Does International Health Aid Follow Recipients’ Needs? Extensive and Intensive Margins of Health Aid Allocation

World Development Vol. 64, pp. 104–120, 2014 0305-750X/Ó 2014 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2014.05.020

Does International Health Aid Follow Recipients’ Needs? Extensive and Intensive Margins of Health Aid Allocation SUEJIN A. LEE a and JAE-YOUNG LIM b,* a Cornell University, Ithaca, USA b Korea University, Seoul, Republic of Korea Abstract. — Using OECD-DAC data covering 112 recipient countries for 1995–2011, this paper examines the responsiveness of health aid to the recipients’ needs in terms of infant mortality, child mortality, and HIV prevalence. This paper fills a gap in the literature by investigating extensive and intensive margins of health aid allocation patterns at project- and donor-levels. We find that when the health status of a recipient country deteriorates, the total value of health aid to the country increases due in large part to an increase in the number of health aid projects and to an increase in the average aid value from each donor country. Ó 2014 Elsevier Ltd. All rights reserved. Key words — health aid, aid allocation, recipients’ needs, extensive and intensive margins

1. INTRODUCTION

aid (Neumayer, 2005), education aid (Christensen, Homer, & Nielson, 2011) and health aid (Esser & Bench, 2011; Fielding, 2011). Over the last 30 years, the health sector has increasingly become an important recipient of external assistance, with the emergence of new health threats—such as HIV/AIDS and pandemic influenza—and with recognition that health is a key determinant of economic growth and poverty reduction (Dodd, Schieber, Cassels, Fleisher, & Gottret, 2007). In this regard, a notable recent trend in the ODA programs is that the donors are shifting their orientation from massive industrialization programs for economic growth toward more poverty-reducing and health-improving objectives. This tendency became more apparent when the United Nations adopted the Millennium Development Goals (MDGs) in 2000. Three of the eight MDGs relate directly to health—to reduce child mortality rates (Goal 4); to improve maternal health (Goal 5); and to combat HIV/AIDS, malaria, and other diseases (Goal 6). Also, other MDGs are indirectly related to health outcomes as health is the basis for or ultimate consequences of poverty reduction, education, gender equality, and environmental sustainability. Specifically, the amount of health aid increased dramatically by 368% during 1995–2011, while the total amount of foreign aid increased by 167% in the same period. Figure 1 illustrates the trend of health aid over the years. Although health aid accounted for only 6.8% of total aid in 1995, the proportion increased to 12% in 2011. While health aid is clearly rising, it is, however, less clear whether health aid is following the needs of the recipient countries. To the best of our knowledge, there have been two studies regarding health aid allocation. First, Esser and Bench (2011) were concerned with the responsiveness of health aid to the needs of the recipient countries, using the sample of

The international community has provided particularly large sums in grants and loans through Official Development Assistance (ODA) programs for developing countries since the 1950s. Accordingly, the assessment of foreign aid has been an important policy issue not only for international organizations and policy makers but also for academic researchers. To assess foreign aid, the aid literature has focused on the question of whether and to what extent the aid is effective in promoting economic growth of the recipient countries. To date, the literature on aid effectiveness is inconclusive. Advocates suggest that massive foreign aid can help poor countries to overcome the “poverty trap” (Clemens, Radelet, Bhavnani, & Bazzi, 2012; Levine, 2004; Sachs, 2005). Skeptics, on the other hand, argue that aid can adversely affect the country’s competitiveness and benefit the political elite (Kanbur, Sandler, & Morrison, 1999; Rajan & Subramanian, 2005; Younger, 1992). In fact, effectiveness of foreign aid is closely linked to the question of how and where the aid is allocated. It seems obvious that the foreign aid is less effective when it is not allocated according to the needs of the recipient countries. There is a general consensus in the literature that foreign aid is allocated mostly according to political, economic, and strategic considerations (Dudley & Montmarquette, 1976; Gounder, 1994; Maizels & Nissanke, 1984; McKinley & Little, 1977). Recent studies, however, have also found that foreign aid has become more responsive to poverty and the institutional conditions of recipient countries than in the past (Claessens, Cassimon, & Campenhout, 2009). The assessment of total foreign aid with respect to its allocation and effectiveness is not straightforward because aggregated aid is complex and multifaceted. We argue, therefore, that it is necessary to take the heterogeneity of aid allocation into consideration. Compared to the vast empirical literature considering the distribution patterns of aggregated foreign aid, there is relatively little systematic empirical evidence on how aid in specific sectors responds to the sector-specific needs of the recipient countries. Only recently, have there been studies disaggregating aid by its sectors to assess how donors prioritize their aid money among different sectors (Nelson, 2010; Thiele, Nunnenkamp, & Dreher, 2007) and studies focusing on the distribution patterns of a specific sector such as food

* We are grateful to Chirok Han, Song-soo Lim, Byeongseon Seo, and Christopher Wolfe for their insightful comments. We also thank the editor and two anonymous referees for their invaluable comments and suggestions. All remaining errors are ours. This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-013S1A3A2043324). Final revision accepted: May 13, 2014. 104

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS?

105 14

25000

12 20000 10 15000

8 6

10000

4 5000 2 0

0 19951996199719981999200020012002200320042005200620072008200920102011 Health Aid (constant 2011 million USD)

Health Aid/Total Sector Aid (%)

Figure 1. Trend of health aid (Unit: constant 2011 million USD, %). Source: Authors’ calculation using OECD DAC CRS online database.

27 developing countries during the period 2005–07. They measured the needs of the recipients by the health priority indicators produced by the Kaiser/Pew Global Health Survey (Kaiser Family Foundation, 2007) and Disability-Adjusted Life-Years (DALY). Their analyses show that patterns of DALY explain neither public nor private health aid flows, while health priority indicators are weakly correlated. Also, Fielding (2011) investigated the selectivity issue in health aid by assessing whether the quality of governance in recipient countries affects the allocation of health aid for 109 recipients over the period 1995–2006. He found that countries with greater political rights received significantly more aid, but so did the countries with higher corruption levels. He also found evidence that donors responded to changes in neonatal mortality rates on average. This paper aims to conduct an in-depth empirical analysis on how international health aid responds to the needs of recipient countries in terms of infant mortality, child mortality, and HIV prevalence rate. This paper fills a gap in the literature by differentiating the extensive and intensive margins of health aid distribution patterns at project- and donor-levels, as well as the allocation of the total value of health aid. Specifically, the term “the extensive margin of health aid” refers to the number of projects or the number of donors, and “the intensive margin of health aid” refers to the average value of health aid per project or per donor. 1 An assessment of the extensive margin of the number of projects and donors and the intensive margin of the average value of health aid per project and donor is important because as the needs of the recipient country increase, there might be different health aid distribution patterns at project- and donor-levels. The total value of health aid may increase in a recipient country by increasing the average value of the existing projects in the country or/and by initiating new health aid projects. Also, the total value of health aid may increase in a recipient country not only because the currently active donors increase the value of their health aid in a recipient country but also because new donors join the international community’s efforts to help the country. While this is an issue in all sectors, understanding aid distribution patterns at project- and donor-levels is particularly important in the health sector as the aid architecture in health has become ever more complex with the emergence of large and many global health partnerships. It has been noted in

Dodd et al. (2007) that these foreign actors are playing a substantial and often predominant role in both the financing and delivery of health care services in low- and middle-income countries. There are both merits and demerits of the consequences in providing health aid more extensively or more intensively at project- and donor-levels. Firstly, many authors have pointed out that aid provided by a large number of donors and projects in a recipient country may undermine the efficiency by overwhelming the capacity of recipient governments to manage aid inflows, and further complicating donor harmonization efforts at the global level (e.g., Acharya, de Lima, & Moore, 2006; Kimura, Mori, & Sawada, 2012; Roodman, 2006a, 2006b). On the other hand, there are also possible positive effects of extensively provided health aid. With a large number of projects, foreign aid may reach better the regions and groups of the recipient countries which were largely neglected in the provision of medical services. Also, where there are many aid donors, there may be clear and visible competition among donors for more attractive and better projects (Acharya et al., 2006). Secondly, health aid provided with a large scale of average value per project or donor can also bring about two different consequences. Piva and Dodd (2009) argue that large-scale projects and large sums of donor money have a more significant impact on health and the health delivery system in recipient countries in that they are more likely to attract political attention, receive significant technical input and have economies of scale. They also suggest that, on the other hand, smaller projects are important for piloting new approaches, testing innovations, delivering benefits to individual communities, and covering emerging or unplanned health system needs (p. 933). The key finding of this paper is that when the health status of a recipient country deteriorates, the total value of health aid to the country increases due in large part to an increase in the number of health aid projects and to an increase in the average aid value from each donor country. The remainder of the paper is organized as follows. Section 2 offers an overview of the allocation patterns for the extensive and intensive margins of health aid at project- and donor-levels during the period 1995–2011. In Section 3, we present the empirical methodology and data for investigating the determinants of the different margins of health aid, and in Section 4, we report the estimation results. Section 5 concludes with a

