WPS6998 Policy Research Working Paper 6998 Aid Is Good for the Poor Yumeka Hirano Shigeru Otsubo East Asia and the Pacific Region Office of the Chief Economist August 2014 Policy Research Working Paper 6998 Abstract Aid is good for the poor. This paper uses detailed aid poor or any other income group, beyond their effects on data spanning 60 developing countries over the past two average incomes. The paper finds that trade and foreign decades to show that social aid significantly and directly direct investment tend to benefit the richest segments of benefits the poorest in society, while economic aid increases society more than other income groups. Therefore, the the income of the poor through growth. This new and presented evidence suggests that aid can play a crucial role unequivocal finding distinguishes the current study from in enabling the poor to benefit more from globalization. past studies that only utilized aggregate aid data and These discoveries underscore the need to assist develop- returned ambiguous results. The paper also confirms that ing countries to find the mix of economic and social aid none of the elements of globalization (trade, foreign direct that jointly promotes the participation of the poor in the investment, remittances), policies (government expenditure, development process under globalization. In this manner, inflation management), institutional quality, nor other aid can make greater strides in spurring development. plausibly pro-poor factors have systematic effects on the This paper is a product of the Office of the Chief Economist, East Asia and the Pacific Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at yhirano@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Aid Is Good for the Poor Yumeka HIRANO (World Bank) & Shigeru OTSUBO (Nagoya University) Keywords: Aid, Poverty, Inequality, Globalization, Institutions, Development Effectiveness JEL Classification: F35, I32, O11 ------------------------------------------ Yumeka HIRANO is an Economist in The Chief Economist Office of the World Bank’s East Asia and Pacific Region (EAPCE). Shigeru Otsubo is a Professor at the Graduate School of International Development, Nagoya University. Financial support from the Japan Society for the Promotion of Science (JSPS) Research Fellowships for Young Scientists (Research#: 24-3494, Representative: Hirano) and the JSPS Scientific Research (A) & Challenging Exploratory Research (Research#: 22252005/23653066, Representative: Otsubo) is gratefully acknowledged. We would like to thank Bert Hofman, Debra Saito, Kiyoshi Fujikawa, and Naoko Shinkai for their valuable comments. 1 1. Introduction Enhancing the effectiveness of aid has long been the international development community’s core agenda, given the limited resources available for the fight against poverty. With the establishment of the Millennium Development Goals (MDGs) in 2000 and the implementation of the Paris Declaration (PD) on Aid Effectiveness in 2005, the international community has continued to improve the impact of aid on development. However, poverty still persists despite drastic changes in the development landscape. While progress has been made toward the MDGs, as well as great strides in poverty reduction over the past few decades, globalization has introduced economic structural changes, and huge income inequality and development disparities persist across and within regions and countries (UN, 2013). As such, while the eradication of poverty remains a challenge in the post-2015 development agenda, inequality has entered the equation. This led the World Bank to establish in 2013 the twin goals of “poverty eradication” and “shared prosperity” (World Bank, 2013). In the mid-1990s, private capital flows, especially foreign direct investment (FDI) and overseas remittances, surpassed by far the level of Official Development Assistance (ODA), which used to be the main financial flow to developing countries. Nevertheless, as highlighted in the Monterrey Consensus for Financing for Development (FfD) adopted in 2002, ODA is still expected to play a key role in assisting developing countries to fully utilize the opportunities presented by globalization. Against this backdrop, the effectiveness of development aid has been debated for decades. While aid is generally considered to have a positive impact on economic growth, which is believed to be a vital force in poverty reduction, cross-country empirical studies on aid effectiveness for growth have shown ambiguous results. Some studies have confirmed a positive relationship (Gulati, 1978; Hansen & Tarp, 2000; Clemens, Radelet, & Bhavnani, 2004; Minoiu & Reddy, 2010), while others have argued that there is no significant impact of aid on growth (Mosley, Hudson & Horrell, 1987; Boone, 1996; Easterly, Levine, & Roodman, 2003 & 2004). Moreover, some studies have suggested that the positive impact emerges only with certain prerequisite conditions such as good policies (Burnside & Dollar, 2000; Collier & Dollar, 2001, 2002), climate-related geographical environments (Dalgaard, Hansen, & Tarp, 2004), and only in certain forms or categories of aid (Sawada, Kohama, & Kono, 2003; Clemens et al., 2004; Minoiu & Reddy, 2010). Although the current mainstream discourse asserts the importance of good policies for development aid to promote growth effectively, the controversy begs further research. In order to investigate development effectiveness, the impact of globalization, in the forms of trade, FDI, and remittances, for example, should not be neglected. In the literature on globalization and inequality, there has been heated debate about whether the poor benefit from globalization. Numerous past studies confirm that, on average, globalization, or economic integration, contributed to poverty reduction through higher growth. However, some have argued that globalization did not disproportionally benefit the poor (Fischer, 2003; Milanovic, 2005; Ravallion, 2005; Harrison, 2006). That is, globalization did not help the poor more than other segments of society. One of the most influential papers by Dollar and Kraay, Growth 2 is Good for the Poor (2001, 2002), revealed that the poor benefited from growth, with empirical evidence that the income of the bottom quintile increased equiproportionally to that of the national average. At the same time, they did not find any factors of openness, globalization, policy, or institutions that had systematic effects on the poor other than through growth. In the end, Dollar and Kraay emphasized the importance of growth, as well as policies (including openness policies) and institutions that contribute to growth, for poverty reduction. Their latest paper, Growth Still is Good for the Poor (Dollar, Kleineberg, & Kraay, 2013), reaffirms the importance of growth, while they end up concluding that there is no robust evidence that certain policies, such as openness, education, and health expenditures, are particularly “pro-poor” or conducive to promoting “shared prosperity” other than through their direct effect on overall economic growth (p. 18). The challenge remains to find out if there is any factor that leads to poverty-reducing distributional changes. With their end results and assertions, we came up with a question: “Is aid good for the poor?” With the debate raging on, the current study revisits the discussion of development effectiveness by examining whether aid indeed contributes to promoting growth and reducing poverty. This study undertakes further analysis of the impact of aid, by applying the conceptual framework of globalization and the Poverty-Growth-Inequality (P-G-I) relationship as illustrated in Figure 1, with a focus on each leg of the triangular structure: aid and growth, aid and inequality, and growth/inequality and poverty. Figure 1 The Conceptual Framework and Structure of the Study Source: Modified, based on Otsubo (2009, p. 58) The following questions are addressed in this study, with the corresponding conclusion section in parentheses: 1.1 Does aid promote growth? (Section 4.1) 1.2 Do institutions matter for development effectiveness on growth? (Section 4.2) 1.3 Does aid mitigate inequality? (Section 4.3) 1.4 Can policies, institutions, and globalization be good for the poor? (Section 4.7) 1.1 Does Aid Promote Growth? This question bears repeating, regardless of the past literature addressing it. The current authors suspect 3 that one of the reasons for the ambiguous results in past aid-growth literature comes from the usage of aggregate aid data in their analyses. The insignificant, even negative, results should be regarded as natural because aid is not necessarily, or directly, targeted to produce economic growth. The purposes of aid are diverse depending on the programs or projects. The different effects of aid should be taken into consideration in the specification of the empirical analyses. Unlike most of the previous literature, this study employed sector-level data on economic aid, social aid, as well as aggregate aid, to test the impact of aid. In addition to this more granular view, a re-examination of the aid-growth relationship with the latest data (up to 2010 in this study) is warranted by recent changes in practice. Burnside and Dollar (2004) elucidated these changes, suggesting, “past aid had been allocated indiscriminately with regard to institutions and policies that were critical for growth,” and “now aid is more systematically allocated to countries with sound institutions and policies.” (p. 21) Bourguignon and Sundberg (2007) also discussed changes in international aid architecture and aid allocation since the mid-1990s. 1.2 Do Institutions Matter for Development Effectiveness on Growth? Another unique aspect of this research is the focus on institutional quality. We argue that institutional quality, and capacity in particular, are key factors in determining the effectiveness of development aid (Hirano, 2014). Unlike past literature that focused on policy as the control factor, this study pays more attention to institutions (further explained in section 3.3) in the aid-growth empirical analysis. Having a good policy stance (in terms of monetary, fiscal, and trade policies) does not necessarily mean that countries are capable of creating good policy impacts or managing their respective economies effectively.1 Further, policy stances can be changed or improved in a relatively short period of time, with or without external help. On the other hand, building institutions takes longer than policy changes to bring about effects on economic growth (Williamson, 2000), similar to the impact of aid in the long-run growth theory. However, it is precisely this institutional quality that is a more critical determinant of the effectiveness of aid, than policy, as the quality of institutions reflects long-term characteristics of countries that also affect policies and growth (Burnside & Dollar, 2000). Even with good policies, constraints remain, if there is an absence of good institutions. Further, institutions cannot be changed or improved easily until a country becomes self-reliant. In sum, while institutional quality is more challenging to change than policy, it is the more consequential when it comes to making aid effective. 