List of Abbreviations
List of Tables
List of Figures
List of Appendices
2 The Debate on Aid Effectiveness
2.1 The aid-growth literature
2.2 Looking beyond the effect of growth
2.3 Evidence from donor investigations
2.3.1 The aid allocation literature: Deriving donor interests
2.3.2 Determining donor differences
3 Research objectives
4 Preliminary Empirical Analyses
4.1 Data sources and definitions
4.2 Disaggregating aid - descriptive statistics
5 A first approximation: Cross-Country Evidence
5.1 The effect on growth
5.2 The effect on social indicators
5.3 Post-Cold War changes
6 Insights from a Dynamic Panel Data Model
6.1 Reinvestigating the effect on growth
6.2 Reinvestigating the effect on social indicators
6.3 Robustness tests
7 Policy Implications
List of Abbreviations
Abbildung in dieser Leseprobe nicht enthalten
List of Tables
1 Group assignment of bilateral donors
2 Aid allocation of bilateral donor groups
3 OLS cross-sectional estimations: Aid and growth
4 OLS cross-sectional estimations: Aid and infant mortality
5 OLS cross-sectional estimations: Aid and primary completion
6 First evidence of post-Cold War changes in aid effectiveness
7 Aid and growth: GMM panel data evidence
8 Aid and social indicators: GMM panel data evidence
9 Robustness of panel data estimations - the effect on growth
10 Robustness of panel data estimations - the effect on social indicators
List of Figures
1 The development of aid payments by donor
2 Conditional relationships between aid and growth
2a) G1 aid and growth
2a) G2 aid and growth
List of Appendices
Appendix A: The UN Development Goals
Appendix B: Data description and sources
Appendix C: Patterns of aid allocation
Appendix D: Cross-sectional estimation results
Appendix E: GMM estimation results: Aid and growth
Appendix F: GMM estimation results: Aid and social indicators
Appendix G: Robustness tests - main regression
Appendix H: Additional tests for robustness
Appendix I: Hausman specification tests
No Poverty, Zero Hunger, Good Health, Well-being and Quality Education - these are the first priorities of the Sustainable Development Goals (SDGs) that were launched jointly by all UN Member States on 1 January 2016. The agenda of this agreement contains 17 main goals with a total of 169 targets and is dedicated to improving global living conditions and to address issues of environmental and economical sustainability with a planning horizon through to 2030. Development assistance from economically advanced countries, also referred to as aid, is one of the major means to provide financing for countries with less developed economies that face severe social problems and which often cannot handle these problems alone.
Half a century ago, several major industrialised countries began to launch large aid programmes after establishing the Development Assistance Committee (DAC) that serves as a platform for dialogue and discussion about the way international donors can contribute most effectively to the development of economically poorer countries. Since 1964, aid flows from DAC member countries are recorded as Official Development Assistance (ODA) and corresponding data is made publicly available. The number of donors, as well as the amount of funds, is consistently growing, reaching a maximum of almost USD 95 billion spent by 30 member countries in 2014. Over certain periods of time, aid payments accounted for a large share of the domestic economy of single recipient countries. In the most extreme cases, aid flows from various donors added up to more than 50% of the government budget - often spanning several years.
Such extensive monetary flows that aim to improve economic and social conditions in a broad range of developing countries were unknown until that time, leaving open the question of the resulting effect. Because the topic also implies a moral importance due to the donors’ commitment to eliminate global poverty and inequality, the topic immediately attracted world-wide interest among the research society. The first studies dealing with the subject appeared almost simultaneously as the first aid flows were disbursed. These studies assess the possible impact of aid, theoretically based on former growth models. As soon as first data became available, researchers began to directly estimate the effect of aid on economic parameters. By now, the topic of aid effectiveness covers an immense amount of contributions with a variety of different research fields.
Estimation methods, as well as quality and availability of data on developing countries have improved over the time. However, despite the emergence of more accurate methodology and an increasing number of scientific contributions, the entire history of research on this topic is accompanied by a lack of consensus. On the one hand, several studies find that aid has a significant positive impact on growth, whereas an equally large share of contributions cannot find an effect at all, or even demonstrate a negative correlation between aid and growth. The disagreement among researchers is not only consistent over time, but also across different estimation methods and even between studies that apply the same approaches based on identical data. The perennial debate that has emerged over the past twenty years particularly leaves the reader with the impression that research in this area is going around in circles. Studies that take up the finding of a preceding investigations, apply small changes to the original approach and end up with a different result largely shape the image of the latest aid effectiveness literature - one can therefore conclude that the question whether aid as a positive impact on growth or not still remains unanswered.
It is certain, however, that aid in general is falling short of the initial expectations that donors and first researchers had. Therefore, a wide range of the literature aims to find out why aid does not show to have the desired effect. Several studies find that aid increases non-productive government expenditures instead of benefitting the poor. Others argue that aid only shows to be effective when directed to recipient countries with a high level of institutional quality and good policies, and conclude that funds have largely been transferred to the wrong countries. A further avenue of research that has developed over the past decades pursues a different objective than estimating the effect of aid: By analysing the pattern of aid flows, the aid allocation literature determines the main driving forces behind the donors’ choice of recipient countries. Resulting insights prove to be plausible explanations for the missing effectiveness of aid. The parameters that presumably shape the allocation decisions are either related to the recipient, such as humanitarian needs, institutional quality or the level of economic openness; or to the donor, e.g. political and economic interests or ties with recipients that have formerly been colonised by the donor.
One of the major conclusions of one strand of the aid allocation literature, most of which published during the past 15 years, is that there are striking differences among donors regarding their prioritisation of recipient needs or their own objectives. Several studies find that major industrialised countries, such as the US, Japan or France, put less emphasis on recipient needs and more on their own objectives. In contrast, aid from multilateral donors or smaller bilateral donors, such as Scandinavian countries or the Netherlands, is considered to be far less driven by economic or power-political interests and proves to be distributed with a high regard for the different recipient-related parameters.
Taking this observation into consideration, it appears evident that the effectiveness of aid, equal to the pattern of allocation, differs among donors. Yet, nearly all investigations focus on the effect of aggregated aid - i.e. total aid funds from all donors. Only in recent years, research begins to consider this issue, mostly distinguishing between the impact of bilateral and multilateral aid. Empirical evidence on potential differences in aid effectiveness between single bilateral donors, however, remains very limited. Because bilateral aid accounts for most of the total supplied funds, further research on this topic is indispensable.
Consequently, the main research objective of this thesis is to investigate whether the insight of the aid allocation literature, i.e. bilateral donors differ significantly regarding the motives behind allocating aid, can be applied to determine measurable differences in aid effectiveness. In other words, are differences in the way distinct donor countries allocate aid large enough to result in quantitative differences in the impact on develop- ment of recipient countries? The answer to this question could offer an explanation for why aid research often fails to find a positive effect of aid. And more importantly, it could provide insight on the types of aid that appear to be unrelated with development and those that show a positive impact on recipient countries.
In order to estimate differences between bilateral donors, this study forms two groups that are assumed to allocate aid in a significantly different manner - an approach that is not applied up to the present. The first group comprises Denmark, Finland, the Netherlands, Norway and Sweden; all of which seem to exclude personal interest and to put great emphasis on recipient needs. France, Germany, Japan, the United States of America (US) and the United Kingdom (UK) form the second group as bilateral donors that are considered to allocate aid with high regard to their own objectives. To determine the possible difference in aid effectiveness between these two donor groups, the aim of this thesis is to follow the latest and established strategies of estimating the effect of aid on growth in a first place.