106

WORLD DEVELOPMENT

summary of the empirical findings and discussion on policy implications. 2. EXTENSIVE AND INTENSIVE MARGINS OF HEALTH AID Following the classification of the OECD Development Assistance Committee (DAC) Creditor Reporting System (CRS) database, this paper defines health aid comprised of General health (CRS121), Basic health (CRS122), and Population policies/programs and reproductive health (CRS130) sector. 2 Figure 2 illustrates the trend of health aid by its three subsectors. Overall, health aid appears to have increased persistently during the period. Specifically, CRS122 and CRS130 sector show an increasing trend, but CRS121 sector remains at more or less the same level. Notably, CRS130 sector appears to be the one that leads the increasing trend of total health aid. CRS130 sector accounts for a large proportion of total health aid because HIV/AIDS control is attracting much attention from the international community. In our dataset, STD control including HIV/AIDS (CRS13040), a sub-sector of CRS130, increased its proportion in total health aid from 9.46% in 1995 to about 40% in 2011. Country i’s aggregate health aid from the world in period t (Tit) can be decomposed into the extensive margin in terms of the number of projects (PNit) and the intensive margin in terms of the average value of each project (PAit = Tit/PNit): T it ¼ PN it  PAit : With respect to the data used in this paper, a “project” is referred to as an “aid activity” which can take many forms: “It could be a project or a program, a cash transfer or delivery of goods, a training course or a research project, a debt relief operation, or a contribution to a non-governmental organization (DCD/DAC, 2014).” 3 Figure 3 illustrates the extensive margin of the number of projects and the intensive margin of the average value of health aid per project. Although the extensive margin of the number of projects initiated in a country increased significantly from 8 to 89 projects during the period 1995–2011, the intensive margin of the average value of health aid per project shows a decreasing trend. Average project size dropped from USD 2.5 million in 1996 to about USD 0.5 million in

2011. This implies that small-size projects have proliferated in many recipient countries. Figures 4 and 5 present the extensive margin of the number of projects and the intensive margin of the average value of health aid per project, respectively, by five different income groups: Least Developed Countries (LDCs), Other Low-Income Countries (Other LICs), Lower Middle-Income Countries (LMICs), Upper Middle-Income Countries (UMICs), and High-Income Countries (HICs). 4 Except for the HICs, the other four income groups show a similar trend: the extensive margin of the number of projects is increasing but the intensive margin of the average value of health aid per project is decreasing. It is worth noting that LDCs and Other LICs are the groups with the most rapidly increasing trends of the project extensive margin, followed by LMICs and UMICs. Country i’s aggregate health aid from the world in period t (Tit) is decomposed into the extensive margin in terms of the number of donors (DNit) and the intensive margin in terms of the average aid value from each donor (DAit = Tit/DNit): T it ¼ DN it  DAit : Figure 6 illustrates the extensive margin of the number of donors and the intensive margin of the average value of health aid per donor. The average number of donors giving health aid to a single recipient country increased from an average of three donors per recipient country in 1995 to 10.5 donors in 2011. 5 The increase in the extensive margin of the number of donors might partially explain the reason why the extensive margin of the number of projects increased as shown in Figures 3 and 4. However, the average amount of aid per donor does not seem to have changed much, shifting around USD 5 million for the whole period. Figures 7 and 8 present the extensive margin of the number of donors and the intensive margin of the amount of health aid by income group. The number of donors shown in Figure 7 increased in all income groups except the HICs. In Figure 8, the average amount of aid per donor increased in the LDCs and Other LICs since 2000, but the other income groups do not show such an increasing trend. Until 2007, the LMICs received the largest amount of health aid per donor but the LDCs and Other LICs have recently overtaken them. Overall, per recipient country on average, small-size projects have proliferated and the number of donors increased in the health sector during 1995–2011. This tendency evidences that

25000

20000

15000

10000

5000

0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 General health (CRS121) Basic health (CRS122) Population and reproductive health (CRS130)

Figure 2. Trend of health aid by sub-sector (Unit: constant 2011 million USD). Source: Authors’ calculation using OECD DAC CRS Online Database.

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS? 100

107 3

90 2.5

80 70

2

60 50

1.5

40 1

30 20

0.5

10 0

0 19951996199719981999200020012002200320042005200620072008200920102011

Number of Projects (Project-Extensive) - Left Average Amount of Aid per Project (Project-Intensive) - Right

Figure 3. Project-extensive and intensive margins of health aid per recipient country (Unit: number of projects, constant 2011 million USD). Source: Authors’ calculation using OECD DAC CRS Online Database.

180 160 140 120 100 80 60 40 20 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 LDCs

Other LICs

LMICs

UMICs

HICs

Figure 4. Project-extensive margin of health aid by income group (Unit: number of projects). Source: Authors’ calculation using OECD DAC CRS Online Database. Notes: LDCs, Other LICs, LMICs, and UMICs indicate Least Developed Countries, Other Low-Income Countries, Lower Middle-Income Countries, and Upper Middle-Income Countries respectively. Recipients are divided into income groups according to the criteria of OECD DAC CRS database.

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 LDCs

Other LICs

LMICs

UMICs

HICs

Figure 5. Project-intensive margin of health aid by income group (Unit: constant 2011 million USD). Source: Authors’ calculation using OECD DAC CRS Online Database.

108

WORLD DEVELOPMENT 12 10 8 6 4 2 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Number of Donors (Donor-Extensive) Average Amountof Aid Per Donor (Donor-Intensive)

Figure 6. Donor-extensive and intensive margins of health aid per recipient country (Unit: number of donors, constant 2011 million USD). Source: Authors’ calculation using OECD DAC CRS Online Database.

20 18 16 14 12 10 8 6 4 2 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 LDCs

Other LICs

LMICs

UMICs

HICs

Figure 7. Donor-extensive margin of health aid by income group (Unit: number of donors). Source: Authors’ calculation using OECD DAC CRS Online Database.

14 12 10 8 6 4 2 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 LDCs

Other LICs

LMICs

UMICs

HICs

Figure 8. Donor-intensive margin of health aid by income group (Unit: constant 2011 million USD). Source: Authors’ calculation using OECD DAC CRS Online Database.

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS?

health aid has been disaggregated into a number of small projects with limited and measurable targets for clearer results during the period. Similarly, in terms of total foreign aid, Kimura et al. (2012) find that the average number of bilateral aid donors and the average number of projects per recipient country show an increasing trend from the 1970s to the early 2000s, while the average project size exhibits a downward trend from the mid-1980s. 6 3. EMPIRICAL METHODOLOGY AND DATA (a) Model specification This paper aims to examine the responsiveness of health aid to the recipients’ needs in terms of infant mortality, child mortality, and HIV prevalence rate. As a benchmark model, we employ Ordinary Least Squares (OLS) with time fixed effects as specified in Eqn. (1). 7 Our regression includes three groups of explanatory variables in addition to the control variables. The benchmark model is Aid it ¼ a þ b1 Needsit þ b2 Interestsit þ b3 Selectivity it þ dX it þ ht þ eit ;

ð1Þ

where the dependent variable Aidit is health aid provided in recipient i in period t. For the dependent variable, five alternative variables are used: (1) total value of health aid, (2) extensive margin of the number of projects, (3) intensive margin of the average value of health aid per project, (4) extensive margin of the number of donors, and (5) intensive margin of the average value of health aid per donor. Needsit is our focus variable representing the needs of recipient country i in period t, such as infant mortality rate, child mortality rate, or HIV prevalence rate; Interestsit is a vector of variables capturing the donors’ commercial motivations in giving aid, such as US value of total imports of a recipient country and natural resource reserves; Selectivityit is a vector of variables capturing the characteristics of recipient country i in period t which may influence the effective use of health aid, such as democracy level, control of corruption, and internal/external armed conflict; Xit is a vector of other control variables; ht is the time dummy; and eit is the idiosyncratic error term. Our “aid” variable may exhibit persistence over time. That is, donors prefer to keep giving aid to the same countries and many aid projects tend to last for several years: the current level of aid to recipient i in year t, Aidit, may be determined by its past level, Aidit1. Therefore, in our preferred specification, we introduce one-lagged health aid in the explanatory variable to analyze the dynamics of adjustment in health aid allocation over the sample period. We also introduce country fixed effects in the model to control for unobserved country-specific and time-invariant factors determinants of health aid. Thus, the fixed effects regression with one-lagged health aid in the explanatory variable is specified as Aid it ¼ a þ qAid it1 þ b1 Needsit þ b2 Interestsit þ b3 Selectivity it þ dX it þ ui þ ht þ eit ;

