Journal Issue
Share
Article

Does Foreign Aid Reduce Poverty? Empirical Evidence from Nongovernmental and Bilateral Aid

Author(s):
Boriana Yontcheva, and Nadia Masud
Published Date:
May 2005
Share
  • ShareShare
Show Summary Details

I. Introduction

Fifty years since the first official development assistance (ODA) programs were instituted, the question of the effectiveness of foreign aid remains an unresolved issue. Many papers have been written on the macroeconomic impact of aid, but mixed results have been reported and those papers that have identified significantly positive effects face heavy methodological criticism.

In trying to assess the effectiveness of foreign assistance, most studies focus on the impact of aid flows on GDP growth and other macroeconomic variables, such as investment or public consumption, implicitly referring to the notion that aid is meant to bridge the savings-investment gap that poor countries face. There has been much less research conducted on the impact of foreign aid on the evolution of human development indicators (HDIs). This is surprising, because the objectives announced by the donor community have evolved from intensive industrialization programs advocated in the 1950s to more recent poverty-reducing objectives such as the Millennium Developments Goals (MDGs) (see Appendix D for a detailed description of these goals). If the donors’ objective is to reach the MDGs, then assessing their assistance’s effectiveness should examine whether aid flows have a positive impact on selected HDIs.

This paper investigates the hypothesis that aid is meant to improve HDIs and assesses whether foreign aid can help recipient countries to reach some of the MDGs. We follow Boone (1996) and conduct an empirical study of the effects of aid on two human development indicators, infant mortality and education. We will use two measures of foreign aid, official bilateral aid flowing from a donor government to a recipient one, which is the standard measure in the literature, and aid projects led by international non-governmental organizations (NGOs) in developing countries. We choose the latter measure of aid for two reasons: first, as NGOs play an increasingly prominent role in the development scene and channel a growing share of development assistance, it becomes necessary to verify whether NGOs are effective in reducing poverty. So far, evaluations of NGO aid have been conducted solely at the project level. Second, as NGOs have been shown to allocate aid according to the “right” incentives and distribute aid directly at the grassroots level, this type of aid flow should avoid the two pitfalls of misallocation and misuse commonly attributed to official bilateral aid.

Our results show that NGO aid reduces infant mortality and does so more effectively than official bilateral aid. The impact on illiteracy is less significant. We also test whether foreign aid reduces government efforts in achieving developmental goals and find some evidence of a substitution effect between bilateral aid and public social sector expenditures while NGO aid does not affect social spending in the recipient country.

The remainder of the paper is organized as follows. Section II presents the literature on aid effectiveness. Section III introduces the MDGs, some selected HDIs, and some other variables of interest. Section IV describes our data set and the econometric methods we use. Section V presents the effectiveness of NGO aid on infant motality and on adult illiteracy.

Section VI assesses the impact of foreign aid on the recipient’s government effort. Section VII concludes.

II. Aid Effectiveness: A Literature Review

The literature on the effectiveness of aid focuses almost exclusively on the macroeconomic impacts of aid, measuring the effects of aid on economic growth, savings, and investment. It lacks a strong analytical framework and therefore relies heavily on empirical work. However, empirical evidence is ambiguous at best. Even though methodological issues have been refined, this literature presents rather inconclusive results.

The focus on whether aid improves GDP growth can be traced back to the two-gap model (Chenery and Strout, 1966), which remains the most influential theoretical underpinning of the aid effectiveness literature. In this model, developing countries face constraints on savings and export earnings that hamper investment and economic growth. Aid flows are meant to fill the gap between investment needs and domestic savings. Even though this model has been the target of severe criticism almost since its inception, it has provided the underlying principles both for early aid policies (Easterly 1999) and for regression specifications of most empirical papers, which focused on the aid-growth and aid-savings relationships. Most early authors concluded that aid had no significant impact on growth, savings or investment. Aid was shown to increase unproductive public consumption (Mosley and others, 1992) and to fail to promote investment. The latter point is confirmed by Boone (1996) and Reichel (1995) who find a negative relationship between savings and aid, and point to a substitution effect. This result is amended by Hadjimichael (1995), who notes that the relationship between aid and domestic savings is negative in most countries, but positive for good adjusters. The latter point is confirmed by Burnside and Dollar (2000) (henceforth BD), who showed that aid can be effective when policies are good. The BD paper has elicited abundant comment from researchers (to cite only a few, see Guillaumont and Chauvet (2001), Collier and Dehn (2001) and Easterly, Levine, and Roodman (2003)), but their results have been challenged as being “extremely data dependent” (Dalgaard and Hansen, 2001, Clemens, Radelet and Bhavnani, 2004).

Three main arguments have been advanced to explain the disappointing results of most aid effectiveness studies; aid is misallocated (donors give aid for strategic reasons to the wrong recipients), aid is misused (recipient governments pursue non-developmental agendas) and GDP growth is not the right measure of aid effectiveness. First, while all aid effectiveness papers implicitly define the donors’ objective as solely the promotion of economic growth or the reduction of poverty in the recipient countries, a parallel strand of literature on aid allocation2 has shown that most donors often pursue a different underlying agenda and allocate aid also according to their own strategic interest. If a significant part of aid is allocated for strategic purposes, no positive impact in terms of growth or poverty alleviation should be expected. Second, most studies on aid effectiveness assume that the recipient government shares the donor’s officially altruistic objective. This need not be so. As argued by Svensson (2000) and Murshed and Sen (1995), a recipient government and a perfectly altruistic donor can have conflicting objectives, as the former represents a variety of stakeholders, including wealthy individuals who might influence the aid distribution. If foreign aid is misallocated and misused, then it cannot be expected to have a significant impact on growth. Third, as suggested by Boone (1996), aid effectiveness should not be measured by its impact on GDP growth. Aid could be increasing consumption rather than investment, which would explain the disappointing results of studies on growth, but still reduce poverty through either “higher consumption of the poor or greater provision of services to the poor.” Boone tested for this by examining the impact of aid on changes in basic indicators of human development such as infant mortality, primary schooling ratios, and life expectancy. Given the evolution in the avowed objectives of the donor community from industrialization programs to poverty reduction as reflected in the adoption of MDGs, we will follow Boone and measure the impact of aid on social indicators instead of macroeconomic variables.

Moreover, all of the above-mentioned empirical analyses of aid effectiveness are concerned exclusively with official development aid. However, the last 20 years have seen the emergence of a new category of actors on the foreign aid stage. As noted by Meyer (1995), participation by NGOs in foreign aid has intensified significantly during the last two decades. 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 to the general public has never been higher. NGOs are perceived as having two distinctive features that differentiate them from other donors. First, they are advocates of the most vulnerable populations and their motivation is widely perceived as mainly altruistic. Second, their actions at the grassroots level are seen as conducted at private-sector levels of cost control and efficiency, while they achieve development objectives and serve the needs of many people (Rose-Ackermann, 1996).

While NGO activities are expanding, commentators are increasingly expressing reservations about the actual contribution that NGOs make to development. Edwards and Hulme (1997) analyze in detail how the donor-state and NGO relationships could compromise the work of civil organizations or modify their approaches. Robinson (1997) suggests that NGOs will become mere implementers of donors’ policies. Gauri and Fruttero (2003) analyze and test NGO motivation and show that these organizations are influenced in allocating aid by a concern for donor funding.

Despite these concerns about NGO activities, no empirical study has been undertaken on the effectiveness of NGO aid at the macro level. The debate has so far focused on the nature of the NGOs and their relationship with donors, relying only on project evaluations to assess the effectiveness of NGO aid (see Cox and Koning, 1997, for a useful evaluation of European NGOs).

The contribution of this paper is twofold. First, we complement the literature on the effectiveness of foreign aid by assessing the impact of foreign aid flows on social indicators instead of on GDP growth. Second, we contribute to the debate on the merits of NGOs by conducting an empirical evaluation of their actions. Our aim is to verify empirically whether foreign aid has had a positive impact on two human development indicators whose improvement is part of the MDGs. We will test the impact of aid on infant mortality and adult illiteracy. Because they differ in both motivation and implementation, we distinguish between bilateral aid and nongovernmental aid. We also test whether foreign aid reduces the efforts a recipient country government makes in fighting illiteracy and infant mortality. We therefore assess the impact of aid on the share of social spending of recipient countries.

