AUTHOR ACCEPTED MANUSCRIPT

AUTHOR ACCEPTED MANUSCRIPT FINAL PUBLICATION INFORMATION Heterogeneity in the Allocation of External Public Financing : Evidence from Sub-Saharan African Post-MDRI Countries The definitive version of the text was subsequently published in Applied Economics Letters, 20(7), 2012-10-22 Published by Taylor and Francis THE FINAL PUBLISHED VERSION OF THIS MANUSCRIPT IS AVAILABLE ON THE PUBLISHER S PLATFORM This Author Accepted Manuscript is copyrighted by World Bank and published by Taylor and Francis. It is posted here by agreement between them. Changes resulting from the publishing process such as editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this version of the text. You may download, copy, and distribute this Author Accepted Manuscript for noncommercial purposes. Your license is limited by the following restrictions: (1) You may use this Author Accepted Manuscript for noncommercial purposes only under a CC BY-NC-ND 3.0 IGO license http://creativecommons.org/licenses/by-nc-nd/3.0/igo/. (2) The integrity of the work and identification of the author, copyright owner, and publisher must be preserved in any copy. (3) You must attribute this Author Accepted Manuscript in the following format: This is an Author Accepted Manuscript by Kinda, Tidiane; Le Manchec, Marie-Helene Heterogeneity in the Allocation of External Public Financing : Evidence from Sub-Saharan African Post-MDRI Countries World Bank, published in the Applied Economics Letters20(7) 2012-10-22 CC BY-NC-ND 3.0 IGO http://creativecommons.org/ licenses/by-nc-nd/3.0/igo/ 2018 World Bank

Heterogeneity in the Allocation of External Public Financing: Evidence from Sub-Saharan African Post-MDRI Countries Tidiane Kinda International Monetary Fund Marie-Helene Le Manchec World Bank Abstract Using a homogenous subset of 18 Sub-Saharan African countries, all recipients of sizeable debt relief, this paper consistently accounts for debt relief across countries and analyzes heterogeneity in the allocation of external public financing through quantile regression methods. The results show that donors were more selective in terms of institutional quality when allocating aid to countries with the highest income per capita in the sample.

2 I. INTRODUCTION The major challenges facing most of the low-income countries (LICs) is to accelerate and sustain economic growth and make progress in reducing poverty. Many LIC governments rely on external resources to finance their economic development because of weak domestic resource mobilization. Excessive volatility in external financing can therefore disrupt the implementation of public investment programs and undermine economic prospects. Options for external financing have also expanded over the years, complicating the monitoring process. In particular, debt relief represents de facto additional financing and is difficult to consistently record across countries because each vehicle to deliver the relief has its own statistical treatment. The Heavily Indebted Poor Country (HIPC) Initiative gave rise to the most complex set of transactions (debt forgiveness, flow rescheduling, or the provision of grants to meet part of future debt service) that have affected public accounts differently. Indeed, the relief could be recorded either in the current account or the capital account. It could also be explicitly shown (as for grants or exceptional financing) or implicitly built-in through a reduction of projected debt service. The approach to aid policy has shifted over time from a strategy of aid-financed investment in the 1970s towards a strategy of aid-induced economic reforms in the 1980s. Since the publication of the World Bank s strong critique of aid in Assessing Aid in the late 1990s, the policy shifted to the concept of selectivity. The effectiveness of aid allocation is now deemed higher when aid is directed to countries with better policy (selectivity) because this strengthens its impact on growth.

3 Empirical evidence is mixed. A number of studies on aid selectivity highlight recent improvements in aid allocation with more attention to developmental concerns than to strategic and political interests. 1 Dollar and Levin (2006) find that multilateral donors, and to a lesser extent bilateral donors, improved their selectivity between 1984-89 and 2000-03. According to the authors, donors allocate more aid to poorer countries and countries with better institutions. Claessens et al. (2007), using a sample of 147 recipient countries over 1970-2004, highlight changes in aid allocation in the beginning of the 1990s that have intensified over time. The authors find that countries economic needs and the quality of their policies and institutions are important determinants of aid allocation, while factors such as debt level, country size, and colonial linkages are less important. Another part of the recent literature, however, argues that no significant improvement in aid allocation has occurred during the recent years. Nunnenkamp and Thiele (2006) illustrate that bilateral and multilateral aid responsiveness to changes in recipient countries institutions and policies is weak. Easterly (2007), with a broad analysis of aid agency practices, finds weak evidence of increased aid selectivity with respect to the quality of recipient countries policies and institutions. After consistently accounting for debt relief across countries, this paper reassesses the determinants of external flows to the public sector using quantile regression methods, which help control for potential heterogeneity in public flows allocation. The analysis covers 18 1 See Claessens et al. (2007) for a comprehensive review of aid allocation literature.

4 homogenous countries, 2 recipients of full debt relief under the Multilateral Debt Relief Initiative (MDRI), for the period 2001-07. The results highlight that donors applied some selectivity in allocating aid. In particular, donors paid attention to institutional quality when allocating aid to countries with the highest income per capita in the sample. II. ESTIMATIONS AND RESULTS Our definition and coverage of external flows to the public sector is comprehensive and presented on a net basis, encompassing both the assets and liabilities for capital flows and the credit and debit for current transfers. The aggregated indicator includes 15 variables (Figure 1). 2 The 18 countries in the analysis are Benin, Burkina Faso, Cameroon, Ethiopia, The Gambia, Ghana, Madagascar, Malawi, Mali, Mozambique, Niger, Rwanda, São Tomé y Príncipe, Senegal, Sierra Leone, Tanzania, Uganda, and Zambia. These countries share common economic and institutional features. Their economies are not diversified and their access to international capital markets is very limited.

5 Figure 1. Schematic Diagram of Aggregated Net External Public Financing Net External Public Financing Net capital flows Current public transfers Net Transfers on debt Project grants Budget grants Other public transfers Gross lending Debt service payments Debt relief explicitly shown in the BOP (grants + exceptional financing) Private sources Public sources Amortization repayments Interest payments Commercial banks Budget loans Project loans Bonds and notes IMF purchases Net external resources directed to the public sector of the selected sample of post-mdri SSA countries increased by 2.4 percentage points of GDP between 2001 and 2007 (Table 1). This illustrates that before the financial and economic crisis, creditors and donors increased external resources to the 18 post-mdri countries. Funding for debt relief appears to be in addition to resources already committed for development cooperation.

6 Table 1. Estimates of Net External Public Financing of Selected Post-MDRI African Countries (In percent of GDP) 2001 2002 2003 2004 2005 2006 2007 Estimates Total net flows 6.1 6.8 6.7 7.0 7.2 7.1 8.5 Net transfers on debt 0.8 2.2 1.8 1.7 1.6 2.3 2.8 Total gross lending 4.2 4.3 3.7 3.7 3.2 3.2 3.3 Debt service repayments (reflecting all debt relief) 3.4 2.2 1.9 1.9 1.6 0.9 0.5 Grants (excl. HIPC Initiative-related grants) 4.7 3.7 4.2 4.6 4.8 4.2 5.3 Other net public transfers 0.7 1.0 0.8 0.7 0.8 0.6 0.4 Source: Authors' estimates. Instead of using aid, which is a donor-based concept, our dependent variable is the per capita real public external public financing. The explanatory variables, in line with Dollar and Levin (2006), capture the financing needs of recipient countries, their population size, and the quality of their governance. GDP per capita is expected to negatively affect public flows, with richer countries receiving fewer resources. The effect of population varies across studies with a larger number of them concluding that smaller countries receive more aid. A two-year lag of real per capita GDP ( GDPit 2 ) is introduced to reduce potential endogeneity of this variable due to the possible impact of public flows on GDP. Lastly, it is expected that countries with better governance would attract higher public flows if selectivity applies. Appendix 1 gives the list, definitions, and sources of the variables. The estimated model is the following: LTFP GDP Population Governance French MDRI SaoT it 1 it 2 2 it 3 it 1 i 2 t 3 i it LTFP it represents per capita real net external financing to the public sector of country i during year t. This model includes three dummies to control for important fixed factors. The

7 MDRI dummy controls for any significant change in net public flows after the implementation of debt relief under the MDRI. French is a colonial ties dummy, which takes one for ex-french colonies and zero for the other countries (almost all the others are ex- British colonies). This helps control for historical links that may affect public flows allocation. The last dummy controls for the specificity of São Tomé y Príncipe, which is the smallest country in the sample receiving the highest amount of net external public flows relative to its GDP. 3 All empirical studies on aid allocation are based on linear regressions or models that estimate the mean value of the amount of aid for given levels of explanatory variables (such as income per capita, population, and governance). These models estimate how countries characteristics affect the average amount of aid they received. A contribution of this paper is to analyze heterogeneity in the allocation of public flows through the quantile method. As opposed to OLS regressions, which model the relationship between predictor variables and the conditional mean of a dependent variable, quantile regressions model the relationship 3 This OLS model is an alternative to panel estimations using fixed or random effects models. The random effects model is based on the assumption that the explanatory variables are not correlated with individual specific effects. However, such a model may not be appropriate since some historic factors have been found to be important determinants of countries current institutions and governance quality. The fixed effects model controls for all invariant factors such as colonial ties. However, given the relatively short time span of the analysis (2001-2007), it would be difficult to identify the effect of structural factors with low variability such as governance quality. Changes in real GDP per capita would also be relatively limited over six years.

8 between predictor variables and the conditional quantile of the dependent variable. 4 Quantile methods could help understand the heterogeneity across countries in the allocation of external public flows. Consistent with part of the literature, the results based on OLS estimations indicate that the public sector of poorer countries and countries with better governance receive more external flows (Table 2). However, there is no evidence that smaller countries receive more external public financing. Analyzing heterogeneity in the allocation of public financing through quantile regressions highlights that donors tend to apply selectivity criteria. 5 Unlike the results of the OLS method, per capita income becomes highly significant for countries ranked in the first quintile of the quantile regression. For these higher income countries, donors have allocated more net resources to the poorest among them. In contrast, donors have paid less attention to income level when allocating financing to the poorest countries (the last quintile). Similar to the results of the OLS and random effects models, countries with stronger governance received higher levels of external funding. However, donors focused less on governance 4 Standard errors and confidence intervals of the coefficients obtained with the quantile regressions can be estimated by asymptotic methods or by bootstrapping. Results obtained with these two methods are robust (Koenker and Hallock 2001), with the bootstrap method being the more practical one. 5 The relatively small and homogeneous sample limits, to some extent, the scope of the quantile analysis than in a broader sample of countries.

9 when allocating resources among the poorest countries. The marginal effect of governance on external flows increases with countries income level (Figure 1), confirming the results. 6 Table 2. Regressions with OLS, Random Effects and Quantile Regression Lag Per Capita Real GDP Dependent variable: Real Net Public Flows Per Capita OLS Random Effects Quantile Regression 25% 50% 75% [1] [2] [3] [4] [5] -0.012-0.01-0.021-0.02-0.001 (1.78)* -0.91 (3.86)*** (1.83)* -0.05 Log (population) 0.002 0.002 0.012-0.004-0.032-0.19-0.13-0.78-0.25-1.62 Governance index 0.197 0.19 0.172 0.167 0.112 (3.72)*** (3.50)*** (5.40)*** (3.44)*** (1.70)* Ex-French colony -0.074-0.075-0.052-0.078-0.071 (2.41)** (2.06)** (2.44)** (2.96)*** (2.22)** Sao Tome and P. 0.516 0.513 0.491 0.481 0.636 (3.35)*** (4.49)*** (1.97)* (2.35)** (2.79)*** MDRI 0.096 0.095 0.099 0.088 0.083 (3.12)*** (3.13)*** (4.47)*** (3.01)*** (2.07)** Constant 0.336 0.329 0.247 0.345 0.434 (7.30)*** (4.84)*** (5.09)*** (5.29)*** (3.89)*** Observations 126 126 126 126 126 Number of countries 18 18 18 18 18 R-squared 0.5 0.5 0.25 0.26 0.38 * Significant at 10%; ** significant at 5%; *** significant at 1% Absolute value of t statistics in parentheses For quantile regression, bootstrapped (with 500 replications) t statistics in parentheses 6 The marginal effects are obtained after controlling for other determinants of public external flows, as in the previous regressions.

-.2 0 Marginal Effect of Governance.2.4.6 10 Figure 2. Marginal Effect of Governance as Real GDP Per Capita Changes (Dependent Variable: Real Net Public Flows) Marginal Effect of Governance 95% Confidence Interval 1 2 3 4 5 6 7 8 9 Real GDP per capita in Hundred of US Dollar III. CONCLUSIONS This paper has analyzed external financing to the public sector of 18 Sub-Saharan African countries that have benefited from debt relief under the Multilateral Debt Relief Initiative. After consistent accounting for debt relief across countries, the paper reassessed the determinants of external flows to the public sector. The results show that institutional quality is an important determinant of external flows to the public sector. Analyzing heterogeneity in aid allocation through quantile regressions shows that donors were more selective when allocating aid to countries with the highest per capita income in the sample. For the poorer countries in the sample, the quality of their institutions had a lower degree of influence on external financing to their public sector.

11 References Claessens S., Cassimon D., and Campenhout B. V., 2007, Empirical Evidence on the New International Aid Architecture, IMF Working Paper No. 07/277. Dollar D. and Levin V., 2006, The Increasing Selectivity of Foreign Aid, 1984-2003. World Development, Vol. 34(12), pp. 2034-2046. Easterly W., 2007, Are aid agencies improving? Economic Policy, Vol. 22(52), pp. 633-678. Koenker, R. and Hallock, K., 2001, Quantile Regression: An Introduction, Journal of Economic Perspectives, Vol. 15(4), pp. 143-156. Nunnenkamp P., and Thiele R., 2006, Targeting Aid to the Needy and Deserving: Nothing But Promises? The World Economy, Vol. 29(9), pp. 1177-1201. World Bank, 1998, Assessing Aid: What Works, What Doesn't and Why? Oxford University Press.

12 Appendix 1. Definition and Sources of the Parameters Variables Definitions Sources GDP Two years lag of the Real GDP per capita World Economic Outlook Population Log of Population database Governance French Arithmetic mean of the following index: voice and accountability, political stability, governance effectiveness, regulatory quality, rule of law, and degree of corruption French ex-colonies dummy World Bank s Kaufmann Index MDRI SaoT Dummy taking 1 during the two years of MDRI implementation (2000-2007) and 0 otherwise. Sao Tome and Principe dummy