ð2Þ

where ui is a vector of country fixed effects which denotes timeinvariant differences in health aid across countries. It is noted that in Eqn. (2), Aidit1 is correlated with eit and hence the OLS results may be biased. It is also noted that although countries with greater needs (or with bad health outcomes) may receive more health aid, health aid may also

109

improve health outcomes. That is, the “needs” variables may cause a reverse causality problem, being correlated with the unobserved component of health aid. Therefore, we have two endogenous variables to control in our model: the lagged dependent variable and the “needs” variable. 8 We use the the System Generalized Method of Moments (GMM) of Blundell and Bond (1998), Blundell and Bond (2000) to estimate Eqn. (2). The system GMM solves endogeniety bias by treating lagged differences of the endogenous variables as instruments in the level Eqn. (2) and lagged levels of the endogenous variables as instruments in the differenced Eqn. (3). To illustrate, for the needs variables, D(Needsit  1) will be a valid instrument for Needsit in the level equation, because as long as eit is serially uncorrelated, D(Needsit  1) will be orthogonal to eit. Likewise, Needsit  1 will be valid instrument for D(Needsit) in the differenced equation, because Needsit  1 will be orthogonal to D(eit). DAid it ¼ qðDAid it1 Þ þ b1 ðDNeedsit Þ þ b2 ðDInterestsit Þ þ b3 ðDSelectivity it Þ þ dðDX it Þ þ Dht þ Deit

ð3Þ

System GMM obtains estimated coefficients of Eqn.(2) by simultaneously solving the appropriately weighted set of the moment conditions defined by both level and differenced Eqns. (2) and (3). We apply a two-step robust system GMM with the Windmeijer (2005) correction and use only two lags of instruments to avoid the possibility of over-fitting the model. 9 The inclusion of an excessive number of instruments for a fixed number of groups may result in a finite-sample bias in the estimates. A large number of instruments also weaken the power of Hansen’s J test for over-identifying restrictions and may cause the two-step standard errors of the estimator to be severely downward biased (Roodman, 2009a). (b) Dependent variables Our health aid data are drawn from the OECD DAC CRS database (http://stats.oecd.org/), which collects on an ongoing basis aid flows to developing countries from DAC countries, some multilateral organizations, and other donors. The OECD DAC CRS database covers 181 recipient countries in total, but we exclude 29 countries which are classified as the HICs and 40 countries whose data are not available for some explanatory variables. Thus, our dataset for regression analysis includes 112 recipient countries with 17 years of observations from 1995 to 2011. 10 As explained in the previous section, for the dependent variable, five different measures of health aid are alternatively estimated: (1) total value of health aid, (2) number of projects, (3) average value of health aid per project, (4) number of donors, and (5) average value of health aid per donor. We use aid commitment data for computing different measures of health aid because actual disbursement data are less complete for some donors in the earlier years. 11 Also, health aid is expressed in constant US dollars, using the DAC deflator from the OECD website, so as to account for both inflation and exchange rate variations in terms of 2011 base year (DCD/DAC, 2014). Table 1 presents summary statistics of the five alternative dependent variables. On average, a recipient country receives 58.6 million dollars with 50 projects and four donors. The average amount of health aid per project is USD 1 million and the average amount of health aid per donor is USD 4.3 million. In the regression, health aid variables are all expressed as natural logarithms. 12

110

WORLD DEVELOPMENT Table 1. Definitions and summary statistics of the dependent variables

Variables Total Project extensive Project intensive Donor extensive Donor intensive Total Project extensive Project intensive Donor extensive Donor intensive

Total value of health aid (constant 2011 million USD) Number of projects Average value of health aid per project (constant 2011 million USD) Number of donors Average value of health aid per donor (constant 2011 million USD) Total value of health aid (log) Number of projects (log) Average value of health aid per project (log) Number of donors (log) Average value of health aid per donor (log)

(c) Explanatory variables Following the specification of previous studies, our allocation model includes four categories of explanatory variables: recipients’ needs, donors’ interests, selectivity, and other control variables. (i) Recipients’ needs Needs of the recipient countries are captured in this model by infant mortality rate, child mortality rate, and HIV prevalence rate, which are all taken from the United Nations Millennium Development Goals Indicators Database (http:// mdgs.un.org/). We selected infant mortality and child mortality rates as the “needs” variables capturing the primary health indicators of recipient countries for the following reasons. First, infant and child mortalities are noted as the measurable health indicators for MDGs Goal 4 which calls for reduction in under-5 mortality by two-thirds. Second, infant and child mortality are known as the important indicators of health for whole populations, reflecting the intuition that structural factors affecting the health of entire populations have a significant impact on the mortality rate of infants and child (Reidpath & Allotey, 2003). Third, infant mortality and child mortality are more sensitive than life expectancy to changes in economic conditions (Boone, 1996). Finally, data on infant mortality and child mortality are available for a large set of countries and are more reliable than other indicators, such as maternal mortality, life expectancy, and DisabilityAdjusted Life-Years (DALY). In addition, we also include the prevalence of HIV as another “needs” variable because of its rising importance in health aid as shown in Figure 2. (ii) Donors’ interests We include two measures of donors’ interests: value of total imports and crude oil reserves in a recipient country. 13 The total imports variable is measured in constant 2000 million US dollars taken from United Nations Conference on Trade and Development (UNCTAD) statistics (http://unctadstat. unctad.org/) and is included as an indicator of how donors’ commercial interests influence aid allocation. The crude oil reserves variable (in billion barrels) is taken from US Energy Information Administration (http://www.eia.gov/) and is used to capture the strategic motivation of donors for energy resources. (iii) Selectivity It is expected that donors will allocate more aid to those countries which can use development aid more effectively. Burnside and Dollar (2000) suggest that development aid is more effective when given to countries with good policies. Therefore, we use an indicator of “control of corruption” which reflects “perceptions of the extent to which public power

Obs.

Mean

Std. Dev.

Min

Max

2584 2584 2584 2584 2584 2308 2308 2308 2308 2308

58.552 49.995 1.027 8.520 4.297 2.600 3.227 -0.627 1.933 0.667

133.042 64.855 1.820 6.762 7.328 2.208 1.475 1.343 0.912 1.568

0 0 0 0 0 9.790 0 -9.790 0 9.790

1890.193 533 26.195 30 117.077 7.544 6.279 3.266 3.401 4.763

is exercised for private gain, including both petty and grand forms of corruption, as well as ‘capture’ of the state by the elite and private interests (Kaufmann, Kraay, & Mastruzzi, 2010).” Data are taken from the World Bank Worldwide Governance Indicators (http://www.govindicators.org/). Since the control of corruption variable is only available for 1996, 1998, 2000, and 2002–11, we use interpolation with cubic polynomials for the missing years (1995, 1997, 1999, and 2001). 14 We also use two more “selectivity” measures. One is the level of democracy, using the Polity IV dataset taken from the Center for Systematic Peace (http://www.systemicpeace.org/), which provides a rating between minus 10 and plus 10 to indicate the level of political openness, with higher values indicating more democratic regimes. 15 The other is a dummy variable for armed conflict to account for recipient countries’ internal and external fragility. An armed conflict dummy is taken from the Uppsala Conflict Data Program (UCDP, http://www. pcr.uu.se/). Armed conflict is defined in UCDP as “a contested incompatibility which concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths (Wallensteen & Sollenberg, 2001).” (iv) Other control variables In addition to the three categories of variables explained above, we include other control variables such as GDP per capita, GDP per capita squared, and population. We include both GDP per capita and its squared term as many studies refer to middle-income bias of foreign aid (e.g., Gounder, 1995). According to Gounder (1995), middle-income bias is the tendency that per capita foreign aid rises as per capita income of a recipient country rises, and begins to fall when the recipient’s per capita income reaches a relatively high level. Thus, we expect the coefficient for GDP per capita to have a positive sign and its squared term to have a negative sign. Also, population is included as a measure of the size of the recipient country. Data for GDP per capita and for population are taken from the World Bank’s World Development Indicators (http://data.worldbank.org/). Summary statistics for the explanatory variables are reported in Table 2. 4. EMPIRICAL RESULTS AND DISCUSSION (a) Results for total value of health aid Table 3 reports the estimated results for the determinants of the total value of health aid allocation. The estimates of columns (1), (2), and (3) are made by OLS with country-clustered robust standard error, and the estimates of columns (4), (5), and (6) are made by two-step robust System GMM. As