III. Millennium Development Goals, Human Development Indicators, and Other Variables

Many of the targets of the MDGs were first discussed at international conferences and summits held during the 1990s. They were later compiled and became known as the International Development Goals.3 In September 2000, the member states of the United Nations unanimously adopted the Millennium Declaration, and the General Assembly recognized the Millennium Development Goals as part of the road map for implementing the Millennium Declaration. Appendix D lists all MDGs. We choose to focus on two selected human development indicators whose improvement is part of the MDGs, infant mortality and adult illiteracy. We chose these indicators for the following reasons: (i) health and education indicators are more concrete measures than poverty; (ii) as indicated by Boone (1996), infant mortality indicators respond quickly to improved health services and can therefore be considered as “flash indicators of improvement in the conditions of the poor”.

Table 1 below presents average levels of the above-mentioned indicators4 and aid flows for countries with the highest levels of infant mortality and countries with the lowest levels of infant mortality in a sample of 87 countries.5 For a more detailed description of our sample, see section IV.

Table 1.Infant Mortality
Infant mortalityFemale illiteracyICRGAgricultural value added per workerReal GDP per capitaPoverty headcountUrbanizationPopulation growth rateReal NGO aid per capitaReal bilateral aid per capitaHealth expenditures per capita
Mean for 10 countries with highest infant mortality135.1274.9549.89265.39222.2758.9828.692.610.2130.874.70
Overall mean66.1338.0061.501693.371822.9639.8242.812.220.1936.6356.34
Mean for 10 countries with lowest infant mortality13.028.7071.375329.107074.5721.4166.851.470.1555.33247.05

Our objective is to test the impact of aid on those two indicators. To identify other variables that could be determinants of infant mortality and illiteracy, we will follow Caldwell (1986) and Dreze and Sen (1989), as well as Gupta and others (2003), for evidence on improvement of human development indicators and on the success of health and education programs. For infant mortality, besides testing for bilateral and nongovernmental aid, we control for the level of development represented by per capita GDP, the poverty headcount, the level of rural development, as indicated by per worker agricultural value added, and female illiteracy. We assess the impact of government efforts in reducing infant mortality represented by the per capita health expenditure. We also verify the impact of institutional variables, such as a governance index represented by the Freedom House policy indicator (ICRG) and the degree of urbanization.

Comparing the average female illiteracy rate in the ten countries that have the highest infant mortality rate with the average female illiteracy rate in the entire sample suggests that higher rates of female illiteracy are associated with higher levels of infant mortality. For the ten countries with the lowest levels of infant mortality, the average rate of female illiteracy is far below the average for the 87 countries, suggesting that low levels of infant mortality are associated with low levels of female illiteracy. As expected, countries with high infant mortality rates also seem to be less urbanized on average, whereas those with low infant mortality rates have a higher than average level of urbanization. Infant mortality rates appear to be negatively associated with rural development as measured by agricultural value added. The poverty headcount is higher than average in countries with a higher mortality rate and lower in countries with a lower mortality rate.

One might expect that aid flows to countries with higher rates of infant mortality would be higher than average. Although this seems to hold true for NGO aid, it does not appear to be true in the case of total bilateral aid flows. NGO aid per capita is, on average, higher in countries with higher rates of infant mortality and lower in countries with lower rates of infant mortality. Bilateral aid per capita, on the other hand, is lower than average in countries with high infant mortality rates and higher on average for countries with low infant mortality rates. Government effort in the area of health, as measured by public health expenditure per capita, is far lower than average in countries with low levels of infant mortality and much higher than average in countries with low levels of infant mortality.

Table 2 below presents some interesting findings for adult illiteracy based on a sample of 76 countries.6

Table 2.Illiteracy
IlliteracyICRGAgricultural value added per workerReal GDP per capitaPoverty headcountUrbanizationPopulation growth rateNGO aid per capitaBilateral aid per capitaEducation expenditure per capita
Mean for 10 countries

with highest illiteracy

rates
70.0356.16295.86318.0148.5728.762.660.2930.4510.65
Overall mean31.0661.662124.011733.1039.4644.342.160.2035.7381.96
Mean for 10 countries with lowest illiteracy rates4.1469.275063.095182.0025.3365.681.470.1651.32258.77

As is the case with infant mortality, higher illiteracy rates appear to be associated positively with higher poverty levels and negatively with the level of urbanization and rural development. NGO aid per capita is higher than average in countries with high levels of illiteracy and lower than average in countries with lower levels of illiteracy. Bilateral aid, on the other hand, is lower than average in countries with high illiteracy rates and higher in countries with high levels of illiteracy. Government effort (measured by education expenditure per capita) appears to be far lower than average in countries with the highest levels of illiteracy and much higher than average in countries with high levels of illiteracy.

Figure 1a below shows average per capita NGO aid and average per capita bilateral aid as deviations from their sample means for a sub-sample of 20 countries sorted by their infant mortality levels. For expositional purposes, figure 1a covers only 20 countries from our sample; it includes the 10 countries with the highest infant mortality rates and the 10 countries with the lowest infant mortality rates. The countries are sorted in descending order from left to right going from Niger (country with the highest infant mortality) to Hong Kong SAR (lowest rate). As can be seen from the trend lines, countries with lower infant mortality rates receive relatively less NGO aid, but relatively more bilateral aid. Conversely, countries with the highest mortality ratios (left side of the figure) receive relatively more NGO aid, but less-than-the-mean bilateral aid.

Figure 1aPer capita NGO and bilateral aid expressed as deviations from sample mean

Figure 1b below presents a similar table with a sub-sample of 20 countries sorted by their illiteracy levels. From left to right, we first have countries with very high illiteracy rates (Niger), and then toward the right, countries with the lowest illiteracy rates. Again, it can be seen that NGO aid is allocated relatively more to countries with high illiteracy and relatively less to countries with better HDIs. No such pattern emerges from bilateral aid allocation. A rather interesting result, to which we will come back later, is that aid allocation seems quite different between bilateral donors and NGOs. This observation is in line with the results identified by the literature on aid allocation.7 Many authors have indeed indicated that bilateral donors allocate aid mostly according to their strategic interests and do not respond to humanitarian concerns. The figure reveals that although most donors have adopted the MDGs as their official policy, their aid allocation pattern may not reflect it. The way NGOs allocate aid seems in line with their image as advocates of the poor and vulnerable.

Figure 1bPer capita NGO and bilateral aid expressed as deviations from sample mean

IV. Data sources and Methodology Issues

Data sources

Given the private and extremely diversified nature of these organizations, there are no datasets reporting on all aid flows stemming from all international NGOs. Therefore, we chose to limit our study to European NGOs and utilize data from the European Commission representing projects proposed by European NGOs and cofinanced by the European Union (EU)8. We are aware that these projects are only a partial view of all projects implemented by European NGOs but we assume (i) that NGOs getting financed by the EU are representative NGOs in Europe and (ii) that those projects financed by the EU are representative of all the projects that the NGO implements. Our data were in current euros which we converted to constant dollars. Our sample covers a 12-year period from 1990 to 2001. The number of countries actually used varies according to the regressions owing to missing observations for key exogenous variables. Bilateral aid is obtained from the OECD. We created a dummy variable “IMF” that assumes the value 1 if the recipient country has a structural adjustment program supported by the IMF (Enhanced Structural Adjustment Facility -ESAF or Poverty Reduction and Growth Facility -PRGF), and 0 otherwise. Exogenous variables included in the regressions on human development indicator include female illiteracy, the level of governance proxied by the ICRG Index, the level of rural development proxied by a measure of agricultural value added per worker, the level of development measured as real GDP per capita, the poverty headcount and population growth rate, which have been taken from various sources, including the World Development Indicators compiled by the World Bank. Data on expenditures have been taken from IMF staff estimates and the World Bank databases. For a more detailed description of our data sources, refer to Appendix A.

