact on growth could also be due to the fact that aid has not primarily been given to low-corruption countries, making it more difficult to realise its supposed macroeconomic impact. This conclusion is supported by the significant positive coefficient of the interaction term between aid and the absence of corruption (ODA.Corr).
It should also be noted that the positive and significant coefficient of ODA in model 1a.3 would suggest that ODA itself can have a positive impact on growth; however, this is not supported by most of the remaining specifications. This outlier result is due to the interactive variable introduced in this model (ODA.Corr) having a high negative correlation with ODA, which could lead to the latter’s coefficient yielding spurious results. To confirm the validity of the interaction between ODA and corruption, detailed robustness checks for this result are provided in the sub-section below.
The results of the present study do not confirm the conclusions of some empirical studies that aid has a positive impact on growth in all circumstances (e.g., Karras [4]), nor do they confirm the conclusion that aid has no positive impact on growth in any circumstances (e.g., Rajan & Subramanian [19]).
However, as already mentioned in Section 1, this article’s methodology differs from previous studies, which relied on analysing the interaction between aid and quality policy indices (involving subjective policy assessments) or composite institutional indices (masking the effects of recipients’ basic components). For example, Burnside & Dollar [1] found that the impact of aid on growth is greater in a favourable policy environment than in an unfavourable one, whereas Salisu & Ogwumike [10] found that a macroeconomic environment that is perceived as sound is essential for aid to contribute effectively to sustainable growth.
The direct effect of the absence of corruption (Corr) on growth is also positive and significant in most of the models specified. In the baseline model (1), an improvement in the level of corruption by one standard deviation is associated with an average increase in GDP growth of 3.1 percentage points. This finding suggests a potential association between countries with fewer economic distortions due to corruption, and higher levels of growth. It could be argued that corruption is not a causal factor but rather a consequence of the initial level of development of a country. However, the effects of such a possibility are covered by the inclusion in the models of the variables relating to the initial values of GDP per capita and life expectancy of the recipient countries.
The above conclusions regarding the corruption variable and its interaction with ODA lend weight to the institutionalist approach. According to this approach, the main factor in growth is the existence of an institutional framework that contributes to economic progress by providing an orderly system and reducing uncertainty and transaction costs in trade. Furthermore, this is consistent with recent empirical studies suggesting an inverse relationship between economic growth and the level of corruption (e.g., Gründler & Potrafke [37]).
With regard to the effects of the remaining control variables, it is noteworthy that the logarithmic transformation of the initial GDP per capita (Ln_IniGDPpc) has a significant negative sign, which appears to provide confirmation of the assumption made by various currents that, as low-income countries adopt the more productive technologies of the leading countries, those at earlier stages of development should grow faster than those at more advanced stages, and thus a convergence effect in terms of economic growth should occur. This finding is also consistent with much of the empirical literature (e.g., Burnside & Dollar [1]). However, it should be noted that the scope of this study is limited to low- and middle-income developing countries. Therefore, the results presented here suggest convergence within this category of countries, but not necessarily between them and high-income economies.
On the contrary, the initial value of life expectancy (IniLE) has a significant positive coefficient, thus indicating a situation of non-convergence. This can be tentatively explained by the fact that life expectancy may reflect favourable conditions for subsequent growth that are not captured by GDP. For example, some countries may have a high GDP but low human development, which is detrimental to future growth.
The variable reflecting the number of battle-related deaths (BRD) is only significant in some of the specifications, bearing the expected negative sign. The weak effect of this variable may be due to the time span covered by the periods considered, which may cancel out the effects of conflicts, especially when accompanied by substantial aid inflows.
With regard to the democracy index (Dem) variable, the coefficients obtained are all highly non-significant, implying that the proposed models do not empirically confirm a direct relationship with growth. Such a conclusion is not in line with some recent empirical literature (e.g., Kabir & Alam [38]), which suggests that growth is positively associated with the level of democracy. However, it could be argued that this discrepancy is related to the significant difference in time periods considered (the data used by Kabir & Alam [38] only go back to 2006, whereas this study considers data going back to 1990). It could also be argued that the non-significant result is due to the fact that more democratic countries tend to have higher initial income levels and are therefore more prone to lower growth rates. However, the correlation between the democracy index (Dem) and initial GDP per capita (Ln_IniGDPpc) does not exceed 0.28.
Ultimately, the explanation for the results concerning the democracy index may lie in the fact that the mere existence of formal democratic institutions (such as elections) does not guarantee economic growth; the critical factor may be state capacity. While a democracy can generate incentives for good policies, weak state capacity can result in the government failing to implement these policies effectively. In fragile countries, a combination of limited capacity, pressure for immediate redistribution and state capture by elites can result in a dysfunctional system.
It is important to note that the results obtained for the democracy variable differ significantly from those obtained for the corruption variable, despite both being considered generic components of institutional quality indicators from the outset. These results emphasise the importance of the methodology employed in this study, which prioritises indicators that capture the fundamental characteristics of recipient countries over more complex composite indicators that obscure the impact of their respective components.