WB world development indicators WB world development indicators 9.51 21.02 6.95 15.46

1.13 2.16 2232 2380 GDP per capita, PPP (constant 2005 international $) (Log) Total population (Log)

4.00 9.13

Uppsala conflict data program (UCDP) 1

Other control variables Log of GDP per capita Log of population

Control of corruption

Armed conflict dummy

0.38 0.17 2518

0

WB worldwide governance indicators (interpolated) 1.57 0.64 0.50 2320

2.29

Center for systematic peace 10 10 6.11 1.87 1945

US EIA 211.17 0 3.63

21-Point scale ranging from 10 (hereditary monarchy) to +10(consolidated democracy) Ranging from 2.5 (weak) to 2.5 (strong) governance performance Dummy variable for 1 if armed conflict, 0 otherwise Selectivity Democracy level

Crude oil reserves

2355

16.13

UNCTAD 14.08 7.44 Donors’ interests Log of import

Imports of good and services in constant 2000 million USD (log) Crude oil proved reserves in billion barrels

2317

2.17

2.84

UN MDGs indicators 3.31 2.30 -0.29 Log of prevalence of HIV

1753

1.62

5.05 5.62 1.36 1.72 3.57 3.85 2397 2397

Infant mortality rate (0-1 year) per 1,000 live births (Log) Children under five mortality rate per 1,000 live births (Log) People living with HIV, 15-49 years old, percentage (Log)

0.78 0.89

Source Max Min Std. Dev. Mean Obs.

Recipients’ needs Log of infant mortality Log of child mortality

Variables

Table 2. Definitions and summary statistics of the explanatory variables

UN MDGs indicators UN MDGs indicators

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS?

111

mentioned in the earlier section, the lagged dependent variables and the “needs” variables are treated as endogenous in the GMM specifications. In Table 3, all specifications show that the total value of health aid is allocated responsively to the needs of the recipients. The OLS estimates imply that a 10% increase in infant mortality and in child mortality induces about 6.89 and 6.26% in health aid, respectively. Also, a 10% increase in HIV prevalence rate raises 2.50% in health aid. In the System GMM specification, a 10% increase in infant mortality, in child mortality, and in prevalence of HIV raises 4.69%, 4.45%, and 1.92% in health aid, respectively in the short run. Taking the lagged dependent variables into account, a 10% increase in infant and child mortality, and prevalence of HIV increase by 6.40, 6.14, and 2.21 in the long run, which are similar to the OLS estimates. 16 Among the “donors’ interests” variables, the volume of total imports (in log of million USD) does not reveal any relevance to the allocation of total health aid in all specifications. 17 We also find that the coefficients for crude oil reserves (in billion barrels) are negative and statistically significant. 18 Thus, we find no evidence that donors, taken together, provide more aid to the countries of their commercial interests. It is noted, however, that this study uses the aggregate health aid provided by various bilateral and multilateral donors, and therefore, further studies using bilateral health aid flows would be necessary to assess the heterogeneous interests of individual donors. Among the “selectivity” variables, only the level of democracy shows a positive relationship with the total value of health aid in columns (1), (2), (4), and (5). In contrast, the corruption variable is found to be insignificant in all equations. Thus, we have only limited evidence that donors “select” countries and provide more health aid to those countries that could use foreign aid more effectively. Indeed, this finding is consistent with Alesina and Weder (2002) which find similar results for the allocation of total foreign aid, with respect to the level of democracy and corruption. 19 Meanwhile, the armed conflict variable does not appear to be associated with the allocation of health aid. This finding may suggest that the armed conflict variable is not only a “selectivity” variable but also a “needs” variable because of casualties in armed conflicts. For other control variables, as we expected, the coefficients of log of GDP per capita and its squared term have a positive and a negative sign, respectively, indicating that there exists a middle-income bias in the allocation of total health aid. Total health aid is also influenced by population size. All GMM specifications presented in Table 3 pass the Arellano–Bond test of serial independence in the error terms. The hypothesis that the errors are not correlated is rejected in the case of the AR(1) error-component specification and is not rejected for AR(2). In addition, all GMM specifications pass the Hansen J test by not rejecting the null hypothesis that the over-identification restrictions are valid. Thus, we can say that the instruments for the lagged dependent variables and the “needs” variables are exogenous. (b) Results for project-extensive and project-intensive margins of health aid As explained in the previous section, we aim to assess how the needs of the recipient countries are related not only to the total value of health aid but also to the extensive margin of the number of health aid projects and the intensive margin of the average value of each project. In fact, as the needs of the recipient country increase, donors may respond to this by increasing the value of the existing projects or/and by initiating new health aid projects. Table 4 presents the determinants of

112

WORLD DEVELOPMENT Table 3. Total value of health aid allocation Total value of health aid (log) OLS (1)

Recipients’ needs Log of infant mortality rate

(2)

(3)

0.689*** (0.158)

(4)

(5)

(6)

0.469*** (0.157) 0.626*** (0.131)

Log of child mortality rate

0.445*** (0.131) 0.250*** (0.044)

Log of prevalence of HIV Donors’ interests Log of import

SYS GMM

0.192*** (0.044)

0.060 (0.158) 0.019*** (0.004)

0.041 (0.156) 0.019*** (0.004)

0.258 (0.170) 0.019*** (0.003)

0.014 (0.121) 0.014*** (0.003)

0.009 (0.111) 0.013*** (0.003)

0.145 (0.113) 0.015*** (0.003)

0.034*** (0.012) 0.125 (0.145) 0.196 (0.160)

0.035*** (0.011) 0.101 (0.141) 0.192 (0.158)

0.016 (0.011) 0.060 (0.160) 0.118 (0.157)

0.024** (0.010) 0.074 (0.130) 0.177 (0.134)

0.025*** (0.009) 0.090 (0.127) 0.164 (0.125)

0.010 (0.009) 0.024 (0.162) 0.093 (0.101)

3.441*** (0.877) 0.275*** (0.059) 0.652*** (0.140)

3.606*** (0.881) 0.287*** (0.059) 0.635*** (0.137)

4.658*** (1.001) 0.362*** (0.069) 0.854*** (0.151)

21.105*** (4.035) Yes 1,741 112 0.572

21.539*** (4.023) Yes 1,741 112 0.575

24.128*** (4.519) Yes 1,535 97 0.602

2.606*** (0.758) 0.212*** (0.053) 0.415*** (0.111) 0.267*** (0.059) 13.263*** (3.256) Yes 1,632 112

3.005*** (0.801) 0.237*** (0.056) 0.423*** (0.102) 0.276*** (0.062) 14.828*** (3.407) Yes 1,632 112

4.369*** (0.877) 0.346*** (0.063) 0.666*** (0.112) 0.131*** (0.049) 19.347*** (3.747) Yes 1,436 97

Arellano–Bond Test AR(1) p-value AR(2) p-value

0.000 0.479

0.000 0.464

0.000 0.912

Overidentification Test (Hansen) p-value

0.328

0.332

0.933

Crude oil reserves Selectivity Democracy level Control of corruption Armed conflict dummy Others Log of GDP per capita Log of GDP per capita, Squared Log of population Lagged dependent variable (t1) Constant Year dummies Number of observations Number of countries R2

Notes: Estimates of columns (1), (2) and (3) are made by Ordinary Least Squares with country-clustered robust standard errors, and estimates of columns (4), (5) and (6) are made by two-step robust System GMM with the Windmeijer (2005) correction. Lagged dependent variables and the “needs” variables are treated as endogenous in the GMM specifications. Shown in parentheses are standard errors. *** Significance at 1%. ** Significance at 5%. * Significance at 10%.

the number of health aid projects and the average value per project, obtained by the System GMM, which is our preferred specification. We find that a 10% increase in the three “needs” variables of infant mortality, child mortality, and prevalence of HIV is significantly associated with a 1.25%, 1.12% and 0.35% increase in the number of projects, respectively. In the equations for the project-intensive margin, however, infant and child mortality are not statistically significant, implying that the average value of health aid per project does not respond to the child health of the recipient countries, on average. In contrast, a 10% increase in prevalence of HIV increases the average value per health aid project by 0.81%. Looking at the results of both Tables 3 and 4, we can infer that when a recipient country’s health status in terms of infant

or child mortality deteriorates, the total value of health aid committed to this country increases because of the increase in the number of projects committed to the country but not in the projects’ average value. In contrast, an increase in the prevalence of HIV raises the total value of health aid, in terms of both number of projects and the average project value. This finding may reflect the recent global attention on combating HIV/AIDS as discussed in Section 2, but merits further investigation in future research. Among the control variables, the results are similar for both the project-extensive and project-intensive margins in terms of the signs and the significance levels. An exception is the democracy level, which shows marginally positive results for the average value of health aid per project in columns (4)

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS?

