Does NGO Aid Go to the Poor? Empirical Evidence from Europe
  • 1, International Monetary Fund
  • | 2, International Monetary Fund

Contributor Notes

Author(s) E-Mail Address:,

This paper studies the aid allocation of European nongovernmental organizations (NGOs). Once population is controlled for, poverty consistently appears as the main worldwide determinant of NGO aid allocation. NGOs do not respond to strategic considerations. Their funding source does not seem to exert a great influence on their aid allocation decision. We also find differences across regions. Militarization and the political nature of the regime of the recipient country affect aid allocation in the Middle East. Life expectancy influences aid allocation in countries in the Western Hemisphere and the Middle East.


This paper studies the aid allocation of European nongovernmental organizations (NGOs). Once population is controlled for, poverty consistently appears as the main worldwide determinant of NGO aid allocation. NGOs do not respond to strategic considerations. Their funding source does not seem to exert a great influence on their aid allocation decision. We also find differences across regions. Militarization and the political nature of the regime of the recipient country affect aid allocation in the Middle East. Life expectancy influences aid allocation in countries in the Western Hemisphere and the Middle East.

I. Introduction

Aid is back on the international agenda. After more than a decade of aid fatigue, the international community envisages substantial increases in aid flows to poor countries in Africa and parts of Latin America and Asia.2 While a generous scaling-up of foreign aid seems an indispensable condition for reducing poverty and achieving global prosperity, history has shown that it is not sufficient.

Why foreign aid has failed to attain development objectives can be approached from two different but not incompatible perspectives. On the one hand, some argue that foreign aid is partially wasted by recipient governments and inherently poses incentive problems and produces negative externalities, which, in turn, diminish its effectiveness. On the other hand, the allocation of aid by donor countries has also been questioned. If donors allocate aid according to some strategic considerations and do not respond to the need of beneficiaries, why should we expect improved growth or poverty reduction in recipient countries?

With regard to aid allocation, there is a general consensus in the literature that official development assistance is allocated mostly according to political, economic, and strategic considerations.3 Most of the empirical analyses of aid patterns conducted in the last 30 years have concluded that bilateral donors distribute aid mainly to former colonies, to countries with which they have economic ties, or to political partners (Alesina and Dollar, 2000). Multilateral aid seems to give more weight to the needs of recipients, but it is not immune from political considerations.

Meanwhile, in the last 20 years, new actors have emerged on the development scene. Private, nongovernmental organizations (NGOs) are channeling an increasing share of development assistance. According to their proponents, NGOs care about the most vulnerable populations and represent the voices of the poor. Their motivation is meant to be humanitarian and not to follow any reasons of state. As NGOs expand, however, they are increasingly funded by institutional donors, and concerns have been voiced about the impact of donor funding on NGOs’ behavior. While many authors consider this last issue, to the best of our knowledge, there is not yet any cross-country empirical evidence on the actual determinants of nongovernmental aid flows. While the literature on official aid shows that the distribution of ODA is biased by the strategic interests of donor countries, we need to verify if NGOs do better.

This paper complements both the literature on aid determinants and that on nongovernmental aid by examining to whom and why NGOs distribute aid. We look for factors explaining an NGO’s decision to allocate aid to a given beneficiary. In particular, we check whether NGOs respond to variables representing the recipients’ needs such as poverty and life expectancy. We also examine whether the political and institutional environment matters and consider whether NGOs are influenced by strategic considerations. In particular, we are interested in whether NGOs’ decision making could be affected by their funding sources. We also investigate whether there are regional differences in the aid allocation decision.

Our results show that NGOs allocate aid mostly according to the needs of their beneficiaries. Once population is controlled for, poverty consistently appears as the main worldwide determinant of NGO aid allocation. NGOs do not respond to strategic considerations. Their funding source does not seem to exert a great influence on their aid allocation decision. We also find differences across regions. Militarization and the political nature of the regime of the recipient country affect aid allocation in the Middle East. Life expectancy influences aid allocation in countries in the Western Hemisphere and the Middle East.

The paper is organized as follows. Section II presents the literature on foreign aid allocation. Section III describes our dataset and the econometric methods we use. Section IV presents NGO aid allocation across countries. Section V concludes.

II. What do we Learn from the Literature?

The question about the underlying motivation of donors has been posed since the early years of aid flows. McKinley and Little (1977, 1978a, 1978b) test the “recipient needs” model. This model adopts, as a starting point, the officially proclaimed objective of financial assistance by industrial nations, which is mostly to promote economic development and reduce poverty in needy countries. Therefore, the model assumes that aid should flow to receiving countries according to their needs, reflected by their degree of underdevelopment. Accordingly, geographical aid distribution is expected to follow objective development requirements of these countries. Typical need indicators used to explain country allocations of aid are per capita income levels (Dudley and Montmarquette, 1976), life expectancy, literacy rates, caloric intake (McKinley and Little, 1977) or combinations of various such indicators aimed at establishing summary measures of the quality of life in recipient countries.4 Unfortunately, irrespective of the variety of indicators taken into consideration, most empirical tests of the recipient needs model failed to explain the aid distribution of most bilateral donors.

As trying to explain aid flows through the needs of beneficiary countries alone gave poor results, Dudley and Montmarquette (1976) took into account strategic variables. They questioned the assumption that donors’ concerns are only of an altruistic nature. Their seminal paper develops a hybrid model where donors are also concerned with their own strategic interest and provide aid according to a geographical distribution that takes into account not only the needs of the beneficiaries, but also the donors’ political, commercial, or other strategically minded interests. Their work showed that “among the variables which determine [the geographical distribution of aid by donor countries] the economic links, represented by lagged exports from the donor to the recipient proved to be very important” (Dudley and Montmarquette, 1976, p. 141).

Strategic variables can be measured by various indicators. Dudley and Montmarquette (1976) adopt the donor’s exports to the recipient country, while Maizels and Nissanke (1984) and Gounder (1994) take the donor’s investment in the recipient country as a proxy for its interests. McKinley and Little (1978a and 1978b) consider the political association between donors and recipient countries. More recent work confirms the importance of strategic variables. Grilli and Riess (1992) specify a hybrid model that takes into account both types of variables (strategic and altruistic) and apply it to European Community (EC) aid distribution where aid from the EC depends positively both on the need of the recipient and the commercial interest of the European donor. They determine that bilateral European aid depends mostly on exports from the EC to developing countries, while multilateral aid distribution is more influenced by the needs of recipients, proxied by a human development index and the debt level. Alesina and Dollar (2000) find evidence that bilateral aid patterns are dictated by political and strategic considerations, such as colonial history and UN votes, and that donor governments differ substantially in their degree of altruism. Collier and Dollar (2002) conclude that actual bilateral aid allocation is not efficient from a poverty-reduction point of view.

All the above-mentioned empirical analyses of aid allocation are concerned with official aid, considering mostly aid from bilateral donors or, more rarely, from multilateral institutions to governments of recipient countries. However, in the last 20 years, we have seen the emergence of a new category of actors on the foreign aid scene. As noted by Meyer (1995), there has been an intensified participation of NGOs in ODA. NGOs have prospered with the “associetal revolution” (Salamon, 1994, p. 109). Their number has grown exponentially; the size of some of them makes them significant players in social welfare and employment markets at the national level; the funding they attract has increased enormously and their visibility with the general public has never been higher. Even though available data are fragmented, all sources point toward an increase in NGOs’ actions. A report from the OECD indicates that “resources channeled through NGOs in all OECD Member countries rose from 0.2 percent of the total bilateral ODA of DAC [Development Assistance Committee] Members in 1970 to 17 percent in 1996 — to reach, in absolute terms, an amount equal to twice the total 1996 ODA of the United Kingdom, the DAC’s sixth largest donor by volume” (Woods, 2003, p. 9).

The rise of NGOs is not an accident. While it reflects private initiative and voluntary action, it also follows an increase in popularity of NGOs with governments and official aid agencies and the willingness of donors to make funds available to them. Indeed, as NGOs are expanding, many of them have switched from being primarily funded by private donors to being essentially funded by institutional donors. Two main reasons can explain why more and more public funds are channeled through NGOs. First, official agencies support NGOs in providing welfare services because of their assumed cost-effectiveness in reaching the poorest (Meyer, 1995). Second, NGOs are also seen as representative of the poor and most vulnerable. The relationship with the “people” is seen as giving them greater public legitimacy than some governments, while their managerial features are seen as permitting private sector levels of cost-control and efficiency.

However, while NGOs’ actions are expanding, many authors are concerned by the actual contribution that NGOs are making to development; in particular, questions have been raised about their legitimacy and about their relations with their funding sources. Edwards and Hulme (1994 and 1996) present a detailed analysis of how the donor and NGOs’ relationships could compromise the work of civil organizations or modify their approaches. Smillie (2000) fears that NGOs become mere implementers of donors’ policies. Jack (2001) conceptualizes the relationships between NGOs, states, and donors. Gauri and Fruttero (2003) analyze and test NGO project allocation across Bangladesh and show that NGOs’ spatial project localization across Bangladeshi provinces is influenced by a concern for obtaining donor funding.

We contribute both to the empirical aid allocation literature and to the above-mentioned debate through a cross-country analysis of NGO aid allocation, verifying whether NGO allocation decisions depend on the need of their beneficiaries or on strategic variables. We test for the potential influence exerted by the institutional donor funding them. We also assess the impact of the institutional setting of the recipient country as we control for militarization and democracy.

III. Dataset, Stylized Facts, and Methodology

A. Data on Nongovernmental Aid

The objective of this paper is to identify the determinants of NGO development aid.5 Our first task was to identify and collect data that would reflect the aid allocation decision of NGOs. We needed a dependent variable that represents the decision taken by international NGOs on to whom to give aid. However, since NGOs are private and extremely diversified organizations, there is no global database reporting the geographical destination of nongovernmental aid flows; comprehensive and precise data on all NGOs are simply not available. Ideally, we would have liked to conduct a survey among NGOs but such a compilation would have been beyond the scope of our research. We chose instead to build a database based upon data collected from institutional donors supporting NGO work. Specifically, we use data obtained from the European Commission and representing the projects proposed by European NGOs and cofinanced by the European Union (EU). While those aid flows do not cover all actions implemented by all European NGOs, we believe that analyzing this subset of NGO aid should help unveil their preferences in aid allocation. We are aware that this limits the conclusion of our study to European NGOs funded by the EU, but given the lack of general data, this is a first step in conducting empirical analysis of nongovernmental aid allocation. As the main institutional donor funding NGO activities, the EU has a long history of collaboration with nongovernmental organizations and has designed a specific budget line (B 6-7000) to fund projects proposed by NGOs.6 Despite the fact that funds are provided by the EU, this budget line should be a good indicator of NGO choices as cofinancing is a cooperation mode between the donor and the NGO based on letting the NGO decide about aid destination in terms of project choices and countries. According to Cox and Koning’s, 1997 study on the relationship between NGOs and the European Commission, NGO aid allocation does not follow any specific EU official policy. However, even if we believe that the European Commission does not impose its vision on the proposed projects, there could be a self-selection bias where NGOs applying for funding propose only projects likely to receiving the appropriate financing. Gauri and Fruttero (2003) show that, in Bangladesh, NGOs decide to implement their project in the regions likely to please the institutional donors that are funding them. Therefore, we need to control for potential EU influence. We do so by including EU official multilateral aid to governments as a regressor.

We use data reported by the European Commission. We convert the flows to constant U.S. dollars. Our sample covers the 1990 to 1999 period and has been restricted to 78 recipient countries due to missing observations for key exogenous variables (see Appendix A for the country list). As we are interested in NGOs’ intentions, we chose aid commitments as our dependent variable.7 Commitments truly describe the allocation process while being untainted by the substantial volatility and instability found in aid disbursements.8

Our dependent variable is aid in absolute terms rather than per capita. While the literature is divided on this subject, with many studies focusing on per capita aid, we follow Feeny and McGillivray (2001), focusing on the decision variable most widely used by practitioners. Looking at annual reports, we can see that NGOs report aid in absolute terms. Obviously, this does not preclude that population should be controlled for, as it is an evident measure of the size of recipients’ need.

B. Explanatory Variables

As our aim is to understand NGOs’ decisions on aid allocation across beneficiary countries, we apply a hybrid model integrating three different sets of variables. On the one hand, we want to verify whether need variables explain to whom NGOs give aid. On the other hand, we want to assess the impact of the political and institutional environment of the countries, trying to determine whether NGOs work more in democratic regimes. Finally, we will control for two strategic variables: (i) whether the recipient countries are economically close to NGOs’ region of origin (here, Europe); and (ii) the influence of NGOs’ funding source.

We adopt the following measures of the development needs of receiving countries: (i) the national poverty level, measured by the percentage of people living below the national poverty line; and (ii) life expectancy at birth in years, which reflects the level of health care. We use data from the World Bank (World Development Indicators).

Following Alesina and Dollar (2000), we assess whether the aid allocation decision is affected by recipient country institutions and governance. Our indicator of good governance is a variable ‘Political Score’ based on an index called Overall Polity Score of the Country Indicators for Foreign Policy (CIFP) developed by the Canadian Department of Foreign Affairs and International Trade. This is a global rank-based index of the democracy score (see Appendix B). On a nine-point scale, 1 is considered “strongly democratic,” and 9 “strongly autocratic.” Militarization of the country is measured by a ratio of military personnel as a percentage of labor force (World Development Indicators); this variable reflects many possible causes of potential domestic and foreign instability.

To integrate an indicator of strategic interest of the donors, we consider the most commonly used index of commercial partnership: the imports of the recipient from the donor country or region, here the EU. Our explanatory variable is the ratio of the recipients’ imports from the EU to total imports. To measure the potential influence of the EU as a funding source, we use official multilateral aid from the EU. We add four regional dummies distinguishing between countries from Africa, the Western Hemisphere, Asia, and the Middle East. For more details on the definition of the variables and their sources, see Appendix B. We control for population.

C. Stylized Facts

Table 1 below indicates, per region, the mean over the period and the standard deviation of each exogenous variable. For the list of countries in the sample and in each region, see Appendix A.

Table 1.

Mean and Standard Deviation of Exogenous Variable by Region

article image

Note that the four regions present rather different characteristics. Compared to the average level in our sample, Africa is a poorer region, with lower life expectancy, a slightly worse political score and a lower militarization level. The least number of people living below the national poverty line can be found in the Middle East, a region that has the highest militarization ratio, the worse political score, but also the longest life expectancy. As regions differ, we will consider NGOs’ allocation decision not only for the entire sample, but also within each region.

Table 2 below shows the regional repartition of NGO aid in our sample. In terms of aid per capita, both Africa and the Western Hemisphere receive a share of aid bigger than their share of world population. Asia received less aid per capita than the rest of the world, while representing more than 72 percent of the sample’s population. On a per country basis, the Western Hemisphere region includes fewer countries than Africa, but receives a bigger aid share.

Table 2.

Average NGO Aid by Region As a Percentage of Total Sample Aid

article image

D. Methodology

We examine why and to whom NGOs give aid by estimating variants of the following regression:

Aidit = β0 + β1NEEDjit + β2POLit + β3MILit + β4STRATit + popit + μi + εit(1)

where i indexes countries; t indexes time; NEEDj is a measure of the recipient country development needs (j∈[poverty, life expectancy]); POL is an index of democracy; MIL measures the militarization of the recipient country, STRATj is a measure of strategic interest (j∈ [import from EU, official multilateral EU aid]); μI is the unobserved country-specific effect; and εit is a time-varying error term.

In practice, in allocating their aid budgets or in applying for funding, decision makers often dispose of information from the year prior to that for which the aid is allocated. Indeed, except for humanitarian operations, which can benefit from emergency procedures, project approval and budgetary decisions can take many months. Moreover, information from the field can also take time to become accessible to aid agencies. Therefore, we use lagged right-hand side variables, except for EU multilateral aid, assuming that within the same organization, information about aid allocation from the department in charge of multilateral aid is rapidly available to the department in charge of NGO support.

We first estimate equation (1) with a one-way error component static random effects model. We test our equation for serial correlation and found none. We choose to characterize the individual-specific component as each individual’s realization of a random variable that is specific to him. As our dependent variable is a decision by an NGO to allocate aid in a specific way to a specific destination, we follow Nerlove and Balestra (1996) who emphasize Haavelmo’s (1994) view that a population “consists not of an infinity of individuals, in general, but an infinity of decisions” that each individual may take. In our case, as we are, in effect, drawing 700 individual decisions randomly from the larger population of all NGO decisions, we consider the random effect model to be an appropriate specification.

However, as the random effects model assumes that there is no correlation between individual effects and the regressors, it may suffer from inconsistency. Therefore, in addition, we also consider consistent albeit less efficient estimators. The fixed effect estimator might be consistent but is a poor choice for our specification, as it does not allow estimating time invariant variables of interest such as the average poverty headcount in a country. Moreover, as many of our regressors are development indicators, they vary little over time and estimators based on first-differencing are inappropriate and inefficient. Hence, we follow Arrellano and Bover (1995), adopting a Generalized Method of Moments (GMM) system estimator that not only use the equation in first differences but also instruments the equation in levels. We present one-step and two-step system GMM for our base specification. However, for our regressions over smaller regional sub-samples, we keep only one-step GMM results for which inference based on the asymptotic variance matrix has been found to be more reliable than for the asymptotically more efficient two-step estimator.9 Simulation evidence suggests that the “asymptotic standard errors tend to be much too small or asymptotic t-ratios much too big for the two-step estimators in sample sizes where the equivalent tests based on the one-step estimator are quite accurate.”10

IV. Results

A. Determinants of NGO Aid Allocation: Worldwide and Regional Results

Table 3 below presents our base specification estimated over our entire sample. The dependent variable is the log of constant NGO aid flow per country. Our base specification explains NGO aid by the militarization of the recipient country, its political score (a high score indicates a more autocratic regime), life expectancy, poverty, population, multilateral EU aid to governments of recipient countries and the imports of the recipient country from the EU. We add regional and year dummies.

Table 3.

Determinants of NGO Aid Allocation: Worldwide Results

article image
Robust p values in brackets.

Columns 1 and 2 present our base specification without and with year effects using a random effect estimator. Column 3 (FE) reports results obtained by the fixed effect estimator. Column 4 presents two-step GMM results and column 5 reports one-step GMM results. The fixed effect estimator is a poor estimator given the little or no time-variance among many of our regressors.

Once population is controlled for, NGO aid appears as driven mostly by poverty. Poverty explains aid allocation regardless of the estimator used. If poverty increases by 1 standard deviation, NGO aid is raised by more or less 1 standard deviation as well—this result is consistent across estimators. Our dummy “Western Hemisphere” is significant, as could be expected by Table 2.

Both strategic variables are mostly insignificant—except for the fixed effect estimation where it appears to be mildly significant and presents a very small and negative coefficient. Countries that receive 1 standard deviation less official European aid will receive 0.06 standard deviations more NGO aid. It appears that, while getting EU financing, NGOs remain rather independent in their allocation decision or even allocate their aid to countries that receive less official aid. This is an important result as it downplays the concerns expressed in the literature about the negative impact of a close relationship between institutional donors and NGOs. It indicates that, even if funded by official sources, NGOs do not become mere subcontractors, implementing donors’ policies.

As can be seen from Table 3, all estimators tend to point to the same results. While we cannot rule out a bias in the random effect estimation, we are reassured that consistent estimators give us similar results. For the expositional clarity, all tables below report only the results obtained using a random effect estimator.

Table 4 below presents the results we obtained when estimating our model on regional sub-samples.

Table 4.

Determinants of NGO Aid Allocation: Regional Results

article image
Robust p values in brackets.* significant at 10%; ** significant at 5%; *** significant at 1%.

As for our entire sample, poverty explains aid allocation in all regions but in the Western Hemisphere, where life expectancy takes over as the explanatory need variable. If poverty increases by 1 standard deviation, NGO aid is raised by more or less 1 standard deviation as well—this result is consistent across regions.

Both strategic variables (imports from the EU and EU official aid) remain insignificant for all regional sub samples.

Life expectancy is a determinant for aid allocation within the Western Hemisphere and the Middle East. The elasticity of aid with respect to life expectancy is rather large as a reduction in life expectancy of 1 percent increases NGO aid by more than 5 percent in the Western Hemisphere and by more than 8 percent in the Middle East. As seen in Table 1, those two regions have already achieved relatively long life expectancy and a further improvement in health conditions seems to lead to a resource reallocation.

Militarization is significant both in the Middle East and the Western Hemisphere but the coefficients have a different sign. This variable is difficult to interpret, as it can be the outcome of many different conditions. A high level of militarization can reflect domestic or cross border troubles that can potentially be accompanied by a worsening of civilians’ living conditions and therefore justify NGOs’ decision to allocate more aid. It could also reflect a military regime, a condition that would lead us to expect a lower level of NGO involvement. The sign of the coefficient could reflect either condition. In order to distinguish between possible situations, we created a composite variable MIL-POL to capture the joint effect of militarization and political conditions. A higher MIL-POL would indicate a highly militarized and autocratic regime and, therefore, we unambiguously expect less NGO involvement and a negative sign.

Population behaves as expected, except in the Middle East, where it presents an unexpected negative sign, but this can be explained by the weight of Lebanon in the aid allocation. As can be seen in Figure 1 below, Lebanon receives the most NGO aid within the region while being a relatively small country in terms of population; it also has the most inhabitants living below the national poverty line in the region.

Figure 1.
Figure 1.

NGO Aid and Poverty in the Middle East

Citation: IMF Working Papers 2006, 039; 10.5089/9781451862997.001.A001

Source: World Development Indicators and European Commission.

B. The Joint Effect of Militarization and Political Institutions

Table 5 below presents the previous regression augmented by the MIL-POL variable.

Table 5.

The Joint Effect of Militarization and Political Institutions

article image
Robust p values in brackets.* significant at 10%; ** significant at 5%; *** significant at 1%.

All results obtained from our base specification remain. Poverty remains consistently a driving force for NGO aid allocation. All variables behave as in our above-presented base specification. The MIL-POL variable is significant for allocation of aid to the Middle East. It presents the expected sign, indicating that NGOs tend to avoid working in highly autocratic and militarized countries. Moreover, the impact of those conditions on aid allocation is rather large. A 1 percent increase in MIL-POL will reduce NGO aid allocation to the Middle East by more than 3 percent. On the basis of the entire sample, MIL-POL is mildly significant with the expected sign.

V. Conclusion

This paper brings forward one appropriate response to the important issue of whether NGO aid allocation is development oriented or whether it is influenced by strategic considerations, notably by their relationship with institutional donors. On the one hand, research on the motivations underlying aid allocation of donors has so far taken into account only bilateral and multilateral aid flows. On the other hand, the literature on nongovernmental aid has expressed concerns about the potential consequence of increased NGO funding by official donors as this could modify NGOs’ priorities. This paper complements both literature strands as we conduct an empirical analysis of European NGO aid allocation over 9 years and 78 recipient countries.

We test NGOs’ aid distribution for variables representing the needs of the recipient country, such as poverty levels and life expectancy as well as for variables representing some potential strategic interest. In particular, we are interested in identifying whether NGOs are influenced by the funding source. We also test for militarization and the political state of the recipient country. Our findings indicate that NGOs are more likely to intervene in poor countries with low life expectancy. NGOs appear rather immune to strategic considerations, as the influence of the institutional donor that funds them is very weak, and as they don’t really react to the commercial closeness between the recipient country and their region of origin. Less democratic and highly militarized countries are less likely to benefit from NGO aid. We identified differences in the motivating factors across regions.

On the whole, NGOs seem relatively immune to strategic interests and seem to keep up their promise of being advocates of the poor and vulnerable.

Does NGO Aid Go to the Poor? Empirical Evidence from Europe
Author: Gilles Nancy and Boriana Yontcheva