Endogeneity issues

As do the authors of most papers in the aid effectiveness literature, we share a concern about the issue of a potential simultaneity bias. If aid is dependent on some social indicators, then it might not be exogenous with respect to the human development index taken into consideration. The same is true for the government’s effort. We have therefore three regressors that might be potentially endogenous; NGO aid, bilateral aid, and social spending. We use the Davidson and MacKinnon test (1993) which is similar to the Durbin-Wu-Hausman test to check for the endogeneity of our regressors. The test involves an OLS regression of the original dependent variable on the original regressors augmented by the residuals of the first stage instrumental variable regressions. A rejection of the null hypothesis indicates that the instrumental variables fixed effects estimator should be employed. This test may be employed to a subset of endogenous variables and help identify which variable can cause endogeneity.

V. Foreign Aid and MDGs

A. The infant mortality regression

Methodology

In this section we try to answer the following question: does aid reduce infant mortality and illiteracy? We examine this question by estimating variants of the following regressions:

where i indexes countries, t indexes time, HDIj is a Human Development Indicator (j∈[Infant Mortality, Illiteracy]), NGOAidit is per capita NGO aid, BAidit represents per capita bilateral aid, zitis a vector of other exogenous variables that might affect the human development indicators, and GEit measures the recipient government’s effort in promoting human development. The last is proxied here by either per capita health spending or per capita education spending; μI is the unobserved country specific effect, and εit is a time-varying error term.

We estimate equation (1) for infant mortality with a one-way error component static panel model.9 We tested our equation for serial correlation and found none. A random effects regression model has two main advantages compared with the fixed effect estimation. First, it gives more efficient estimates, and, second, it allows us to explore the effects of time invariant variables of interest, such as the average poverty headcount in a country. However, as the random effects model assumes that there is no correlation between individual effects and the regressors, it may suffer from inconsistency. Therefore, we use the random effects model when our estimations pass the Hausman specification test, i.e., the test does not reject the null hypothesis of no systematic differences between the estimates obtained from fixed and random effects. We used least squares estimation.

Results

Our base sample for infant mortality is an unbalanced panel of 58 countries from 1990 to 2001 including 233 observations.10Table 3 below presents results from our estimations. The three sets of results reported in columns I, II, and III include health expenditure, NGO aid per capita, and bilateral aid per capita as explanatory variables, but the sets of exogenous variables included in the three regressions are slightly different. In column IV, we present results without health expenditure per capita. These have been included for illustrative purposes. The addition of GDP per capita causes the coefficient on health expenditure per capita to lose its significance, possibly owing to the high correlation between the two variables.

Table 3:Infant Mortality Regressions
Infant Mortality
IIIIIIIV
Health expenditure per capita (1)-0.0527

[0.034]**
-0.0607

[0.021]**
-0.0145

[0.597]
Bilateral aid per capita (1)0.0063

[0.648]
0.0086

[0.514]
0.0064

[0.611]
0.0171

[0.160]
NGO aid per capita (1)-0.0152

[0.027]**
-0.0126

[0.086]*
-0.0102

[0.147]
-0.0131

[0.037]**
Female illiteracy0.5272

[0.000]***
0.3314

[0.000]***
0.3167

[0.000]***
0.336

[0.000]***
Urbanization0.0593

[0.710]
0.0641

[0.543]
0.1055

[0.308]
0.1286

[0.185]
Poverty headcount0.2694

[0.001]***
0.1921

[0.023]**
0.1696

[0.045]**
Population growth rate0.0293

[0.453]
0.0877

[0.151]
0.0322

[0.590]
0.0384

[0.500]
ICRG-0.0027

[0.030]**
-0.0022

[0.067]*
-0.0027

[0.012]**
IMF0.0057

[0.761]
0.0109

[0.611]
-0.0025

[0.905]
0.0025

[0.897]
GDP per capita (2)-0.3083

[0.000]***
-0.2915

[0.000]***
Agricultural value added per worker-0.0611

[0.343]
-0.162

[0.001]***
-0.0291

[0.611]
-0.071

[0.187]
Constant2.4439

[0.008]***
2.952

[0.000]***
4.1838

[0.000]***
4.223

[0.000]***
Observations233202202253
Number of countries58494950
Note: p values in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. All variables except ICRG and IMF are in natural logs. Specifications II, III, and IV have been modeled as random effect regressions whereas I has been modeled as a fixed effect regression based on results of Hausman tests for consistency between fixed and random effect models.

Health expenditure per capita, NGO aid per capita, and bilateral aid per capita are in real US dollar terms. The base year being used is 2000.

GDP per capita is in constant US$ terms with base year 1995.

Note: p values in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. All variables except ICRG and IMF are in natural logs. Specifications II, III, and IV have been modeled as random effect regressions whereas I has been modeled as a fixed effect regression based on results of Hausman tests for consistency between fixed and random effect models.

Health expenditure per capita, NGO aid per capita, and bilateral aid per capita are in real US dollar terms. The base year being used is 2000.

GDP per capita is in constant US$ terms with base year 1995.

Our main findings are that increased health expenditure per capita reduces infant mortality as does greater NGO aid per capita11. We do not find any significant impact of total bilateral aid on infant mortality. Our result is in line with Boone (1996) who found “no significant impact of aid on improvement in infant mortality, primary schooling rations nor life expectancy.” As expected, higher levels of development (measured by GDP per capita) and possibly greater rural development (measured by agricultural value added per worker) lead to lower levels of infant mortality. If GDP per capita increases by 1 percent, infant mortality decreases by 0.3 percent. Reducing female illiteracy can have the strongest positive impact on infant mortality as a decrease in female illiteracy by 1 percent decreases infant mortality by 0.52 percent in our first regression. As expected, an increase in poverty leads to higher infant mortality as well; the elasticity of infant mortality with respect to poverty ranges from 0.16 to 0.26. We also find a significant impact of the level of governance (proxied by ICRG) in reducing infant mortality, albeit with lower elasticities. Population growth rate, urbanization, and the presence of an IMF-supported structural adjustment program are found to be insignificant. A number of reasons could be proposed to explain why NGO aid seems to work better than bilateral aid in reducing infant mortality. First, as can be seen in section III, Figure 1b, NGOs and bilateral donors allocate aid differently: on average and over the entire period, countries with lower infant mortality rates receive relatively less NGO aid, but relatively more bilateral aid. Conversely, countries with the highest mortality ratios receive relatively more NGO aid, but less-than-the-mean bilateral aid. There seems to be a certain inconsistency between the avowed objectives of bilateral donors and their actual aid allocation, as indicated by Alesina and Dollar (2000). Second, maybe NGOs have direct links to the poor and vulnerable and are therefore more efficient. Third, as argued by Boone (1996), aid transiting through recipient governments could be diverted for the benefit of wealthy elites linked to those in power.

B. The illiteracy regression

Methodology and results

Our base sample for adult illiteracy is an unbalanced panel of 76 countries from 1990 to 2001 including 420 observations.12We use two stage least squares regression technique, as we found some endogeneity in regressors.13 Results are presented in Table 4 below:

Table 4.Adult Illiteracy
Adult IlliteracyIIIIIIIV
Education expenditure per capita (1)-0.055

[0.001]***
-0.031

[0.067]*
-0.029

[0.117]
-0.146

[0.002]***
Bilateral aid per capita (1)0.0456

[0.597]
0.077

[0.307]
0.072

[0.378]
0.173

[0.384]
Urbanization-1.035

[0.000]***
-1.036

[0.000]***
-0.898

[0.000]***
-0.486

[0.303]
Population growth rate0.175

[0.000]***
0.242

[0.000]***
0.153

[0.000]***
0.210

[0.001]***
Real NGO aid per capita (1)0.001

[0.891]
0.004

[0.470]
0.000

[0.956]
-0.006

[0.658]
IMF-0.018

[0.161]
-0.002

[0.884]
-0.020

[0.110]
-0.021

[0.280]
ICRG-0.000

[0.548]
GDP per capita (2)-0.151

[0.010]***
Poverty headcount0.108

[0.732]
Constant6.794

[0.000]***
6.666

[0.000]***
7.173

[0.000]***
4.340

[0.133]
Observations420358420319
Number of countries76627654
Note: p values in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%.All variables except ICRG and IMF are in natural logs. The specification has been modeled as a random effect model, using instrumental variable regression. The instrument set includes the lag of education expenditure as an instrument for education expenditure in the current period and the lag of past NGO aid per capita as an instrument for bilateral aid per capita. The choice of the instrument set has been determined using the Davidson MacKinnon test for exogeneity.