113

Table 4. Project-extensive and intensive margins of health aid allocation Project-extensive: number of projects (log) (1) Recipients’ needs Log of infant mortality rate

(2)

(3)

0.125*** (0.037)

(4)

(5)

(6)

0.185 (0.133) 0.112*** (0.032)

Log of child mortality rate

0.155 (0.100) 0.035*** (0.010)

Log of prevalence of HIV Donors’ interests Log of import

Project-intensive: average value of health aid per project (log)

0.081** (0.041)

0.007 (0.025) 0.002*** (0.001)

0.003 (0.025) 0.002*** (0.001)

0.036 (0.024) 0.001*** (0.000)

0.013 (0.087) 0.007*** (0.002)

0.006 (0.080) 0.007*** (0.002)

0.051 (0.096) 0.007*** (0.002)

0.003 (0.002) 0.028 (0.019) 0.099*** (0.036)

0.003 (0.002) 0.021 (0.019) 0.089** (0.035)

0.002 (0.002) 0.001 (0.021) 0.067 (0.042)

0.010* (0.006) 0.052 (0.097) 0.038 (0.096)

0.011* (0.006) 0.065 (0.097) 0.014 (0.104)

0.008 (0.007) 0.048 (0.120) 0.087 (0.087)

0.303** (0.127) 0.025*** (0.009) 0.100*** (0.028) 0.532*** (0.041) 0.134*** (0.042) 0.075*** (0.028) 1.505** (0.654) Yes 1,427 112

0.324** (0.127) 0.026*** (0.009) 0.097*** (0.027) 0.534*** (0.041) 0.125*** (0.042) 0.075*** (0.027) 1.519** (0.654) Yes 1,427 112

0.415*** (0.129) 0.034*** (0.009) 0.120*** (0.032) 0.579*** (0.039) 0.101** (0.047) 0.063** (0.031) 1.499** (0.642) Yes 1,255 97

2.381*** (0.495) 0.190*** (0.037) 0.172** (0.077) 0.171*** (0.044)

2.329*** (0.520) 0.186*** (0.038) 0.182** (0.075) 0.176*** (0.047)

3.416*** (0.955) 0.267*** (0.067) 0.244** (0.098) 0.121*** (0.044)

11.415*** (2.092) Yes 1,632 112

11.279*** (2.173) Yes 1,632 112

14.806*** (4.017) Yes 1,436 97

Arellano–Bond Test AR(1) p-value AR(2) p-value

0.000 0.514

0.000 0.613

0.000 0.660

0.000 0.531

0.000 0.509

0.000 0.794

Overidentification test (Hansen) p-value

0.443

0.480

0.766

0.560

0.466

0.920

Crude oil reserves Selectivity Democracy level Control of corruption Armed conflict dummy Others Log of GDP per capita Log of GDP per capita, squared Log of population Lagged dependent variable (t-1) Lagged dependent variable (t-2) Lagged dependent variable (t-3) Constant Year dummies Number of observations Number of countries

Notes: Estimates are made by two-step robust System GMM with the Windmeijer (2005) correction. Lagged dependent variables and the “needs” variables are treated as endogenous in the GMM specifications. Shown in parentheses are standard errors. In columns (1), (2) and (3), second and third lagged dependent variables are additionally included because System GMM with the first lagged dependent variable did not pass the Arellano–Bond’s AR(2) Test. *** denote significance at 1 percent respectively. ** denote significance at 5 percent respectively. * denote significance at 10 percent, respectively.

and (5). Another exception is the armed conflict dummy, which is negatively associated with the number of projects in columns (1) and (2). (c) Results for donor-extensive and donor-intensive margins of health aid As the health needs increase in a developing country, total health aid to the country may increase not only because the currently active donors increase the value of their health aid but

also because new donors join the international community’s efforts to help the country. Table 5 presents the estimated results when the extensive margin of the number of donors and the intensive margin of the average value of health aid per donor are the dependent variables. In the equations for both margins, the recipients’ needs variables enter with positive coefficients, but in terms of the significance level, the donor-intensive margin appears to be more responsive to the needs of the recipients. 20 Similar to the results in the project-extensive and projectintensive margins in Table 4, the coefficients for the democracy

114

WORLD DEVELOPMENT Table 5. Donor-extensive and intensive margins of health aid allocation Donor-extensive: number of donors (log) (1)

Recipients’ needs Log of infant mortality rate

(2)

(3)

(4)

(5)

(6)

0.383*** (0.128)

0.046 (0.029) 0.041* (0.023)

Log of child mortality rate

0.317*** (0.116) 0.017** (0.008)

Log of prevalence of HIV Donors’ interests Log of import

Donor-intensive: average value of health aid per donor (log)

0.142*** (0.040)

0.001 (0.018) 0.001*** (0.000)

0.000 (0.018) 0.001*** (0.000)

0.020 (0.015) 0.001*** (0.000)

0.015 (0.092) 0.009*** (0.002)

0.002 (0.092) 0.008*** (0.002)

0.126 (0.125) 0.009*** (0.002)

0.001 (0.001) 0.017 (0.017) 0.031* (0.018)

0.001 (0.001) 0.016 (0.016) 0.031* (0.018)

0.001 (0.001) 0.009 (0.016) 0.020 (0.018)

0.015** (0.007) 0.054 (0.102) 0.088 (0.105)

0.016** (0.007) 0.063 (0.109) 0.085 (0.110)

0.009 (0.008) 0.022 (0.137) 0.006 (0.098)

0.134 (0.090) 0.013* (0.006) 0.046** (0.018) 0.350*** (0.038) 0.203*** (0.032) 0.166*** (0.028) 0.403 (0.418) Yes 1,427 112

0.145 (0.092) 0.013** (0.007) 0.046** (0.018) 0.347*** (0.038) 0.201*** (0.032) 0.165*** (0.028) 0.421 (0.422) Yes 1,427 112

0.273*** (0.090) 0.022*** (0.007) 0.071*** (0.018) 0.360*** (0.039) 0.202*** (0.033) 0.132*** (0.031) 0.878** (0.394) Yes 1,255 97

2.321*** (0.662) 0.184*** (0.048) 0.287*** (0.083) 0.221*** (0.049)

2.438*** (0.636) 0.193*** (0.045) 0.280*** (0.082) 0.231*** (0.054)

4.146*** (1.087) 0.320*** (0.075) 0.452*** (0.131) 0.123*** (0.042)

12.226*** (2.642) Yes 1,632 112

12.292*** (2.591) Yes 1,632 112

18.195*** (4.870) Yes 1,436 97

Arellano–Bond Test AR(1) p-value AR(2) p-value

0.000 0.916

0.000 0.913

0.000 0.814

0.000 0.416

0.000 0.391

0.000 0.836

Overidentification test (Hansen) p-value

0.507

0.503

0.904

0.428

0.426

0.952

Crude oil reserves Selectivity Democracy level Control of corruption Armed conflict dummy Others Log of GDP per capita Log of GDP per capita, squared Log of population Lagged dependent variable (t-1) Lagged dependent variable (t-2) Lagged dependent variable (t-3) Constant Year dummies Number of observations Number of countries

Notes: Estimates are made by two-step robust System GMM with the Windmeijer (2005) correction. Lagged dependent variables and the “needs” variables are treated as endogenous in the GMM specifications. Shown in parentheses are standard errors. In columns (1), (2) and (3), second and third lagged dependent variables are additionally included because System GMM with the first lagged dependent variable did not pass the Arellano–Bond’s AR(2) Test. *** Significance at 1%. ** Significance at 5%. * Significance at 10%.

level were found significantly positive in the equations for the intensive margins. These results in Tables 4 and 5 suggest that the positive and significant coefficient for the democracy level in the equation for the total value of health aid is due to the fact that countries with a high degree of democracy receives a greater amount of health aid per project, as well as per donor country. It is also interesting to note that the armed conflict dummy enters with negative and significant coefficients in the equations for the donor-extensive margin of columns (1) and (2). Thus,

findings in Tables 4 and 5 jointly suggest that countries under internal or external conflict receive a smaller number of health aid projects and donors when other conditions are the same. (d) Robustness checks 21 We perform a number of robustness checks. First of all, we are concerned with whether the increase in the number of reporting donors in the CRS database has affected our results. During our sample period, the number of donors increased

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS?

from 21 to 46. Therefore, as a robustness check, we restrict our sample to 21 donors which were already reporting to the CRS in 1995. 22 In Table 6, we report the results for the equations when the “needs” variable is infant mortality rate. 23 For the sake of brevity, reported are only the estimated coefficients for the infant mortality in various equations. For comparison purpose, reported are also the corresponding results obtained from the unrestricted sample, as originally presented in Tables 3–5. As shown in Panel B of Table 6, the results with the original 21 donors are remarkably similar to those with the full sample (Panel A). Secondly, we are concerned with whether different sample periods may result in different results. We have worked with the sample period from 1995 to 2011. DCD/DAC (2014) reports that the coverage ratio of CRS commitment data in 1995 was 70% and has become over 90% since 2000. Therefore, as another robustness check, we restrict our sample to the period 2000–2011. As reported in Panel C of Table 6, we find that our benchmark results are robust to the different sample periods. We also report the corresponding results for HIV prevalence as an indicator of recipients’ needs in Panels B and C of

115

Table 7, together with the benchmark results in Panel A taken from Tables 3–5. The results for the HIV prevalence as a needs variables are also robust to the restricted samples of donors and periods. A minor exception is the equation for the project-intensive margin when the number of donors is restricted to 1995 original donors. For the HIV prevalence, we also conduct additional robustness checks with the sub-category of health aid. As explained in Section 2, the health aid data that we use in the analysis are obtained from combining three health-related sectors: General health (CRS121), Basic health (CRS122) and Population policies/programs and reproductive health (CRS130). Particularly, STD control including HIV/AIDS (CRS13040) aid is under CRS130. Therefore, we repeat the estimations for Panels A, B, and C with the use of the data under CRS130 only, instead of aggregating the data under CRS121, 122, and 130. 24 In terms of statistical significance, the results are almost identical. It is interesting to note, however, that the size of estimated coefficients with data from CRS130 is generally greater than with data from CRS121, 122, and 130. This seems reasonable because CRS130 is more directly related to the control of HIV/AIDS.

Table 6. Robustness checks with restricted samples—infant mortality Recipients’ needs Log of infant mortality

Total (1)

Project-extensive (2)

Project-intensive (3)

Donor-extensive (4)

Donor-intensive (5)

A. Original: unrestricted sample

0.469*** (0.157)

0.125*** (0.037)

0.185 (0.133)

0.046 (0.029)

0.383*** (0.128)

B. Restricted number of donors (21)

0.387** (0.177)

0.152** (0.064)

0.177 (0.149)

0.040 (0.029)

0.391** (0.156)

C. Restricted sample period (2000–11)

0.258** (0.107)

0.106*** (0.036)

0.104 (0.105)

0.037 (0.028)

0.206** (0.101)

Notes: Estimates are made by two-step robust System GMM with the Windmeijer (2005) correction. Lagged dependent variables and the “needs” variables are treated endogenous in the GMM specifications. Shown in parentheses are standard errors. For the sake of brevity, reported are only the estimated coefficients for the infant mortality in various equations. All equations pass the Arellano-Bond AR(2) test and Hansen’s overidentification test. *** Significance at 1%. ** Significance at 5%. * Significance at 10%.

Table 7. Robustness checks—HIV prevalence rate Recipients’ needs Log of HIV prevalence

Total (1)

Project-extensive (2)

Project-intensive (3)

Donor-extensive (4)

Donor-intensive (5)

A. Original: unrestricted sample, CRS120+130

0.192*** (0.044)

0.035*** (0.010)

0.081** (0.041)

0.017** (0.008)

0.142*** (0.040)

B. Restricted number of donors (21), CRS120+130

0.183*** (0.049)

0.071*** (0.020)

0.054 (0.037)

0.028** (0.011)

0.107** (0.042)

C. Restricted sample period (2000–11), CRS120+130

0.130*** (0.038)

0.023** (0.010)

0.078** (0.032)

0.015* (0.008)

0.097*** (0.030)

D. Unrestricted sample, CRS 130

0.230*** (0.047)

0.072*** (0.019)

0.134*** (0.033)

0.099*** (0.018)

0.158*** (0.033)

E. Restricted number of donors (19), CRS130

0.112** (0.051)

0.062*** (0.022)

0.114** (0.051)

0.126*** (0.028)

0.098** (0.042)

F. Restricted sample period (2000–11), CRS130

0.264*** (0.043)

0.067*** (0.015)

0.162*** (0.036)

0.113*** (0.018)

0.169*** (0.029)

Notes: Estimates are made by two-step robust System GMM with the Windmeijer (2005) correction. Lagged dependent variables and the “needs” variables are treated endogenous in the GMM specifications. Shown in parentheses are standard errors. For the sake of brevity, reported are only the estimated coefficients for the HIV prevalence rate in various equations. All equations pass the Arellano-Bond AR(2) test and Hansen’s overidentification test. *** Significance at 1%. ** Significance at 5%. * Significance at 10%.

116

WORLD DEVELOPMENT

5. SUMMARY AND CONCLUDING REMARKS Using OECD-DAC data covering 112 recipients over the period 1995–2011, this paper distinguishes the extensive and intensive margins of health aid allocation patterns at projectand donor-levels and assesses whether health aid follows the needs of the recipient countries. We find a number of interesting stylized facts about the allocation of health aid. First, the total amount of health aid is allocated responsively to the needs of the recipients, as proxied by infant mortality, child mortality, and prevalence of HIV. Specifically, a 10% increase in infant mortality, in child mortality, and in prevalence of HIV raises 4.69%, 4.45%, and 1.92% in health aid in the short run, and 6.40%, 6.14%, and 2.21% in the long run, respectively. This finding is encouraging, as many researchers find that the marginal effectiveness of aid is highest where there is greatest need (e.g., Collier & Dollar, 2002). Second, a rise in infant or child mortality of a recipient country results in an increase in the total value of health aid committed to this country not because it receives more

valuable projects, but rather because it receives more health aid projects. In contrast, an increase in the prevalence of HIV raises the total value of health aid in terms of both number of projects and the average project value. Third, when there is an increase in a recipient country’s health needs in terms of all three indicators, the total value of health aid committed to this country increases because both the number of donors and the average committed dollar amount of each donor increase. But we find that the average committed value of each country is more responsive than the number of donors. Overall, when the health outcome of a recipient country deteriorates, the total value of health aid to the country increases due in large part to an increase in the number of health aid projects and to an increase in the average aid value from each donor country. As mentioned in the introduction, there are both merits and demerits of allocating more extensively or more intensively at project- and donor-levels. Our findings indicate that there is a great need of an assessment of how different margins of health aid result in differential impacts on the health outcomes of the recipient countries.

NOTES 1. This paper follows the recent trend of the international trade literature, focusing more on the micro-level data—such as the “extensive margin of the number of firms and products” and the “intensive margin of the amount traded per firm and product”—than on the aggregated country-level data (Please refer to Bernard, Jensen, Redding, & Schott, 2007; Lee, Park, & Wang, 2013; Melitz, 2003). 2. Appendix Table 8 lists sub-categories under CRS code 121, 122 and 130. 3. It should be noted that the project-level health aid data may have measurement errors, for there may be different criteria for defining “project” across donors. When there is a need to increase health aid, donors may find it administratively more convenient to set up new projects rather than extending existing projects. On the other hand, donors may also find it more convenient to increase more funds to existing projects. Overall, we believe that the measurement error is rather neutral. 4. The list of countries in the sample is reported in Appendix Table 9. 5. The increasing trend of the extensive margin of the number of donors per recipient might be influenced by the increase in the total number of donors reported in the OECD DAC CRS database. In the database, the total number of health aid donors increased from 21 in 1995 to 46 in 2011. See Appendix Table 10 for the list of donor which provided health aid in 2011. 6. This tendency is not a new finding: Morss (1984) pointed out that “the most important feature distinguishing foreign aid in the 1970s from earlier programs was the proliferation of donors and projects.” He states that in the 1950–60s, foreign aid was given in a big scale involving large infrastructure investment or provided general budget support to a specific sector, such as health and agriculture; while in the 1970s, foreign aid disaggregated into many smaller packets of projects, for there were growing doubts about aid effectiveness. 7. In the case of OLS using a panel dataset, the unobservable countryspecific error term ui and the idiosyncratic error term eit are not separately estimated. Thus, we use country-clustered robust standard errors which are heteroskedasticity- and autocorrelation-consistent.

8. One may suspect that other variables such as income and governance are also endogenous. However, health aid accounts for only about 10% of total foreign aid, and thus, direct and indirect effects of health aid on income and governance may be negligible. 9. We use the ‘xtabond2’ command of Stata provided by Roodman (2009b) 10. With bilateral aid data, period averages are often used to smooth the volatility of aid commitments (e.g., Alesina & Dollar, 2000; Neumayer, 2003). However, we prefer to use annual data so as not to lose statistical efficiency by averaging the data. Also, we are using aggregated aid data by recipient country and thus there is less annual fluctuation than in an individual donor data. 11. Yontcheva and Nancy (2006) also argue that using aid commitments is preferable to aid disbursements because the latter are tainted by substantial volatility and instability. 12. We may lose zero observations by taking the logs. Some authors account for sample selection bias, which may occur when we lose zero observations by taking logarithms on health aid, by employing methods such as Tobit, Heckman two-step, or two-part models (e.g., Neumayer, 2005). However, the selection bias in our log-linear model should be marginal, for we are only concerned with the recipient country’s total health aid flows from the world, not individual donors’ own commitment; and zero observations account for about 10% in our sample. 13. In bilateral models which concern both donor- and recipient-specific characteristics, donor’s interest could be captured in many dimensions such as colonial relationship, bilateral trade, linguistic/ethnic similarity, distance, and UN vote similarity. (e.g., Alesina & Dollar, 2000; Berthe´lemy, 2006). 14. We found very similar estimation results even when we use the original data with sub-sample excluding the years for which “control of corruption” data are not available. 15. As an indicator of democracy level, Fielding (2011) used the “voice and accountability” index reported by Kaufmann et al. (2010). We also tried the same index in place of the Polity IV index, but found similar

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS? results. However, we prefer to report the results for the Polity IV index because the “voice and accountability” index is only available annually from 2003, whereas the Polity IV index is available for the entire sample period. b1 (b1: “needs” variable coefficient, d: lagged16. Long-run effect ¼ 1d dependent variable coefficient).

17. We also estimated the model alternatively with total trade (exports plus imports), FDI flows, and FDI stock, and found no statistically significant results with any of them. 18. Similar results were found with natural gas reserves (in trillion cubic feet) in place of crude oil reserves. 19. Alesina and Weder (2002) also find that there exist significant differences across donors, where Scandinavian countries reward less corrupted recipients while the United States gave more aid to relatively corrupted recipients, ceteris paribus. Similarly, Isopi & Mattesini (2008) present a theoretical model and an empirical analysis, suggesting that the aid–corruption relationship depends on the preferences of the donor.

117

20. It is noteworthy that the relatively less responsive reaction in the number of donors to greater need may be due to the fact that the number of donors is a bounded variable which cannot exceed the number of CRS reporters. However, it is also worth noting that, in our sample, the average number of donors per recipient country is only about 8.5 as reported in Table 1 while the maximum possible number of donors is 48. 21. We are very grateful to one of the referees for suggesting a number of constructive robustness checks. 22. Donors that initially reported health aid in 1995 include AsDB Special Funds, Australia, Belgium, Canada, Denmark, EU Institutions, Finland, France, Germany, IDA, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, UK, and US. 23. We also performed the robustness checks when the “needs” variable is child mortality in place of infant mortality, and found no qualitative difference. 24. Under CRS130, 19 donors were already in 1995, and 36 donors in 2011.

REFERENCES Acharya, A., de Lima, A., & Moore, M. (2006). Proliferation and fragmentation: Transactions costs and the value of aid. The Journal of Development Studies, 42(1), 1–21. Alesina, A., & Dollar, D. (2000). Who gives foreign aid to whom and why?. Journal of Economic Growth, 5(1), 33–63. Alesina, A., & Weder, B. (2002). Do corrupt governments receive less foreign aid?. The American Economic Review, 92(4), 1126–1137. Bernard, A. B., Jensen, J. B., Redding, S. J., & Schott, P. K. (2007). Firms in international trade. Journal of Economic Perspectives, 23(3), 105–130. Berthe´lemy, J.-C. (2006). Bilateral donors’ interest vs. recipients’ development motives in aid allocation: Do all donors behave the same?. Review of Development Economics, 10(2), 179–194. Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(1), 115–143. Blundell, R., & Bond, S. (2000). GMM estimation with persistent panel data: An application to production functions. Econometric Reviews, 19(3), 321–340. Boone, P. (1996). Politics and the effectiveness of foreign aid. European Economic Review, 40(2), 289–329. Burnside, C., & Dollar, D. (2000). Aid, policies, and growth. The American Economic Review, 90(4), 847–868. Christensen, Z., Homer, D., & Nielson, D. L. (2011). Dodging adverse selection: How donor type and governance condition aid’s effects on school enrollment. World Development, 39(11), 2044–2053. Claessens, S., Cassimon, D., & Campenhout, B. V. (2009). Evidence on changes in aid allocation criteria. The World Bank Economic Review, 23(2), 185–208. Clemens, M. A., Radelet, S., Bhavnani, R. R., & Bazzi, S. (2012). Counting chickens when they hatch: Timing and the effects of aid on growth. The Economic Journal, 122(561), 590–617. Collier, P., & Dollar, D. (2002). Aid allocation and poverty reduction. European Economic Review, 46(8), 1475–1500. DCD/DAC (2014). Technical guide to terms and data in the Creditor Reporting System (CRS) aid activities database, http://www.oecd.org/ dac/stats/crsguide.htm, Accessed Online in March 4, 2014. Dodd, R., Schieber, G., Cassels, A., Fleisher, L., & Gottret, P. (2007). Aid effectiveness and health. Making Health Systems Work: Working Paper No. 9. World Health Organization. Dudley, L., & Montmarquette, C. (1976). A model of supply of bilateral foreign aid. American Economic Review, 66(1), 132–142. Esser, D. E., & Bench, K. K. (2011). Does global health funding respond to recipients’ needs? Comparing public and private donors’ allocations in 2005–2007. World Development, 39(8), 1271–1280. Fielding, D. (2011). Health aid and governance in developing countries. Health Economics, 20(7), 757–769.

Gounder, R. (1994). Empirical results of aid motivations: Australia’s bilateral aid program. World Development, 22(1), 99–113. Gounder, R. (1995). Population and middle-income biases in Australia’s bilateral aid: Some empirical results. Development Policy Review, 12(1), 29–44. Isopi, A., Mattesini, F. (2008). Aid and corruption: Do donors use development assistance to provide the "right" incentives?, CEIS Working Paper No. 121, Center for Economic and International Studies. Kaiser Family Foundation (2007). A global look at public perceptions of health problems, priorities, and donors: The Kaiser/Pew global health survey. Menlo Park, CA: The Kaiser Family Foundation and Pew Global Attitudes Project. Kanbur, R., Sandler, T., & Morrison, K. (1999). The future of development assistance: Common pool and international public goods. Washington, DC: Overseas Development Council. Kaufmann, D., Kraay, A., & Mastruzzi, M. (2010). The worldwide governance indicators: a summary of methodology, data and analytical issues. Policy Research Working Paper 5430. World Bank. Kimura, H., Mori, Y., & Sawada, Y. (2012). Aid proliferation and economic growth: A cross-country analysis. World Development, 40(1), 1–10. Lee, H.-H., Park, D., & Wang, J. (2013). Different types of firms, different types of products, and their dynamics: An anatomy of china’s imports. China Economic Review, 25, 62–77. Levine, R., & the What Works Working Group (2004). Millions saved, proven successes in global health. Washington, DC: Center for Global Development. Maizels, A., & Nissanke, M. K. (1984). Motivations for aid to developing countries. World Development, 12(9), 879–900. McKinley, R. D., & Little, R. (1977). A foreign policy model of U.S. bilateral aid allocation. World Politics, 30(1), 58–86. Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6), 1695–1725. Morss, E. R. (1984). Institutional destruction resulting from donor and project proliferation in Sub-Saharan African countries. World Development, 12(4), 465–479. Nelson, R. (2010). Does aid follow need?. Humanitarian motives in aid allocation: Manuscript, Harvard University. Neumayer, E. (2003). The determinants of aid allocation by regional multilateral development banks and United Nations agencies. International Studies Quarterly, 47(1), 101–122. Neumayer, E. (2005). Is the allocation of food aid free from donor interest bias?. The Journal of Development Studies, 41(3), 394–411. Piva, P., & Dodd, R. (2009). Where did all the aid go? An in-depth analysis of increased health aid flows over the past 10 years. Bulletin of the World Health Organization, 87, 930–939.

118

WORLD DEVELOPMENT

Rajan, R. & Subramanian, A. (2005). Aid and growth: What does the cross-country evidence really show? IMF Working Papers 05/127. Reidpath, D., & Allotey, P. (2003). Infant mortality rate as an indicator of population health. Journal of Epidemiology Community Health, 57(5), 344–346. Roodman, D. (2006). Aid project proliferation and absorptive capacity. Center for Global Development Working Paper, 75. Roodman, D. (2006). Competitive proliferation of aid projects: A model. Center for Global Development Working Paper, 89. Roodman, D. (2009a). Practitioner’s corner: A note on the theme of too many instruments. Oxford Bulletin of Economics and Statistics, 71(1), 135–158. Roodman, D. (2009b). How to do xtabond2: An introduction to “difference” and “system” GMM in Stata. Stata Journal, 9(1), 86–136. Sachs, J. (2005). The end of poverty: Economic possibilities for our time. New York: Penguin Press. Thiele, R., Nunnenkamp, P., & Dreher, A. (2007). Do donors target aid in line with the Millennium Development Goals? A sector perspective of aid allocation. Review of World Economics, 143(4), 596–630.

Wallensteen, P., & Sollenberg, M. (2001). Armed conflict 1989–2000. Journal of Peace Research, 38(5), 629–644. Windmeijer, F. (2005). A finite sample correction for the variance of linear efficient two-step GMM estimators. Journal of Econometrics, 126(1), 25–51. Yontcheva, B., & Nancy, G. (2006). Does NGO aid go to the poor? Empirical evidence from Europe. IMF Working Papers 06/39. Younger, S. (1992). Aid and Dutch disease: Macroeconomic management when everybody loves you. World Development, 20(1), 1587–1597.

APPENDIX A see Tables 8–10

Table 8. Description and coverage of health aid CRS CODE

Description

121 12110

General health Health policy and administrative management

12181 12182 12191

Medical education/training Medical research Medical services

122 12220

Basic health Basic health care

12230

Basic health infrastructure

12240

Basic nutrition

12250

Infectious disease control

12261

Health education

12262 12263 12281 130

Malaria control Tuberculosis control Health personnel development Population policies/programs and reproductive health Population policy and administrative management

13010

13020

Reproductive health care

13030

Family planning

13040

STD control including HIV/AIDS

13081

Personnel development for population and reproductive health

Source: OECD DAC Statistics

Clarifications/additional notes on coverage Health sector policy, planning and programs; aid to health ministries, public health administration; institution capacity building and advice; medical insurance programs; unspecified health activities Medical education and training for tertiary level services General medical research (excluding basic health research) Laboratories, specialized clinics and hospitals (including equipment and supplies); ambulances; dental services; mental health care; medical rehabilitation; control of non-infectious diseases; drug and substance abuse control [excluding narcotics traffic control (16063)] Basic and primary health care programs; paramedical and nursing care programs; supply of drugs, medicines and vaccines related to basic health care District-level hospitals, clinics and dispensaries and related medical equipment; excluding specialized hospitals and clinics (12191) Direct feeding programs (maternal feeding, breastfeeding and weaning foods, child feeding, school feeding); determination of micronutrient deficiencies; provision of vitamin A, iodine, iron etc.; monitoring of nutritional status; nutrition and food hygiene education; household food security Immunization; prevention and control of infectious and parasite diseases, except malaria (12262), tuberculosis (12263), HIV/AIDS and other STDs (13040). It includes diarrheal diseases, vectorborne diseases (e.g. river blindness and guinea worm), viral diseases, mycosis, helminthiasis, zoonosis, diseases by other bacteria and viruses, pediculosis, etc Information, education and training of the population for improving health knowledge and practices; public health and awareness campaigns; promotion of improved personal hygiene practices, including use of sanitation facilities and handwashing with soap Prevention and control of malaria Immunization, prevention and control of tuberculosis Training of health staff for basic health care services

Population/development policies; census work, vital registration; migration data; demographic research/analysis; reproductive health research; unspecified population activities Promotion of reproductive health; prenatal and postnatal care including delivery; prevention and treatment of infertility; prevention and management of consequences of abortion; safe motherhood activities Family planning services including counseling; information, education and communication (IEC) activities; delivery of contraceptives; capacity building and training All activities related to sexually transmitted diseases and HIV/AIDS control e.g. information, education and communication; testing; prevention; treatment, care Education and training of health staff for population and reproductive health care services

DOES INTERNATIONAL HEALTH AID FOLLOW RECIPIENTS’ NEEDS?

119

Table 9. List of recipient countries No.

Country name

Income group

No.

Country name

Income group

No.

Country name

Income group

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Angola Bangladesh Benin Bhutan Burkina Faso Burundi Cambodia Central African Rep. Chad Comoros Congo, Dem. Rep. Djibouti Equatorial Guinea Eritrea Ethiopia Gambia Guinea Guinea-Bissau Haiti Laos Lesotho Liberia Madagascar Malawi Mali Mauritania Mozambique Nepal Niger Rwanda Senegal Sierra Leone Solomon Islands Sudan Tanzania Timor-Leste Togo Uganda Yemen Zambia

LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs LDCs

41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76

Kenya Kyrgyz Republic Tajikistan Zimbabwe Armenia Bolivia Cameroon Cape Verde Congo, Rep. Egypt El Salvador Fiji Georgia Ghana Guatemala Guyana Honduras India Indonesia Iraq Moldova Mongolia Morocco Nicaragua Nigeria Pakistan Papua New Guinea Paraguay Philippines Sri Lanka Swaziland Syria Turkmenistan Ukraine Uzbekistan Vietnam

Other LICs Other LICs Other LICs Other LICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs LMICs

77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112

Albania Algeria Argentina Azerbaijan Belarus Botswana Brazil Chile China Colombia Costa Rica Dominican Republic Ecuador Gabon Iran Jamaica Jordan Kazakhstan Lebanon Libya Macedonia, FYR Malaysia Mauritius Mexico Montenegro Namibia Panama Peru Serbia South Africa Suriname Thailand Tunisia Turkey Uruguay Venezuela

UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs UMICs

Notes: LDCs, Other LICs, LMICs, and UMICs indicate Least Developed Countries, Other Low Income Countries, Lower Middle Income Countries, and Upper Middle Income Countries respectively. Recipients are divided into income groups according to the criteria of OECD DAC CRS database.

Table 10. List of donors No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Donor name African Development Bank (AfDB) Arab Bank for Economic Development in Africa (BADEA) Asian Development Bank (AsDB) Australia Austria Belgium Canada Czech Republic Denmark European Commission Finland France GAVI Alliance Germany Global Fund to Fight Aids, Tuberculosis and Malaria Greece

Donor type Multilateral Multilateral Multilateral DAC Country DAC Country DAC Country DAC Country DAC Country DAC Country Multilateral DAC Country DAC Country Multilateral DAC Country Multilateral DAC Country

120 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

WORLD DEVELOPMENT Iceland Inter-American Development Bank (IDB) International Development Association (IDA) International Fund for Agricultural Development (IFAD) Ireland Islamic Development Bank Italy Japan Joint United Nations Programme on HIV/AIDS Korea Kuwait (KFAED) Luxembourg Netherlands New Zealand Norway OPEC Fund for International Development (OFID) Portugal Spain Sweden Switzerland United Arab Emirates United Kingdom United Nations Children’s Fund (UNICEF) United Nations Development Programme (UNDP) United Nations Economic Commission for Europe United Nations Peacebuilding Fund (UNPBF) United Nations Population Fund (UNFPA) United Nations Relief and Works Agency (UNRWA) United States World Health Organization (WHO)

Notes: Listed are donors with any activity reported in the health sector in 2011.

Available online at www.sciencedirect.com

ScienceDirect

DAC Country Multilateral Multilateral Multilateral DAC Country Multilateral DAC Country DAC Country Multilateral DAC Country Non-DAC Country DAC Country DAC Country DAC Country DAC Country Multilateral DAC Country DAC Country DAC Country DAC Country Non-DAC Country DAC Country Multilateral Multilateral Multilateral Multilateral Multilateral Multilateral DAC Country Multilateral