With regard to the weight of exports in GDP (Expo), the obtained coefficients are positive and significant for all the specifications. An increase of one standard deviation in the weight of exports (equivalent to 19.3 percentage points of GDP) is associated with an additional 3.6 percentage points in terms of average GDP growth (Model 1). This result suggests that countries focused on foreign trade may have grown faster than those focused on domestic markets, lending some credibility to approaches that emphasise the importance of external demand and attribute growth differentials to the quantity of natural resources. Therefore, it can be argued that issues arising from Dutch disease or excessive variation in terms of trade, particularly relevant for countries that mainly export commodities and import manufactured products, have not been evident within the scope of the present study. However, this result contrasts with the assumption that many developing countries will face situations in which the export sector becomes too large in proportion to the overall economy, which, combined with an incipient industry, could lead to poorer growth outcomes.
With regard to the relation between ODA and child mortality, the evidence suggests a generally stronger association, with a positive sign, when compared to the findings for economic growth. However, it should be noted that the significance of this association varies according to the specific model under consideration. For instance, in regression 2 of Table 2, an increase of one percentage point in the average ODA as a share of GDP is associated with a reduction of 0.024 permille points in the annual average child mortality rate, which could be considered somewhat modest and may only be relevant for countries receiving substantial aid inflows. Note that the average annual decline in child mortality for the entire sample of recipients is 0.955 permille points, whereas the initial average rate is 26.4 permille. Furthermore, the interactive terms between aid and the various socio-political or economic indicators (ODA.X) do not show a significant positive sign of the respective coefficients. In contrast to the findings concerning growth, the impact of aid on child mortality does not appear to be amplified in countries with lower corruption levels.
Although few studies have analysed the effectiveness of aid on health indicators and there is a paucity of analysis of the impact of aid on human development, as Nwude et al. [27] and Yiheyis & Woldemariam [15] point out, it can be concluded that the results of the present study do not support the conclusions of Nwude et al. [27]. They found that ODA has a negative and significant effect on life expectancy and a positive and significant relationship with the under-5 mortality rate, which is in line with the pessimistic view of aid. However, caution should be exercised when comparing results, given that the analysis period in Nwude et al. [27] was significantly different (starting in 1999 instead of 1990) and the number of recipients covered was significantly lower (81 instead of 132).
Moreover, although somewhat limited in scope, other recent studies of social outcomes offer a different view. For instance, Yiheyis & Woldemariam [15] found that ODA has a positive, albeit modest, impact on the Human Development Index (HDI) of recipient countries (based on a sample of just 35 African countries), whereas Lee et al. [14] found that the impact of total aid on the HDI of recipient countries is statistically significant, although based on a sample of only 15 Asian countries. Previously, Lin Moe [6] determined that ODA exhibits a substantial positive correlation with human development within a limited sample of eight middle-income Asian countries.
The results for the other independent variables differ greatly from those for growth. Regarding initial GDP per capita (Ln_IniGDPpc) and initial life expectancy (IniLE), the high explanatory power of these variables is largely replaced by initial child mortality (IniCM), which is now the most significant factor. An increase in initial child mortality of one standard deviation (30.7 permille points) is associated with an additional average annual decrease in child mortality of 0.7 permille points. This result suggests a convergence mechanism within the category of developing countries, possibly arising from technology and knowledge diffusion. Similarly to the growth results, this conclusion appears to support the convergence hypothesis, but only in terms of convergence among recipient countries and not necessarily with advanced economies.
With regard to the remaining control variables, in contrast to the results obtained for economic growth, only the democracy index (Dem) generates a positive and significant association with child mortality. In the baseline model (2), an increase in the level of democracy by one standard deviation (5.9 points on a scale with a range of 20 points) is associated with an additional 0.3 permille points in the annual average reduction in child mortality. Therefore, it can be concluded that while an absence of corruption is important for growth, the democratic quality of recipient countries appears to be more closely linked to improvements in social indicators, such as child mortality, than to economic factors. This conclusion lends some support to the principles of the institutionalist approach, which holds that the quality of institutions is crucial for economic and social progress.
Additionally, it can be postulated that aid particularly exposed to fungibility, such as general budget support for health, is more likely to be used appropriately in democratic systems, where a country’s political and institutional structures may provide greater incentives for the effective allocation of resources to health services. This would avoid one of reasons why sectoral health aid often appears to fail to deliver positive health outcomes (i.e., fungibility).
It can thus be posited that the improvement of child mortality may be contingent, above all, on the capacity of a nation’s institutions to deliver services, a capacity that may be particularly fostered in a robust democratic environment. In contrast, a goal such as economic growth is likely to be influenced by a variety of factors that, at least in the short- and medium-term, may lie beyond the purview of the recipients’ institutions.
However, regression 2a.3 in Table 2 suggests that the lack of significance of the interaction term between aid and democracy (ODA.Dem) does not allow the assumptions above to be empirically corroborated. This conclusion is consistent with that of other studies, such as Wilson [40].

Robustness Checks

In order to confirm the conclusions drawn earlier in this section—namely, with regard to the main variables of interest (ODA and its interactive terms)—this subsection presents the results of four types of robustness checks: statistical tests (see also Table 1 and Table 2), alternative regression methods, alternate exclusion of explanatory variables, and consideration of the possible endogeneity of ODA interactive terms.
With regard to the verification of the instrumental variables used in this study (described in Section 3), it should be noted that an instrument is considered weak if it exhibits low correlation with the endogenous variable once the exogenous independent variables have been taken into account. The validity of the instrumental variables can be tested using the Cragg–Donald F-statistic, which is then assessed against the Stock–Yogo relative bias critical values. These values determine the maximum acceptable relative bias of the two-stage least squares estimator compared to OLS. The four thresholds identified for the present analysis correspond to biases of 5%, 10%, 20% and 30%, respectively. It is important to note that higher Cragg–Donald F-statistics are indicative of superior fits. The assessment is conducted in accordance with the criteria outlined in the work of Skeels & Windmeijer [44]. The Cragg–Donald F-statistics obtained for the period under consideration (1990–2015) are 12.35 for the economic growth model and 11.99 for the child mortality model, which are above the corresponding Stock–Yogo relative bias critical values of 10% (11.49). Consequently, the instrumental variables proposed above can be regarded as valid.
In addition, a J-test for over-identifying restrictions was conducted in order to assess whether all instruments are exogenous under the assumption that at least one of them is indeed exogenous. The purpose of the test is to ascertain whether the specified model contains sufficient restrictions that can be checked to ensure that they are all consistent. For a highly precise model, all time restrictions could be met simultaneously. Therefore, higher J-statistics (i.e., with lower probability values) are associated with a higher probability of rejecting the model. Additionally, an Arellano–Bond serial correlation test was conducted to evaluate the potential presence of serial correlation. A lower chance of serial correlation is associated with higher probability values of the corresponding m-statistic.
The results show that the estimations presented in Table 1 can be considered valid since the overidentification restrictions hypothesis is rejected. This is evidenced by the J-statistic p-values being above 0.1 in models 1 to 1a.3 and above 0.05 in models 1a.4 and 1a.5. Furthermore, the hypothesis of serial correlation is rejected based on the p-values for the m-statistic of the Arellano–Bond serial correlation tests being above 0.1. As mentioned earlier in this subsection, the validity of the instruments was also assessed using the Cragg–Donald F-statistic.
With regard to the estimates presented in Table 2, they also provide a clear rejection of the hypotheses of serial correlation. With respect to the hypotheses concerning over-identification of restrictions, models 2a.1, 2a.2 and 2a.4 pass the J-statistic test at the 5% level (p-value above 0.05), and the remaining models at the 10% level (p-value above 0.1).
In order to further confirm the robustness of the results, alternative estimations were carried out using a methodology other than the five-period panel data GMM (applied to the models in Table 1 and Table 2); namely, a cross-sectional regression with averages for the entire period 1990–2015. The coefficients of the main variables of interest are shown in Table 3.
Table 3. Coefficients of aid and its interactive term with corruption in an alternative model (GMM cross-sectional static regression, 1990–2015, with instrumental variables).
The results of this alternative specification, in which the negative correlation between ODA and ODA.Corr is weaker, confirm that ODA alone does not have a positive and significant association with growth (in fact, it has a significantly negative coefficient), while the interactive term ODA.Corr continues to reveal a significantly positive coefficient. Furthermore, the results for the ODA variable and its other interactive terms are broadly consistent with those obtained using the dynamic panel data methodology. Therefore, the results of this alternative specification are consistent and support the conclusions obtained with the original methodology.
Further confirmation of this result was obtained by removing the ODA variable (along with its associated instrumental variables) from the regression, while retaining only the ODA.Corr interactive term. The results still showed a positive and significant interaction effect between aid and corruption.
An additional robustness check was carried out by excluding various regressors alternately from the initial models (Table 1 and Table 2). The regressors excluded were: the democracy variable (Dem), which was expected to be particularly related to corruption; the variable expressing conflict-related deaths (BRD); and the variable expressing the weight of exports in the recipient economies (Expo). The results showed that none of these exclusions caused a meaningful change to the conclusions regarding the sign and significance of the main variables of interest, confirming the validity of the original models.
Finally, an analysis was carried out by rerunning the models based on the premise that the ODA interactive terms could also be endogenous. This exercise did not significantly change the conclusions regarding the behaviour of the main variables of interest in the present study.

5. Conclusions

This study assesses the potential macro effects of ODA on two socio-economic indicators of aid recipients: GDP per capita and child mortality. Its main contribution to the existing literature is its examination of the potential impact of ODA in conjunction with certain social, political and economic conditions in recipient countries, which were not singled out in previous studies.
The models use GMM analysis within an instrumental variables approach to address the issue of ODA endogeneity. The main independent variables of interest are the amount of ODA received by each recipient, expressed as a ratio of GDP, and its interaction terms with the other economic and socio-political explanatory variables.
The results for the period 1990–2015 show a positive, although mostly statistically non-significant, impact of ODA on growth. The interactive terms between aid and various socio-political or economic indicators also suggest that the respective coefficients are not significant, except for the interaction between aid and the absence of corruption. The non-significant sign of ODA does not support the hypothesis of a generally positive effect of aid. On the other hand, the significant interaction between aid and corruption suggests that ODA has been less effective in boosting economic growth in corrupt countries. A one standard deviation increase in this interactive term is associated with an additional 4.9 percentage points of average GDP growth. Conversely, there is no evidence that the effectiveness of aid for growth differs according to the recipient’s income level, the frequency of conflicts, the level of democracy, or the importance of exports to the economy.
The effects of ODA on child mortality appear to be more significant, with a positive sign, than those obtained for economic growth, albeit still modest. A one percentage point increase in ODA as a proportion of GDP is associated with a reduction in the child mortality rate of 0.024 permille point, while the interactive terms between aid and various socio-political or economic indicators do not show a significant positive effect. Unlike the results for growth, the impact of aid on child mortality does not appear to be greater in countries with lower levels of corruption.
Although the conclusions presented in this paper do not allow for the immediate rejection of the micro-macro paradox regarding the impact of aid on growth, they do provide a comprehensive understanding of the reasons that could lead to such contradictory results. Since aid is not primarily allocated to the least corrupt countries and aid is more effective for growth in the least corrupt recipient countries, its possible macro effects are more difficult to realise. This is an alternative explanation to the more straightforward idea that aid is simply ineffective.
The policy implications of the present study’s findings and subsequent recommendations for policymakers, authorities and researchers can be summarised as follows:
−
The conclusion that the level of corruption of recipients is clearly non-significant for the allocation of aid (obtained by examining the reduced form equation of ODA allocation) is particularly surprising, and the reasons that led to this situation should be duly considered by the authorities responsible for the delivery of aid;
−
Having concluded that foreign aid does not generally have a significant positive association with economic growth, except in situations where it is associated with lower levels of corruption, it is therefore advisable that, in order to maximise its impact on recipient growth, aid should be prioritised to less corrupt countries, thereby improving its potential macroeconomic effectiveness;
−
It should be noted that the conclusion in the previous paragraph does not necessarily imply that donors should not provide aid to countries with higher levels of corruption, but rather that instead of focusing initially on large-scale aid, the reproductive effect of which in terms of economic growth may not be guaranteed due to the prevailing institutional framework, donors should instead focus in an initial phase on activities and programmes that contribute to improving the level of governance and the institutional framework of the recipient countries (i.e., reducing the risk of corruption), with the aim of creating the conditions for future growth;
−
The implementation of the recommendations mentioned in the previous paragraphs necessarily implies a change in the behaviour of donors with regard to their criteria for the allocation of development aid, since, as already mentioned, at an aggregate level they have not taken into account the level of corruption of the recipients.
However, it should be noted that the present study has certain limitations, mainly due to the fact that the number of observations is not particularly high, due to the combination of the number of beneficiaries studied (132) and the small number of periods analysed. Nevertheless, this is not an easy problem to solve, since, in addition to the lack of detailed data for earlier periods, it is not advisable to increase the number of observations by using panel data with many short sub-periods, as they are both prone to high short-term volatility in the dependent variable and in the measurement of ODA, which in turn could significantly affect any conclusions. An additional problem to be considered is the very limited statistical capacity of many developing countries and their lack of accuracy in producing data such as GDP or mortality statistics.
A future line of research could be to examine the impact of specific sub-categories of ODA, such as institutional technical cooperation versus project aid, or general budget support versus specific sectoral support. However, the use of such aggregates is also limited by the lack of detailed statistical data for all recipients and for the whole period considered in this study, as well as by the fact that a significant proportion of aid is reported as general aid when in fact it is allocated to specific sectors or activities. Another problem to be considered is the issue of fungibility of aid, in the sense that the granting of aid to a specific sector may not necessarily imply an increase in the recipient’s expenditure in that sector, but rather the maintenance of the originally planned expenditure and the diversion of the additional amounts to other sectors. Another line of future research could be to further explore the existence of a possible non-linear relationship between aid and socio-economic indicators (as in Moreira [3], Clemens et al. [21], or Wagner [45]), under the premise that there is reason to expect that too much aid is negative for growth due to the limited absorptive capacity of recipients (as predicted, for example, by some critics of the dual-gap approach), which accordingly raises the possibility of diminishing returns to aid.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/world7060089/s1.

Author Contributions

Conceptualization, P.F., S.B.M. and J.C.; Methodology, P.F., S.B.M. and J.C.; Software, P.F., S.B.M. and J.C.; Validation, P.F., S.B.M. and J.C.; Formal analysis, P.F., S.B.M. and J.C.; Investigation, P.F., S.B.M. and J.C.; Resources, P.F., S.B.M. and J.C.; Data curation, P.F., S.B.M. and J.C.; Writing—original draft, P.F., S.B.M. and J.C.; Writing—review & editing, P.F., S.B.M. and J.C.; Visualization, P.F., S.B.M. and J.C.; Supervision, S.B.M. and J.C.; Project administration, P.F., S.B.M. and J.C.; Funding acquisition, P.F., S.B.M. and J.C. All authors have read and agreed to the published version of the manuscript.

Funding

Jorge Caiado was supported by national funds through FCT—Fundação para a Ciência e a Tecnologia, I.P., in the framework of the unit UID/06522/2025.

Data Availability Statement

The data used in this study is available as Supplementary Material. These data were based on publicly available datasets, which are described in Appendix B.

Acknowledgments

Sandrina. B. Moreira gratefully acknowledges MARE–Marine and Environmental Sciences Centre/ARNET–Aquatic Research Network for its support as research community. AI-based tools were used in a limited capacity for language refinement; the authors reviewed all outputs and take full responsibility for the manuscript.

Conflicts of Interest

The authors have no relevant financial or non-financial interests to disclose. The authors have no competing interests to declare that are relevant to the content of this article.