1 Market-oriented policies were promoted through the implementation of structural adjustment as a form of aid for the debt crisis in the early 1980s. However, the structural adjustment was considered a failure in most countries, especially in Sub Saharan Africa, as the implementation of “good policies” did not trigger expected development. The failure of the structural adjustment in the 1980s led to greater importance being placed on institutions, which represent the internal efforts and management capabilities of a country, rather than policy stance, when providing development aid. 4 1.3 Does Aid Mitigate Inequality? Another challenge of this research is to empirically investigate whether aid has mitigated inequality. If the ultimate goal of aid is to reduce or end poverty, aid needs to disproportionally benefit the poor and reduce inequality. If the benefits of aid are not extended to the poor, the gap between the rich and the poor cannot be narrowed. Analyzing the impact of aid on poverty reduction from both growth and distributional perspectives can be considered another original contribution of this study, as there is little past empirical research on aid that has paid attention to the distributional impact of aid for achieving poverty reduction. 1.4 Can Policies, Institutions, and Globalization Be Good for the Poor? We look at the effects of aid on the poor in comparison with other factors including policy, institutions, and the facets of globalization. Past literature (Dollar & Kraay, 2001, 2002; Dollar et al., 2013; Milanovic, 2005) seemed to agree that there was no factor of policy, institutions, or globalization that significantly benefitted the poor other than through growth. Our earlier study provides a literature survey and empirical stocktaking of the impacts of (1) globalization (via trade, FDI, and remittances) on growth, (2) inequality on the poor, and (3) inequality on the Gini Index (Hirano & Otsubo, 2012).2 Our earlier study also found that no factor of globalization has a significant impact on the poor other than through growth. Taking these assertions into account, we reexamine the impact of policies, institutions, and globalization on the poor with the new dataset, while proposing that development aid could be a key factor with significant impact on the poor. 2. Methods and Empirical Models In order to examine the effectiveness of aid on growth, inequality and poverty reduction, we conducted a series of cross-country analyses using the two-stage least squares (TSLS) estimator. We used the unbalanced panel data of worldwide income, aid, poverty, distribution, policy and institutions. 2.1 Leg 1: Aid and Growth (Growth Regressions) We conducted the following regressions, employing the Barro-type ad hoc growth equations (Barro, 1991, 1997) for conditional convergence, with elements of exogenous conditions such as the fertility rate and terms of trade (X); a set of important factors to consider such as policies, institutions, globalization, and other potential determinants (Z); and aid (economic aid, social aid, and aggregate aid) (A): (lnyct – lnyc0)/t = α + β1lnyc0 + γlnXct + δlnZct + θlnAct + λt’ + εct’ (1) 2 The results of our earlier study will be compared with the results of this study in section 4. 5 where y is the average per capita income, c and t indicate countries and years, respectively, and λt’ + εct’ is a composite error term. In other words, the dependent variable is the period average growth rate of growth in real per capita income. The natural logarithm of real per capita income of the initial year, lnyc0, is included as an explanatory variable to examine the conditional β-convergence. Furthermore, we examined the impact of institutional quality 3 on development effectiveness by dividing the samples into two different groups, by (a) level and (b) change in institutional quality respectively, in order to show the differences in development effectiveness among different sets of countries: (i) countries whose initial level of institutional quality is higher than the sample mean; (ii) countries whose initial level of institutional quality is lower than the sample mean; (iii) countries whose institutional quality is improving (i.e., the change ratio of institutional quality is positive); and (iv) countries whose institutional quality is not improving (i.e., the change ratio of institutional quality is zero or negative).4 2.2 Leg 2: Aid and Inequality (Inequality Regressions) We examined the possible impact of aid on reducing inequality, employing three different equations: a) level regressions for the average income of the poorest quintile; b) growth regressions for the average income of the poorest quintile; and c) growth regressions for the Gini indices. We give most attention to b), as we consider that looking at the impact of aid to change in the poverty situation of the bottom quintile, rather than levels, is more appropriate and important for assessing development effectiveness. Our interest in a) conducting level regressions, is to identify pro-poor country fixed effects, which consist of various aspects of a country and make a country pro-poor, but are difficult to quantify. Lastly, we tested the impact of aid on changes in Gini indices, or c), for a robustness check, as Gini indices are often used in inequality analyses. However, this study pays more attention to changes in the inequality situation of the poor than that of the entire country, in order to assess development effectiveness. a) Estimation by the level relationship for the average income of the poorest quintile. We examined the level relationship between average income of the poorest quintile and that of the country, as well as other variables similar to Dollar and Kraay (2002, Eq. 1) by employing a fixed-effects regression model. We looked at the parameter of θ as an indicator of the pro-poorness of aid. Impact of aid to the poor, indicated by the parameter of θ, can be compared with that of policies and institutions, 3 We employed indices for the quality of institutions extracted from the International Country Risk Guide (ICRG) indicator (PRS Group, 2012). 4 Splitting samples based on certain criteria is a common approach in empirical growth literature. For instance, the sample can be divided by positive/negative growth of a country (Dollar and Kraay, 2001), or income groups (Mosely, 1980; Burnside & Dollar, 2004). We also tested interaction terms in order to see the joint effects and nonlinearity, i.e., whether policy/institutional quality is a key determinant of development effectiveness. Nevertheless we found in estimating coefficients of cross terms, aid and institutions were not statistically significant. This arises from the fact that estimating marginal effects (i.e., marginal elasticity parameters) of aid heavily depends on the sample selection, time duration, and sets of variables, as has been expounded in much of the literature. Kraay (2005, Figure 5) showed the interaction term of aid and policy (Burnside & Dollar, 2000) accounted for only four percent contribution to growth. 6 indicated by the parameter of δ: ln ypct = α + β ln yct + δlnZct + θlnAct + μc + λt + εct (2) where yp is the average per capita income of the poorest quintile, μc is the unobserved country fixed-effects, λt is the time fixed effects, and εct is a random error term. While looking for the pro-poor explanatory variables, we also intended to identify the country fixed effects, i.e., country-specific factors, that make the country pro-poor, which we call pro-poor country fixed effects, or pro-poor country specific factors. These include the aspects of a country that comprise the degree of its support for lower socioeconomic classes, such as the strength of leadership and degree of dedication to poverty reduction, as well as the strength of traditional socio-economic institutions, in terms of, for example, the civil society network, efficacy of service delivery to the poor, and robustness of the social safety net, which cannot be measured by universal institutional indices.5 b) Estimation for the growth rates of average income of the poorest quintile. We, in turn, estimate Eq. 2 utilizing the growth over change ratio in order to see the impact on the growth rates of average income of the poorest quintile, by variations of average income of the country and other variables. We also explore this examination in more depth, by conducting regressions for the first to the fifth quintiles respectively (quantile regressions): (ln ypct – ln ypc0)/t = β(ln yct – ln yc0)/t + δlnZct + θlnAct + λt’ + εct’ (3) We looked into the key parameter β, which measures the elasticity of income growth of the poorest quintile with respect to that of mean income, similar to Dollar and Kraay (2002, Eq. 4). Our main interest is to see the parameter of θ as an indicator of a pro-poor orientation of aid. c) Estimation with possible convergence in inequality. We regressed changes in Gini indices on the initial levels of Gini and on the changes in income levels, and (changes in) Z and A variables: (ln Ginict – ln Ginic0)/t = α + β1 ln Ginic0 + β2[(ln yct – ln yc0)/t] + δlnZct + θlnAct + λt’ + εct (4) where Gini is a proxy of inequality as measured by Gini coefficients. 5 Other typical examples of fixed effects include capacity, motivation, and productivity, which cannot be controlled by explanatory variables (Higuchi, Ota, & Shimpo, 2006, chap. 8). 7 2.3 Leg 3: Poverty and Growth/Inequality (Poverty Regressions)6 We next regressed changes in poverty headcount ratios on the initial levels of poverty, average income growth, and changes in inequality measurements with control variables of A: (ln Povct – ln Povc0)/t = α +β1 ln Povc0 +β2 [(ln yct – ln yc0)/t] +β3 [(ln Ginict – ln Ginic0)/t] + θlnAct + λt’ + εct’ (5) where Pov is a poverty index. We used poverty headcount ratios in this study. 