Education expenditure per capita, NGO aid per capita and Bilateral aid per capita are in real US$ terms; the base year being used is 2000.

GDP per capita is in constant US$ terms with base year 1995.

Note: p values in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%.All variables except ICRG and IMF are in natural logs. The specification has been modeled as a random effect model, using instrumental variable regression. The instrument set includes the lag of education expenditure as an instrument for education expenditure in the current period and the lag of past NGO aid per capita as an instrument for bilateral aid per capita. The choice of the instrument set has been determined using the Davidson MacKinnon test for exogeneity.

Education expenditure per capita, NGO aid per capita and Bilateral aid per capita are in real US$ terms; the base year being used is 2000.

GDP per capita is in constant US$ terms with base year 1995.

Our main findings include a significant and negative impact of education expenditure on adult illiteracy. We do not find any significant impact of NGO aid per capita and total bilateral aid per capita. The coefficient on urbanization is negatively signed and is significant in three out of four estimations. More urbanized countries with slower population growth present lower levels of adult illiteracy. The presence of an IMF-supported structural adjustment program does not seem to be detrimental to literacy. We do not find any significant impact of the level of governance or the average poverty level in a country on its level of illiteracy.

VI. Government Efforts and Foreign Aid

The results from the regressions on infant mortality and on illiteracy find that government effort measured by social sector expenditures plays an important role in improving human development indicators. In this section, we explore the relationship between government effort and official and nongovernmental aid flows. We want to verify how foreign aid affects government efforts, measured as the share of health or education expenditure in total public spending. We expect that, in absence of fungibility (Pack and Pack, 1993), bilateral aid would increase social spending as it would augment the size of the budget devoted to health and education because donors often earmark funds to social sectors. Bilateral aid could also be insignificant, if it is used for other purposes or if aid is fungible. With regard to NGO aid, funds do not transit through government programs, and the issue becomes whether an increase in NGO activity has a catalytic effect on government action or if, on the contrary, a substitution effect occurs. If a school is already built by an NGO, will the government have the incentive to provide more teachers or will it allocate a corresponding budget to non-education expenditures?

Methodology

We examine the above question by estimating variants of the following regressions:

where i indexes countries; t indexes time; GEexit measures the recipient country effort proxied by the government’s education/health expenditure as a percentage of total government expenditure; NGOAidit is NGO aid as a percentage of expenditure; BAidexitis bilateral aid as a percentage of government expenditure; xit is a vector of exogenous variables that might affect social spending. IMFit indicates the presence of an IMF-supported program; μi is the unobserved country specific effect; μt is a year dummy, and εit is a time-varying error term.

We have used the dynamic specification for social expenditures in equation (2) for our estimations. We tested the above equation and found serial correlation. It seemed reasonable to assume that there would be some degree of persistence in the share of government expenditure that is allocated to different sectors in a single country. An ordinary least squares regression in this context would be inconsistent owing to the violation of the assumption of lack of correlation between the error term and regressors. This correlation is introduced because the error term contains the country specific effects, and the lagged dependent variable is also a function of country specific effects. Similarly, the within transformation would not solve the problem as it would introduce a negative correlation between the lagged dependent variable and the transformed error. The random effects model cannot be used as there is a correlation between the error term, μi + εit, and the lagged dependent variable.

The general approach to estimation of the dynamic specification as above is to use the Generalized Method of Moments (GMM) approach. Whereas the first difference GMM uses past levels of the dependent variable as instruments for the equation in first differences, the system estimator is an extended version of the linear GMM estimator that also includes lagged levels of the dependent variable as instruments for the equation in levels (Blundell and Bond, 1998). We present first difference and system GMM results for our specifications for social expenditure. However, in dynamic panel data models where the autoregressive parameter is moderately large and the number of time series small, the first difference estimator has been found to have large finite sample biases and poor precision in simulation studies as lagged levels provide weak instruments for first differences (Alonso-Borrego, 1996). Blundell and Bond (1998) provide Monte Carlo simulation results to compare the performance of system and first difference and find that, in case of the coefficient on the lagged dependent variable being >0.8, we can expect the first difference estimate of the coefficient to be seriously downward biased and imprecise. They show that this bias remains substantial even for panels with 11 periods per individual.

OLS and Within Group estimators of the model can be used to assess a range for the coefficient on the lagged dependent variable, as the former is likely to be biased upward whereas the latter would be expected to be biased downward (Bond, 2002). We therefore also present the OLS and Within Group estimates of the lagged dependent variable for the social spending regressions.

We report GMM difference and system one-step 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.14 Simulation evidence suggests that there are modest gains in efficiency from using the optimal two-step weight matrix.15 It also 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.”16

Results

The base sample for regressions on government effort in the area of health is an unbalanced panel of 50 countries over the period 1990-2001, consisting of 305 observations17. The results from estimating equation (2) are presented in Table 5 below:

Table 5.Health Expenditure
Health ExpenditureGMM-SYSGMM-SYSGMM-DIFFGMM-DIFF
Health expenditure (t-1)0.958

[0.000]***
0.938

[0.000]***
0.459

[0.000]***
0.463

[0.000]***
Military expenditure0.019

[0.488]
0.010

[0.643]
-0.077

[0.370]
-0.101

[0.240]
Bilateral aid0.012

[0.441]
0.019

[0.180]
-0.001

[0.967]
-0.002

[0.919]
GDP per capita0.011

[0.741]
0.022

[0.383]
-0.184

[0.383]
-0.075

[0.645]
Current revenue0.149

[0.003]***
0.111

[0.015]**
0.398

[0.000]***
0.426

[0.000]***
ICRG0.000

[0.989]
0.006

[0.072]*
NGO aid-0.006

[0.407]
-0.005

[0.458]
-0.006

[0.619]
-0.014

[0.256]
IMF0.070

[0.045]**
0.062

[0.031]**
0.046

[0.173]
0.041

[0.249]
Constant-0.728

[0.000]***
-0.617

[0.007]***
Hansen test[1.000][1.000][1.000][1.000]
Arellano Bond AR(1) test[0.013][0.004][0.007][0.003]
Arellano Bond AR(2) test[0.999][0.814][0.255][0.163]
Observations264305226259
Number of countries43504147
Note: p values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. One-step GMM results are reported with robust standard errors. A full set of year dummies is included in both specifications. The instrument set includes lags of health expenditure dated t-2 and earlier as instruments in level and system equations. Predetermined variables include NGO aid and presence of an IMF-supported program, the implied moment conditions are included in the instrument matrix. A full set of year dummies has been included in both specifications. The lags of health expenditure and NGO aid have been used as regressors. All variables except ICRG, IMF and GDP per capita are expressed as a percentage of expenditure. The OLS estimate for the lagged dependent variable is 0.963 whereas the within groups estimate is 0.539, suggesting that there is indeed a significant downward bias in the estimates obtained by the first difference estimates.
Note: p values in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. One-step GMM results are reported with robust standard errors. A full set of year dummies is included in both specifications. The instrument set includes lags of health expenditure dated t-2 and earlier as instruments in level and system equations. Predetermined variables include NGO aid and presence of an IMF-supported program, the implied moment conditions are included in the instrument matrix. A full set of year dummies has been included in both specifications. The lags of health expenditure and NGO aid have been used as regressors. All variables except ICRG, IMF and GDP per capita are expressed as a percentage of expenditure. The OLS estimate for the lagged dependent variable is 0.963 whereas the within groups estimate is 0.539, suggesting that there is indeed a significant downward bias in the estimates obtained by the first difference estimates.