Appendix A. List of 132 Developing Countries and Territories Included in the Research

AfghanistanGabonNiger
AlbaniaGambiaNigeria
AlgeriaGeorgiaPakistan
AngolaGhanaPanama
Antigua and BarbudaGrenadaPapua New Guinea
ArgentinaGuatemalaParaguay
ArmeniaGuineaPeru
AzerbaijanGuinea-BissauPhilippines
BangladeshGuyanaRwanda
BelarusHaitiSaint Lucia
BelizeHondurasSaint Vincent and the Grenadines
BeninIndiaSamoa
BhutanIndonesiaSao Tome and Principe
BoliviaIranSenegal
Bosnia and HerzegovinaIraqSerbia
BotswanaJamaicaSeychelles
BrazilJordanSierra Leone
Burkina FasoKazakhstanSolomon Islands
BurundiKenyaSomalia
Cabo VerdeKiribatiSouth Africa
CambodiaKyrgyzstanSri Lanka
CameroonLao People’s Democratic RepublicSudan
Central African RepublicLebanonSuriname
ChadLesothoSwaziland/Eswatini
ChileLiberiaSyrian Arab Republic
China (People’s Republic of)LibyaTajikistan
ColombiaMadagascarTanzania
ComorosMalawiThailand
CongoMalaysiaTimor-Leste
Costa RicaMaldivesTogo
Côte d’IvoireMaliTonga
CubaMauritaniaTunisia
Democratic Republic of the CongoMauritiusTurkey
DjiboutiMexicoTurkmenistan
DominicaMicronesiaUganda
Dominican RepublicMoldovaUkraine
EcuadorMongoliaUruguay
EgyptMontenegroUzbekistan
El SalvadorMoroccoVanuatu
Equatorial GuineaMozambiqueVenezuela
EritreaMyanmarViet Nam
EthiopiaNamibiaYemen
FijiNepalZambia
FY Rep. Macedonia/North MacedoniaNicaraguaZimbabwe

Appendix B

Table A1. Descriptive Statistics for the Period of 1990–2015 and Sources of Data.

Appendix C

Table A2. Matrix of Correlations Between Regressors.