2.4 Description of Explanatory Variables Aid Variables (A) Aid variables are our main interest for assessing the effectiveness of aid in reducing poverty. This study used two different sectoral data of economic aid and social aid, in addition to aggregate aid relative to GDP. It is assumed that economic aid has a positive impact on growth, as it is expected to expand supply capacity. On the other hand, social aid, which includes projects for basic human needs such as education and health, is expected to exhibit a pro-poor orientation in inequality regressions as it is more targeted to the poor. Aggregate aid is also expected to show pro-poorness, as other categories (aggregate aid except economic aid and social aid) which include commodity aid and humanitarian aid, do not have an immediate impact on growth, but assist the poorer segments of society. (Details will be given in Section 3.2.) Factors and Conditions (Z) 7 Policy Variables: This study employed three policy variables (fiscal, monetary, and trade policies), commonly used in the empirical aid-growth literature: government consumption relative to GDP (G/GDP), the period average inflation rate, and trade (exports plus imports) relative to GDP (T/GDP). The G/GDP was included to determine how the size of government affects growth, and whether fiscal policy affects poverty reduction. While excessively large government expenditure is negatively associated with growth, it may have a positive effect on poverty if the expenditure is targeted at the poorer people in society. Inflation was also included as a policy variable to determine how monetary policy affects holistic growth as well as the poor. Excessive inflation tends to have a strong negative influence on household consumption. We suspect that the negative impact of inflation is much more severe on the poor, whose 6 We had another research question: “How does aid work for poverty reduction, all in all?” We intended to confirm the existence of poverty convergence and the elasticity of poverty reduction with respect to growth and income distribution. Furthermore, we attempted to test whether aid has additional systematic effects for poverty reduction, other than its impact through growth and changes in inequality. We found that significance of the other channels was very marginal. For brevity, the details of the results are not shown in this paper. The model is explained here in order to show our entire research framework. 7 The named categories for each variable are not mutually exclusive. 8 consumption rate in comparison to the entire income tends to be relatively higher than that of other income groups. The T/GDP, a measure of trade openness, was included to determine whether economic policies on international integration would affect growth in aid recipients. Numerous studies suggest that openness of trade is highly correlated with economic growth. Trade is believed to promote industrialization and enhance technological progress, providing learning opportunities for developing countries, which eventually positively influence economic growth. At the same time, trade volume tends to become larger in some developing countries mainly because of the increasing importation of consumption goods. How much of the benefit of liberalized trade can be derived for growth varies among countries. Taking this into consideration, the impact of trade is expected to be neutral; it depends on many factors.8 Institutional Variables: Institutional quality was incorporated as a variable of interest. Many previous studies used an institutional quality index, where particular indices were taken from the International Country Risk Guide (ICRG) indicators (PRS Group, 2012) to define their institutional quality, such as quality of bureaucracy, rule of law, and expropriation risk. Considering the capacity of the recipient country key for development effectiveness, we used the variable of institutional quality which indicates capacity of the government. A country with good institutions, especially capacity, can manage both public and private activities efficiently, which brings about positive effects on growth, as well as development effectiveness. With many previous studies having confirmed the positive relationship between institutional quality and growth, the expected sign of the coefficient is positive. Exogenous Variables (X) The fertility rate and average rate of change in terms of trade (TOT) were added as exogenous variables. These variables are considered to be exogenous, or not correlated with other selected variables, while they have explanatory power for growth. Regional Dummies: The current study (for Eq. 1) used three regional dummies:9 East Asia and the Pacific (EAP), Sub-Saharan Africa (SSA), and the Former Soviet Union (FSU),10 with other regions treated as the base. EAP (or only the East Asian countries) and SSA regional dummies are the ones most often used together in aid-growth empirics (Burnside & Dollar, 2000, 2004; Dalgaard et al., 2004; Rajan & 8 The high correlation between trade and institutional quality should be carefully considered in cross-country analyses. Rodriguez and Rodrik (2001) claimed that trade might simply act as a proxy for a variety of other important policy and institutional variables. As a consequence, once the institutions are controlled for, integration/trade has no direct effect on income. 9 We assembled a full set of regional dummies in the database corresponding to the World Bank’s classification: East Asia and Pacific (EAP), Europe and Central Asia (ECA), Latin America and the Caribbean (LAC), Middle East and North Africa (MNA), South Asia (SAR), Sub-Saharan Africa (SSA), and Former Soviet Union (FSU). These are used for robustness check. 10 The current study used the FSU as a regional designation rather than the ECA that includes the FSU, as the FSU had more significance. This is consistent with the judgment of Burnside and Dollar (2004). 9 Subramanian, 2008). In addition, we considered it important to include the FSU dummy, as the countries of FSU have experienced rapid reform with economic downturns and remarkable growth thereafter. Time Dummies: We used the crisis dummy as a time designation, rather than normal time dummies such as annual or decadal time dummies. To have a crisis dummy attached, the period has to contain one of the following: Debt Crisis (1982-83), Asian Financial Crisis (1997-98), or the World Financial Crisis (2008-09). This dummy takes care of time-fixed effects (λt and λt’)11 caused mostly by crises. Instruments Endogeneity of aid is one of the inevitable limitations of aid-growth studies. 12 In light of the shortcoming, instrument variable (IV) estimators are commonly used in the regression analyses. The current study used standardized instruments of lagged values for the levels (including instruments of lagged period averages for explanatory variables of period averages), and initial value and lagged rate of changes for growth spells. Instrumented explanatory variables are identified in the tables of the regression results. In growth regressions for the national average income (Eq. 1), variables of trade, institutions, and aid, except for their initial values, were instrumented.13 In regressions for the poorest quintile income (Eqs. 2 & 3), only national average per capita income was instrumented. In the same manner, only income growth spells were instrumented in the Gini convergence regressions (Eq. 4). In the poverty change (P-G-I) regressions (Eq. 5), Gini changes were instrumented. 3. Data The medium-term data set was constructed as a base for the analyses in this study, since our interest is in the impact seen over the medium to long run. The growth spells of the different time durations are compiled from the available data points of bottom-quintile income shares during the period from 1978 to 2010.14 This medium-term database contains 242 growth spells of the five to nine-year period, each with the average duration of 5.72 years, for 98 countries. Out of these, the growth spells of aid sectoral analyses were obtained in accordance with data availability for sectoral aid. The number of observations for sectoral aid is lower, as it was available only after the year 1995. The data for sectoral aid analysis contains 183 growth spells for 60 countries. The data for aggregate aid, which was available during the period of 1978-2010, contains 216 growth spells for 87 countries. When comparing the regression results of sectoral aid and aggregate aid, the sample period of aggregate aid was restricted to match that of sectoral aid. Actual 11 The compound error term can be decomposed into three components of cross-country fixed effects, time-fixed effects, and a random error term. 12 For further discussion see Hansen & Tarp, 2001, pp. 554-561; Clemens et al., 2004, pp. 26-27; Dalgaard et al, 2004, pp. 201-208; Hudson, 2004, p. 187; and Rajan and Subramanian, 2008, pp. 645-650. 13 We posited that there were no significant endogeneity issues for other variables when assessing their impact on inequality. We took the judgment on the usage of IV similar to Dollar and Kraay (2002, pp. 204-205). 14 If the data was not available for the full spell, the period average ratio was calculated with the data that was available. 10 numbers of observations were further reduced in IV estimations, when one spell was used as an instrument when necessary. Summary statistics for selected variables used in the regressions analyses are listed in Appendix 1.The details for the sample of spells are summarized in Appendix 2. 3.1 Data Sources Most of the data are taken from the secondary data of various international institutions: major socio-economic data from World Development Indicators (WDI) of the World Bank (World Bank, 2012), institutional indicators from the ICRG of the PRS Group (PRS Group, 2012), and aid data from OECD Stat Extracts (OECD-DAC, 2013a). The variables and their data sources are listed in Appendix 3. 