GMM system and difference results are reported. Our preferred results are the system results. The coefficient on the lagged dependent variable is estimated to be close to 1, and we therefore expect that there is a significant downward bias in this estimate using the difference method. For each regression, we report the Hansen J test of over-identifying restrictions. It does not reject the joint null hypothesis that the instruments are valid, i.e., uncorrelated with the error term, and that the excluded instruments are correctly excluded from the estimated equation. Examining p-values of the AR(1) and AR(2) tests of autocorrelation, we can see that, as expected, we reject the null of no first order serial correlation in residuals and do not find evidence for second order serial correlation.

We find a significant and positive impact of the past year’s share of health in total expenditure on the percentage of expenditure devoted to health in the current period. We do not find any evidence that either bilateral aid or NGO aid has an impact on the share of spending on health care in total expenditure. The fact that bilateral aid does not change government spending on health means that there is no additionality suggesting that bilateral aid is fungible and that if aid increases non-aid-financed expenditures decline. This possible fungibility could explain the lack of effect of bilateral aid on infant mortality. On the other hand, the fact that NGO aid does not affect negatively government spending implies additionality of the public sector’s efforts and NGO initiatives. The presence of IMF-supported program in a country leads to a higher share of expenditure being allocated to health. This result could reflect the fact that, while IMF programs generally call for fiscal tightening, the Poverty Reduction and Growth Facility has attempted to limit the adverse impacts on priority sectors such as health care and education. As expected, current revenue is also found to have a significant and positive impact on public spending on health. The level of governance is found to be significant and positive in the first difference result, but is insignificant using the system estimation.

The base sample for regressions on government effort in the area of education is an unbalanced panel of 51 countries over the period 1990-2001 and consists of 313 observations.18 The results are presented in Table 6 below.

Table 6.Government Effort in Education
Education ExpenditureGMM-SYSGMM-SYSGMM-DIFFGMM-DIFF
Education expenditure (t-1)0.836

[0.000]***
0.803

[0.000]***
429

[0.000]***
0.368

[0.000]***
Military expenditure-0.019

[0.423]
0.010

[0.579]
-0.038

[0.493]
0.000

[0.997]
Total bilateral aid0.006

[0.586]
0.009

[0.439]
0.026

[0.058]*
0.019

[0.168]
GDP per capita0.000

[0.989]
-0.009

[0.656]
0.190

[0.224]
0.088

[0.428]
Current revenue0.184

[0.004]***
0.244

[0.000]***
0.465

[0.000]***
0.452

[0.000]***
ICRG0.000

[0.827]
0.001

[0.594]
NGO aid-0.009

[0.259]
-0.005

[0.490]
0.000

[0.988]
0.002

[0.763]
IMF0.070

[0.048]**
0.056

[0.064]*
0.055

[0.099]*
0.039

[0.184]
Constant-0.363

[0.092]*
-0.502

[0.016]**
Hansen test1.001.001.001.000
Arellano Bond AR(1) test0.000.000.000.00
Arellano Bond AR(2) test0.130.130.080.08
Observations272313234267
Number of countries44514147
Note: Robust p values in brackets * significant at 5%; ** significant at 1%. One-step GMM results are reported with robust standard errors. A full set of year dummies is included in both specifications. The instrument set includes lags of education expenditure dated t-2 and earlier as instruments in level and system equations. Predetermined variables include NGO aid and presence of an IMF-supported program; the implied moment conditions are included in the instrument matrix. A full set of year dummies has been included in both specifications. The lags of education expenditure and NGO aid have been used as regressors. All variables except ICRG, IMF and GDP per capita are expressed as a percentage of expenditure. The OLS estimate for the lagged dependent variable is 0.878 whereas the within groups estimate is 0.412, suggesting that there is indeed a significant downward bias in the estimates obtained by the first difference estimates.
Note: Robust p values in brackets * significant at 5%; ** significant at 1%. One-step GMM results are reported with robust standard errors. A full set of year dummies is included in both specifications. The instrument set includes lags of education expenditure dated t-2 and earlier as instruments in level and system equations. Predetermined variables include NGO aid and presence of an IMF-supported program; the implied moment conditions are included in the instrument matrix. A full set of year dummies has been included in both specifications. The lags of education expenditure and NGO aid have been used as regressors. All variables except ICRG, IMF and GDP per capita are expressed as a percentage of expenditure. The OLS estimate for the lagged dependent variable is 0.878 whereas the within groups estimate is 0.412, suggesting that there is indeed a significant downward bias in the estimates obtained by the first difference estimates.

As in the case of the previous exercise involving the share of public expenditure devoted to health care, we prefer the system results because the coefficient on the lagged dependent variable is close to unity. For each regression, we report the Hansen J test of over-identifying restrictions. It does not reject the joint null hypothesis that the instruments are valid. Examining p-values of the AR(1) and AR(2) tests of autocorrelation, we can see that as expected we reject the null of no first order serial correlation in residuals, and do not find evidence for second order serial correlation at the 5 percent level.

We find that the share of public spending devoted to education during the previous period had a significant and positive impact on the share of expenditure devoted to education in the current period. We do not find any evidence that NGO aid has an impact on the share of health care in total expenditure. The first difference results do find a mildly significant and positive impact of bilateral aid on education expenditure. However, as indicated above, first difference results are probably biased, and we can therefore conclude only that there is no substitution effect of increased bilateral aid on education expenditures. The presence of an IMF-supported program in a country seems to lead to a higher share of expenditure being allocated to education. Increased current revenue is also found to have a significant and positive impact on government effort in education.

VII. Conclusion

The objective of this paper was to reassess the effectiveness of foreign aid in terms of its impact on selected human development indicators. We chose to evaluate how aid affects infant mortality and illiteracy because improvements in both indicators are official objectives of all donors who have adopted the Millennium Development Goals. We based our empirical study on two separate measures of aid – bilateral aid flows to the governments and nongovernmental aid. We choose to distinguish between those two aid types as official aid and NGO aid flows differ both in their allocation pattern and in their implementation.

Our results show that NGO aid significantly reduces infant mortality while bilateral aid does not. A number of reasons could explain this result. First, as their proponents claim, NGO aid may be more effective than government actions in reaching out to the poor. Improving infant mortality may be more efficiently done at the grassroots level. Second, NGO aid is allocated more toward countries with high infant mortality while bilateral aid favors countries with already lower infant mortality. This demonstrates that, while official donors have adopted reaching MDGs as their official policy, their aid allocation pattern is not consistent with their avowed objectives. Third, bilateral aid seems fungible and increases in aid don’t seem to be reflected in health expenditures. The lack of additionality implies that bilateral aid increases lead to declines in non-aid-financed expenditures, canceling the potentially positive effect of bilateral aid on infant mortality. Fourth, our measure of official aid as total bilateral aid per country may not be the appropriate indicator as it covers all types of projects and programs19. Fifth, Boone’s (1996) claim that recipient governments divert aid to benefit a wealthy elite may also be an explanation for our results.

Concerning illiteracy, our results are less significant. Apart from the usual control variables that behave as expected, only government education expenditures are shown to reduce illiteracy. A possible explanation for these results is that a 10-year period is too short for aid to have an impact as improvements in illiteracy take much more time to be recorded. Research should be conducted again when longer time series will be available. Finally, we also examined the impact of foreign aid on government efforts and found that NGO aid does not reduce the recipient governments’ efforts.

From a policy perspective, if we take infant mortality as a “flash indicator” of the living conditions of the poor (Boone, 1996), in our study, NGO aid appears more effective in reaching out to the poor and vulnerable populations, and therefore donors who have chosen to channel aid through NGOs have made the right choice. This paper is the first empirical study on the effectiveness of NGO aid at the macro level and confirms the legitimacy of NGO actions. Whether NGO aid can be easily scaled up is a different question that should be answered in future research.

Appendix A. Data Description and Sources

1) NGO AID

Source: European Commission, budget line B7-6000.

The data represent commitments of European Commission to cofinancing NGO projects in a given country in a given year.

2) Bilateral Aid

Source: OECD

Total official development assistance committed to a country in a given year by the countries that are part of the Development Assistance Committee.

3) Infant mortality rate

Source: World Development Indicators

The number of infants who die before reaching one year of age, per 1,000 live births in a given year.

4) Total adult illiteracy rate

Source: World Development Indicators

The proportion of adults aged 15 and above who cannot, with understanding, read and write a short, simple statement on their everyday life.

5) Female adult illiteracy rate

Source: World Development Indicators

The proportion of female adults aged 15 and above who cannot, with understanding, read and write a short, simple statement on their everyday life

6) Public education expenditure

Sources: World Development Indicators for education data and IMF staff estimates for total expenditures. See Appendix B contains for on the level of government that is relevant for each country.

7) Public health expenditure

Source: IMF staff estimates. See Appendix B for the relevant level of government for each country.

8) Health expenditure per capita

Source: World Development Indicators and staff estimates

9) Education expenditure per capita

Source: World Development Indicators and staff estimates

10) Agriculture value added per worker (constant U.S. 1995 dollars)

Source World Development Indicators

Agriculture value added per worker is a measure of agricultural productivity. Value added in agriculture measures the output of the agricultural sector (ISIC divisions 1-5) less the value of intermediate inputs. Agriculture comprises value added from forestry, hunting, and fishing as well as cultivation of crops and livestock production. Data are in constant 1995 U.S. dollars.

11) GDP per capita

Source: World Development Indicators.

The data are in constant 1995 U.S. dollars.

12) ICRG

Source: International Country Risk Guide/Political Risk Services (IBC USA Publications, Inc.). Composite index based on political financial and economic risk ratings compiled by ICRG. Political risk accounts for 50 percent of the composite rating while financial and economic risk ratings account for 25 percent each. The highest overall rating (theoretically 100) indicates the lowest risk, and the lowest rating (theoretically 0) indicates the highest risk.

13) IMF

Source: IMF

Dummy that assumes value 1 if there is a structural adjustment program supported by the IMF in a country in a given year and 0 if there is no program.

14) Urbanization

Source: World Development Indicators

Urban population as a percentage of total population.

15) Population growth rate

Source: World Development Indicators

16) Military expenditure

Source: World Development Indicators

Military expenditure as a percentage of central government expenditures.

17) National poverty

Source: World Development Indicators

Percentage of the population living below the national poverty line. National estimates are based on population-weighted sub-group estimates from household surveys.

Appendix B. Level of Government Relevant for Public Expenditure Data
Angol aGeneralIndiaGeneralDominican RepublicCentral
BeninCentralIndonesiaGeneralEcuadorCentral
BotswanaCentralKiribatiCentralEl SalvadorCentral
Burkina FasoCentralKoreaCentralGrenadaCentral
BurundiCentralLao PDRGeneralGuatemalaCentral
Central African RepublicCentralMalaysiaCentralGuyanaCentral
Cape VerdeCentralMaldivesCentralHondurasCentral
ChadCentralMarshall IslandsCentralJamaicaCentral
CameroonCentralMongoliaGeneralMexicoCentral
ComorosCentralMyanmarCentralNicaraguaCentral
Congo, Dem. Rep. OfCentralNepalCentralPanamaCentral
Congo, Republic ofCentralPapua New GuineaCentralParaguayCentral
Cotê d’IvoireCentralPhilippinesCentralPeruPublic Sector
Equatorial GuineaCentralSamoaCentralSt. Kitts and NevisPublic Sector
EritreaCentralSolomonCentralSt. LuciaCentral
EthiopiaGeneralSri LankaCentralSt. Vincent & the GrenadinesCentral
GabonCentralThailandCentralSurinameCentral
Gambia, TheCentralTongaCentralTrinidad & TobagoCentral
GhanaCentralVanuatuCentralUruguayCentral
GuineaCentralVietnamGeneralVenezuelaCentral
Guinea BissauCentralAlgeriaCentralAlbaniaGeneral
KenyaCentralBahrainCentralBosnia & HerzegovinaGeneral
LesothoCentralDjiboutiCentralBulgariaGeneral
LiberiaCentralEgyptGeneralCroatiaCons. Central
MadagascarCentralIranGeneralCyprusCentral
MalawiCentralJordanCentralCzech RepublicGeneral
MaliCentralKuwaitCentralHungaryGeneral
MauritiusCentralLebanonCentralMacedoniaGeneral
MozambiqueCentralLibyaCentralMaltaCentral
NamibiaCentralMauritaniaCentralNetherlands AntillesCentral
NigerCentralMoroccoCentralPolandGeneral
NigeriaCentralOmanCentralRomaniaGeneral
RwandaCentralQatarCentralSlovak RepublicGeneral
São Tomé & PríncipeCentralSaudi ArabiaCentralTurkeyCentral
SenegalCentralSyriaCentralArmeniaCentral
SeychellesCentralTunisiaCentralAzerbaijanGeneral
Sierra LeoneCentralUnited Arab EmiratesCentralBelarusGeneral
South AfricaGeneralWest Bank & GazaCentralEstoniaGeneral
SwazilandCentralYemenCentralGeorgiaGeneral
TanzaniaCentralArgentinaGeneralKazakhstanGeneral
TogoCentralBahamas, TheCentralKyrgyz RepublicGeneral
UgandaCentralBarbadosCentralLatviaGeneral
ZambiaCentralBelizeCentralLithuaniaGeneral
ZimbabweCentralBoliviaPublic SectorMoldovaGeneral
BangladeshCentralBrazilFederalRussiaGeneral
BhutanCentralChileCentralTajikistanCentral
CambodiaGeneralColombiaCentralTurkmenistanGeneral
ChinaGeneralCosta RicaCentralUkraineGeneral
FijiCentralDominicaCentralUzbekistanGeneral
Appendix C. Countries
Countries in sample for stylized facts on infant mortality
AlgeriaCote d’ IvoireLao PDRPeru
ArgentinaDominican RepublicLesothoPhilippines
BarbadosEcuadorLiberiaRwanda
BangladeshEgypt, Arab Rep.MadagascarSenegal
BelizeEl SalvadorMalawiSouth Africa
BeninEquatorial GuineaMaliSri Lanka
BoliviaEritreaMauritaniaSudan
BotswanaEthiopiaMalaysiaSwaziland
BrazilFijiMauritiusSyrian Arab Republic
Burkina FasoGambia, TheMexicoTajikistan
BurundiGhanaMongoliaTanzania
CambodiaGuatemalaMoroccoThailand
CameroonGuinea-BissauMozambiqueTogo
Cape VerdeHaitiNamibiaTrinidad and Tobago
Central African RepublicHong Kong, ChinaNepalTunisia
ChadHondurasNicaraguaUganda
ChileIndiaNigerUruguay
ChinaIndonesiaNigeriaVenezuela, RB
ColombiaIsraelPakistanVietnam
ComorosJamaicaPanamaZambia
Congo, Rep.JordanPapua New GuineaZimbabwe
Costa RicaKenyaParaguay
Countries in base sample for infant mortality regressions
AlgeriaChinaGhanaMalawiPakistanTanzania
BangladeshColombiaGuatemalaMaliPanamaThailand
BeninCosta RicaHondurasMauritaniaPapua New GuineaTrinidad and Tobago
BoliviaDominican RepublicIndiaMongoliaParaguayTunisia
BrazilEcuadorIndonesiaMoroccoPeruUganda
Burkina FasoEgypt, Arab Rep.JamaicaMozambiquePhilippinesVietnam
CambodiaEl SalvadorJordanNepalRwandaZambia
CameroonEritreaKenyaNicaraguaSenegalZimbabwe
ChadEthiopiaLao PDRNigerSri Lanka
ChileGambia, TheMadagascarNigeriaSwaziland
Countries in the base sample for regressions on adult illiteracy and stylized facts
AlgeriaChinaIndonesiaMoroccoSwaziland
ArgentinaColombiaIsraelMozambiqueSyrian Arab Republic
BangladeshCongo, Rep.JamaicaNamibiaThailand
BarbadosCosta RicaJordanNepalTogo
BelizeCote d’ IvoireKenyaNicaraguaTrinidad and Tobago
BeninDominican RepublicLao PDRNigerTunisia
BoliviaEcuadorLebanonNigeriaUganda
BotswanaEgypt, Arab Rep.LesothoPakistanUruguay
BrazilEl SalvadorMadagascarPanamaVenezuela, RB
Burkina FasoEthiopiaMalawiParaguayVietnam
BurundiFijiMalaysiaPeruZambia
CambodiaGambia, TheMaliPhilippinesZimbabwe
CameroonGhanaMauritaniaSenegal
Central African RepublicGuatemalaMauritiusSouth Africa
ChadHondurasMexicoSri Lanka
ChileIndiaMongoliaSudan
Countries in the base sample used for regressions on government effort in health
AlgeriaColombiaJordanNamibiaSierra Leone
ArgentinaCyprusKenyaNepalSouth Africa
BoliviaEcuadorLebanonNicaraguaSri Lanka
BotswanaEgypt, Arab Rep.LesothoPanamaSwaziland
BrazilEl SalvadorMadagascarPapua New GuineaThailand
Burkina FasoEthiopiaMalaysiaParaguayTunisia
BurundiFijiMauritiusPeruUganda
CameroonGuineaMexicoPhilippinesUruguay
ChileIndiaMongoliaSenegalYemen, Rep.
ChinaIndonesiaMoroccoSeychellesZimbabwe
Countries in the base sample used for regressions on government effort in education
AlgeriaChinaGuineaMauritiusParaguayThailand
ArgentinaColombiaIndiaMexicoPeruTunisia
BoliviaCyprusIndonesiaMongoliaPhilippinesUganda
BotswanaEcuadorJordanMoroccoSenegalUruguay
BrazilEgypt, Arab Rep.KenyaNamibiaSeychellesYemen, Rep.
Burkina FasoEl SalvadorLebanonNepalSierra LeoneZimbabwe
BurundiEthiopiaLesothoNicaraguaSouth Africa
CameroonFijiMadagascarPanamaSri Lanka
ChileGhanaMalaysiaPapua New GuineaSwaziland
Appendix D. Millennium Development Goals

