Wor King Papers. Economics Working Papers. Types of Foreign Aid Christian Bjørnskov - PDF

Wor King Papers Economics Working Papers Types of Foreign Aid Christian Bjørnskov TYPES OF FOREIGN AID Christian Bjørnskov* Aarhus University Department of Economics and Business, Fuglesangs Alle

Please download to get full document.

View again

of 24
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.


Publish on:

Views: 16 | Pages: 24

Extension: PDF | Download: 0

Wor King Papers Economics Working Papers Types of Foreign Aid Christian Bjørnskov TYPES OF FOREIGN AID Christian Bjørnskov* Aarhus University Department of Economics and Business, Fuglesangs Alle 4, DK-8210 Aarhus V, Denmark. Abstract: Foreign aid is given for many purposes and different intentions, yet most studies treat aid flows as a unitary concept. This paper uses factor analysis to separate aid flows into different types. The main types can be interpreted as aid for economic purposes, social purposes, and reconstruction; a residual category captures remaining purposes. Estimating the growth effects of separable types of aid suggests that most aid has no effects while reconstruction aid has direct positive effects. Although this type only applies in special circumstances, it has become more prevalent in more recent years. Keywords: Foreign aid, dimensionality, development, economic growth JEL Codes: O11, F35 *. Maria Birch Møller provided excellent research assistance. I thank Mike Tierney for help with the AidData database, Peter Nannestad for valuable input to the factor analysis, and Axel Dreher, Martin Gassbebner, Jakob de Haan, Niklas Potrafke, Andrew Young and participants at the 2012 Beyond Basic Questions workshop and the 2012 meetings of the Southern Economic Association for helpful comments on earlier versions. All remaining errors are of course mine. 1 1. Introduction The likely consequences of foreign aid have led to heated discussions among economists since the 1950s (Friedman, 1958; Bauer, 1976). While the intentions from the very beginning were that foreign aid would finance productive investments in order to help developing countries achieve take-off (cf. Rosenstein-Rodan, 1957; Rostow, 1960), the first studies to assess the returns to aid yielded mixed results (Griffin and Eno, 1970; Papanek, 1972, 1973). Since Mosley, Hudson and Horell (1987), a long series of studies that have estimated the effectiveness of aid, either on growth, investments or a set of social outcomes, has found no robust effects. Small parts of the profession continue to argue either that aid in general works or that aid is always harmful (Hansen and Tarp, 2000, 2001; Chauvet and Guillamont, 2001; Minoiu and Reddy, 2010; Ovaska, 2003; Kourtellos, Tan and Chang, 2007; Djankov, Montalvo and Reynal-Querol, 2008). However, systematic surveys document that the converging consensus in the literature is that aid overall has no significant growth effects (Rajan and Subramanian, 2008; Doucouliagos and Paldam, 2008, 2010, 2011; Nowak-Lehman et al., 2012). The aid literature proposes two main explanations for this result while the present paper follows an emerging literature in exploring a third option. As an example of the first explanation, Roodman (2008) tries to settle the discussion by arguing that there is an effect of aid, but that it is so small that econometric problems prevent clear identification. Basically, this type of explanation assumes that aid and central variables are measured with so much noise in developing countries that real effects and noise are statistically indistinguishable. The alternative explanation is that foreign aid has positive direct effects but also comes with several negative indirect effects yielding an average net development effect of zero. Studies in the second strand of the literature first of all document that aid causes Dutch Disease, which undermines competitiveness and the export and manufacturing sector (Arellano et al., 2009; Werker, Ahmed and Cohen, 2009; Bjerg, Bjørnskov and Holm, 2011). In particular, aid inflows and 2 larger projects may induce inflation and relative price changes that distort the economy (Tornell and Lane, 1999; Torvik, 2001; Acharya, Fuzzo de Lima and Moore, 2006; Doces, 2011; Rajan and Subramanian, 2011). Second, inflows of aid allow governments and politicians to spend aid on popular, as opposed to productive, purposes, and to ignore structural problems for substantially longer than if benefits as well as costs accrued only to the country and its own decision-makers (Boone, 1996; Moss, Pettersson and van De Walle, 2007). As such, aid may undermine a number of political incentives and institutional reforms that would be beneficial to long-run growth (Remmer, 2004; Knack, 2001, 2004; Djankov, Montalvo and Reynal-Querol, 2008; Heckelman and Knack, 2008; Bjørnskov and Schröder, 2010). The problems of political incentives also relate to the issue of fungibility: that as aid inflows finance activities within the set of priorities of the government, politicians can rationally decide to reprioritize the budget allocation as aid