References

  1. Burnside, C.; Dollar, D. Aid, Polices, and Growth. Am. Econ. Rev. 2000, 90, 847–868. [Google Scholar] [CrossRef] [Scilit]
  2. Burnside, C.; Dollar, D. Aid, Policies, and Growth: Revisiting the Evidence. In World Bank Policy Research Working Paper 3251; Word Bank Group: Washington, DC, USA, 2004; Available online: http://www-wds.worldbank.org/external/default/WDSC...d/PDF/wps3251Aid.pdf (accessed on 1 January 2022).
  3. Moreira, S. Evaluating the Impact of Foreign Aid on Economic Growth: A Cross-Country Study. J. Econ. Dev. 2005, 30, 2. [Google Scholar]
  4. Karras, G. Foreign aid and long-run economic growth: Empirical evidence for a panel of developing countries. J. Int. Dev. 2006, 18, 15–28. [Google Scholar] [CrossRef] [Scilit]
  5. Bobba, M. Aid and Growth: Politics Matters. In Inter-American Development Bank (January); EconPapers: Örebro, Sweden, 2007. [Google Scholar]
  6. Lin Moe, T. An empirical investigation of relationships between official development assistance (ODA) and human and educational development. Int. J. Soc. Econ. 2008, 35, 202–221. [Google Scholar] [CrossRef] [Scilit]
  7. Arndt, C.; Jones, S.; Tarp, F. Aid, Growth, and Development: Have We Come Full Circle? J. Glob. Dev. 2010, 1, 5. [Google Scholar] [CrossRef] [Scilit]
  8. Kilby, C.; Dreher, A. The impact of aid on growth revisited: Do donor motives matter? Econ. Lett. 2010, 107, 338–340. [Google Scholar] [CrossRef] [Scilit]
  9. Minoiu, C.; Reddy, S.G. Development aid and economic growth: A positive long-run relation. Q. Rev. Econ. Financ. 2010, 50, 27–39. [Google Scholar] [CrossRef] [Scilit]
  10. Salisu, A.A.; Ogwumike, F.O. Aid-Macroeconomic Policy Environment and Growth: Evidence from Sub-Saharan Africa. Pak. J. Appl. Econ. 2010, 20, 1–12. Available online: https://ideas.repec.org/a/pje/journl/article10iii.html (accessed on 1 January 2023). [CrossRef] [Scilit]
  11. Byron, S. Examining Foreign Aid Fungibility in Small Open Economies. Open Econ. Rev. 2011, 23, 675–712. [Google Scholar] [CrossRef] [Scilit]
  12. De Matteis, A. Relevance of poverty and governance for aid allocation. Rev. Dev. Financ. 2013, 3, 51–60. [Google Scholar] [CrossRef] [Scilit]
  13. Minasyan, A.; Nunnenkamp, P.; Richert, K. Does Aid Effectiveness Depend on the Quality of Donors? World Dev. 2017, 100, 16–30. [Google Scholar] [CrossRef] [Scilit]
  14. Lee, E.; Jung, K.; Sul, J. Searching for the Various Effects of Subprograms in Official Development Assistance on Human Development across 15 Asian Countries: Panel Regression and Fuzzy Set Approaches. Sustainability 2019, 11, 1152. [Google Scholar] [CrossRef] [Scilit]
  15. Yiheyis, Z.; Woldemariam, K. Remittances, Official Development Assistance, and Human Development in Africa: An Empirical Analysis. J. Afr. Dev. 2020, 21, 189–212. [Google Scholar] [CrossRef] [Scilit]
  16. Easterly, W. Can Foreign Aid Buy Growth? J. Econ. Perspect. 2003, 17, 23–48. [Google Scholar] [CrossRef] [Scilit]
  17. Dalgaard, C.J.; Hansen, H.; Tarp, F. On the empirics of foreign aid and growth. Econ. J. 2004, 114, F191–F216. [Google Scholar] [CrossRef] [Scilit]
  18. Rajan, R.G.; Subramanian, A. What Undermines Aid’s Impact on Growth? Exch. Organ. Behav. Teach. J. 2005, 126, 54. [Google Scholar]
  19. Rajan, R.G.; Subramanian, A. Aid and Growth: What Does the Cross-Country Evidence Really Show? Rev. Econ. Stat. 2008, 90, 643–665. [Google Scholar] [CrossRef] [Scilit]
  20. Stockemer, D.; LaMontagne, B.; Charette, J. Panacea, placebo, or poison? The impact of development aid on growth. Can. J. Dev. Stud. 2011, 32, 3–16. [Google Scholar] [CrossRef] [Scilit]
  21. Clemens, M.A.; Radelet, S.; Bhavnani, R.R.; Bazzi, S. Counting Chickens when they Hatch: Timing and the Effects of Aid on Growth. Econ. J. 2012, 122, 590–617. [Google Scholar] [CrossRef] [Scilit]
  22. Jin, S.K.; Oh, C.H. Revisiting effects and strategies of official development assistance (ODA): A panel analysis. Int. Rev. Public Adm. 2012, 17, 139–167. [Google Scholar]
  23. Benmamoun, M.; Lehnert, K. Financing Growth: Comparing the Effects of FDI, ODA, and International Remittances. J. Econ. Dev. 2013, 38, 43–65. [Google Scholar] [CrossRef] [Scilit]
  24. Driffield, N.; Jones, C. Impact of FDI, ODA and Migrant Remittances on Economic Growth in Developing Countries: A Systems Approach. Eur. J. Dev. Res. 2013, 25, 173–196. [Google Scholar] [CrossRef] [Scilit]
  25. Lessmann, C.; Markwardt, G. Aid, Growth, and Devolution: Considering Aid Modality and Different Types of Decentralization. World Dev. 2016, 84, 118–130. [Google Scholar] [CrossRef] [Scilit]
  26. Wako, H.A. Aid, institutions and economic growth in sub-Saharan Africa: Heterogeneous donors and heterogeneous responses. Rev. Dev. Econ. 2017, 22, 23–44. [Google Scholar] [CrossRef] [Scilit]
  27. Nwude, E.C.; Ugwoke, R.O.; Uruakpa, P.C.; Ugwuegbe, U.S.; Nwonye, N.G.; McMillan, D. Official development assistance, income per capita and health outcomes in developing countries: Is Africa different? Cogent Econ. Financ. 2020, 8, 1774970. [Google Scholar] [CrossRef] [Scilit]
  28. Avom, D.; Kamguia, B.; Njangang, H. Understand growth episodes in Sub-Saharan Africa: Do exogenous shocks matters? J. Int. Trade Econ. Dev. 2021, 30, 596–624. [Google Scholar] [CrossRef] [Scilit]
  29. Dalgaard, C.-J.; Hansen, H. The Return to Foreign Aid. J. Dev. Stud. 2017, 53, 998–1018. [Google Scholar] [CrossRef] [Scilit]
  30. Mosley, P. Aid-effectiveness: The Micro-Macro Paradox. IDS Bull. 1986, 17, 22–27. [Google Scholar] [CrossRef] [Scilit]
  31. Riddell, R.C. Foreign Aid Reconsidered; James Currey/Overseas Development Institute: Oxford, UK, 1987. [Google Scholar]
  32. Thorbecke, E. The evolution of the development doctrine and the role of foreign aid, 1950–2000. In Foreign Aid and Development—Lessons Learnt and Directions for the Future; Tarp, F., Ed.; Routledge: London, UK, 2000; pp. 12–35. [Google Scholar]
  33. Chenery, H.B.; Strout, A.M. Foreign Assistance and Economic Development. Am. Econ. Rev. 1966, 56, 679–733. [Google Scholar]
  34. Ghatak, S. Introduction to Development Economics; Psychology Press: London, UK, 2010. [Google Scholar]
  35. Degnbol-Martinussen, J.; Engberg-Pedersen, P. Aid: Understanding International Development Cooperation; Zed Books: London, UK, 2005. [Google Scholar]
  36. Bird, G. World Finance and Adjustment: An Agenda for Reform; Springer: Berlin/Heidelberg, Germany, 1985. [Google Scholar]
  37. Gründler, K.; Potrafke, N. Corruption and economic growth: New empirical evidence. Eur. J. Political Econ. 2019, 60, 101810. [Google Scholar] [CrossRef] [Scilit]
  38. Kabir, M.A.; Alam, N. The Efficacy of Democracy and Freedom in Fostering Economic Growth. Emerg. Econ. Stud. 2021, 7, 76–93. [Google Scholar] [CrossRef] [Scilit]
  39. Mishra, P.; Newhouse, D. Does health aid matter? J. Health Econ. 2009, 28, 855–872. [Google Scholar] [CrossRef] [Scilit]
  40. Wilson, S.E. Chasing Success: Health Sector Aid and Mortality. World Dev. 2011, 39, 2032–2043. [Google Scholar] [CrossRef] [Scilit]
  41. Arellano, M.; Bond, S. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 1991, 58, 277–297. [Google Scholar] [CrossRef] [Scilit]
  42. Blundell, R.; Bond, S. Initial conditions and moment restrictions in dynamic panel data models. J. Econom. 1998, 87, 115–143. [Google Scholar] [CrossRef] [Scilit]
  43. OECD. Converged Statistical Reporting Directives for the Creditor Reporting System (CRS) and the Annual DAC Questionnaire—Chapters 1–6. DAC; OECD: Paris, France, 2013. [Google Scholar]
  44. Skeels, C.L.; Windmeijer, F. On the Stock-Yogo Tables. Econometrics 2018, 6, 44. [Google Scholar] [CrossRef] [Scilit]
  45. Wagner, L. Identifying thresholds in aid effectiveness. Rev. World Econ. 2014, 150, 619–638. [Google Scholar] [CrossRef] [Scilit]
  46. Allansson, M.; Melander, E.; Themnér, L. Organized violence, 1989–2016. J. Peace Res. 2017, 54, 574–587. [Google Scholar] [CrossRef] [Scilit]
  47. Gleditsch Petter, N.; Wallensteen, P.; Eriksson, M.; Sollenberg, M.; Strand, H. Armed Conflict 1946–2001: A New Dataset. J. Peace Res. 2002, 39, 615–637. [Google Scholar] [CrossRef] [Scilit]
  48. Transparency International. Corruption Perceptions Index 2022; Transparency International: Berlin, Germany, 2023. [Google Scholar]
  49. Center for Systemic Peace. Integrated Network for Societal Conflict Research. Available online: http://www.systemicpeace.org/inscrdata.html (accessed on 1 January 2018).
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.