3.2 Aid Data Aid data by sector are obtained from the OECD’s Creditor Reporting System (CRS). The CRS provides an aid activity database, which contains detailed quantitative and descriptive data on individual aid projects and programs. The CRS data has made it possible to analyze the sectoral and geographical breakdown of aid for selected years and donors to examine development effectiveness. In the CRS, data on the sector of destination are recorded using purpose codes. There are eight main categories: (I) social infrastructure and services; (II) economic infrastructure and services; (III) production sector; (IV) multi-sector/cross-cutting; (V) commodity aid and general program assistance; (VI) action relating to debt; (VII) humanitarian aid; and (VIII) unallocated/unspecified.15 We used the sectoral data of (I) social infrastructure and services, social aid as we called it, and (II) economic infrastructure, as economic aid, in comparison with the aggregate aid of (I)-(VIII). According to the aid classification of CRS, social aid covers efforts to develop the human resource potential and ameliorate living conditions in aid recipient countries (OECD-DAC, 2007, 2013b). It includes, but is not restricted to: education such as educational infrastructure, services and investment in all areas; health and population such as assistance to hospitals and clinics, other medical services, including disease control and vaccination programs, and reproductive health and family planning; and water supply, sanitation and sewerage. Economic aid covers assistance for networks, utilities and services that facilitate economic activities. It includes, but is not restricted to: transportation, that is, equipment or infrastructure for road, rail, water and air transport; energy, production and distribution of energy; communication for television, radio and electronic information networks; and banking, financial, and business services. In addition, we used aggregate aid, which is the sum of the eight main categories mentioned above. The current study used post-1995 economic and social aid commitment data, due to data limitations in prior years. The completeness of CRS commitments for DAC members has improved from 70% in 1995 to over 90% in 2000, and reached nearly 100% by 2003 (OECD-DAC, 2013c). On the other hand, 15 The contributions of (V)-(VIII) cannot be broken out by sector and are reported as non-sector allocable aid. 11 disbursement data completeness was below 60% prior to 2002, improved to 90% that year, and reached nearly 100% starting with 2007 flows. We felt that only using post-2002 data was not sufficient for time-series cross-country analyses, while including the data before 1995 would bias the results to some extent.16 Since our intention was to examine the significance of the (directional) impact of aid rather than measuring the degree of elasticity in the aid coefficients, this study simply used commitment data. We judged that this would not change the significance of the impact of sectoral aid.17 As to the measurement of aid, this study used the period average gross economic and social aid, as well as net aggregate aid received.18 3.3 Constructing Institutional Quality Indices The meaning, measurement and implications of institutions or “institutional quality” can differ considerably among researchers. We used specified institution indices to examine the impact on our research focus as the importance of using different components of institutions for growth regression has been discussed in the literature. Dalgaard et al (2004, p. 200), pointed out that a composite index of policies might encapsulate some components that enhance the return to aid while others diminish this impact, with the net effect possibly insignificant. Abramovitz (1986) argued that the growth-enhancing effects of economic integration depend on the absorptive capacity of the host economies and that institutional quality was one of the main factors underlying such capacity. We constructed indices for the quality of institutions extracted from the ICRG of the PRS Group. The ICRG indicator is given by the ICRG Rating System, whose rating comprises 22 variables in three subcategories of risk: political (ICRG-P), financial (ICRG-F), and economic (ICRG-E) (PRS Group, 2012). Out of the various composites and individual indices contained in the ICRG, we formed the capacity sub-composite (ICRG Capacity) by compiling five specific components from the ICRG-P bracket: government stability, investment profile, corruption, law and order, and bureaucracy quality. Among the 12 indicators in the ICRG-P, these five indicators denote management capabilities of recipient countries, while the seven other indicators represent the security and stability of a country in a different form of “institutional” capacity. We consider institutional quality of a recipient country, and government managerial capacity in particular, to be key for development effectiveness. As for the measurement, we placed greater importance on change in institutional quality rather than the 16 For instance, some earlier studies employed sectoral disbursement data before 1995 compiled from the CRS bulk data. They calculated the share of the sectoral aid to total disbursement in the CRS database, which was then divided by total ODA (gross). However, it should be noted that in addition to the limited supply of data in the CRS, it was not necessarily quiproportional by sector. In light of these constraints, this study simply used the commitment data, which OECD defines as “a firm obligation to furnish assistance of a specified amount under agreed financial terms and conditions and for specific purposes, for the benefit of a recipient country (OECD-DAC, 2007, p. 11).” OECD-DAC also advises that data on a commitment basis is of better quality than that based on disbursement. 17 The period average ratios of disbursement to commitment data are approximately 96% for the sample countries and periods used in this study. 18 The net value for sectoral data cannot be obtained from the CRS database system. 12 level of institutional quality. While the majority of past research used levels of institutional quality with the assumption that institutional quality does not change much, we consider (directional) changes in institutional quality to matter for assessing development effectiveness. Rapid poverty reduction in low-income countries depends primarily on these countries improving their own policies and institutions (Collier & Dollar, 2001, p. 1787). We emphasize “improving,” that is changing, institutions in this study. 3.4 Cross-Correlations Between Variables Table 1 shows cross-correlations between dependent variables in the top-left rectangle, between independent/explanatory variables in the bottom-right rectangle, and between dependent and explanatory variables in the bottom-left non-shaded rectangle for the period from the late 1990s to 2000s. We considered correlation among variables in order to avoid a possible bias by multicollinearity when selecting the sets of variables. All three aid variables, e.g., economic, social, and aggregate aid, demonstrate positive correlation with the period average growth rate of per capita income of the poorest quintile. However, social and aggregate aid is negatively correlated with the period average growth rate of a nation’s average per capita income while economic aid has a positive correlation. These comparisons of simple correlation coefficients indicate that the effects of aid could vary between different sectors of aid. All aid variables demonstrate negative correlation with Gini. The impact of social and aggregate aid is significant, indicating that aid may contribute to reducing inequality. In addition, change in institutional quality has a positive correlation with the period average growth rate of per capita income, while the sign of the coefficient turns negative in the relationship with that of the poorest quintile. Among the correlations between aid variables and other explanatory variables, the aid variables are negatively correlated with the initial level of institutional quality (ICRG Capacity): Economic aid and aggregate aid are correlated at the 1% significance level and social aid at the 5% significance level. This suggests that aid is given to lower-income countries, whose initial condition of institutional quality is usually low at early stages of development. On the other hand, the correlation between aid variables and change in institutional quality is positive, but insignificant. The relationship between aid and change in institutional quality shows wide dispersion, unlike the relationship in levels. This suggests that the allocation of aid in the past did not necessarily depend on whether the institutional quality of recipient countries was changing, i.e. improving. The period average growth rate of trade has a significant positive correlation with only economic aid at the 10% level. Economic aid has facilitated the improvement of various economic infrastructure, which is key for promoting trade. For instance, Aid-for-Trade Initiatives have been implemented to help developing countries overcome trade-related issues, such as transportation management and custom regulations. In addition, both initial levels and change in institutional quality are significantly correlated with the initial level of trade. We found that institutional quality variables did not have any significant correlation with the 13 growth rate of trade in our sample. 4. Regression Results 4.1 Economic Aid Promotes Growth Table 2 shows the regression results of Equation 1—growth regressions—with variables of policies, institutions, and aid. The series of regression analyses showed that economic aid indeed promotes growth. First, we examined the impact of three policy variables and an aid variable on growth (Columns 3-5). A negative coefficient for initial G/GDP (fiscal policy) implies that large government is, on average, negatively associated with growth performance. A strong negative coefficient is continuously attached to the inflation variable, indicating that high inflation can be an impediment to growth. This suggests the importance of inflation control (monetary policy). The result of changes of trade/GDP (trade policy) was mixed, and it was statistically insignificant. That is, whether trade promotes growth depends on how each country manages it. This tendency is robust, as it occurred in the same manner in our earlier study (Hirano & Otsubo, 2012). All the coefficients of economic, social, and aggregate aid were positive at the 5% significance level. However, these positive results happened to stem from the relatively high correlation between trade and aid variables, especially for economic aid. To avoid a possible distortion in the results, we decided to use only the inflation variable as a policy variable.19 The estimation results with one policy (inflation) and aid variables are shown in Columns 6-8. A significant negative impact of inflation did persist. We found that the coefficient of economic aid (Column 6) was a positive and significant impact while