merely finances something that would otherwise have been partly covered by ordinary budget means. Fungibility thus has the consequence that aid may contribute to activities well outside the interest of donors (Feyzioghlu, Swaroop and Zhu, 1998; Swaroop and Devarajan, 2000; Collier and Hoeffler, 2007; Werker, Ahmed and Cohen, 2009). However, recent studies have sketched a third argument: that foreign aid is given with such different purposes in mind that though precisely measured, the sheer diversity of disbursements makes identification of effects almost impossible (cf. Calderisi, 2006; Dreher, Nunnenkamp and Thiele, 2008b; Wright and Winters, 2010). A few studies have therefore begun to examine the likely effects of specific types of foreign aid. 1 Clemens et al. (2012) argue that early-impact aid, which includes budget and balance of payments support and aid intended to support infrastructure and industrial development, affects growth within a time horizon detectable in standard regression design. Kilby and Dreher (2010) instead attempt to separate aid inflows depending on the motives of different donors, suggesting that aid given with political motives is less likely to contribute to development. Other studies focus 1 It is worth mentioning that this point was a staple of the earliest critique of foreign aid brought up by both Friedman (1958) and Morgenthau (1962). 3 specifically on the effects of aid for education and health (Michaelowa and Weber, 2007; Dreher, Nunnenkamp and Thiele, 2008b; Mishra and Newhouse, 2009; Christensen, Homer and Nielson, 2011). The purpose of this paper is to address the particular problem of aid diversity and to explore the dimensionality of foreign aid disbursements and the consequences of treating foreign aid as a multidimensional international transfer. The paper addresses one particular potential problem stressed by Roodman (2007): that substantial multicollinearity can create the appearance of significance, which may be a substantial problem if specific types cannot meaningfully be separated. If types of aid in general are disbursed together, studies of the effectiveness of specific types may not count all relevant disbursements and thus underestimate the effect. To alleviate this problem, I use the AidData database, which is becoming a standard alternative to OECD / World Bank data in the aid literature (Nielson et al., 2010; Nielsen et al., 2011). Doing so allows me to separate aid into different types, based on the 24 purpose codes in which AidData reports all development projects. This choice implies that aid effectiveness can be estimated without the heterogeneity problem inherent in most previous studies. The analysis shows that most foreign aid disbursements to developing countries between 1970 and 2005 can be split into three clearly identifiable and separable groups and a residual group. Separating types of foreign aid thus allows for substantially more precise estimates of the consequences of aid. When accounting for endogeneity problems, GMM panel estimates suggest that most aid is without consequences. Yet, aid with the purpose of reconstruction exhibits a positive significant effect on growth, suggesting that aid is only effective under such specific circumstances. The rest of the paper is structured as follows. Section 2 briefly outlines the statistical problem and describes the data and estimation strategies. Section 3 reports the results of estimating the dimensionality of foreign aid and outlines the structure of the separated data. Section 4 re-estimates two claims from the existing literature on aid effectiveness while section 5 concludes. 4 2. Data and estimation strategy 2.1. Main data The data on aid used in the following are all from the recent PLAID (Project Level Aid) database, as reported by AidData (Nielson et al., 2010), which offers a more inclusive account of global aid flows. The organization, which was originally set up as a partnership between Brigham Young University, the College of William and Mary, and the non-profit development organization Development Gateway, aims at counting all aid flows regardless of their source. Relative to the common OECD-DAC database, AidData offers nearly twice as large flows of aid when recorded as commitments and excluding concessional loans, not least by adding projects from NGOs and other additional sources not included in OECD statistics. All data are recorded in project form, categorized in 24 purpose codes listed in Table I and a 25 th purpose code, Administrative costs of donors. This feature of AidData allows the separation of different types of aid in a more comprehensive way than in previous studies. I elect to use data on actual disbursements instead of commitments, as some commitments are known not to be fulfilled. While AidData thus counts larger promised inflows of aid, it also allows exclusively measuring actual, documented inflows. For similar reasons, I only include aid given to specific countries and thus exclude regional aid flows that cannot be assigned to one country. Using AidData comes with a further benefit: compared to non-oil developing countries with full national accounts data in the Penn World Tables and aid flows recorded by the OECD, AidData allows adding 76 observations (10 % of the sample). Many of these observations stretch further back in time, yielding a sample that is more balanced in terms of time, geography and level of development than previous datasets. This means more data from countries such as Congo (Brazzaville), Djibouti, Fiji and Vietnam, as well as substantially more data from small countries, including Dominica, Guyana, Kiribati, Micronesia and Sao Tomé and Principe. The use of a large sample size including more countries than previous studies alleviates the inherent problem in much literature that the inclusion of countries in 5 datasets is not random (Hollyer, Rosendorff and Vreeland, 2011; Bjerg, Bjørnskov and Holm, 2011). Indeed, Roodman (2007) finds that the major source of fragility in the aid effectiveness literature is sample expansion, which in many cases renders previous results insignificant (cf. Easterly, Levine and Roodman, 2004). By allowing a large and more diverse dataset, using this source is a priori also likely to yield more reliable results. As a first example of the potential insights from using AidData, the database enables researchers to assess donors administrative loss from delivering foreign aid. Although many commentators and several politicians claim that a major part of aid disbursements is lost in administration in donor countries before being disbursed, the reported data suggest otherwise. Administrative costs have almost entirely been reported after 2000, but the available evidence shows that in the total AidData database, only.32% of total reported aid since 2000 consists of administrative costs. Only 14 country-year observations on administrative costs are above 2% of total aid to the country in a given year, and only two of those are not small island states. 2 However, these costs clearly do not necessarily relate to the administrative costs borne by recipient governments and organizations. Apart from administrative costs, other purpose codes vary considerably. Disaster prevention and preparedness, support to women, and support to NGOs and civil society are the smallest posts. The major posts are general budget support, agriculture, forestry and fishery, transport and storage and the general other category; the average country within the sample period received 2.2% of GDP as foreign aid in one of these four categories. Yet, while the smallest disbursement categories remain the same, the 2 The 14 country observations are the Cook Islands in 2000 (2.88%), Equatorial Guinea in 2004 (2.28%), French Polynesia in 2004 (5.47%), Mayotte in 2005 (34.36%), Mozambique in 2001 (2.12%), Sao Tomé and Principe in 2005 and 2006 (2.18% and 2.02%), the Seychelles in 2004 and 2005 (8.26% and 4.35%), Saint Lucia in 2004 and 2006 (3.54% and 3.27%), Trinidad and Tobago in 2004 (3.12%), Turks and Caicos Islands in 2003 (9.39%) and Vanuatu in 2004 (2.58%). The observation from Mayotte is almost certainly a reporting error since adjacent years do not exhibit particular administrative costs. 6 largest change over time, suggesting that even in a purely dynamic perspective, aid types can be separated. Having total aid disbursements separated into purpose codes allows testing one additional potential source of effect heterogeneity. In the following, I also include the Herfindahl-Hirschmann index of aid in the 24 categories, which effectively measures the degree of concentration of aid to one purpose. This also to some extent captures the problems of aid proliferation, although across purposes instead of donors. As Kimura, Mori and Sawada (2012) show that the bureaucratic and administrative difficulties increase with the number of donors to report to, a similar problem is likely to occur when aid is spread across more purposes, each of which necessitates its own reports (Moss, Pettersson and van de Walle, 2007). While AidData represents an improvement over existing data sources, reporting nonetheless remains a potential problem. Holding total aid disbursements in AidData up against commitments of net official development assistance (ODA) data from the World Bank (2011), i.e. comparing actual, documented flows with official promises including concessional loans, there is an average discrepancy of 4.8% of GNI. In other words, a comparison suggests that AidData on average may underestimate aid inflows by about one half, if the World Bank had accurately reported official aid disbursements. One should, however, be careful of interpreting the difference as pure underreporting since the standard data from the World Bank and OECD not only report commitments instead of disbursements, but also includes concessional loans with a grant element of up to 25% and debt relief in the concept of net ODA. According to World Development Indicators, the average grant element of new loans extended to developing countries since 1980 has been approximately 38%, which is likely to substantially inflate OECD aid inflows (World Bank, 