Goal 1: Eradicate extreme poverty and hunger

Target 1: Halve, between 1990 and 2015, the proportion of people whose income is less than $1 a day

Target 2: Halve, between 1990 and 2015, the proportion of people who suffer from hunger

Goal 2: Achieve universal primary education

Target 3: Ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling

Goal 3: Promote gender equality and empower women

Target 4: Eliminate gender disparity in primary and secondary education preferably by 2005 and in all levels of education no later than 2015

Goal 4: Reduce child mortality

Target 5: Reduce by two-thirds, between 1990 and 2015, the under-five mortality rate

Goal 5: Improve maternal health

Target 6: Reduce by three-quarters, between 1990 and 2015, the maternal mortality ratio

Goal 6: Combat HIV/AIDS, malaria, and other diseases

Target 7: Have halted by 2015 and begun to reverse the spread of HIV/AIDS

Target 8: Have halted by 2015 and begun to reverse the incidence of malaria and other major diseases

Goal 7: Ensure environmental sustainability

Target 9: Integrate the principles of sustainable development into country policies and program and reverse the loss of environmental resources

Target 10: Halve, by 2015, the proportion of people without sustainable access to safe drinking water and basic sanitation

Target 11: Have achieved, by 2020, a significant improvement in the lives of at least 100 million slum dwellers

Goal 8: Develop a global partnership for development

Target 12: Develop further an open, rule-based, predictable, nondiscriminatory trading and financial system (includes a commitment to good governance, development, and poverty reduction—both nationally and internationally)

Target 13: Address the special needs of the least developed countries (includes tariff-and quota-free access for exports enhanced program of debt relief for HIPC and cancellation of official bilateral debt, and more generous ODA for countries committed to poverty reduction)

Target 14: Address the special needs of landlocked countries and small island developing states (through the Program of Action for the Sustainable Development of Small Island Developing States and 22nd General Assembly provisions)

Target 15: Deal comprehensively with the debt problems of developing countries through national and international measures in order to make debt sustainable in the long term

Target 16: In cooperation with developing countries, develop and implement strategies for decent and productive work for youth

Target 17: In cooperation with pharmaceutical companies, provide access to affordable, essential drugs in developing countries

Target 18: In cooperation with the private sector, make available the benefits of new technologies, especially information and communication

References

    AlesinaAlberto and DavidDollar2000Who Gives Foreign Aid to Whom and Why?Journal of Economic GrowthVol. 5 (March) pp. 3363.

    • Search Google Scholar
    • Export Citation

    Alonso-BorregoCesar and ManuelArellano1996Symmetrically Normalized Instrumental Variable Estimation Using Panel DataCEMFI Working Paper No. 9612 (Madrid: Centro de Estudios Monetarios y Financieros).

    • Search Google Scholar
    • Export Citation

    BarroRobert J. and XavierSala-i-Martin1995Economic Growth (New York: McGraw-Hill)

    BlundellRichard and StephenBond1999GMM Estimation with Persistent Panel Data: An Application to Production FunctionsIFS Working Paper W99/04 (London: Institute for Fiscal Studies).

    • Search Google Scholar
    • Export Citation

    BlundellRichard and StephenBond1998Initial Conditions and Moment Restrictions in Dynamic Panel Data ModelsJournal of EconometricsVol. 87 (August) pp. 11543.

    • Search Google Scholar
    • Export Citation

    BondStephen2002Dynamic Panel Data Models: A Guide to Micro Data Methods and PracticeCeMAPP Working Paper CWP09/02 (London: Institute for Fiscal Studies).

    • Search Google Scholar
    • Export Citation

    BondStephen and FrankWindmeijer2002Finite Sample Inference for GMM Estimators in Linear Panel Data ModelsCeMAPP Working Paper CWP04/02 (London: Institute for Fiscal Studies).

    • Search Google Scholar
    • Export Citation

    BooneP.1996Politics and the Effectiveness of Foreign AidEuropean Economic ReviewVol. 40 (February) pp. 289329.

    BurnsideCraig and DavidDollar1998Aid, the Incentive Regime, and Poverty ReductionPolicy Research Working Paper 1937 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    BurnsideCraig and DavidDollar2000Aid, Policies and GrowthAmerican Economic ReviewVol. 90 (September) pp. 84767.

    BurnsideCraig and DavidDollar2004Aid, Policies and Growth: Revisiting the EvidencePolicy Development Working Paper 3251 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    CaldwellJ. C.1986Routes to Low Mortality in Poor Countries,Population and Development Review pp. 171-220.

    CheneryHollis B. and Alan M.Strout1966Foreign Assistance and Economic DevelopmentAmerican Economic Reviewvol. 56 (September) pp. 679733.

    • Search Google Scholar
    • Export Citation

    ClemensMichael A.StevenRadelet and RikhilBhavnani2004Counting Chickens When They Hatch: the Short-Term Effect of Aid on GrowthCenter for Global Development Working Paper 44 (Washington).

    • Search Google Scholar
    • Export Citation

    CollierPaul and JanDehn2001Aid, Shocks, and GrowthPolicy Research Working Paper 2688 (Washington: World Bank).

    CollierPaul and DavidDollar2002Aid Allocation and Poverty ReductionEuropean Economic ReviewVol. 46 (September) pp. 14751500.

    • Search Google Scholar
    • Export Citation

    ClarkJohn1995The State, Popular Participation, and the Voluntary SectorWorld DevelopmentVol. 23 (April) pp. 593601.

    CoxAidan and AntoniqueKoning1997EC Cooperation With NGO: An Evaluation” (Brussels: European Commission).

    DalgaardCarl-JohanHenrikHansen and FinnTarp2004On the Empirics of Foreign Aid and GrowthEconomic JournalVol. 114 (June) pp. F191F216.

    • Search Google Scholar
    • Export Citation

    DalgaardCarl-Johan and HenrikHansen2001On Aid, Growth and Good PoliciesJournal of Development StudiesVol. 37 (August) pp. 1741.

