Portal de Periódicos da CAPES

Aiding Middle-income Countries? The Case of Spain

2008; Taylor & Francis; Volume: 36; Issue: 4 Linguagem: Inglês

10.1080/13600810802455104

1469-9966

Sergio Tezanos Vázquez,

Religion, Society, and Development

Abstract The geographical allocation of Spanish aid has been little studied, despite its unusual concentration on middle-income countries. This paper develops a theoretical model in which aid allocation depends on a combination of recipient needs, donor interests and performance criteria, and estimates it econometrically for Spain. The results show that the allocation of Spanish aid has been influenced both by Spain's own foreign policy interests and by recipient needs for poverty reduction and development (although not by the quality of recipient governance or recipient absorptive capacity). Former Spanish colonies received a disproportionate share of Spain's aid (as is true mutatis mutandis for other European countries), but aid is allocated among them with greater regard to recipient need than is Spain's aid to other developing countries. Notes I would like to thank especially José Antonio Alonso and Valpy FitzGerald for their careful assistance in this piece of research; also José Félix Tezanos, Juan Manuel Moreno, Pedro José Gómez, Paul Mosley, Adrian Wood, John Toye, Christopher Adam, Howard White, Oliver Morrisey, Rogelio Madrueño and two anonymous referees for their comments. I also gratefully acknowledge the financial support of the Spanish Ministry of Education and the institutional support of the Instituto Complutense de Estudios Internacionales (ICEI) and Queen Elizabeth House (University of Oxford). The views expressed in this paper, however, remain solely those of the author. Of course, the usual caveats apply. 1 McGillivray & White (Citation1993), Tarp et al. (1999), McGillivray (Citation2003a), Jones et al. (Citation2005) and Tezanos (Citation2008) review the economic contributions to the studies of aid allocation. 2 Alesina and Weder measured the historical links by means of the number of years that the developing countries were colonies of the metropolises during the 20th Century, thus excluding all Spanish colonies. They also used a variable of political alliances (the frequency of cases in which the receiving country voted in the UN in the same way as the donor) that could not be used in the case of Spain owing to a lack of information. 3 Isopi and Mavrotas used the World Bank's Operations and Evaluations Department calculations of the aid projects' rates of return, assuming that the levels of effectiveness obtained by the bilateral donors analysed are identical to those of the World Bank (inter alia effectiveness). 4 Other theoretical models that followed Dudley and Montmarquette are: Trumbull & Wall (Citation1994), Tarp et al. (Citation1999), Feeny & McGillivray (Citation2002) and Feeny (Citation2003). Based on this model, several empirical applications have been carried out, using increasingly sophisticated econometric techniques, from the initial regression analyses with cross-section data (see, e.g. Levitt, Citation1968; McKinlay & Little, Citation1977; Maizels & Nissanke, Citation1984; Alonso, Citation1999), to the most complex panel data models with limited dependent variables (such as Tarp et al., Citation1999; Berthélemy & Tichit, Citation2002; Alesina & Weder, Citation2002; Neumayer, Citation2003; Isopi & Mavrotas, Citation2006). 5 Only Tarp et al. (Citation1999) developed a theoretical model adapted to the singular characteristics of the donor analysed: the Danish state. 6 Neither Alonso (Citation1999) nor Sánchez Alcázar (Citation1999) considered in their studies the censored nature of aid. 7 There are also, however, extra-budgetary items, such as debt forgiveness, which is internationally co-ordinated. 8 It should be recalled that the Spanish ODA/GNI ratio has been increasing in the last two decades, from 0.08% in 1986–87 to 0.27% in 2005. Moreover, there is a recent political commitment that determines the aid budget: achieving an ODA/GNI ratio of 0.5% by 2008 (as foreseen by the Aid Plan) and of 0.7% by 2012. 9 There are also aid resources committed to finance horizontal co-operation strategies that cannot be geographically allocated to specific recipient countries. 10 See Tezanos (Citation2007, pp. 8–10) for further explanations on the Spanish multilateral-bilateral trade-off, and the different patterns of geographical aid allocation between multilateral and Spanish ODA. 11 Donor countries, however, can make voluntary subscriptions to multilateral institutions. They can also voluntarily contribute to funds and programmes, which are recorded by the DAC as "multi-bilateral aid"—therefore, as the recipient countries are identifiable these resources are included in thiss analysis. 12 Nevertheless, it is possible that Spain conceives of the pattern of multilateral ODA allocation as complementary to its own geographical preferences. In this way, the large share of Spanish assistance received by Latin America would compensate for the lower attention received by this region on the part of the multilateral co-operation. 13 The use of a "selection threshold" follows the approach of Tarp et al. (Citation1999). 14 The existence of decreasing marginal returns guarantees that the donor will not concentrate all its resources on one recipient: the one with the highest score in the attraction index. 15 We could also define cross-elasticities so as to reflect the fact that the allocation to a particular partner country does not depend only on its RN–DI scores, but also on the scores of the K – 1 remaining recipients. For reasons of simplicity, the model considers only the direct elasticities indicated in equation (Equation17). 16 That is, there is not an a priori reason for the parameters of these two equations to be the same. 17 Good reviews on aid effectiveness literature can be found in Alonso (Citation2003) and McGillivray (Citation2003b). 18 Policy statement by DAC aid ministers and heads of aid agencies on development co-operation in the 1990s, reprinted in the 1989 DAC Development Co-operation Report, OECD (1989). Available at: http://www.oecd.org/LongAbstract/0,2546,en_2649_34435_2755285_119814_1_1_1,00.html. 19 See, among others, the studies of Dudley & Montmarquette (Citation1976), Lensink & White (Citation2001) and Hansen & Tarp (Citation2000). 20 Most of these countries already have in situ technical co-operation offices and country strategic plans. 21 In this respect, the current Aid Plan is committed to making progress in aid planning and management procedures, based on previous results, in order to increase the effectiveness levels. 22 Three alternative econometric models have been used previously in aid allocation analysis: the Tobit model; the type 2 Tobit model (Heckman or sample selection model); and the two-part model. Neumayer (Citation2003) offers a good review of the econometrics of these models within the context of aid allocation analysis. 23 See Tezanos (Citation2007, pp. 17–18) for further explanations on the selection of the econometric model. 24 Specifically, equation (Equation16) is estimated by means of a logit regression model. As there is not an easy routine implemented in STATA 9.2 for logit (unconditional) fixed-effect estimation, it uses a random-effects model. 25 In accordance with the results of the Hausman specification tests, the allocation equation for the ex-colonial countries is estimated by means of a fixed-effects panel data model and the equation for countries without historical links uses a random-effects model (results available upon request). 26 This fact explains the inconsistencies between the Spanish geographical priorities defined in the Aid Plan and the aid eventually disbursed, e.g. there is a number of "preferential countries" that have received negative net disbursements, such as Mexico in the last 7 years. 27 Especially notable were the debt relief of Guatemala in 2001, and Iraq, Madagascar and the Republic of Congo in 2005, which turned these countries into the main recipients of Spanish ODA. 28 Different threshold values change the probability of being selected as an aid partner. However, they do not considerably affect the magnitudes and signs of the estimated parameters. Therefore, the model remains consistent. 29 As the model is specified in natural logarithms (both the dependent variable and the independent variables), it thus facilitates the interpretation of the coefficients in terms of elasticities. 30 The use of the infant mortality rate raises serious concerns, as complete time series data are not available, only 5-year values. 31 High rates of ODA/GNI may stem from a "bandwagon effect" among donors' allocations; however, this variable in the case of Spanish aid is not significantly correlated with the ODA received by the rest of the donors (r 2 = − 0.0304), ruling out the existence of a simultaneity problem. 32 In the Spanish context, there has been a high year-by-year variation in aid quotas, in contrast to the relative stability of the list of partner countries. In fact, the average coefficient of variation of these quotas between 1993 and 2005 was 0.655, i.e. on average the inter-annual variation of a partner country's quota was 65.5%. In this sense, the aid's inertia has been especially important in the selection stage, but not as much in the aid-quota stage, which varies considerably year by year. Author calculations with OECD: DAC (Citation2007) data: developing countries' quotas on Spanish ODA gross disbursements. 33 The only exception is the POLITY2, which is expressed in its original rank units, as it is not amenable to reasonable interpretations in terms of elasticities. 34 In fact, the Aid Plan explicitly points out that less-developed countries will be identified by means of the socio-economic indicators elaborated by international organizations. 35 The share on global ODA is not lagged. The specification test pointed out that the Spanish aid allocation is specially related to the current year aid disbursements of the rest of the donors; moreover, the estimation results do not change upon the number of lags included in this variable. 36 Formally, when variable x i increases one unit, ceteris paribus, the odds ratio is multiplied by a factor equal to e xi . 37 Cuba is not included in the analysis owing to a lack of information. 38 Particularly: Argentina (1999–2001), Colombia (1993), Costa Rica (1992–95, 1998, 2001, 2004 and 2005), Cuba (1992–94), Chile (1997–2005), Dominican Republic (1992–94), El Salvador (1992–94), Equatorial Guinea (2003), Guatemala (1992–94), Honduras (1992), Mexico (1996, 1997, 1999–2001, 2003 and 2004), Nicaragua (1992), Panama (1992–95 and 2002–05), Paraguay (1992–94, 1996, 2000, 2002, 2004 and 2005), Peru (1992–94), Philippines (1992 and 1996), Uruguay (1996–2002 and 2004) and Venezuela (1992–95, 1998, 2004 and 2005). 39 As this estimation stage only analyses whether a developing country is chosen or not as an aid partner, it does not consider the amount of resources finally disbursed. 40 Results available upon request. 41 As the binary variable of colonial past does not vary over time, the prediction errors stem from the time changes in the rest of the explanatory variables, mainly in the previous year's Spanish aid quota (which is the variable that exerts the greatest influence in the selection process, after colonial past). 42 In 1998, Spanish aid to Zimbabwe was especially high due to the financing of two NGO projects that amounted to €504.779 (i.e. 78% of the aid). 43 These four especially influential outliers, however, do not significantly affect the estimates (results available upon request). 44 Spain's ex-colonial countries are: Argentina, Bolivia, Colombia, Costa Rica, Chile, Dominican Republic, Ecuador, El Salvador, Equatorial Guinea, Guatemala, Honduras, México, Morocco, Nicaragua, Panama, Paraguay, Peru, Philippines, Uruguay and Venezuela. 45 Calculations based on accumulated 1993–2005 aid gross disbursements—net of emergency aid and debt relief. 46 Equatorial Guinea, however, has gradually reduced its aid-dependency ratio since the late 1990s. 47 See Tezanos (Citation2007, pp. 35–38) for further explanations on the outliers. 48 See the studies on aid fungibility of Feyzioglu et al. (Citation1998), Devarajan & Swaroop (Citation1998) and Pack & Pack (Citation2003). 49 Lahiri & Raimondos-Møller (Citation2000) developed and aid allocation model focusing on the influence exerted by the different immigrant nationalities present in the donor country. Nevertheless, in the case of Spain, the information that facilitates the State Secretary of Immigration and Emigration does not offer complete time series data on the immigrants' countries of origin, which is limited to the most recent years.