As several recent studies argue that growth is not the only appropriate measure of development, this thesis additionally investigates the impact of aid on the two most often researched social indicators. These are infant mortality, as a key indicator for heath standards, and primary education to measure the quality of the educational system. Both of these indicators are explicitly included in the agenda of the eight Millennium Development Goals (MDGs) of 2000 and are concrete targets of the current SDGs.
Thus, an analysis of the effect on the two indicators also covers a more concrete answer to the question whether aid contributes to sustainable development goals. Moreover, because the literature does not agree on whether or not multilateral aid is more effective than bilateral aid, the analyses of this study are accompanied by a further differentiation between these two types of aid. In addition to the separate analysis of two bilateral donor groups, this kind of disaggregating aid offers a further opportunity to decompose aid into potentially ineffective or beneficial types. As bilateral aid comprises both donor groups of the main investigation, this additional analysis enables a direct comparison of aggregated and disaggregated aid flows, which possibly achieves more valuable results.
The investigations of this study consider two further insights of the previous literature. The first one is that aid is assumed to affect development with a partially large time lag. The rationale behind this assumption is that development assistance largely comprises projects in areas such as infrastructure, energy, education or health, which require a long implementation period until they may achieve benefits. Second, several studies suggest that the aid allocation of major donors has significantly changed since the end of the Cold War, implying changes in the underlying motives. As the latter finding indicates that aid effectiveness increases after 1989, additional estimations shall reduce the observation period to 1990-2014.
In order to gradually answer these research questions, the study first presents a detailed investigation of the existing literature, beginning with a section that outlines the origin and history of the research field. This section aims at identifying problems and solutions associated with the estimation of aid effects and additionally addresses the current debate on the impact of aid on growth. The closer inspection of the recent research covers studies from the last 20 years; the aid effectiveness literature has seen a remarkable upswing during this period, which has led to a persistent discussion. In a next step, the literature review shifts the focus away from the aid-growth debate to the usage of alternative development indicators for measuring the impact of aid.
A further section then discusses those investigations that determine the pattern of aid allocation. Studies of this strand of literature provide an important theoretical foundation for this thesis, i.e. the indication that aid effectiveness might differ between donors.
Investigations of this kind also demonstrate changes in motives for providing aid after the end of the Cold War. The insights of the aid allocation literature lead to a final section that discusses the approaches and results of previous studies that attempt to measure differences in aid effectiveness among donors. Following the literature review, a separate chapter utilises the obtained insights in order to draw implications for the subsequent analysis and to precisely define the research objectives of this study.
A descriptive analysis initiates the empirical part of this study. After describing the applied data and corresponding issues, the main section of this chapter attempts to confirm the findings of the aid allocation literature in the case of the donors that are analysed in this study. An investigation of the temporal development of each donor’s aid flows, followed by a more detailed analysis of the individual pattern of aid allocation shall amplify the evidence achieved from the literature. Subsequent cross-sectional estimations of the actual aid effects provide reader with a first approximation of the results among different donors. The analysis of cross-country data based on ordinary least squares (OLS) method is among the first approaches for estimating the effect of aid, and is still applied by a number of recent studies. The initial assessment is divided into three parts. The first section identifies donor-specific aid effects on growth as the indicator that is most-often applied for estimating the impact on development in recipient countries. The second part observes both infant mortality and primary completion as social indicators. A final section approximates how aid effects vary in the post-1989 period after the end of the Cold War.
The application of cross-sectional data, however, is for several reasons afflicted with a certain inaccuracy, compared to more sophisticated estimation strategies based on panel data that have emerged over the past decades. The main estimation method of choice, therefore, is a generalised method of moments (GMM) based dynamic panel analysis.
Chapter 6, as the main of the empirical analysis, begins by discussing the benefits of the GMM dynamic panel data approach in contrast to cross-sectional investigation. Subsequently, the effect on growth and on both social indicators is reassessed. As the application of GMM estimators is not free from bias either, a third section of chapter 6 tests for robustness of the panel data approach. This is based on simple OLS methodology and a fixed effects (FE) estimator, both of which are also applied in recent research. A further step is to derive policy implications from the obtained results, and chapter 8 concludes.
2 The Debate on Aid E ectiveness
Following the establishment of the Development Assistance Committee in 1960, nearly all the major developed countries started extensive aid programmes with the objective of alleviating global poverty (cf. Doucouliagos and Paldam 2006: 227). Shortly afterwards, first researchers began to assess the effects of aid against the high expectations of the donors. The first studies were published at the end of the 1960s and the early 1970s.
Since then, a great variety of researchers have been dealing with the question whether and to which extent development assistance impacts economic growth of recipient countries. Underlying theories and the methodical procedure to filter out the effect of aid have been reconsidered and adapted to more sophisticated empirical approaches several times up to the present.
More surprisingly, but very characteristic for the aid research, the numerous outcomes for the effect of aid vary to a high extent and are inconsistent in sign and degree up until now. For this reason, and for the great number of contributions in this area, multiple studies attempt to categorise the research field by classifying contributions by their findings or the methodical framework put to use. Scientific approaches from the first decades of research have repeatedly been refined and further developed and hence are less relevant for the empirical part of this thesis. Yet, their contributions are useful for understanding the nature of the aid effectiveness debate and shall, thus, be summarised briefly. Because the most influential and still discussed studies have been published over the past 20 years, their procedure and findings shall be analysed in more detail.1
At the beginning of the debate in this field, aid effectiveness was generally seen in an economical context. After initial attempts to analyse the impact on more precise macroeconomic determinants such as savings and investment, the approach of analysing the direct impact on economic growth has prevailed so far in this strand of the literature.
A first section explains this development and the predominant strand of research, the aid-growth literature. Recently, several contributions consider the impact of aid on social indicators in addition to growth. A second section illustrates the reasons behind this choice, as well as the approaches and outcomes of influential studies in this field.
2.1 The aid-growth literature
The first contributions from the early 1960s mostly focus on the estimated capital required by developing countries. Theoretically based on the Harrod-Domar growth model, these studies address the relationship between foreign assistance and total savings with the underlying assumption that filling up the saving gap implies higher investments and therefore an increase in growth. With a positive expectation towards the effect of aid on growth, they assume that aid payments contribute in an equal amount to the developing countries’ savings and thus investment, such as Rosenstein-Rodan (1961: 136), to name one of the most influential studies of this period.
The arguments of a subsequent strand of the literature likewise draw on the effect on savings. However, questioning the high expectations of the previous literature, these studies find out a negative impact of foreign assistance on domestic savings of developing countries. Since they attempt to determine the extent of aid for the first time by applying cross-sectional or time series analyses, they are often regarded the first contributions to the aid effectiveness literature (cf. Clemens et al. 2012: 591). The earliest publication of this strand of the aid-savings literature, Griffin and Enos (1970: 320), states that the former expected relationship between aid and savings would be too simple and that a general negative tendency between foreign assistance and domestic savings points towards a negative effect on growth. Weisskopf (1972: 37) even concludes that the effect of aid on ex ante domestic savings would be significantly negative.
Soon afterwards, Papanek (1972, 1973) marks an early turning point in the literature by shifting away the focus from analysing the effect on savings towards directly measuring the impact of aid on growth. He argues that savings had so far been determined by subtracting inflows from investment, which would lead to a self-explanatory negative correlation (cf. Papanek 1972: 945-947). Considering the issue of causality between low savings in least developed countries (LDC) and high aid payments for the first time, he questions the validity of previous results. Furthermore, he disputes the approaches of his predecessors because of their aggregation of all foreign inflows. He attempts to separate these, resulting in a regression for predicting the various effects of aid, savings, investment and other components of foreign inflows on growth (Papanek 1973: 121) and concludes that there is a strongly significant correlation between aid and growth (ibid.: 129).
The approach of Papanek is taken up by a series of subsequent studies that investigate the effect of aid on either investment. Amongst these is Mosley (1980: 81-84), who introduces a further methodical advancement to the aid literature by instrumenting for aid in a two stage least squares regression. He comes to the conclusion that the overall relation between aid and growth is negative and insignificant, but positive when only considering the poorest countries (ibid.: 89-90). Although several subsequent publications take up this procedure, the actual impact of aid remains controversial: Levy (1988: 1781-1787), for instance, finds a strongly significant positive impact on both investment and growth in African countries. Snyder (1993: 484) even reports an overall positive effect of aid after controlling for population growth of the recipient countries.
On the opposite side, various studies find that aid and growth prove are unrelated, such as Mosley et al. (1987: 636).
The recent literature, considering the current status, is introduced by the work of Boone (1996). With the help of more complete data, he performs panel data analyses with five-year averages while controlling for country specific effects (ibid.: 305). Moreover, he uses three new instruments for dealing with potential endogeneity in the regressors and analyses aid effects on growth, infant mortality, life expectancy and primary school attendance (ibid. 308-309). Boone (1996: 322-323) arrives at the conclusion that aid does not show to have any significant effect on neither of the parameters. Instead, he observes a connection between aid and higher total consumption. He states at the same time, however, that there is no evidence that higher consumption would benefit the poor (ibid.: 315). Boone’s findings have been addressed by many studies up to the present.
The literature that has emerged since then can be classified into three strands of studies - in the following also termed studies of strand 1-3.2 A first line of studies finds a positive impact of aid under certain circumstances, such as the existence of specific policies or relations between the donor country and the recipient. A second series of studies, strand 2, confirms the results of Boone, i.e. aid and growth are uncorrelated. On the contrary, a third strand of the literature determines a positive effect of aid that holds regardless of certain conditions.
The first strand, which Clemens et al. (2012) term the “conditional strand” (p. 592), comprises several of the most influential studies of the aid-growth literature. These studies reflect Boone’s results of an absent effect of aid on growth in a total sample, but determine a positive effect of aid only under certain circumstances and, thus, imply op- portunities of (re)structuring aid flows in an effective manner. The arguably best-known example is Burnside and Dollar (2000), which extend the instrumentation strategy by including an interaction term of aid and economic policies. After determining a signifi- cantly positive coefficient of this variable, Burnside and Dollar (2000: 857) conclude that aid works better in countries that feature strong economic policies and institutions.
Many subsequent papers, for example Easterly (2003: 24) or Clemens et al. (2004: 8), highlight the great influence of Burnside and Dollar (2000) on both research and the policies of major aid organisations, such as the World Bank and other multilateral development banks. Furthermore, there are various papers that confirm the relationship between good policies and the effectiveness of aid on growth, as for example Alvi et al. (2008: 702). Other studies belonging to this strand of the literature identify a positive effect of aid on growth in countries that are, for instance, geographically located outside the tropics (cf. Dalgaard et al. 2004: 201) or in a longer period of peace (cf. Collier and Doller 2002: 1135).
A further number of publications can be assigned to a second strand due to their common conclusion that aid and growth are unrelated, regardless of certain circumstances. To this category belongs Roodman (2007: 266), who tests the robustness of several influential first strand studies - amongst them Burnside and Dollar (2000) and Dalgaard et al. (2004). He concludes that all outcomes of the analysed studies are characterised by fragility and that aid “is probably not a fundamentally decisive factor for development” (ibid.: 275). One of the most prominent publications of the past decade, as many studies highlight (cf. Clemens et al. 2012: 595), can also be found in this category: Rajan and Subramanian (2008) analyse the causal relation between aid and growth based on both cross-sectional and panel data analyses in a comprehensive way. They test for several assumptions, such as the decisive role of good policies and the geographical location, as well as the timing of the impact of aid; and conclude that aid over all their findings has no significant positive effect on growth (ibid.: 660).
Opposed to the findings of the second cluster, the third strand of the recent literature reveals an overall positive impact of aid on growth. An early influential work is Hansen and Tarp (2001), who also address the insight of Burnside and Dollar (2000) that aid would only work in good policy environments. They are the first within the aid literature to apply a GMM estimator. This procedure has attracted particular attention within the aid literature and is adopted by many subsequent studies, such as the aforementioned Rajan and Subramanian (2008: 658-660).3 Hansen and Tarp (2001: 552) retain the strategy of instrumenting for policy, but modify the original set of instruments and include lagged values of the aid regressors for modelling the assumption of non-linear returns of aid. Accounting for both original least squares and GMM-estimations, Hansen and Tarp (2001: 563) conclude that aid has a positive and significant effect on growth via investment.
Gomanee et al. (2005A: 1068) arrive at the same result in a sample of 25 countries in Sub-Saharan Africa (SSA). Several publications of the past years can also be classified as studies of the third strand. Arndt et al. (2010: 19-21), as well as Clemens et al. (2012: 608) draw on the original specification of Rajan and Subramanian (2008) demonstrate a modest but general positive effect of aid on growth. Both studies have in common that they turn away from GMM-estimations, although this method has been recognised as a sound practice in the aid literature since the work of Hansen and Tarp (2001). The robustness check, carried out in section 6.3, revisits this recent doubt about the applicability of GMM-estimators. Finally, Brückner (2013: 131-135) also achieves a significantly positive effect of aid. Remarkable for his approach is that he tackles the endogeneity problem by explicitly estimating the effect of an increase in gross domestic product (GDP) per capita growth in recipient countries on the amount of foreign aid in a first step.
2.2 Looking beyond the e ect of growth
The investigation of the impact of aid on economic growth in developing countries may be considered the classical strand of the aid effectiveness literature. As indicated in the introduction, however, there are several reasons why the growth effect should not be the only criteria to use for measuring the effectiveness of aid. Several researches of the past two decades shift their focus towards investigating the effect of aid on different indicators of human development. These studies include key indicators in the areas of health and education, as well as poverty indices. In the following, this section analyses research in this rather novel field with a focus on the underlying theoretical assumptions for turning towards a focus on social indicators.
A fundamental objective of investigating the effectiveness of aid regarding its impact on social indicators is to evaluate whether it contributes to achieving the goals that have jointly been defined by the United Nations (UN) member countries. At the UN Millennium Summit in 2000, both developing and developed countries met for the first time to establish a common target catalogue for promoting international development, labelled Millennium Development Goals. These targets include improvements in the fields of global health, education, poverty reduction, equality, as well as peace and freedom promotion with an objective horizon of 15 years (cf. UNDP 2003: 31). After the expiration of the 15 years planning horizon, the common goals were redefined in 2016 and laid down in the 17 Sustainable Development Goals of the UN (cf. DESA: 2016).4 Because aid is expected to be an essential instrument for achieving these goals, several studies devote to explicitly analysing the correlation between aid and the achievement of MDGs, e.g. Dreher et al. (2008: 292) or Masud and Yontcheva (2005: 3).
Apart from the rather self-evident objective of assessing the ability of aid to contribute to commonly defined goals, there are more reasons why indicators for social welfare are suitable for measuring aid effectiveness. First, several studies point out that aid targeted towards improvements in social areas may only have an impact on growth in the very long run. Arndt et al. (2015: 9) demonstrate this time-lagged growth effect of aid based on the example of the education field. They consider only those kinds of aid which are directed at improving educational quality and show that higher school attendance rates, caused by these aid flows, may only have an observable influence on growth after a considerable share of beneficiaries has passed through the education system. Boone (1996: 293) already stresses that infant mortality, as a key indicator for health, would respond quickly to improved conditions. In line with this, Dreher et al. (2008) argue that social indicators would be “more specific outcome variables than growth” (292). A further line of arguments emphasises the high explanatory power of social indicators in measuring welfare gains of the poor. Gomanee et al. (2005B: 300) point out that increases in aggregate economic growth do not necessarily reflect better conditions of the poor; which is, however, a main target of aid. For this reason, they attempt to capture a broader measure of total welfare benefits by focusing on improvements in both infant mortality and the Human Development Indicator (HDI). Chong et al. (2009: 60), as well as Alvi and Senbeta (2012: 955) share the same position. In contrast to their predecessors, however, they both rely on poverty headcount and poverty gap indices. They argue that this procedure gives indication on income distribution and equality, representing additional important factors of development besides a countries’ average economic growth (ibid.: 957).
Corresponding to the overall conclusion of the classical aid-growth literature, the findings of this strand of the aid effectiveness research are also diverse. However, a majority of the studies report positive effects of aid. Recent findings, furthermore, show a clear trend towards more optimistic findings. Boone (1996: 312-313) is the first to incorporate social indicators into his estimations with the purpose of measuring improvements in the living conditions of the poor. For this purpose, he chooses infant mortality, life expectancy, as well as primary schooling. He finds out that aid does not significantly influence neither of the variables. As mentioned in the previous section, Boone rather observes a higher consumption rate correlating with aid payments and concludes that these results “are consistent with a model where politicians maximize welfare of a wealthy elite” (ibid.: 322).
In line with this finding, Chong et al. (2009: 79) are not able to detect a significant impact of aid on their measures for poverty. This negative overall picture of the relation between aid and social indicators, however, is only supported in a small number of studies. Gomanee et al. (2005B: 305), as well as Masud and Yontcheva (2005: 13) find that aid reduces infant mortality. While Gomanee et al. (2005B, 302) use total aggregated aid flows, Masud and Yontcheva (2005: 9) distinguish between bilateral aid (BA) and aid from non-governmental organisations (NGOs) and demonstrate that only the latter has a significant effect.
Particularly in recent years, a predominantly positive perception of aid towards im- proving social welfare has become apparent. Dreher et al. (2008: 299), for instance, show that aid specifically assigned to the education sector contributes significantly to higher school enrolment rates. In a similar approach, Mishra and Newhouse (2009: 865) consider aid in the health sector and find that this type of aid contributes to alleviating infant mortality. The two latter mentioned investigations apply a separation of aid flows by their purpose. Because their target is to investigate only the assistance for education and health respectively, they separate these aid types from the total aid numbers. Section 3.2 discusses the approach of disaggregating aid according to assignment in more detail.
Even without separating aid, Alvi and Senbeta (2012: 965) determine a negative and significant impact of total aid flows on both poverty gap and poverty rate. Finally, Arndt et al. (2015: 10-13) apply a two-step analysis with intermediate outcomes of aid, amongst them measures for health and education, that contribute accumulatively to final outcomes, represented by GDP growth and poverty indices. They conclude that aid has an average positive long-run effect on the final outcomes by stimulating the intermediate outcomes, such as better health and education conditions (ibid.: 15).
2.3 Evidence from donor investigations
So far, one major conclusion can be drawn from the literature analysis: A larger share of the research on aid effectiveness fails to find a significantly positive impact of aggregated aid flows (cf. Rajan and Subramanian 2008 and 2011, Burnside and Dollar 2000 or Easterly et al. 2004). For this reason, a major current research goal in this field is to gain insight into the reasons for the apparent failure of total aggregated aid. Section 2.1 already describes one empirical procedure for this purpose; by instrumenting for certain characteristics of the developing countries, such as their institutional quality or climatic circumstances, scholars try to find out under which circumstances aid generates positive effects and when it appears to not have any significant impact. Studies of this type have already had a strong influence on policy, such as the shift of aid allocation of major agencies towards countries with a better policy environment (cf. Easterly et al. 2004: 774).
A further possibility of understanding why overall aid apparently does not achieve its goals is to look at the single development assistance flows separately, i.e. disaggregating aid numbers. The underlying assumption behind this approach is that different aid, or aid flows to different countries, can vary in their effectiveness. The disaggregation, again, can be done in different ways. The two major differences of these approaches are the following: One way is to consider the underlying objective behind the different aid flows and, consequently, to distinguish between the different purposes of aid, which are defined by the donor. Many studies apply such a procedure, amongst them Headey (2008: 169) and Clemens et al. (2012: 594), who subtract the share of humanitarian aid from total ODA with the purpose of only considering the types of aid that are targeted at increasing economic growth, or Dreher et al. (2008: 300), who find a positive impact of aid and education indicators while considering only those aid flows that were assigned to the education sector. This thesis, however, does not apply such a procedure. Section 4.2 provides the argumentation behind this choice.
Beside this kind of disaggregation, there is a second established way of breaking down aggregated aid: Various studies subdivide aid flows according to the way they are allocated to recipient countries. This type of studies is often referred to as the aid allocation literature (cf. Harrigan and Wang 2011: 1282). One objective of several of these investigations is to determine a pattern of aid allocation that is most effective regarding the aggregated increase of welfare in developing countries. Another major target of this kind of research is to determine factors that prove to be decisive for the choice of the individual donor’s allocation. Such factors, on the one hand, can be derived at the recipient country level - humanitarian needs or the degree of poverty are examples for this. But they can also be investigated from the donor country perspective and include factors such as political or commercial interests. Finally, these approaches are expanded by several studies in order to compare the aid allocation of different donors. These studies argue that the patterns of aid allocation vary significantly between different donors. Some of them conclude that inadequate aid allocations, presumably caused by personal donor motives, are a major cause for the ineffectiveness of total aggregated aid. The following section depicts this strand of the literature and their implications for this thesis in more detail, with a focus on those studies that devote to determining differences between different donors.
The conclusions of the aid allocation literature, however, do not provide information on the eventual effects of different patterns of aid flows. Or put differently: They provide insights about differences in aid allocations of different donors (qualitatively), but do not quantify these. Only a few studies so far attempt a quantification of aid impacts separated by donors. Their measurements are mostly based on implications of the aid allocation literature. A further section depicts these approaches and their results.
2.3.1 The aid allocation literature: Deriving donor interests
To begin with, this section outlines the main approaches and findings of the aid allocation literature. Implications derived from this strand of research serve as the main theoretical foundation of this thesis. Similar to the aid-growth literature depicted in section 2.1, there are numerous studies contributing to this research field with a history that goes back to the 1970s.5
McKinlay and Little (1977) already elaborate on explaining the allocation of official bilateral aid by the example of the US development assistance. They develop five models for each possible type of donor interest and analyse, based on a cross-sectional analysis of the period 1960-1970, which of the models correlates most with the allocation of US aid over this period (ibid.: 68-72). The interests represented by the five models include fostering development, political stability and democracy in developing countries, but also (own) economical, security and power-political interests of the US. The authors find (ibid.: 78) that the model of the last-mentioned set of interests is the most appropriate one for describing the choice of allocation. They arrive at the drastic conclusion that the high explanatory potential of US power-political and security interests behind its aid allocation shows that aid is mainly used to maintain the balance of power between donor and recipient and that it “permits aid to be seen as a dimension of imperialism” (McKinlay and Little 1977: 80).
Many previous studies come to similar conclusions. In their influential work, Collier and Dollar (2001: 1791-1793) calculate a poverty efficient allocation of aid and conclude that the current allocation, driven by other motives than total poverty alleviation, is far from this optimum. They also demonstrate that the actual aid allocation would be more efficient with a stronger focus on the quality of policies and institutions of the recipient countries (ibid.). According to their conclusions, a stronger allocation emphasis towards countries with a high poverty rate and sound policies would be necessary for achieving the MDGs (ibid.: 1800). This is in line with the findings of those studies that instrument for policy, as depicted in section 2.1. Equally responding to this result, Harrigan and Wang (2011: 1291) calculate that only 14,8% of the total US aid is allocated on the basis of good policies. Kilby and Dreher (2010: 340) likewise state that donor motives are able to impact the effectiveness of aid significantly. In this context, Alesina and Dollar (2000: 45-46) show that bilateral donors provide more aid to countries that have been former colonies, suggesting that geopolitical and economic interest play a determinant role for their aid allocation. Berthélemy (2006A: 188) adds that, beside the geopolitical and post-colonial variables, the export intensity of bilateral donors is also correlated positively with the bilateral aid allocation.
Let us now turn to those studies of the aid allocation literature that attempt to outline differences between the allocation of certain donor countries. As an early example, Dollar and Levin (2006: 2036) analyse differences in aid allocation between multilateral agencies and bilateral donors. Among their findings is that bilateral aid appears to have a weaker relationship with both democracy and rule of law compared to aid from multilateral donors. Furthermore, they point out that MA is more effectively targeted to poorer countries (ibid.). As already pointed out by their predecessors, they emphasize the important role of former colonial ties between rich industrialised countries and the recipient countries, as well as the high correlation between aid and exports of bilateral donors (cf. Dollar and Levin 2006: 2044). These results are echoed by Harrigan and Wang (2011: 1288), who investigate the aid allocation of the US compared to other multilateral and bilateral donors. They find that bilateral donors put less emphasis on good policy environments and recipient needs than multilateral agencies and illustrate this by the example of the two countries with the highest aid donations, the US and Japan, which allocate only a low proportion of their aid according to recipient needs and good policy (ibid.).
Looking more closely at the causes for the distribution of US aid flows in the region of Middle East and North Africa (MENA), they find a high positive correlation between their measurement for donor interest and aid directed towards Israel and Jordan - the main allies of the US in this region. Opposed to this, their coefficients for Iran, Sudan and Syria, as countries that are typically considered hostile to US policy over the observation period of this study, show a significantly negative and sign. By this example, they demonstrate the high relevance of strategic interests in the pattern of bilateral aid allocation. However, Berthélemy (2006B) adds to this discussion that “multilateral aid allocations are not themselves immune to the influence of donor self- interest variables [. . . ]” (107). His estimations show that that the commercial interest of several important bilateral donors influences the allocation pattern of major multilateral agencies significantly (ibid.: 99-101). Several subsequent studies depict the orientation of major bilateral aid donors towards own goals instead towards recipients’ needs or their achievements in good governance.
Dreher et al. (2011: 1955-1956), e.g., compare the aid allocation of countries that have launched their first aid programmes during recent years, such as Brazil or Korea, with those of the traditional DAC countries. Against the background that the newly emerged donors are often criticised as being self-centred, they illustrate that DAC donors show a very similar pattern to these, characterised by a correlation between aid flows and donor exports, as well as their disregarding of the level of corruption in recipient counties. Alesina and Dollar (2000: 46-47) demonstrate the geopolitical interest behind aid flows of main industrialised countries. By analysing the UN voting pattern of recipient countries, they find a strong correlation between these votes and the aid allocation of all the major industrialised donor countries, i.e. France, Germany, Japan, the UK and the US. According to their argumentation, this result reflects that aid is a measure to pursue strategic goals (ibid.).
At a different point, Alesina and Dollar (2000: 42) argue that the four Scandinavian countries Denmark, Finland, Norway and Sweden show a great similarity in their allocation of aid. The authors find that these countries, opposed to the industrialised countries, put a high emphasis on the needs of poor countries and the rewarding of good policies (ibid.: 50). At the same conclusion arrives Berthélemy (2006A, 190-191).
After analysing the trade intensity of bilateral donors with the individual recipient countries, he groups donors according to their orientation towards either own interests, or those of the developing country. In addition to the Scandinavian countries, he classifies Switzerland and the Netherlands as altruistic donors, i.e. they decide about the allocation of assistance regardless their bilateral relation with the recipient. The distribution of the beforementioned industrialised, he depicts further, differs strongly from this. Especially for France, Italy, Japan and the US, he finds a more egoistic pattern compared to other donors (cf. Berthélemy 2006A: 193). Dreher et al. (2010: 60), who conduct an analysis of both Swedish bilateral and NGO aid, find that Swedish BA is not correlated with increases in exports. Furthermore, they show that the recipient countries’ resource endowments correlate negatively with both the Swedish NGO aid and BA (ibid.). One would therefore assume a strong orientation of the Swedish aid towards recipient needs.
In the context of different donor motives and their connection to aid effectiveness, one further finding of the aid allocation shall be highlighted at this point: In their analysis of aid allocated during the 1980s, Schraeder et al. (1998: 321-322) already expect that donor motives would change significantly after the end of the Cold War. As a reason for this, they assume that the great role of political ideology in the allocation of aid would decrease with tensions relieving between the Eastern and the Western bloc. Dollar and Levin (2006: 2044) take this argument into consideration and find that the correlation between aid flows and economic governance of recipient countries turns for both bilateral and multilateral aid (MA) from negative in the 1980s to positive in an observation period of 2000-2003. Harrigan and Wang (2011: 1290) confirm this outcome only partially: They determine that all donor countries, apart from Canada and the US, have placed a stronger focus on recipient needs after 1989 than during the Cold War. However, they also show that only a smaller share of the donors has changed the emphasis on good policies after the end of 1989. Among the countries who do not appear to have changed in this regard are Japan, the US as well as multilateral donors.
To sum up this section: A closer examination of the aid allocation literature reveals that there are considerable differences in the way different donors distribute aid amongst recipient countries. Individual donor motives, colonial ties and historical circumstances are conceivable reasons for this. Moreover, several studies distinguish between those donors that feature a recipient oriented “fair” allocation of aid and others that tend to distribute their aid with a strong orientation towards own goals. Studies of this type are conducted against the background of the aid effectiveness debate and attempt to provide an explanation of the apparent failure of aggregated aid.
2.3.2 Determining donor di erences
Insights of the aid allocation literature indicate that effects resulting from development assistance can strongly differ among donors. One possibility to quantitatively test this argument is to investigate the impact of aid on final outcomes while looking at the aid flows from distinct donors separately. Some studies so far aim at identifying such donor-related aid effectiveness. The most commonly analysed type of donor differences, in this context, is the one between multilateral and bilateral donors. Headey (2008: 170), for instance, finds that the effect of MA on growth is nearly twice as high as the effect of BA. Beside the separation between the two kinds of donors, he also uses a different measure for aid, as mentioned before: He focuses on aid that is intended to increase production by subtracting flows of humanitarian aid and, additionally, excluding the repayments flows of aid loans (ibid.: 164). This empirical strategy differs from the most commonly used approach of considering total net flows of aid.
Rajan and Subramanian (2008: 656), on the contrary, do not apply this kind of disag- gregation and do not find any differences between BA and MA regarding the impact on GDP per capita growth. They even observe that neither of the two aid types has a significant positive effect over a variety of different observation periods, both in cross- sectional and dynamic panel data analyses. In addition to the observation of growth effects, there are also first contributions analysing differences between BA and MA with respect to social indicators. Following an analysis of the impact of aggregated aid on poverty measures, Alvi and Senbeta (2012: 968) distinguish between aid of bilateral and multilateral donors in a second step. Whereas total aid appears to be negatively but only moderately correlated with poverty, they cannot find any significant effect of bilateral aid on poverty alleviation. On the contrary, their estimation for the separated multilateral aid proves to be higher than the initial coefficient of aggregated aid. This suggests that the negative sign of their first estimation is exclusively determined by the capability of MA in alleviating poverty (cf. ibid.: 967-968).
In this line, Masud and Yontcheva (2005: 11-14) investigate the correlation between aid and two social indicators - children mortality for measuring health conditions, as well as adult illiteracy as an indicator for educational quality. In contrast to the above- mentioned studies, they distinguish between aid from bilateral donors on the one hand, and assistance provided by NGOs as a comparative measure. With respect to infant mortality, they find that NGO aid significantly reduces the mortality rate, while BA proves to have no influence. Regarding the illiteracy rate, they find no significant effect by neither of the two aid variables. In a next step (ibid. 17-18) they demonstrate that both NGO and bilateral aid are not negatively correlated with the recipient countries’ government spending. Their conclusion is that NGO aid creates additional value to the government’s efforts in the health sector, whereas the missing effectiveness of BA becomes more apparent.
Up to this point, the studies presented in this section attempt to determine differences in aid effectiveness between bilateral donor countries and non-bilateral donors, such as multilateral agencies or NGOs. The aid allocation literature, however, also suggests that there are significant differences in allocation between single donor countries. Yet, the existing literature on measuring aid effectiveness does not address this finding to a larger extent. There are nevertheless first attempts that shall be examined closely in the following. A series of estimations in Rajan and Subramanian (2008:655-656) draw on the conclusion of the aid allocation literature that some bilateral donor countries provide more effective aid than others. They only take the aid flows from the five countries Denmark, Finland, Iceland, Norway and Sweden into account. Among all their regressions for this group of donors, which include for both the growth and aid horizon variations from 10 to 40 years, Rajan and Subramanian find only one to be significantly positive (cf. ibid.).
Minoiu and Reddy (2010)6 reach a different conclusion after investigating the dif- ferences between three different groups of bilateral donor countries. They devote to determining distinctions in aid effectiveness by systematically distinguishing between aid that is expected to promote growth on the one hand, and such types of aid that are typically regarded to have a smaller or non-existing effect on development on the other hand (28). For this purpose, they draw on the conclusions from the aid allocation literature and analyse the aid flows from donors that are supposed to allocate aid in a different manner. For the purpose of quantifying differences between donor countries, they compare one group of donor countries that are supposed to allocate aid in an altruistic manner with two control groups. The group of donors that are regarded fair comprises the four Scandinavian countries Denmark, Finland, Norway and Sweden plus the Netherlands. The two comparison groups are, first, all countries of the first group plus five further countries that are predominantly regarded to allocate their aid based on recipient needs and, second, all countries of the first group plus five countries that are assumed to allocate their aid with a focus on geostrategic aspects (30). They find that both groups of countries that are considered fair donors show a high and significant effect on the growth of recipient countries, whereas the effect of the third group that also comprises strategically oriented donors is considerably smaller (33). This result is based on both cross-sectional and dynamic panel analyses (36).
Finally, there are several investigations that quantitatively asses the above-mentioned finding of changing donor motives after the end of the Cold War. Headey (2008: 172), as well as Bearce and Tirone (2010: 843) both find that aid effectiveness increases after the end of the Cold War. They achieve this result based on different empirical approaches. While Headey (2008: 171) applies a dummy variable for the post-Cold War era to his GMM-estimations, Bearce and Tirone (2010: 842) split their sample into two separate OLS panel data estimations to analyse the pre- and post-Cold War periods respectively.
Summing up the insights from this section, the following patterns of estimated donor differences become apparent: Based on evidence from the aid allocation literature, several studies distinguish between aid flows from bilateral and multilateral donors. Regarding the difference in the effect on economic growth, there appears to be no common ground in the literature. Multilateral aid, however, turns out to be more effective when determining the effect on social indicators. The same finding becomes apparent when comparing aid flows from bilateral donors with those from NGOs. There is also first evidence for differences between single bilateral donor countries, which confirms the conclusions from the previous literature that different donors allocate aid in a significantly different manner based on different donor objectives.
3 Research objectives
Against the background of the remaining controversy over the impact of development assistance, this thesis aims at contributing to the aid effectiveness puzzle in the following way: The aid allocation literature suggests outcomes of aid programmes largely depend on how donors distribute aid and, consequently, on the underlying reasons and motives behind the pattern of aid allocation. Several studies also point out that there are significant differences in the way different donors allocate aid. Considering these findings, one can assume that a possibly explanation for the fact that many studies fail to find a positive effect of aggregated aid on final outcomes is that the impact of aid from single donors differs significantly. In other words, the positive effect of donors with an aid allocation based on recipient needs and development objectives could be countervailed by the effects of donors that are allocating their aid in more strategic or less development oriented manner.
The objective of this thesis is to test this hypothesis quantitatively by analysing the effects of aid from different donors separately. In doing so, this study investigates the aggregated aid flows of each donor with the objective to determine the respective total effects. This procedure offers the possibility to provide insights about which types of aid are effective in promoting development of recipient countries and which types are not.
Several studies apply a similar approach by analysing the differences in effectiveness between multilateral and bilateral aid. Their conclusions, as depicted in the previous section, lack a common consensus. Therefore, this thesis aims at contributing to this debate with a comprehensive approach. More importantly, it addresses the differences in aid effectiveness between single bilateral donors. The aid allocation literature provides strong evidence to suggest that the extent to which aid effects final outcomes differs significantly from one donor country to another. Moreover, bilateral aid accounts for almost two thirds of the total aid payments, as the subsequent section depicts. For this reason, and because aid from bilateral donors is often considered less effective than such from multilateral agencies, a closer look at the different donor countries could provide important insights. The lessons that could be learned from this approach are not only gaining insights into the puzzle of the failure of aggregated aid. In the case that certain donors prove to provide more effective aid than others, one could also draw conclusions for the effectiveness of different aid allocations and, thus, derive implications for the various aid agencies and bilateral development banks.
To allow for an accurate assessment of aid effectiveness, and to possibly determine aid that effects only a certain indicator, this study shall not only analyse the effect on growth, but also on social indicators for health and education.7 This decision rests upon the above-mentioned fact that positive effects of aid on single social indicators might not be visible when looking at the aggregated outcome on growth, and on the fact that both the MDGs and SDGs comprise several targets regarding education as well as health (see appendix A). Moreover, there is first evidence that aid allocation during the Cold War had significantly been influenced by political interests, and that donor motives changed towards recipient needs after the end of it. Considering this observation with an analysis of the changes in aid effectiveness after the year 1989, this study also attempts to provide evidence for the importance of donor motives.
Finally, an often-discussed topic of aid effectiveness research is the role of timing.
Clemens et al. (2012: 594), for instance, argue that when analysing aid effectiveness, it is crucial to bear in mind that different flows vary in the period it takes until their effects arrive. Following their argumentation (ibid.: 598), examples for aid that is assumed to have an early impact are investments in the transport, energy or agriculture system, as well as budget support. On the other hand, investments in the education or health system are typically considered to only show effects in the very long run (cf. Arndt et al. 2015: 9).
To tackle this issue, several studies look at various time intervals separately when analysing the impact of aid, such as Rajan and Subramanian (2008: 656) or Minoiu and Reddy (2010: 32). As different aid flows can vary widely regarding the period until positive effects may occur, it appears important to pay particular attention to the role of time when constructing estimation models. Therefore, this study follows the objective of differentiating between different possible timings of aid effects. This allows to gain a more complete picture of aid impacts on the one hand and, furthermore, provides insights into the period over which aid shows to have the highest impact.
4 Preliminary Empirical Analyses
To achieve these objectives, the approach of this study is as follows: The research objective of interest is the effect of aid on development, measured by economic growth, as well as indicators for education and health. With the general underlying assumption that aid effectiveness is significantly influenced by donor decisions, the aim is to subdivide aid flows by donors and to look at the respective impacts on the development measures separately. Besides a separation between bilateral and multilateral aid, this study also distinguishes between single bilateral donor countries based on the insights of the aid allocation literature, i.e. that there are considerable differences in motives of donor countries resulting in varying allocation patterns.
To determine the impact of aid on the chosen development indicators, the study applies both cross-sectional and dynamic panel data analyses. The focus, however, is on the latter one. Chapter 6 discusses the reasons why GMM methodology is the preferred approach. Regarding the observation period, the estimations consider the full coverage of available data, i.e. the period 1960-2014. In addition, a 25-year observation period covering 1990-2014 serves to determine differences in aid effectiveness after the end of the Cold War.
Before analysing the actual effect of aid from different donors, however, this chapter aims at providing evidence from a descriptive analysis. In a first step, main data sources, as well as definitions of the chosen parameters are presented. The second section then comprises the main part of the preliminary investigations. Based on the findings of the aid allocation literature, this analysis aims at determining differences in the way the chosen donors distribute aid. The objective here is to underpin the insight of the previous literature, and to verify if the main results, namely that major industrialised countries put a higher priority on own objectives than smaller Nordic bilateral donors, apply to the donors analysed in this study.
4.1 Data sources and de nitions
To begin with, this section provides key definitions of the selected variables and presents the sources. A first issue to be addressed is the choice of the aid variables. The DAC database of the Organisation for Economic Co-operation and Development (OECD), as the most common source of the aid research, offers several different typed of aid measures. The three main types are total aid numbers, aid per capita of the recipient country population and aid as percent of the recipient country GDP (cf. OECD-DAC 2016). This study applies the latter measure, in line with most previous studies (cf. Rajan and Subramanian 2008: 651, Arndt et al. 2010: 8 or Clemens et al. 2012: 603). The straightforward reason for this choice is that several of the other explanatory variables also refer to GDP and, therefore, have a common reference point that facilitates interpretation.
A further aspect concerns the different elements of aid that one can consider for es- timating aid effectiveness. One option is to exclude certain monetary flows, such as humanitarian aid or the repayments of assistance loans, as Headey (2008: 179), for instance, does. The objective of this study, however, is to capture the overall aid ef- fects of each donor. Therefore, the subsequent estimations consider total aid flows, which comprise all grants and development loans with a grant element of at least 25% (cf. OECD-DAC 2016). This decision also corresponds to the approach of Rajan and Subramanian (2008: 662).
Now the focus is on the main aid variables of this study, i.e. those aim at determining differences among bilateral donors. Separating BA is achieved by forming groups of countries with a similar aid allocation, following the approach of Minoiu and Reddy (2010: 30). This procedure allows to observe sufficiently high aid flows and, thus, to potentially achieve better empirical results. In contrast to the separation strategy of Minoiu and Reddy (ibid.), however, this study aims at directly comparing one group of countries considered to allocate their aid effectively (from the recipient point of view), with another group of countries that are predominantly regarded as donors with an aid allocation oriented towards their own interests. The aim of this more direct comparison is to elaborate further in detail on the extent to which aid effectiveness can differ between different bilateral donors. According to Alesina and Dollar (2000: 42) and Berthélemy (2006A: 190-191), the four northern European countries Denmark, Finland, Norway and Sweden, as well as the Netherlands allocate aid strongly based on recipient needs and with a focus on development objectives.
For this reason, and because Minoiu and Reddy (2010: 33) observe a strong and significant positive relation between aid from these countries and growth in recipient countries, the choice of this study for the first group of countries, i.e. those with a “fair” allocation, is the same. The second group, however, differs from the decision of Minoiu and Reddy (2010: 30). As explained above, this group shall comprise countries with an allocation that presumably differs significantly from the one of group 1 donors (G1).
Section 2.3.1. reveals that among such countries are France, Germany, Japan, the UK and the US. These countries form the second group of bilateral donors (G2). The table below gives a summary of the two groups of bilateral donors.
Appendix B1 shows the sources for all selected variables, distinguished by dependent variables, aid variables and the remaining control variables applied by the subsequent estimations. The respective analyses in section 5.1 and 5.2, moreover, explain which variables are applied for which estimation.
Table 1: Group assignment of bilateral donors
Abbildung in dieser Leseprobe nicht enthalten
Appendix B2 presents a summary of the recipient countries considered in this study.
The sample comprises nearly all countries that have received aid since the establishment of the DAC. A different approach would be to consider only a closer selection of aid recipients. Gomanee et al. (2005A: 1071), for instance, take into account only several LDCs in SSA, against the background that those countries have shown a particular low growth over the past decades. As the investigation of the subsequent section shows, however, such an approach would leave out major recipient countries of the analysed donor groups and, consequently, lead to a considerable bias of the eventual aid effectiveness.
4.2 Disaggregating aid descriptive statistics
As some preliminary remarks to the empirical investigation of this study, this section illustrates differences in aid allocation of the chosen donor groups, based on the insights of the aid allocation literature. At this point, the term disaggregation requires a further explanation. As mentioned in section 2.3, aid flows can be separated in various ways.
One option, for instance, is to distinguish between the purpose of aid that has been assigned by the donor. Several studies published over the last two decades apply such a procedure. One purpose behind this way of disaggregating aid can be to particularly analyse the effect of aid that is intended to increase the education level, such as Dreher et al. (2008: 300) do, or to only investigate those aid flows that are expected to have a short-term growth effect, as it is the approach of Clemens et al. (2012: 494). This thesis, however, investigates the total aggregated aid flows of the respectively considered donor.
This has two main reasons.
Firstly, one can assume that aid under certain circumstances or specific types of aid flows are less effective than others. Kimura et al. (2012: 7-8), for instance, show that aid proliferation caused by excessive amounts of budget support or many projects executed by different donors can have a negative effect on growth. Consequently, if one considers only particular types of aid flows and leaves out others, it is possible that those kinds of aid that have no or a negative effect are not entering the final result. The aim of this study, yet, is to determine differences between different donors. And because the aid allocation literature concludes that certain donors distribute higher shares of those types of aid that are regarded more effective than others, leaving out certain kinds of aid could distort the final result significantly.
The same argumentation applies to the use of such type of disaggregation applied by Clemens et al. (2012: 598-599), i.e. if one considers only the kinds of aid that are expected to have a positive effect in the short run. This procedure excludes all kinds of aid that are expected to show an impact in the long run or not at all. When comparing the effectiveness of different donors, the results of this procedure would be of very limited meaningfulness. Given the case that one finds a positive effect of the separated kinds of aid that should have an early impact and. Given that this positive impact would disappear completely when looking at the total aid numbers, one could argue that the share of the separated flows is not sufficiently high to enter the final result. Vice versa, the (potentially negative) influence of the formerly excluded flows could also be large enough to outweigh the other flows.
Secondly, excluding elements of aid that are assigned to certain sectors, such as to the health sector or to humanitarian needs, as Headey (2008: 169) or Clemens et al. (2012: 594) do, is not free from bias either. The rationale behind the strategy of these studies is these types of aid were originally not intended to affect growth and, thus, could be left out for a growth analysis. However, Bloom et al. (2004: 9-10) demonstrate that good health conditions have a significantly positive influence on growth. Therefore, there are no convincing arguments why aid assigned to the health sector, or even humanitarian aid, should not be considered when analysing the long-term effects of aid. Instead of dividing aid according to assigned purposes or sectors, this thesis attempts to capture the absolute flows of each donor and, thereby, to provide evidence for the overall differences amongst them. Figure 1 provides a first look at the total aid numbers of each donor in the course of aid history.
Looking at the development of aid payments over the past five decades, it becomes immediately apparent that donors of the first country group dominate the overall picture.
During the 1960s, almost all aid labelled as ODA has been disbursed by the five countries of group 2. The major part of the payments over this period relates to US aid, accounting for almost 55% of the total aid flows spent in the 1960s.8 With more than 12% of the aid disbursements in this period, France is also one of the first bilateral donors with large-scale aid programmes. During the 1970s, Germany, Japan and the UK additionally started to provide aid on a larger scale. Although the share of disbursements from other bilateral and multilateral donors has been growing constantly up to now, the
Figure 1: The development of aid payments by donor
Abbildung in dieser Leseprobe nicht enthalten
Notes: Data from OECD-DAC. The figures show five-year totals in current US Dollar, billions.
share of G2 countries in total aid disbursements was still 45% over the period 2010-2014 and almost USD 300 billion in total.
Compared to these numbers, G1 countries account for a much lower proportion of total aid flows. However, considering that all donors in this group are relatively small countries by population, they are among the donors with the highest aid per capita payments. In the mid-1970s, Sweden and the Netherlands started larger aid programmes.
In the early 1980s, furthermore, Denmark and Norway essentially extended their aid disbursements. Up to the present, Finland has been a relatively small donor compared to the other countries of this group. Only during the late 2000s, Finland started to increase its aid funds significantly. Over the past five years of the observation period (2010-2014), G1 countries almost accounted for almost a quarter of the aid disbursed by G2, with total disbursements of more than USD 72 billion. Considering the ratio between the aid numbers of the two groups, one can conclude about a first shortcoming of measuring the effect of aggregated aid, or aggregated bilateral aid. Given that the allocation of aid and, consequently, the resulting effectiveness differs between the two donor groups, it is easy to picture that potentially positive effects of a small G1 donors could be alleviated by those of just one big G2 country. This is a further reason why investigating aid effectiveness of different donors separately could provide additional insights.
A similar amount of aid to that of G1 is added by the remaining bilateral donors of the DAC.9 Multilateral donors also disburse a considerable share of aid which has sharply been increasing over the past decades. As mentioned before, several studies show that multilateral aid had worked better over regarding the effect on growth, such as Burnside and Dollar (2000: 863) or Headey (2008: 170). One consequence of such findings could be that one can observe a significant increase in the share of multilateral aid over the past years (as figure 1 shows).
After looking at the total aid numbers divided by donor, a further view on allocation patterns of individual donors aims at providing first evidence for the divergence in aid effectiveness between different donors. Since one of the main targets of this study is to analyse the differences amongst bilateral donors, the following overview (table 2) focuses on the two selected country groups. The table shows the pattern of aid allocation while considering the ten recipient countries with the highest amounts of aid received from the two groups respectively. In doing so, it distinguishes between the two periods 1960-1989 and 1990-2014 with the purpose of capturing changes between the pre- and post-Cold War era. Moreover, infant mortality and income per capita, two of the major indicators for measuring development in the aid effectiveness literature, are included for each recipient country and the two periods, to enable a simplified illustration of differences in the consideration of recipient needs.
When comparing the differences between the two periods, it firstly becomes apparent that both donor groups diversify aid more broadly after the end of the Cold War. Whereas G1 (G2) allocated approximately 47% (36%) of the total aid budget to the first five recipient countries in the period 1960-1989, the respective shares decrease to 31% (23%) in the period afterwards. This could have several reasons. One possibility is that donors have expanded their aid portfolio and established relations with more recipient countries over the time. One could also explain this by drawing on those findings of the aid allocation literature that refer to changes in donor motives after the end of the Cold War, as outlined in section 2.3.1. Thus, the importance of political relations with individual recipient countries could have decreased after the reduction of global tensions. Other motives, such as the quality of the recipients’ governance and policy environments, could in turn have increased. This reasoning would be in line with the results of Dollar and Levin (2006: 2042). Their estimations show an increasing role
Table 2: Aid allocation of bilateral donor groups
Abbildung in dieser Leseprobe nicht enthalten
Notes: Data from OECD-DAC and WDI. The table shows only the recipient countries of the sample (see Appendix B1). Additional countries with high aid receipts might be omitted.
of variables measuring good recipient policies for explaining the allocation pattern of bilateral donors after 1989.
Focusing on the temporal changes in the G1 aid allocation, one can observe that there are no major changes regarding the choice of recipient countries. Out of the ten countries that received the highest amounts of aid by these donors before 1990, six remain among the key recipients during the second period. Yet, one can notice one peculiarity of the G1 allocation: Countries in SSA seem to have a particularly high role among their recipient countries. Five of the recipients during the first period, and even seven during the second period are located in this area. This comprehensive priority of G1 countries for allocating aid towards SSA, moreover, can be recognised when looking at each of the donor separately.10 Alesina and Dollar (2000: 42-44) also find that Scandinavian donors tend to spend high amounts of aid to African recipient countries, and additionally show that this choice of allocation is not motivated by political interest. One can rather assume that a high need for further development in these countries drives their choice.
Countries in SSA are both before and as from 1990 among the countries with the lowest per capita income and the highest infant mortality, as table 2 demonstrates. The allocation of G2 donors shows quite a different picture. Countries that exhibit a relatively high level of development are not only among the donors with the largest aid receipts before 1990, but also during the post-Cold War period. Because each of the donors of this group shows a very characteristic allocation pattern, it is worth taking a closer look at the individual allocation patterns (appendix C3), instead of discussing the group-comprehensive numbers at this point.
1 Extensive investigations of the first decades of research can, for instance, be found in Hansen und Tarp (2000) or the meta-analyses of Doucouliagos and Paldam (2006, 2009).
2 The following categorisation of the recent aid into three strands is inspired by Clemens et al. (2012: 592-593) and extended up to the present.
3 Chapter 6 contains detailed information on the GMM approach and its benefits, as well as its limitations.
4 Annex A shows provides an overview of the agendas of both MDGs and SDGs.
5 For further reading: Harrigan and Wang (2011: 1282-1283) provide a detailed review on the aid allocation literature, including early research.
6 The following page numbers in this paragraph, in parentheses, refer to the study of Minoiu and Reddy (2010).
7 Another possibility of analysing the effect on social outcomes is to apply the HDI of the United Nations Development Programme (UNDP) as a composite statistic covering several different measures. This is done by Gomanee et al. (2005B: 301) or Kosack (2003: 7), for instance. McGillivray (1991: 1467), however, shows that the HDI is positively correlated with each of its components and concludes that it is not appropriate to apply as dependent variable. Therefore, this study focuses on the two areas health and education, as many studies of the aid literature consider either or both of them (see for instance Boone 1996: 303, Masud and Yontcheva 2005: 13-14 or Dreher et al. 2008: 297; 2010: 166).
8 Appendix C1 provides information about the aid payments of each of the investigated donor in five year averages.
9 It is worth mentioning that there are several other bilateral donors that are not part of the DAC. The disbursements of some of them are also listed in the OECD-DAC (2016) database. However, those of other important donors, such as China, are still not reported. For this reason, Figure 1 refers to bilateral donors that are part of the DAC only. For further information, Dreher et al. (2011), for instance, investigate the aid allocation of those donor countries that have recently emerged and which are already listed in OECD-DAC.
10 To provide more accurate information, appendix C2 and C3 additionally shows the aid allocation of each bilateral donor of the two groups separately.