    • Search Google Scholar
    • Export Citation

    DavidsonRussell and James G.MacKinnon1993Estimation and Inference in Econometrics (New York: Oxford University Press).

    DevarajanShantayanan and VinayaSwaroop1998The Implications of Foreign Aid Fungibility for Development AssistancePolicy Research Working Paper 2022 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    DrezeJ. and A.Sen (1989). Hunger and Public Action. OxfordClarendon Press.

    EasterlyWilliam1999The Ghost of Financing Gap: Testing the Growth Model Used in the International Financial InstitutionsJournal of Development EconomicsVol. 60 (December) pp. 42338.

    • Search Google Scholar
    • Export Citation

    EasterlyWilliam2003Can Foreign Aid Buy Growth?Journal of Economic PerspectivesVol. 17 (Summer) pp. 2348.

    EasterlyWilliamRossLevine and DavidRoodman2003New Data, New Doubts: A Comment on Burnside and Dollar’s ‘Aid, Policies and Growth 2000’” American Economic Review.

    • Search Google Scholar
    • Export Citation

    EdwardsMichael and DavidHulme1997Too Close for Comfort? The Impact of Official Aid on NGOs” in NGOs States and Donors: Too Close for Comfort?ed. by Edwards and Hulme (London: Pergamon Press).

    • Search Google Scholar
    • Export Citation

    FeyziogluTarhanVinayaSwaroop and MinZhu1996Foreign Aid’s Impact on Public SpendingPolicy Research Working Paper 1610 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    FeyziogluTarhanVinayaSwaroop and MinZhu1998A Panel Data Analysis of the Fungibility of Foreign AidWorld Bank Economic ReviewVol. 12 (January) pp. 2958.

    • Search Google Scholar
    • Export Citation

    GauriVarun and AnnaFruttero2003Location Decisions and Nongovernmental Organization Motivation: Evidence from Rural BangladeshPolicy Research Working Paper 3176 (Washington: World Bank).

    • Search Google Scholar
    • Export Citation

    GuillaumontPatrick and LisaChauvet2001Aid and Performance: A ReassessmentJournal of Development StudiesVol. 37 (August) pp. 6692.

    • Search Google Scholar
    • Export Citation

    GuptaSanjeev and others 2003Foreign Aid and Revenue Response: Does the Composition of Aid Matter?IMF Working Paper 03/176 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    HansenHenrik and FinnTarp2000Aid Effectiveness DisputedJournal of International DevelopmentVol. 12 (April) pp. 37598.

    • Search Google Scholar
    • Export Citation

    HansenHenrik and FinnTarp2001Aid and Growth RegressionsJournal of Development EconomicsVol. 64 (April) pp. 547570.

    HadjimichaelMichael T. and others 1995Sub-Saharan Africa: Growth, Saving, and Investment, 1986–93IMF Occasional Paper No. 118 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    LensinkRobert and HowardWhite2000Aid Allocation, Poverty Reduction and the Assessing Aid ReportJournal of International DevelopmentVol. 12 (April) pp. 399412.

    • Search Google Scholar
    • Export Citation

    LensinkRobert and HowardWhite2001Are There Negative Returns to Aid?Journal of Development StudiesVol. 37 (August) pp. 4265.

    • Search Google Scholar
    • Export Citation

    McGillivrayMark and OliverMorrisey2001Aid Illusion and Public Sector Fiscal BehaviourJournal of Development StudiesVol. 37 (August) pp. 11836.

    • Search Google Scholar
    • Export Citation

    MeyerCarrie A.1995Opportunism and NGOs: Entrepreneurship and Green North-South TransfersWorld Development23 (August) pp. 127789.

    • Search Google Scholar
    • Export Citation

    MosleyPaulJohnHudson and SaraHorrell1992 “Aid The Public Sector and The Market in Less Developed Countries: A Return to The Scene of Crime “Journal of International DevelopmentVol. 4 (March-April) pp. 13950.

    • Search Google Scholar
    • Export Citation

    MosleyPaulJohnHudson and ArjanVershoor2004Aid, Poverty Reduction and the New ConditionalityThe Economic JournalVol. 114 (June) pp. F217F243.

    • Search Google Scholar
    • Export Citation

    MurshedS. Mansoob and SomnathSen1995Aid Conditionality and Military Expenditure Reduction in Developing Countries: Models of Asymmetric InformationEconomic JournalVol. 105 (March) pp. 498509.

    • Search Google Scholar
    • Export Citation

    OECD and UNDP1999Improving the Effectiveness of Aid Systems: The Case of Mali (Paris: Organization for Economic Cooperation and Development).

    • Search Google Scholar
    • Export Citation

    PackHoward and Janet RothenbergPack1993Foreign Aid and the Question of FungibilityReview of Economics and StatisticsVol. 75 (May) pp. 25866.

    • Search Google Scholar
    • Export Citation

    PeterssonJ.2004Foreign Aid Fungibility, Growth, and Poverty ReductionWorking Paper (unpublished; Stockholm: Stockholm University).

    • Search Google Scholar
    • Export Citation

    ReichelRichard1995Development Aid, Savings and Growth in the 1980s: A Cross-Section AnalysisSavings and DevelopmentVol. 19No. 3 pp. 27996.

    • Search Google Scholar
    • Export Citation

    Robinson1997Privatising the Voluntary Sector: NGOs as Public Sector Contractors” in NGOs, States and Donors: Too Close for Comfort?ed. by MichaelEdwards and DavidHulme (London: Pergamon Press).

    • Search Google Scholar
    • Export Citation

    Rose-AckermanSusan1996Altruism, Nonprofits and Economic TheoryJournal of Economic LiteratureVol. 34 (June) pp. 70128.

    • Search Google Scholar
    • Export Citation

    SnyderDonald W.1993Donor Bias Towards Small Countries: An Overlooked Factor in the Analysis of Foreign Aid and Economic GrowthApplied EconomicsVol. 25 (April) pp. 48188.

    • Search Google Scholar
    • Export Citation

    SvenssonJakob2000Foreign Aid and Rent-SeekingJournal of International EconomicsVol. 51 (August) pp. 43761.

    Tavares José2003Does Foreign Aid Corrupt?Economics LettersVol. 79 (April) pp. 99106.

    TsikataTsidi M.1998Aid Effectiveness: A Survey of the Recent Empirical LiteratureIMF Paper on Policy Analysis and Assessment 98/1 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    WhiteHoward and OliverMorrissey1997Conditionality When Donor and Recipient Preferences VaryJournal of International DevelopmentVol. 9 (June) pp. 497505.

    • Search Google Scholar
    • Export Citation
1Nadia Masud was a summer intern in the IMF Institute when this paper was being prepared; she is a PhD candidate at the University of Oxford, United Kingdom. Boriana Yontcheva is an economist at the IMF Institute. We gratefully acknowledge the assistance of France Marion, Lisbeth Ekelof and César-Luis Valor-Arce, European Commission who provided key data for this research. We are indebted to B. Baltagi for guidance and discussions on the econometrics of this paper. J. Salvati provided invaluable technical assistance. A. Berg, T. Cordella, A. Feltenstein, and participants at an IMF Institute seminar provided many useful suggestions. Errors are solely ours.
3For a review of progress on the International Development Goals, see www.paris21.org/betterworld.
4Refer to Appendix A for a definition of these variables.
5Appendix C lists the countries included in this sample.
6Refer to Appendix C for a list of these countries.
8For a more detailed exposition of the relationship between the EU and NGOs, see Cox and Koning (1997).
9The Davidson and MacKinnon test showed that regressors were all strictly exogenous.
10Refer to Appendix C for the sample.
11NGO aid is defined as projects implemented by foreign NGOs and, hence, is a resource for the recipient country that is outside the budget.
12Refer to Appendix C for the sample.
13We used the Davidson MacKinnon test for exogeneity.
17Refer to Appendix C for the sample.
18Refer to Appendix C for the sample.
19As shown by Clemens, Radelet, and Bhavnani, (2004), decomposing bilateral aid into different structural categories cam yield very different results in terms of its effectiveness.

Other Resources Citing This Publication