the coefficients of social aid (Column 7) and aggregate aid (Column 8) were positive, but smaller and insignificant. These results confirm that economic aid contributes to growth regardless of certain prerequisite conditions20 of a recipient country, such as good policies, institutional quality, and climate-related geographical environments. This is no surprise, as economic infrastructure and services have been provided to developing countries to spur economic development. The insignificant positive coefficients attached to social and aggregate aid should be considered natural, as their immediate purposes are not for increasing per capita income growth. We argue that the reason why past literature often failed to prove the significant positive impact of aid on growth is not because aid did not actually promote growth, but because the authors failed to extract the different effects from each aid sector. By carefully considering the uniqueness of each sector/type, aid effectiveness can be appropriately assessed by measuring per capita income. Next, we controlled for change in institutional quality and aid on growth (Columns 10-12). The variable of change in institutional quality was positive, yet insignificant, in all estimations. The significant and positive coefficient of social aid (Column 11), as well as a moderately significant and positive 19 For simplicity, the G/GDP variable was also dropped in the estimation in order to measure the marginal effect of aid. Nevertheless, the inclusion of G/GDP did not change the results, while only increasing the R2 by 2-4 percentage points. 20 Past literature argued that aid promotes growth if certain conditions are met, using the interaction terms of aid and certain conditions. This study shows that economic aid promotes growth without using any interaction terms. 14 coefficient of aggregate aid (Column 12) was observed. These increasing significance levels of social and aggregate aid could be attributed to joint effects from change in institutional quality. In other words, the impact of aid on growth can be more significant in a recipient country whose institutional quality and capacity in particular are improving. In contrast, the positive coefficient of economic aid lost its significance in this set of variables (Column 10) due to a statistical limitation. This supposedly inaccurate result came from the high correlation between economic aid and change in institutional quality that the current study confirmed in the correlation matrix. On top of these estimations of institutions and aid, we added inflation as a selected policy variable (Columns 14-16). Inclusion of inflation did not change the core results; it only increased the coefficients’ size and significance of all aid variables. Columns 15 and 16 proved the robustness of this key finding with an additional insight: the effectiveness of social and aggregate aid is larger and more significant with improving institutional quality, and it further increases in a country where monetary policy is appropriately managed. The estimation of economic aid (Column 14) did not show the same result, regardless of our hypothesis that economic aid would work better where institutional quality is improving. The insignificance attached to economic aid simply indicates that the statistical problem arising from the high correlation between economic aid and change in institutional quality remains in this estimation. This issue prevents us from adding these two variables together in growth regressions in the current study. The distorted result cannot be directly compared with the results of social and aggregate aid, which do not have the correlation problem with change in institutional quality. Therefore, the regression results of economic aid with change in institution (Columns 10 & 14) are presented only for reference in this paper. We also should not ignore the fact that the negative coefficient attached to the SSA regional dummy persisted. The SSA nations tend to lag behind in promoting growth and effectiveness of aid. This suggests that the characteristics of the region should be taken into consideration in formulating and implementing development strategies. 4.2 Building Institutions Matters for Development Effectiveness21 Table 3 shows the impact of aid on countries with different institutional levels and rates of change in institutions. We estimated the effects of aid by dividing samples based on different conditions of institutional quality: high, low, improving, or not improving, as defined in the section detailing model specification. Using the estimation results of Columns 6, 7, and 8 in Table 2 as a base (these are listed again in Columns 1, 6, & 11 in Table 3 for quick reference), we further examined the impact of economic, social, and aggregate aid in the respective sample groups. When we tested the sample groups divided by initial level of institutional quality, we found that the impact of aid became less significant (Columns 2, 3, 7, 8, & 12) no matter how high or low the institutional 21 Some parts of this chapter are drawn from Hirano (2014). 15 quality a country had at the starting point. The impact of economic aid lost its statistical significance level of 10% while social aid decreased the t-statistic value. Interestingly, this study found more distinct differences when testing the sample groups divided by directional change in institutional quality. The impact of both economic and social aid became larger with statistical significance in a country where institutional quality was improving (Columns 4 & 9). On the other hand, effects of both economic and social aid became much less significant in a country whose institutional quality is not or improving22 or worsening (Column 5 & 10). As for aggregate aid, a similar effect was observed in the case of countries where institutional quality is not improving: the coefficient of aggregate aid turned negative in the same manner as social aid (Column 15). However, the impact did not become larger even in the sample group with improving institutional quality (Column 14). Even with these specifications of samples in two groups, the impact of aggregate aid was not increased. Mutually offsetting effects of various types of aid are operating in both groups, with or without high or improving institutional quality. In summary, this series of results implies that development effectiveness does not depend on the initial condition of institutional quality, but it does depend on whether the institutional quality, or capacity, of a recipient country is improving or not. The tendency is to direct aid to countries with a lower level of institutional quality, believing that they have room to improve their institutions in the course of economic development. Development effectiveness could be maximized if aid was allocated to a country in which institutional quality is improving. The impact of development aid on growth could be effectively increased if aid packages with an institution- or capacity-building scheme were provided to developing countries, as they often face constraints in building capacity by themselves during the early stages of development. 4.3 Aid Mitigates Inequality This study found strong evidence that aid played a role in reducing inequality. This effect is robust, as confirmed by three different methods.23 Aid, indeed, has been pro-poor. Table 4 shows the estimation results of Equation 2. We conducted a country fixed effects regression in order to examine the pro-poor effects of aid, as well as to identify the countries with pro-poor fixed effects. First, we examined the basic specification of Equation 2 in which we regress the logarithm of per capita incomes of the poor on the logarithm of average incomes without other control variables. This study observed that, on average, the income of the poor rose proportionately with average national income (Columns 1 & 2).24 Thus growth is good for the poor, confirming Dollar and Kraay’s assertion using our data. We looked for other factors that might have direct effects on the poor other than their effects through 22 In a country where institutions are of high quality, the marginal effect of aid tends to be relatively low. 23 The results of Equations 2 and 4 are presented in this section, and those of Equation 3 are shown in Section 4.6. 24 We confirmed it in both level and growth regressions. The robustness is confirmed with the rejection of the null hypothesis that the slope of this relationship is equal to one. The robustness held even without a crises dummy. 16 growth. We could confirm that the coefficients of social aid and aggregate aid were positive at a 1% significance level. This strongly supports the evidence that both social and aggregate aid have strong pro-poor effects. That is, aid indeed has the effect of mitigating inequality. Other than these two variables, we did not find any factor that provides benefits to the poor (Columns 3-7). Instead, we found a significant negative effect of institutional quality (as measured by ICRG Capacity) on the poor (Column 6). This indicates that growth-enhancing institutions have a negative impact on the poor, surpassing the impact they bring about through accelerated growth. The effects of trade on the poor were insignificant (Column 5). The coefficient of economic aid was also insignificant with a negative sign (Columns 7). We intended to test a set of variables of policy, institutional quality, and aid. However, we did not do so, given the statistical problem stemming from multicollinearity.25 We simply tested a set of policy (G/GDP) and institutional quality variables (Column 10), and a set of policy (inflation) and aid variables (Columns 11-13). We found that the sign and tendency still held. Table 5 shows the estimation results of Equation 4, where we regressed changes in inequality (as measured by the Gini index) on the initial levels of inequality (Gini), the changes in income, and levels and changes of A and Z variables. All the coefficients of aid were significantly negative. These results also suggest that aid reduces inequality. This result was robust even in the estimations with variables of policy (inflation and G/GDP) and institutional quality (Columns 9-14). We also found the different effects of economic aid on inequality. The coefficients attached to economic aid were strongly negative in the estimations using changes in Gini coefficients as dependent variables (Columns 3, 6, 9, & 12 in Table 5). However, that impact was insignificant in the estimations using levels/changes of the incomes of the poorest quintile as dependent variables (Column 7 in Table 4).26 These contrasting results indicate that economic aid may not benefit the poorest quintile, while it does reduce inequality among richer segments of the society. 4.4 Large Variations in Pro-Poor Country Fixed Effects Table 6 shows the large variation of cross-country fixed effects of pro-poorness. The fixed effects are obtained from the decomposition of compound error terms.27 Using a crisis dummy for the period/time fixed effects, the estimates of the cross-country fixed effects are made by using the fitted equations in Column 13 of Table 4.28 We identified Ethiopia, Tajikistan, Nepal, Pakistan, Bangladesh, Malawi, Burundi, 25 There is high correlation between the level of institutional quality and each aid variable as mentioned earlier. The correlations of G/GDP and social and aggregate aid are also high at the 1% significance level. In any event, estimating an additional effect over average income should be precisely examined, one by one, as it is a marginal and sensitive effect. 26 That impact of economic aid was insignificant in the estimations using changes of income of the poorest quintile as dependent variables (Column 1 in Table 8), which will be shown later. 27 The cross-country fixed effects are demonstrative of invariant uniqueness or characteristics of the country. Nevertheless, it should also be noted that the identification of pro-poor fixed effects depends heavily on the selection of periods in the set of observations. 28 Therefore, institution-related fixed effects are also included in these estimations of country pro-poor fixed effects. 17 Mali, Gambia, and Burkina Faso as the top ten countries with pro-poor country fixed effects.29 We also identified Panama, Brazil, Colombia, Seychelles, Argentina, Bolivia, South Africa, Honduras, Belize, and Venezuela as the countries with the least pro-poor fixed effects. As for the EAP countries, we found that Lao PDR and Vietnam were countries with relatively high pro-poor fixed effects, ranked 12th and 20th among a total of 78 sample countries. Our plan for future case studies is to look into the country-specific factors in those countries that have made growth more pro-poor. 4.5 Aid Works More Effectively in a Country with Pro-Poor Institutions Certain countries have some unique characteristics, or institutions, that make them pro-poor. These pro-poor country-specific factors are essential in making growth benefit the poor, as the impact on the poor is not determined only by select explanatory variables. We, therefore, came up with another question: whether aid could be more effective for the poor in a country which has pro-poor country fixed effects. To answer this question, we examined the effectiveness of aid depending on the pro-poor country fixed effects of a recipient country. We divided the samples into two different groups based on the pro-poor country fixed effects estimated in Column 10, Table 4: (i) countries whose pro-poor country fixed effects are positive; and (ii) countries whose pro-poor country fixed effects are equal to zero or negative. Table 7 shows the results. This study found that the effectiveness of social and aggregate aid is positive and more significant in countries with positive pro-poor fixed effects, while its significance decreased in countries with negative pro-poor fixed effects. The difference in significance was more distinctive for aggregate aid. As for economic aid, the difference in effectiveness of aid was also obvious between the two sample groups. We found that economic aid would render a negative effect on the poor in a country with negative pro-poor country fixed effects. Meanwhile, we could not find a positive and significant impact of economic aid on the poor, as the impact remained negative (and insignificant) even in countries with (positive) pro-poor fixed effects. These results support the idea that pro-poor country-specific factors could influence the way aid works. This study suggests that a country with a pro-poor orientation could manage aid better, by properly targeting the poor and dedicating efforts to achieve poverty reduction. The pro-poor country-specific factors consist of various aspects of a country, including some special institutions that were not captured in explanatory variables (including in the ICRG Capacity index), traditional norms, social factors or systems. We suggest that identifying the possible pro-poor country-specific factors would provide us clues for increasing pro-poor development effectiveness. This idea has something in common with the policy suggestions by Mosley, Hudson, and Verschoor (2004) in their paper Aid, Poverty Reduction and the “New Conditionality.” They argued that providing aid to a developing country based on pro-poor fiscal policy performance greatly affects poverty reduction. In addition to pro-poor policy, this study shows the 29 The high pro-poor country fixed effects of Ethiopia were observed in our earlier study (Hirano & Otsubo, 2012). 18 importance of pro-poor country fixed effects, and pro-poor institutions (which we could not capture in the estimation) in particular, for increasing development effectiveness. If the pro-poor country fixed effect and pro-poor effect of aid have some linkage, a synergetic effect could emerge for better development impact. 4.6 Aid Is Good for the Poor We extended the analysis for estimating the systematic effects of aid (economic, social, and aggregate aid) on quintile 1 to quintiles 2-5. This study confirms that economic aid is good for the poor due to its growth-inducing impact, and that social aid is good for the poor through systematic distributional effects. These results further substantiate the assertion that aid is good for the poor. Table 8 shows the estimated results of Equation 3, in which we examine the variation of average incomes of quintiles 1-5 by variation of average income of the country with economic, social, and aggregate aid respectively, for the late 1990s to 2000s. We found that social and aggregate aid had positive systematic effects on quintiles 1 and 2 at the 1% significance level (Columns 6, 7, 11, & 12) and quintile 3 at the 5% significance level (Columns 8 & 13). Another remarkable finding is that social and aggregate aid had negative systematic effects on quintile 5 at the 1% significance level. These findings prove that aid is indeed good for the poor, and that aid contributes to narrowing the gap between the rich and the poor. As for economic aid, none of the coefficients were significant (Columns 1-5). That is, economic aid does not have systematic effects on the poor, nor any particular income group. Together with the evidence that economic aid contributes to economic growth (the result of Eq. 1), the estimating results for quintile 1-5 reveal that economic aid is good for the poor through growth effects, meaning the poor gain from economic aid because of its spurring of economic growth. While economic aid does not better the lives of the poor more than the wealthier segments of society, it does boost their incomes—as it boosts all incomes. It is social and aggregate aid that indeed reduces inequality. Taken altogether, we can consider aid to be “super-pro-poor,” in accordance with the definition by Dollar and Kraay (2001, p.8). 4.7 Policy, Institutions, and Globalization Do Not Have Systematic Effects on the Poor We next tested if there are any factors of policy, institutions, globalization, or other potential factors, that have systematic effects on the poor and other income groups. The latest study by Dollar et al. (2013) illustrated that there were no macroeconomic determinants that were particularly pro-poor. In light of those findings, we first reexamined if any pro-poor systematic effects of macroeconomic factors could be observed with our own database for the late 1990s to 2000s. Table 9 shows the results of Equation 3, in which we regressed the changes in the incomes of the poorest quintile on the changes in average incomes with variables of policy, institutions, globalization, and other potential pro-poor factors. Next, we extended our analyses in order to investigate systematic distributional effects of each factor on different income 19 groups (Tables 10-13).30 We replaced the dependent variable of quintile 1 with other income groups of quintiles 2-5 respectively. The study reconfirmed that there was no robust evidence that any variables of policy, institutions, globalization, or other potential pro-poor factor (selected for this study) had any systematic effect on the poor, nor on other quintiles. Impact of Policies and Institutions We found a clear tendency, despite a lack of statistical significance, that institutional quality—as measured by our capacity indicator—has a positive effect only on the richest quintile while it has a negative effect on quintiles 1-4 (Table 10) over average incomes. The t-statistic increases as it goes to the poorest, even though it is statistically not significant. This clear tendency supports the findings in our earlier study that even though institutions as measured by ICRG promoted growth and poverty reduction, these were negatively associated with the incomes of the poor (Hirano & Otsubo, 2012). This finding could stem from high-quality institutions possibly focusing more on (or being better able to reach out to) the rich and middle classes, rather than the poorest in society. Institutional indicators including ICRG and other pro-investor indices have been used as conditionality or benchmarks for aid allocation. This study however suggests that both the pro-growth (pro-investment) and pro-poor nature of institutional quality should be taken into consideration for aid allocation. Impact of Globalization In addition to policies and institutions, we analyzed the impact of globalization on the poor. This study found that trade and FDI had positive systematic effects only on quintile 5 (Column 5 in Table 11; Column 5 in Table 12), albeit a statistically insignificant effect, as globalization’s impact is already captured in average income growth. These findings are similar to what Milanovic (2005) found in his study: the rich benefit from openness in a developing country (in very low income countries in particular). Even though the poor can benefit through growth, which is accelerated under globalization, there seems to be a consensus that the rich, on average, tend to benefit more from globalization (trade and FDI). Our new finding regarding globalization, however, is that remittances tend to benefit poorer people more: table 13 shows some evidence (albeit weak) that remittances have positive systematic effects on quintiles 1-4. In contrast, they have a negative association with the richest quintile. This finding is somewhat intuitive, given that remittances have been essential for lower income households in developing countries to generate more income. It does represent an instance, however, where the poor benefit from globalization disproportionally more, compared to richer people. Nevertheless, it should be noted that the significance of the systematic effect of remittances is higher for quintiles 2 and 3. In fact, it is not often the 30 Only the results of selected factors are reported in this paper. 20 case that migrant workers are from the poorest segment of society. This indicates a need to create opportunities for the poorest to work abroad and receive more benefits from remittances. In other words, we need to target the two lowest quintiles. These results should be understood together with our earlier study, which provides a literature survey and empirical stocktaking of the impact of globalization (Hirano & Otsubo, 2012). Out of these empirical analyses using medium-term growth spells and quintile income share data spanning 98 countries over the period of 1978-2010, the most relevant results to this current study are reproduced in Appendix 4.31 The results include a) the impact of globalization on growth, b) the impact of globalization on inequality (on the poor), and c) the impact of globalization on inequality (on the Gini Index). In the past couple of decades, FDI, and therefore the activities of multi-national corporations, have been driving the process of globalization. As the representative results in Appendix 4 (Columns 3, 6, & 9) show, FDI has promoted growth with statistical significance, while widening inequality. This study suggests that both economic and social aid could make opportunities created by FDI more accessible to the poorer segments of FDI-recipient countries. The impacts of international trade have been mixed (Columns 1, 2, 5, & 8). Trade promotes growth if it is complemented with proper macroeconomic policies and institutions, as many earlier studies pointed out. Trade has been largely neutral in domestic income distribution, again on average and with wide variation. Therefore, the impact of trade on poverty reduction has been ambiguous. It depends on many factors. The impacts of rapidly expanding flows of remittances as a result of labor market integration have been consistently positive on growth, inequality (i.e., reducing income inequality), and poverty reduction. Additionally, the direction of impacts has been rather robust, and the impact tends to be positive, regardless of the time period or countries sampled. However, given the current small size of remittances relative to GDP (or Gross National Income) of the developing countries, statistical significance has not yet been found; expansion would suggest auspicious results. International cooperation that facilitates labor-market integration (i.e. that supports migrant workers from developing countries) could have a large impact on the poverty picture in developing countries. In short, there is no single factor of globalization that significantly benefits the poor more than the growth globalization spurs for all. The results laid out in this study suggest that aid could be utilized to enable the poor to benefit more from globalization-derived development. Impact of Other Factors We further examined the systematic effects of the factors that were plausibly considered pro-poor: 31 For further details, see Hirano, Y. & Otsubo, S. (2012). Poverty-Growth-Inequality Triangle under Globalization: Time Dimensions and the Control Factors of the Impacts of Integration. (GSID Discussion Paper No.191). Japan: GSID, Nagoya University and Otsubo, S. & Hirano, Y. Poverty-Growth-Inequality Triangle under Globalization: What do we really know about the pro-poor/anti-poor impacts of economic integration? (Ch. 2) In Otsubo. S. (Ed.). Globalization and Development Vol.1: Leading Issues in Development with Globalization. Singapore: Routledge (forthcoming). 21 public health expenditure,32 changes in terms of trade, and changes in the share of agriculture-related exports to merchandise exports. We found all of the coefficients were positively associated with the poor; nevertheless they were not at the conventional significance level (Columns 7-9 in Table 9), and neither were their systematic effects observed on other quintile groups with any statistical significance. To explain, health expenditure tends to have positive systematic effects on poorer people, while it has negative systematic effects on richer people. It is a fact that the government in a developing country has to spend a certain amount of the budget to maintain general healthcare services, which not all poor people can access. In other words, health expenditure is not always well targeted for the poor in such cases. By contrast, most healthcare development aid projects have clear targets regarding who should be treated, and what the project will accomplish, with great emphasis on poverty reduction. The implementation is monitored with expertise from overseas, transferring know-how to developing countries. This contributes to the difference between general health expenditure by domestic governments and health services supported by development aid from overseas. This shows the importance of supporting developing countries in utilizing both domestic resources and foreign aid to effectively achieve poverty reduction. 5. Conclusion This study affirms that aid is good for the poor, not only because aid benefits the poor in similar proportion to other income groups on average, but also because aid benefits the poor more than other income groups. This result is based on strong empirical analyses, in which we examined aid in promoting economic growth and mitigating inequality for poverty reduction based on the P-G-I framework. The findings suggest that social aid significantly and directly benefits the poorest in society, while economic aid increases the income of the poor through growth. Taken together, the conclusion is that an optimal mix of economic and social aid would maximize benefits for the poor. This study also confirms that none of the elements of globalization (trade, FDI, and remittances), policies (government expenditure and inflation management), institutional quality, or other plausibly pro-poor factors have systematic effects (to a statistically significant level) on the poor or any other income group, once national average incomes are controlled. Regarding globalization, the results showed a weak yet clear tendency that trade and FDI benefited the richest quintile more than any other income group, which indicates that in order to make globalization engender poverty reduction, complementary policies and institutions are needed. Therefore, the collective evidence presented here suggests that aid could play a crucial role in enabling the poor to benefit from development in this globalized world. These discoveries further point to the need for aid that assists developing countries in search of the mix of economic and social aid that jointly promotes participation of the poorest in the development process under globalization. In this manner, aid can make greater strides in spurring development. In addition, this study highlights the essential nature of institutions for development effectiveness. First, 32 As we found that overall government consumption expenditure did not have robust systematic effects on the poor, we tested health expenditure more specifically. 22 building institutions is key for increasing the effectiveness of aid. The empirical results clearly show that development effectiveness does not depend on the initial quality of a recipient country’s institutions, but it does depend on whether that quality, or capacity, is improving or not. It implies that even a country with lower institutional quality has the possibility of achieving higher growth, unlike the continuing debates on policy conditions for aid effectiveness. This suggests that development practitioners and policy makers should take this potential into consideration in their decision-making about aid allocation. This study also proposes that providing aid packages with institutional capacity-building schemes or projects could increase development effectiveness. This is new insight for the voluminous aid-growth literature. Past studies regarded institutional quality in their estimations as something that does not change for a long time (often using one indicator for the entire period); nevertheless, this study indicates the importance of change in institutional quality for development effectiveness. Second, pro-poor country specific factors, or pro-poor institutions, are key for development effectiveness. The empirical evidence shows that the effectiveness of aid is positive and more significant in countries with a positive pro-poor fixed effects. Therefore, the study suggests that identifying pro-poor country fixed effects, and pro-poor institutions in particular, would help to increase development effectiveness. The combination of pro-poor country fixed effects and pro-poor effects of aid could bring about a synergetic effect for development effectiveness. This study also proposes policy makers and people in the development community take into consideration not only growth-enhancing policies and institutions, but also pro-poor institutions in allocating aid. With the optimal mix of pro-growth/investment-conducive institutions and distributional pro-poor/poor-empowering institutions, development effectiveness will be further increased. Here, a comprehensive survey of best practices and best pro-poor institutions could contribute to an enhancement in development effectiveness in our future practices. These findings make contributions to two fields of literature: globalization and inequality, and aid effectiveness. First, the discovery of the significant pro-poorness of social aid and aggregate aid could shed new light on the controversial debate on globalization and inequality, where the current discourse asserts that there is no factor of globalization that has significant systematic effects on the poor (Dollar & Kraay, 2002; Dollar et al., 2013). Second, this study provides new perspective on the aid effectiveness controversy that has continued for the last 60 years. This study shows that aid is good for the poor by extracting the effects on different sectors and by looking at both the growth effects and distributional effects of aid. We argue that the reason why past cross-country empirics could not confirm the significant impact of aid is not because aid is ineffective, but because past efforts overlooked the distinct natures, and the corresponding distributional effects, of sector-specific aid, and thereby did not accurately assess aid effectiveness. We hope this study will offer some evidence for policy makers, researchers, practitioners, and people in the development community to rethink the role of aid in effectively reducing poverty in a globalized world. 23 References: Abramovitz, M. (1986). Catching Up, Forging Ahead, and Falling Behind, The Journal of Economic History, 46(02), 385-406. Barro, R. J. (1991). Economic Growth in a Cross Section of Countries, The Quarterly Journal of Economics, 106(2), 407-443. ______. (1997). Determinants of Economic Growth: A Cross-Country Empirical Study, MIT Press. Boone, P. (1996). Politics and the Effectiveness of Foreign Aid, European Economic Review, 40, 289-329. Bourguignon, F., & Sundberg, M. (2007). Aid Effectiveness: Opening the Black Box, The American Economic Review, 97(2), 316-321. Burnside, C., & Dollar, D. (2000). Aid, Policies, and Growth, American Economic Review, 90(4), 847-868. ______. (2004). Aid, Policies, and Growth: Revisiting the Evidence. (Working Paper No. 3251). Washington DC: The World Bank. Clemens, M. A., Radelet, S., & Bhavnani. R. (2004). Counting Chickens When They Hatch: The Short-term Effect of Aid on Growth (Working Paper No. 44, Revised Nov.2004). Washington DC: Center for Global Development. Collier, P., & Dollar, D. (2001). Can the World Cut Poverty In Half? How Policy Reform And Effective Aid Can Meet International Development Goals, World Development, 29(11), 1787-1802. ______. (2002). Aid Allocation and Poverty Reduction, European Economic Review, 46(8), 1475-1500. Dalgaard, C., Hansen, H., & Tarp, F. (2004). On the Empirics of Foreign Aid and Growth. The Economic Journal, 114(496), 191–216. Dollar, D., Kleineberg, T., & Kraay, A. (2013). Growth Still Is Good for the Poor (Policy Research Working Paper No. 6568). Washington DC: The World Bank. Dollar, D., & Kraay, A. (2001). Growth Is Good for the Poor (Policy Research Working Paper No. 2587). Washington DC: The World Bank. ______. (2002). Growth is Good for the Poor. Journal of Economic Growth, 7(3), 195-225. Easterly, W., Levine, R., & Roodman, D. (2003). New Data, New Doubts: A Comment on Burnside and Dollar’s “Aid, Policies and Growth” (2000) (NBER Working Paper No.9846). Massachusetts: National Bureau of Economic Research (NBER). ______. (2004). Aid, Policies, and Growth: Comment. American Economic Review, 94(3), 774-780. Fischer, S. (2003). Globalization and Its Challenges, American Economic Review, 93(2), 1-30. Gulati, U. C. (1978). Effect of Capital Imports on Savings and Growth in Less Developed Countries, Economic Inquiry, 16 (4), 563-569. Hansen, H., & Tarp, F. (2000). Aid Effectiveness Disputed, Journal of International Development, 12 (3), 375-398. ______. (2001). Aid and Growth Regressions, Journal of Development Economics, 64(2), 547-70. Harrison, A. (2006). Globalization and Poverty (NBER Working Paper No. 12347). Cambridge, MA: NBER. Higuchi, Y., Ota, K., & Shimpo, K. (2006). Panel Data niyoru Keizaibunseki [Economic Analyses with Panel 24 Data], Tokyo: Nippon Hyoron Sha. Hirano. Y. (2014). Aid, Institutions, and Growth: Building Institutions Matters for Development Effectiveness, Journal of International Development Studies, 23(2). (forthcoming in Nov 2014) Hirano, Y., & Otsubo, S. (2012). “Poverty-Growth-Inequality Triangle under Globalization: Time Dimensions and the Control Factors of the Impacts of Integration,” GSID Discussion Paper No.191. Hudson, J. (2004). Introduction: Aid and Development, The Economic Journal, 114(496), 185-190. Krray, A. (2005, March). Aid, Growth, and Poverty. Paper presented at the IMF seminar on Foreign Aid and Macroeconomic Management, Maputo, Mozambique. Milanovic, B. (2005). Can We Discern the Effect of Globalization on Income Distribution? Evidence from Household Survey. World Bank Economic Review, 19(1), 21-44. Minoiu, C., & Reddy, S. G. (2010). Development Aid and Economic Growth: A Positive Long-run Relation, The Quarterly Review of Economic and Finance, 50, 27-39. Mosley, P. (1980). Aid, Savings and Growth Revisited, Oxford Bulletin of Economics and Statistics, 42 (2), 79-95. Mosley, P., Hudson, J., & Horrell, S. (1987). Aid, the Public Sector and the Market in Less Developed Countries, Economic Journal, 97, 616-41. Mosley, P., Hudson, J. & Verschoor, A. (2004). Aid, Poverty Reduction, and the "New Conditionality.” Economic Journal, 114(496), 217-243. Organization for Economic Co-operation and Development, Development Assistance Committee (OECD-DAC). (2007). Reporting Directives for the Creditor Reporting System. Paris: OECD. ______. (2013a). OECD. Stat Extracts [Data file]. Paris: OECD. Retrieved January 14, 2013, from www.oecd.org/dac/stats ______. (2013b). CRS Purpose Codes – Applicable as of Reporting on 2010 Flows. Paris: OECD. Retrieved May 24, 2013, from http://www.oecd.org/dac/stats/purposecodessectorclassification.htm ______. (2013c). User’s Guide to the Creditor Reporting System (CRS) Aid Activities Database. Paris: OECD. Retrieved May 24, 2013, from www.oecd.org/dac/stats/crsguide Otsubo, S. (Ed.). (2009). Globalization to Kaihatsu [Leading Issues in Development with Globalization]. Tokyo: Keisoshobo. Otsubo, S., & Hirano, Y. (forthcoming). Poverty-Growth-Inequality Triangle under Globalization: What do we really know about the pro-poor/anti-poor impacts of economic integration? In Otsubo. S. (Ed.). Globalization and Development: Leading Issues in Development with Globalization, Vol. 1. Singapore: Routledge. PRS Group. (2012). International Country Risk Guide (ICRG) Methodology. Retrieved in May, 2012 from http://www.prsgroup.com/PDFS/icrgmethodology.pdf Rajan, R. G., & Subramanian, A. (2008). Aid and Growth: What Does The Cross-Country Evidence Really Show? The Review of Economics and Statistics, 90(4), 643–665. Ravallion, M. (2005). Inequality is Bad for the Poor (Policy Research Working Paper No. 3677). Washington 25 DC: The World Bank. Rodriguez, F., & Rodrik, D. (2001). Trade Policy and Economic Growth: A Skeptic’s Guide to the Cross-National Evidence, NBER Macroeconomics Annual, 15, pp.261-325. Sawada, Y., Kohama, H., & Kono, H. (2003, October). Aid, Policies, and Growth: A Further Comment. Paper presented at the experts research workshop on “Quantifying the Impact of Rich Countries’ Policies on Poor Countries” at the Center for Global Development (CGD), Washington DC. UN. (2013). The Millennium Development Goals Report 2013. New York: United Nations. Williamson, O. E. (2000). The New Institutional Economics: Taking Stock, Looking Ahead. Journal of Economic Literature, 38(3), 595-613. World Bank. (1998). Assessing Aid: What Works, What Doesn’t and Why. Washington DC: The World Bank. ______. (2012). World Development Indicators (WDI): 1990-2010 [Data file]. Washington DC: The World Bank. Retrieved August 7, 2012, from http://data.worldbank.org/ ______. (2013). The World Bank Group Goals: End Extreme Poverty and Promote Shared Prosperity. Retrieved August 12, 2013, from The World Bank Web site: http://www.worldbank.org/content/dam/Worldbank/document/WB-goals2013.pdf 26 27 28 29 30 31 32 33 34 35 Appendix 1 Summary Statistics for Selected Variables Used in the Study Variable Observations Mean Std. Dev. Minimum Maximum Period Average Growth Rate (AVG) of Per Capita Income 242 0.03 0.03 -0.10 0.15 Period AVG of Per Capita Income of the Poorest Quintile 242 0.03 0.05 -0.16 0.30 Period Average Rate of Change of Gini Index 240 0.00 0.02 -0.09 0.06 Period Average Rate of Change of Poverty Head Count Ratio 224 -0.06 0.16 -0.69 0.65 Ln(Per Capita Income) 242 7.05 1.09 4.74 9.12 Ln(Per Capita Income of the Poorest Quintile) 242 4.13 1.06 1.63 6.73 Initial Ln(Per Capita Income) 242 7.21 1.10 4.71 9.35 Initial Ln(Per Capita Income of the Poorest Quintile) 242 4.31 1.05 1.79 6.85 Initial Ln(Gini) 241 3.72 0.24 2.97 4.15 Initial Ln(Poverty Headcount Ratio) 229 2.15 1.94 -3.91 4.46 Initial Ln(Fertility) 242 1.13 0.49 0.09 1.96 Period Average Rate of Change of TOT 189 0.00 0.04 -0.16 0.16 Initial Government Consumption/GDP 236 13.78 5.06 3.80 33.81 Government Consumption/GDP 234 14.08 5.19 3.80 37.36 Period Average Government Consumption/GDP 242 13.70 5.22 0.00 36.84 Ln(1 + Period Average Inflation/CPI) 223 0.13 0.16 0.00 1.29 Ln (ICRG Capacity) 216 3.15 0.18 2.34 3.50 Initial Trade/GDP 221 4.10 0.65 2.28 5.82 Trade/GDP 220 82.94 50.98 11.25 380.18 Period AVG of Trade/GDP 215 0.02 0.05 -0.25 0.18 FDI/GDP 241 3.75 4.65 -1.80 52.05 Period Average FDI /GDP 242 3.24 2.92 -0.79 17.62 Period Average Remittance/GDP 227 3.87 6.10 0.00 50.21 Regional Dummy EAP 242 0.12 0.32 0 1 Regional Dummy SSA 242 0.20 0.40 0 1 FSU Dummy 242 0.14 0.34 0 1 Crises Dummy 242 0.50 0.50 0 1 Time Dummy 1990 242 0.31 0.46 0 1 Time Dummy 2000 242 0.66 0.48 0 1 (the Late 1990s-2000s) Period Average Growth Rate of Per Capita Income 159 0.03 0.03 -0.02 0.15 Period Average Growth Rate of Per Capita Income of the Quintile 1 159 0.04 0.05 -0.08 0.30 Period Average Growth Rate of Per Capita Income of the Quintile 2 159 0.04 0.04 -0.07 0.21 Period Average Growth Rate of Per Capita Income of the Quintile 3 159 0.03 0.03 -0.07 0.16 Period Average Growth Rate of Per Capita Income of the Quintile 4 159 0.03 0.03 -0.06 0.16 Period Average Growth Rate of Per Capita Income of the Quintile 5 159 0.03 0.03 -0.04 0.15 Period Average Government Consumption/GDP 159 13.77 5.28 0.00 36.84 Ln(1 + Period Average Inflation/CPI) 150 0.08 0.08 0.00 0.51 Period Average Rate of Change of Institutional Quality (ICRG Capacity) 127 0.00 0.03 -0.07 0.09 Period Average Growth Rate of Trade/GDP 139 0.02 0.04 -0.09 0.18 Period Average FDI/GDP 159 3.89 3.11 -0.79 17.62 Period Average Remittance/GDP 154 4.63 6.91 0.01 50.21 Period average Helth Expenditure/G 159 10.59 4.18 2.49 34.30 Period Average Rate of Change of TOT 127 0.00 0.04 -0.09 0.16 Average Rate of Change of Agri-Food exports Share 128 -0.02 0.07 -0.43 0.16 Period Average Economic Aid/GDP 140 1.05 1.31 0.00 7.21 Period Average Social Aid/GDP 140 2.43 4.57 0.02 47.16 Period Average Aggregate Aid/GDP 140 6.04 8.67 -0.12 63.95 Crises Dummy 159 0.53 0.50 0 1 Source: Authors' compilation 36 37 38 39 40