2011). These data first enter into a dimensionality analysis which informs of how one can separate different types of aid in a statistically valid way. In the following, I also report and use the total 7 aggregated aid inflows as well as aid separated according to the typology in Clemens et al. (2012). When separated, I use the data in a set of standard growth regressions Control variables In the following, all data are aggregated into five-year periods in order to avoid noise, spuriously cointegrated relations occurring in annual data, and identification from business cycles and random reporting errors. The seven periods are , , , , , , and The data in these five-year periods form an almost balanced panel with 110 non-oil developing countries and 753 observations with full data. I employ two different sets of control variables: one adds a set of variables used by most studies of growth in developing countries while the other is restricted to the most basic factors. The problem, and the reason for using both, is that several variables standard in the growth literature could proxy for transmission mechanisms connecting foreign aid and growth (cf. Hodler and Knight, 2012). Examples include the budget balance, added by Rajan and Subramanian (2008), which Remmer (2004) suggests is adversely affected by aid inflows; institutional quality, a standard correlate that a series of studies show is negatively associated with aid (Knack, 2001, 2004; Djankov, Montalvo and Reynal-Querol, 2008); political instability that Licht (2010) suggests may be reduced by aid inflows to dictatorships (Nielsen et al., 2011; Savun and Tirone, 2011); and inflation and terms of trade, both of which are associated with the Dutch Disease phenomenon (Doucouliagos and Paldam, 2008; Rajan and Subramanian, 2011). The simple set of control variables thus only includes the initial logarithm to PPP-adjusted GDP per capita, openness to trade, and disasters (per million inhabitants); the former two are from the Penn World Tables, mark 7 (Heston, Summers and Aten, 2011) while the latter is the number of major natural disasters per one million inhabitants within each five-year period, which derives from EM-DAT (2012). In all specifications, I also add a full set of period fixed effects. 8 Contrary, the variables in the full set include government expenditures as percent of GDP, the investment rate, both measured as the GDP share of total trade and population growth, all from the Penn World Tables, mark 7 (Heston, Summers and Aten, 2011). I also add life expectancy at birth, from World Bank (2011), the dichotomous democracy indicator developed by Cheibub, Gandhi and Vreeland (2010), and the number of coups and confirmed coup attempts, taken from Marshall and Marshall (2009), both of which proxy for differences in institutional quality and political instability across countries. The controls thus capture convergence effects, and most other broadly important factors. In total, I estimate growth rates of country i in period t, GR ti, with a vector of common control variables X ti and a set of additional variables Z ti that together make up the full specification in 1). A ti is either total aid or vectors of types of aid following either Clemens et al. (2012) or the typology developed in the following; D i is a full set of period dummies, F t are country fixed effects and e ti is a noise term. GR ti = a + b X ti + c Z ti + d A ti + f D i +g F t + e ti (1) I handle potential endogeneity problems by supplementing a set of GLS regressions with fixed time and country effects with a further set of GMM estimates (Blundell and Bond, 1998) as implemented in Stata by Roodman (2009). In these regressions, additional instruments include country voting patterns from the United Nations General Assembly and whether or not countries are enrolled in the Highly Indebted Poor Countries program (HIPC). The former set includes the shares of all votes within a period in which the country voted with the US or the Soviet Union / Russia and, respectively (Voeten and Merdzanovic, 2009). This choice is dictated by a series of studies showing that aid flows are affected by countries voting patterns and influence in the Security Council (Dreher, Nunnenkamp and Thiele, 2008a, Dreher, Sturm and Vreeland, 2009; Kegley and Hook, 1991; Kuziemko and Werker, 2006). 3. Separating types of foreign aid 9 The aid literature includes several studies that disaggregate aid flows into different types. All are based on some form of intuitive theoretical argument. Although ignored until recently, Morgenthau (1962, 301) starts this literature by arguing that aid should be divided into six types: humanitarian aid, subsistence aid, military aid, bribery, prestige foreign aid, and foreign aid for economic development. Morgenthau s claim is that only humanitarian aid is non-political while subsistence aid may prevent development catastrophes and the type intended to produce development in a developing country may vary with the particular needs and the benevolence of the gove
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks