JES 37,4. Sunielle Stanford Research Department, Central Bank of Barbados, Bridgetown, Barbados

Transcription

1 The current issue and full text archive of this journal is available at JES 438 Received 24 November 2008 Accepted 1 July 2009 Why do some countries have a long-term dependence on food aid? Winston Moore Department of Economics, Cave Hill Campus, University of the West Indies, Bridgetown, Barbados, and Sunielle Stanford Research Department, Central Bank of Barbados, Bridgetown, Barbados Journal of Economic Studies Vol. 37 No. 4, 2010 pp q Emerald Group Publishing Limited DOI / Abstract Purpose The purpose of this paper is to investigate the main determinants of dependency on food aid programmes. Food aid may take the form of a loan, a sale below market price or a current transfer. Food aid programmes across the globe are generally designed to provide short-term assistance to countries. However, many countries have developed a dependence on food aid assistance. Design/methodology/approach The paper estimates cross-sectional regressions using a database containing 116 developing countries over the period Bayesian averaging of classical estimates is employed to identify the robust determinants of dependency. Findings The study finds that the top two determinants of food aid dependence (both cereals and non-cereals) were food production and the frequency of droughts affecting the country. Food inflation, population density, crop yields, the amount of arable land per capita, the rule of law and the number of armed conflicts were also robustly related to aid dependence. Practical implications These findings suggest that international donors should focus primarily on offsetting the substitution effect of aid on local production as well as implement systems to partially offset the negative effects of droughts, if they are to break the cycle of dependence on food aid. Originality/value Previous studies have investigated the effects of a large number of other variables. The findings from these papers were somewhat conflicting due to differences in model specification. To identify the variables that have a robust relationship with food aid dependence, whatever the model specification, the Bayesian averaging of classical estimates approach proposed by Sala-i-Martin et al. is employed. Keywords International aid, International economics Paper type Research paper Introduction The food available to feed a nation originates from domestic production, imports or aid inflows from donor countries and organisations. Food aid can be distributed freely, in the form of a loan or a sale below market price. While there are various motives for giving food aid, the primary goal is to provide short-term assistance to recipient countries. Over time, it is expected that the development process of these countries would be enhanced, thereby reducing the long-term dependency on aid. However, this has not been the case. Table I lists the top ten food aid receiving countries (evaluated using food aid per capita) for three decades: 1970s, 1980s, 1990s, and the period JEL classification F35; O2; C23

2 s 1980s 1970s Armenia Armenia Democratic Republic of Congo Eritrea Armenia Democratic Republic of Congo Democratic Republic of Congo Sudan Haiti Armenia Democratic Republic of Congo Eritrea Haiti Jordan Pakistan Honduras Pakistan Gabon Haiti Pakistan Honduras Haiti Morocco Kenya Jordan Peru Sri Lanka Vietnam Sudan Sudan Tunisia India Sri Lanka Honduras Honduras Sri Lanka Gabon Israel Peru Republic of Korea Long-term dependence on food aid? 439 Table I. Countries with the highest dependence on food aid The table shows a remarkable consistency in terms of the countries on the list. For example, Armenia, the Democratic Republic of Congo, Haiti, Eritrea, Sudan and Honduras have been in the top ten in each decade. Food aid can improve a nation s food security but at the same time, a heavy dependence on this type of aid over time, can lead to economic and social hardships for countries. In fact, critics argue that one of the drawbacks to food aid is its negative (disincentive) impact on local food production (Schultz, 1960). The hypothesis is that food aid, by boosting supplies of cereals, depresses local prices and reduces the returns to local production and thereby limiting the opportunities for producers to make necessary investments in new equipment and techniques (see Tschirley et al. (1996) for empirical evidence from Mozambique). As farmers reduce supply, this creates demand for food aid shipments even after the initial shock. Ocran and Biekpe (2008) also notes that in developing countries that only the producers of export commodities may be responsive to price incentives or support. Knack (2001) argues that this dependence can undermine the quality of governance and public institutions by weakening accountability, encouraging corruption and alleviating pressures to reform inefficient policies and institutions. Hasan (2002), using a database of 27 developing countries, also finds that there is causal independence between aid, domestic savings and income. From a policy perspective, it is first important to determine the social, economic, political and environmental factors that contribute to aid dependency. Once these aspects have been identified, only then, can policy makers adequately address the issue of aid dependency. While there are several studies examining the issue of food availability and food aid dependency, these studies have not utilised an exhaustive list of explanatory variables and additionally, the studies tend to narrow the research to certain regions. In this study, the authors utilise a dataset of 116 developing countries and 49 independent variables in order to identify the most important determinants of food aid dependence. The remainder of this paper is organised as follows: Section 2 reviews the conceptual and empirical literature on food aid dependence. Section 3 presents data, methodology and outlines the findings from the research, while Section examines the robustness of these results. Finally, the conclusions and policy implications of the research are discussed in Section 5.

3 JES Food aid dependence: conceptual and empirical issues The notion of food aid dependency is intimately related to the belief that there is a disincentive effect of food aid on local production. Schultz (1960), one of the earliest studies to investigate these disincentive effects, contends that since food aid increases local food supply, it depresses prices and as a result discourages local production. The author also maintains that food aid may create policy disincentives. If the problem underlying food insecurity is poor policies or the distribution of resources, food aid can allow governments to postpone or avoid politically difficult reforms. In this way food aid can increase a country s dependence on food transfers. The evidence to support the hypothesis that there is a disincentive effect of food aid on production has not been supported by recent empirical studies in the literature. Lavy (1990) estimates panel vector autoregression models for 33 Sub-Saharan countries for to investigate the causal relationship between production and aid. The author accepts the notion that food aid Granger causes food output. In addition, food aid had positive net effects on food production, offsetting any negative disincentive effects from the additional supply of food. Similar results are obtained by Bezuneh et al. (2003), Barrett et al. (1999) and Tschirley et al. (1996). Abdulai et al. (2005) instead use household-level observations for rural Ethiopia to investigate whether there is a disincentive effect of food aid on domestic food production. The authors employ discrete choice panel data techniques to investigate the factors that influence the amount of labour time supplied to agriculture by households. The explanatory variables in the model include a dummy variable that takes a value 1 if the household receives food aid and zero otherwise, as well as various other village and household level characteristics. The study finds that once one accounts for the endogeneity of food aid receipts, there is no evidence of any disincentive effects of food aid. Moreover, Abdulai, Barrett and Hoddinott show, through the use of panel granger causality techniques, that food aid has a statistically significant impact on food production. Food aid dependency may also occur if the needs of the recipient countries come second to those of the donor nation. The motives for food aid can generally be grouped according to economic, moral and humanitarian, environmental and national security reasons. Despite the Food Aid Convention (FAC) guidelines indicating that food aid donations should benefit the needs of recipient countries, this has not always been the reality. Clay (2002), for example, posits that emergency aid (a subcategory of food aid) is intrinsically political while Diven (2001) highlights that the initial US 1954 legislation on food aid was more in line with donor interest. However, by 1974 the focus was geared more towards recipient needs. Research by Cathie (1997) finds that food aid from the European Union (EU) has generally been focused on the needs of recipients. Such assertions have, in part, contributed to the vast number of studies that empirically investigate whether the determinants of food aid reflect donor interests or recipient needs. The results are mixed and some of these studies where the majority of studies dependent variable is based on aid distributed by the USA are outlined below. Eggleston (1987) focuses on the allocation of US food aid to 64 countries for the period The author utilises pooled samples, cross-sectional samples and time series samples for some individual countries. Eggleston s results indicate that donor allocation is influenced by both recipient needs and donor interests (such as, US

4 political and military interests). Similarly, research conducted by Shapouri and Missiaen (1990) for the period , found that both recipient needs and donor interests were important factors in determining food aid allocation. In comparison, Ball and Johnson (1996) and Zahariadis et al. (2000) approached the topic by disaggregating food aid into three major components, that is, Titles I-III. Specifically, the researchers examine whether the different titles under which the program is administered have been driven by different objectives and whether the motivations underlying food aid shipments have changed over time (p. 515). Nine pooled time series and cross section regression were estimated for a selection of African countries over the period The findings indicate that donor interests, in the form of arms exports and voting similarly at the United Nations, were important determinants for Title I aid, whereas recipient needs proved to be important determinants for Title II aid. Additionally, in the 1970s, levels of food aid were allocated according to donor interest, however, this position was reversed a decade later. Zahariadis et al. (2000) expand their research beyond Title I and Title II categories of aid, and examine the factors which would determine the allocation of Titles I and II types of food aid using a two-part model. In the first stage, also referred to as the gate-keeping stage, donors decide whether or not countries will be eligible for aid and in the second stage, the amount of aid to be allocated to countries that pass stage one is then decided. In the two-part model, each stage is treated independently, and thus errors in each stage are assumed to be uncorrelated. The authors find that for Title I aid, recipient needs and donor interest variables, a proxy for trade interests with the US, were important factors for countries to become eligible for food allocation. On the other hand, in the second stage of food allocation, donor interest was found to be an insignificant factor in determining the amount of Title II food aid distributed to eligible countries. More recently, Neumayer (2005) joins the debate and examines whether donor interest impacts on aid distribution. Unlike the abovementioned studies, the author individually estimates the impact of donor distributions from the USA, EU, the United Nations World Food Programme (WFP) and non-governmental organizations (NGOs), using data from the 1990s. In addition, the author separately estimates the effects for total aid and emergency aid. The dependent variable, food aid, is measured in tons of wheat. For the explanatory variables, Neumayer notes the importance of using variables that capture both donor interests and recipient needs in order to avoid the possibility of omitted variable bias. Thus, independent variables to proxy donor interests and recipients needs are: real per capita GDP (PPP), population, average daily per capita calorie supply, an index of food self-sufficiency, number of refugees, number of years of colonial ties between donor and recipient, geographical distance also between donor and recipient, a variable measuring US military aid, share of food exports, a variable representative of converging political viewpoints and a proxy for cultural similarity. The researcher uses a two-part model the first stage is estimated using a probit model and the second stage using generalized estimating equations (GEE). The results across the four various donors suggest some evidence of donor interest bias. However, it should be noted that because the analysis only relied on data in the 1990s and measured food aid only using wheat, different results may have been obtained were a longer time sample used and non-cereal aid included in the analysis. Apart from studies that rely solely on US food allocation, Shapouri and Missiaen (1990) find donor interests and recipient needs to be an important determinant for Long-term dependence on food aid? 441

5 JES 442 Canadian and EU food aid distribution, while Herrmann et al. (1992) reports that recipient needs are of concerns for EU donors. 3. Modelling food aid dependency 3.1 Data In order to calculate the dependent variable food aid dependence (FID), the authors use data on food aid shipments to recipient countries from all donors as reported by the Food and Agriculture Organisation (FAO) of the United Nations. Two estimates of food aid shipments are employed: cereals and non-cereals. As all these measures are in metric tons, the variables are scaled by the country s total cereal production, to arrive at a measure of food aid dependency. The resulting variables are denoted by faspcc and faspcnc for cereals and non-cereals, respectively. The recurring use of food aid may be based on a country s structural economic characteristics, political and institutional environment, geographic conditions and the frequency of natural disasters. Poorer countries are more likely to need food aid because of their lack of access to private capital markets. The income variables used in the analysis are gross national income and gross domestic product from the World Bank s World Development Indicators 2005 (WDI2005), PPP-adjusted international dollars and divided by population. The other structural economic characteristics, also taken from WDI2005, include the age dependency ratio, proportion of the population below 15 years, proportion of the population above 65, life expectancy, the mortality rate, population, population density, food inflation, inflation, general government final consumption, literacy rate, cereal production and crop yields. The dependency ratio, life expectancy, population density, food inflation and the general level of price inflation are anticipated to be positively related to aid dependence. A larger proportion of the population that is not in labour force places greater pressure on farmers to maintain crop levels after a shock. High life expectancy and population density strains land resources and could, as argued by Makenete et al. (1998), require that the country make-up short-falls in food production through aid shipments. Tschirley et al. (1996) have argued that food aid can depress domestic food prices and therefore act as a disincentive for local production. Food inflation is therefore included in the regressions to capture the effects that changes in food prices have on aid dependence. The general level of price change is also included in the regressions to investigate whether high levels of overall inflation can also be a disincentive for domestic production, which then requires the country to draw on food aid shipments. The literacy rate should enable the country to adopt modern farming techniques, increasing crop yields and reducing the dependence on food aid, while cereal production and crop yields are included to capture any disincentive effects of food aid on domestic production as argued by Schultz (1960). Countries with a better governance structure may be more successful in breaking the cycle of dependence since the cohesion of a government and its political strength can affect a country s ability to implement necessary policies. The political and institutional environmental variables include the set of measures provided by Kaufmann et al. (2006). This database combines observations from various rating agencies and research organisations to develop six aggregate indicators of the quality of governance:

6 (1) voice and accountability; (2) political stability and absence of violence; (3) government effectiveness; (4) regulatory quality; (5) rule of law; and (6) control of corruption. Long-term dependence on food aid? 443 Unfortunately, these indicators are only available for a short period, , while this study employs observations from 1970 to An investigation of the indicators shows that there is little variation from year to year. The authors therefore assume that the quality of governance did not change significantly during the period under investigation. Indicators of whether or not the country is involved in armed conflicts are also included to capture the effects that these events can have on security and labour resources. Observations are taken from Armed Conflicts Version b of the Centre for the study of Civil War. The dummy variable takes the value of 2 for a war, 1 for a minor conflict and zero otherwise. Geographic conditions can also be crucial. Gallup et al. (1999) and Mellinger et al. (2000) argue that location and climate can affect economic performance and policy choices. Landlocked countries are disadvantaged due to their lack of access to ocean transport and water for crops. Also, some countries because of their location are more likely to be affected by natural disasters or other economic shocks. For example, tropical regions are more likely to have a higher incidence of diseases such as malaria. The geographic conditions indicators include dummy variables that capture whether or not the country is landlocked, tropical, longitude, latitude as well as a measure of arable land as a percentage of land area and arable land per capita. All series are taken from the WDI2005. The final group of variables account for the effects of the reoccurrence of natural disasters on food aid dependence. The paper classifies natural disasters into 11 groups: drought, extreme temperature, flood, slides, wave, wild fires, wind storm, earthquake, epidemic, insect infestation and volcano. Dummy variables are generated for each one of these events, where the occurrence of the event is denoted by 1 and the non-occurrence by zero. Observations are obtained from the Emergency Database (EM-DAT), available online at of the Centre for Research on the Epidemiology of Disasters, Université Catholique de Louvain. The database employed in this paper includes annual observations for 116 countries over the period 1970 to The countries are listed in the Appendix. 3.2 Methodology All the econometric results presented in this paper are based on variants of: ln faspcc i ¼ a þ XS b s E si þ XR b r P ri þ XQ b q G qi þ XM b m D mi þ u i s¼1 r¼1 q¼1 m¼1 ð1þ where E are the economic and structural factors, P ri are the quality of governance indicators, G qi are geographic variables and D mi are dummy variables that capture the occurrence of natural disasters. The disturbance term u it is assumed to have the usual

7 JES 444 desirable characteristics. This basic cross-section equation is first estimated by ordinary least squares (OLS) using the sample averages for both the dependent and independent variables to smooth out cyclical fluctuations, as well as for three decades: (1) 1970s; (2) 1980s; and (3) 1990s. This initial approach (also known as a baseline estimation) allows one to identify the factors that have an impact on food aid dependence based on the set of explanatory variables normally identified in aid studies. However, if a different combination of the variables is employed the inferences could change. To identify the robust determinants of food aid dependency, the authors also employ the Bayesian Averaging of Classical Estimates (BACE) approach proposed by Sala-i-Martin et al. (2004). If M j is a model of food aid dependence with a specific set of explanatory variables, the prior probability of model j is given as: PðM j Þ¼ Yk j i¼1 M ji k K! Y k j i¼1! k 1 2 M ji 1 2 ð2þ K where k j are the number of included variables in model j, k is the prior mean model size, K is the total number of potential determinants and M ji is the ith element of the vector. Assuming that each variable has an equal probability of inclusion, equation (2) can be simplified to: k PðM j Þ¼ kj k 1 2 K2kj ð3þ K K The number of potential explanatory variables, K, is 49 and the number of variables included in every model, k j, is fixed to 13. The expected coefficient estimates is the posterior mean conditional on model j,orin other words, the weighted average of OLS estimates, where the weights are proportional to a measure of goodness of fit and prior model size: PðM j j ln faspccþ ¼ X 2K i¼1 PðM jþt kj 2T 2 SSE 2 j PðM i ÞT k i 2T 2 SSE 2 i where T is the sample size and SSE is the OLS sum of squared errors. The posterior variance is calculated from the estimated variances from individual models and the variance in the estimates of the coefficients across models. 3.3 Results The results from the basic cross-country food aid regressions (the baseline estimation) are provided in Table II. Four regressions are presented: column two uses averaged data for the entire sample period, column three uses averaged data for the 1990s only, column four averaged data for the 1980s only and finally column five provides the ð4þ

8 finds from using averaged data for the 1970s only. The regressions include five structural and economic variables (gross national income per capita, food inflation, general government final consumption expenditure, population density, crop yield), governance indicators (government effectiveness and British legal origin), three geographic conditions variables (arable land per capita, tropical countries and longitude measurements) and one natural disasters variable (disasters). Table II shows that the signs of the coefficient estimates are fairly robust to changes in the sample period considered. However, the magnitude of the coefficient on each explanatory variable changes somewhat for the different sample periods. The variables gross national income per capita, food inflation, general government final consumption expenditure, population density are all positively associated with a dependence on food aid. The positive coefficient on the income variable agrees with the findings of authors such as Neumayer (2005), Alesina and Dollar (2000) Herrmann et al. (1992) and Shapouri and Missiaen (1990) who argue that donor interest are an important factor when considering the distribution of aid. As a result, aid allocations may not be reaching the poorest countries around the world since these are of little interest to donors. The perverse relationship between food aid and the country s national income was strongest during the 1970s and 1980s, the peak of the cold war. The association between national income and food aid dependence, however, weakened during the 1990s and was therefore statistically insignificant during this period. A country s population density also had a positive impact on a country s food aid dependence. A densely populated country implies that large-scale farming techniques are less likely to be employed due to the competing demands on land for housing and small family farms. In agreement with a priori expectations a higher crop yield reduces a country s dependence on food aid. The coefficient on government size was positive but statistically insignificant. Turning to the effects of the governance variables, government efficiency is inversely related to aid dependence. More efficient governments are better able to manage the aid distribution process and implement policies to reduce their dependence on food aid over time. The variable British legal origin captures the effect that a country s colonial origin has on aid giving by larger countries. Former metropoles may feel obligated to provide support to former colonies and may also capture similarity in international political views. Geographic variables also had a significant impact on aid dependence. The variable arable land per capita indicates that countries that have a smaller amount of arable land per capita are more likely to have a long-term dependence on food aid. A small amount of arable land reduces the ability of the country s farmers to produce enough for the entire country. It also pushes up the price of arable land, making small farmers more common which are less likely to be using crop rotation techniques, which increases crop yields, and also means that farmers are less able to cope to long periods of drought or other natural disasters. The other two variables capture the effect that a country s geographic location has on its aid dependence. On average, tropical countries are likely to be more dependent on food aid shipments. The relationship between these two variables seems to have strengthened particularly in the 1980s and 1990s. In contrast, the negative coefficient on the longitude variable indicates that eastern countries have a lower dependence on food aid. The final variable included in the Long-term dependence on food aid? 445

9 JES 446 Table II. Determinants of food aid dependence (cereals) s 1980s 1970s Standard errors Standard errors Coefficients Standard errors Coefficients Standard errors Coefficients Variable Coefficients GNI per capita, PPP (current international $) * Inflation, food prices (annual %) * ** General government final consumption expenditure Population density * * Cereal yield (kg per hectare) ** ** ** ** Government effectiveness * ** Legal origin: British * Land use, arable land (hectares per person) ** * Tropical Longitude Total disasters ** ** ** ** Constant ** (6.987) ** ** n Adjusted R-squared Notes: * and ** indicates significance at the 10 and 5 percent level of testing, respectively

10 regression captures the effects of naturally occurring disasters on the demand for food aid. In agreement with a priori expectations, countries that are more subject to natural disasters have a greater dependence on food aid shipments. The food aid dependence regressions are also estimated using non-cereal food aid as the dependent variable and the results are provided in Table III. In general, the results are very similar to the cereal food aid regressions: population density, crop yields, government efficiency and the number of disasters are all significant at normal levels of testing. The importance of these variables is consistent across the three decades considered. Long-term dependence on food aid? Robustness of results It is important to note that of the 116 countries in the initial sample, the BACE approach uses 72 countries based on data availability. The potential determinants of food aid dependence are listed in column one of Table IV, while the weighted mean betas and standard errors, obtained from all the regressions, are given in columns two and three, respectively. The fourth column provides the probability that the estimated coefficient is either positive or negative this measures the significance of the variable at the 5 percent level, while the fifth column remarks about the robustness of the coefficient estimates. Out of the 49 potential determinants of food aid considered, 15 are robustly correlated with aid dependence: life expectancy, population density, proportion of the population over 65 years old, literacy rate, cereal production, crop yields, arable land per capita, food inflation, drought, extreme temperature, flood, fire, rule of law, armed conflicts, East Asian regional dummy and Middle East and North Africa regional dummy. The weighted coefficient estimates for all the variables have similar signs to those obtained in Table II, however, the magnitudes of some variables changed somewhat. The mean beta for population density more than doubled, while those for food inflation, crop yields, arable land per capita and total disasters fell compared to Table II. The two top variables were cereal production and droughts. Cereal production was inversely related to food aid, reflective of the disincentive effects of food aid on domestic supply. The result that there is a negative correlation between cereal production and food aid dependency is robust at the 5 percent level of significance. As expected, the higher the frequency of drought conditions in the country the greater the dependence on food aid. Of the remaining variables that have a robust relationship with food aid dependence, life expectancy, population density, the literacy rate, food inflation and the Middle East and North African regional dummy had a positive association with aid dependence. On the other hand, the proportion of the population over 65 years old, crop yields, arable land per capita, the frequency of extreme temperatures, the frequency of flood, the frequency of fire, the rule of law, the number of armed conflicts and the East Asia and Pacific regional dummy were inversely related to aid dependence. The negative relationship between the number of armed conflicts and food aid reflects the difficulty of supply food aid to countries currently engaged in combat, while rule of law dummy captures the importance of governance in aid distribution. Most of the remaining variables were marginally (that is, their sign certainty probability is larger than 70 percent) related to aid dependency, these included the

11 JES 448 Table III. Determinants of food aid dependence (non-cereals) s 1980s 1970s Standard error Standard error Coefficient Standard error Coefficient Standard error Coefficient Variable Coefficient GNI per capita, PPP (current international $) ** Inflation, food prices (annual %) * General government final consumption expenditure Population density ** Cereal yield (kg per hectare) ** ** * * Government effectiveness ** * ** * Legal origin: British (0.547) Land use, arable land (hectares per person) ** Tropical Longitude Total disasters ** ** ** ** Constant ** n Adjusted R-squared Notes: * and ** indicates significance at the 10 and 5 percent level of testing, respectively

12 Mean Beta Mean standard error Sign certainty probability Remark Age dependency ratio Marginal Life expectancy at birth Robust Mortality rate, infant Not robust Population density Robust Population, total Marginal Population ages Marginal Population aged 65 and above Robust Literacy rate, adult total Robust Cereal production (metric tons) Robust Cereal yield (kg per hectare) Robust Land use, arable land (% of land area) Not robust Land use, arable land (hectares per person) Robust GDP per capita, PPP (constant 2000 international $) Not robust GNI per capita, PPP (current international $) Not robust Inflation, food prices (annual %) Robust Inflation, GDP deflator (annual %) Marginal General government final consumption expenditure Marginal Drought Robust Extreme temperature Robust Flood Robust Slides Marginal Wave/surge Marginal Wild fires Robust Wind storm Marginal Earthquake Not robust Epidemic Not robust Insect infestation Not robust Volcano Marginal Total disasters Marginal Voice and accountability Marginal Political stability Not robust Government effectiveness Not robust Regulatory quality Marginal Rule of law Robust Control of corruption Not robust Armed conflict Robust Landlocked Marginal Longitude Not robust Latitude Not robust Legal origin: British Not robust Tropical Not robust Region: East Asia and Pacific Robust Region: East Europe and Central Asia Not robust Region: Middle East and North Africa Robust Region: South Asia Marginal Region: West Europe Not robust Region: North America n/a Marginal Region: Sub-Saharan Africa Marginal Region: Latin America and Caribbean Not robust Long-term dependence on food aid? 449 Table IV. Robustness of the coefficients, Bayesian averaging of classical estimates (cereals)

13 JES 450 dependency ratio (positive), population (negative), the proportion of the population below 15 years old (positive), inflation (positive), government consumption (positive), land slides (negative), surges (positive), storms (negative), volcanoes (negative), total disasters (positive), voice and accountability (positive), regulatory quality (negative), land locked (negative), South Asia regional dummy (positive), North America regional dummy (positive) and the Sub-Saharan regional dummy (negative). The remaining variables show little evidence of robust correlation to aid dependence. It is interesting to note that some economic variables, such as GDP per capita and GNI per capita, and political variables such as, political stability, government efficiency and corruption, are not robustly related to aid dependence. In addition, geographic variables, such as longitude, latitude, land locked and tropical had little or no influence on food aid dependence. The results presented so far in this section focus primarily on cereal aid, however, it is also important to assess non-cereal food aid to identify any differences in the factors that may explain long-term dependency. The results for non-cereal aid, presented in Table V, are quite similar to those for cereal aid. The two top factors are cereal production and frequency of droughts. In terms of the robust variables, arable land per capita, food inflation and floods are now only marginally correlated with dependence. In contrast, the total number of disasters, government efficiency and the rule of law are robust determinants of non-cereal aid dependence. These findings suggest that political factors are more important for non-cereal aid than for cereal aid. 5. Conclusions and policy implications This study presents an empirical assessment of the factors that influences a country s long-term dependence on food aid. A database of 116 developing countries covering the period 1970 to 2003 is employed in the analysis. The cross-sectional regressions are estimated for three decades, 1970s, 1980s, and 1990s as well as for the entire sample period. The empirical results suggested that the main determinants of food aid dependence were food inflation, population density, crop yields, arable land per capita and the frequency of disasters affecting the country. These findings were generally consistent across regressions for cereal and non-cereal aid. Besides the limited set of potential determinants of food aid dependence used in the cross-section regressions, previous studies have investigated the effects of a large number of other variables. The findings from some of these papers were somewhat conflicting due to differences in model specification. To identify the variables that have a robust relationship with food aid dependence, whatever the model specification, the Bayesian Averaging of Classical estimates approach proposed by Sala-i-Martin et al. (2004) is employed. These regressions indicated that the top two determinants of food aid dependence were cereal production and the frequency of droughts affecting the country. These findings suggest that international donors should focus primarily on offsetting the substitution effect that aid has on local production as well as implement systems to partially offset the negative effects of droughts if they are to break the cycle of dependence on food aid. Similar to the results from the basic cross-country regressions, food inflation and population density were found to have a robust positive relationship to food aid dependence. In contrast, crop yields, the amount of arable land per capita, the rule of law and the number of armed conflicts were inversely related to the ratio of food aid to

14 Mean Beta Mean standard error Sign certainty probability Remark Age dependency ratio Marginal Life expectancy at birth Robust Mortality rate, infant Not robust Population density Robust Population, total Marginal Population ages Marginal Population aged 65 and above Robust Literacy rate, adult total Robust Cereal production (metric tons) Robust Cereal yield (kg per hectare) Robust Land use, arable land (% of land area) Not robust Land use, arable land (hectares per person) Marginal GDP per capita, PPP (constant 2000 international $) Not robust GNI per capita, PPP (current international $) Marginal Inflation, food prices (annual %) Marginal Inflation, GDP deflator (annual %) Marginal General government final consumption expenditure Marginal Drought Robust Extreme temperature Not robust Flood Marginal Slides Robust Wave/surge Marginal Wild fires Robust Wind storm Marginal Earthquake Not robust Epidemic Not robust Insect infestation Not robust Volcano Marginal Total disasters Robust Voice and accountability Marginal Political stability Marginal Government effectiveness Robust Regulatory quality Marginal Rule of law Robust Control of corruption Not robust Armed conflict Robust Landlocked Marginal Longitude Marginal Latitude Not robust Legal origin: British Marginal Tropical Not robust Region: East Asia and Pacific Marginal Region: East Europe and Central Asia Marginal Region: Middle East and North Africa Marginal Region: South Asia Marginal Region: West Europe Marginal Region: North America n/a Marginal Region: Sub-Saharan Africa Marginal Region: Latin America and Caribbean Not robust Long-term dependence on food aid? 451 Table V. Robustness of the coefficients, Bayesian averaging of classical estimates (non-cereals)

15 JES 452 total domestic production. All four of these variables can be affected by domestic policies. Crop yields can be influenced through agricultural research aimed at developing crops to suit the particular local climatic conditions, land reform, land use policies, education and water resources management; these policies should also assist in addressing the problem of arable land per capita, possible in conjunction with family planning policies. The rule of law is one of the indicators of quality of governance. A government which provides incentives to agricultural producers, whether price or non-price incentives, can reduce domestic policy distortions in the food market, and thereby reducing the need for food aid imports. The importance of the number of armed conflicts reflect the difficulty of ensuring that aid reaches the neediest individuals, suggesting that aid to these countries should be done in conjunction with measures to sure-up the country s security situation. References Abdulai, A., Barrett, C.B. and Hoddinott, J. (2005), Does food aid really have disincentive effects? New evidence from Sub-Saharan Africa, World Development, Vol. 33 No. 10, pp Alesina, A. and Dollar, D. (2000), Who gives foreign aid to whom and why?, Journal of Economic Growth, Vol. 5 No. 1, pp Ball, R. and Johnson, C. (1996), Political, economic, and humanitarian motivations for PL 480 food aid: evidence from Africa, Economic Development and Cultural Change, Vol. 44 No. 2, pp Barrett, C.B., Mohapatra, S. and Snyder, D.L. (1999), The dynamic effects of US food aid, Economic Inquiry, Vol. 37 No. 4, pp Bezuneh, M., Deaton, B. and Zuhair, S. (2003), Food aid disincentives: the Tunisian experience, Review of Development Economics, Vol. 7 No. 4, pp Cathie, J. (1997), European Food Aid Policy, Ashgate, Aldershot. Clay, E. (2002), Forty years of multilateral food aid: responding to changing realities, book review article, Development Policy Review, Vol. 20 No. 2, pp Diven, P.J. (2001), The domestic determinants of US food aid policy, Food Policy, Vol. 26 No. 5, pp Eggleston, R.C. (1987), Determinants of the levels and distribution of PL 480 food aid: , World Development, Vol. 15 No. 6, pp Gallup, J.L., Sachs, J.D. and Mellinger, A.D. (1999), Geography and economic development, in Pleskovic, B. and Stiglitz, J.E. (Eds), Annual World Bank Conference on Development Economics 1998, World Bank, Washington, DC, pp Hasan, M.S. (2002), Concessional foreign capital inflows and domestic savings across countries: dependency hypothesis re-visited, Journal of Economic Studies, Vol. 29 No. 6, pp Herrmann, R., Prinz, C. and Schenck, P. (1992), How food aid affects food trade and how food trade matters to the international allocation of food aid, Journal of Economic Development, Vol. 17 No. 1, pp Kaufmann, D., Kraay, A. and Mastruzzi, M. (2006), Governance matters V: Governance indicators for , World Bank Policy Research, September. Knack, S. (2001), Aid dependence and the quality of governance: cross-country empirical tests, Southern Economic Journal, Vol. 68 No. 2, pp

16 Lavy, V. (1990), Does food aid depress food production? The disincentive dilemma in the African context, Policy Research Working Paper Series No. 460, World Bank, Washington, DC. Makenete, A., Ortmann, G. and Darroch, M. (1998), Food-aid dependency in Lesotho: issues and policy implications, Development South Africa, Vol. 15 No. 2, pp Mellinger, A.D., Sachs, J.D. and Gallup, J.L. (2000), Climate, coastal proximity, and development, in Clark, G.L., Feldman, M.P. and Gertler, M.S. (Eds), Oxford Handbook of Economic Geography, Oxford University Press, Oxford and New York, NY, pp Neumayer, E. (2005), Is the allocation of food aid free from donor interest bias?, Journal of Development Studies, Vol. 41 No. 3, pp Ocran, M.K. and Biekpe, N. (2008), Agricultural commodity supply response in Ghana, Journal of Economic Studies, Vol. 35 No. 3, pp Sala-i-Martin, X., Doppelhofer, G. and Miller, R. (2004), Determinants of long-term growth: a Bayesian averaging of classical estimates (BACE) approach, American Economic Review, Vol. 94 No. 4, pp Schultz, T.W. (1960), Impact and implication of foreign surplus disposal on underdeveloped economies: value of US farm surpluses to underdeveloped countries, Journal of Farm Economics, Vol. 42 No. 5, pp Shapouri, S. and Missiaen, M. (1990), Food aid: motivation and allocation criteria, Foreign Agricultural Economic Report, No. 240, The World Bank, Washington, DC. Tschirley, D., Donovan, C. and Weber, M.T. (1996), Food aid and food markets: lessons from Mozambique, Food Policy, Vol. 21 No. 2, pp Zahariadis, N., Travis, R. and Ward, J.B. (2000), US food aid to Sub-Saharan Africa: politics or philanthropy?, Social Science Quarterly, Vol. 81 No. 2, pp Long-term dependence on food aid? 453 Further reading Barrett, C.B. (2001), Does food aid stabilise food availability?, Economic Development and Cultural Change, Vol. 49 No. 2, pp Bauer, P.T. (1984), Reality and Rhetoric: Studies in the Economics of Development, Harvard University Press, Cambridge, MA. Cathie, J. (1982), The Political Economy of Food Aid, Wolfson College, Cambridge. Clay, D., Molla, D. and Habtewold, D. (1999), Food aid targeting in Ethiopia: a study of who needs it and who gets it, Food Policy, Vol. 24 No. 4, pp Degnbol-Martinussen, J. and Engberg-Pedersen, P. (2003), Aid: Understanding International Development Cooperation, Zed Books Ltd, New York, NY. Fuhrer, J. and Moore, G. (1995), Inflation persistence, Quarterly Journal of Economics, Vol. 110 No. 1, pp Riddell, R. (1996), Aid dependency, Aid Dependency: Causes, Symptoms and Remedies; Project 2015, SIDA, Stockholm, pp. 24. Srinivasan, T.N. (1989), Food aid: a cause of development failure or an instrument for success?, The World Bank Economic Review, Vol. 3 No. 1, pp Appendix The countries in the sample are: Albania, Algeria, America Samoa, Angola, Antigua and Barbuda, Argentina, Armenia, Azerbaijan, Bangladesh, Barbados, Belarus, Belize, Benin, Bhutan, Botswana, Brazil, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Cape Verde,

17 JES 454 Chad, Chile, China, Colombia, Comoros, Democratic Republic of the Congo, Republic of Congo, Costa Rica, Cote d voire, Cyprus, Djibouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Estonia, Ethiopia, Fiji Islands, Gabon, Gambia, Georgia, Ghana, Grenada, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, India, Indonesia, Iran, Iraq, Israel, Jamaica, Jordon, Kenya, Dem Korea, Rep Korea, Laos, Lebanon, Lesotho, Liberia, Madagascar, Malawi, Malaysia, Maldives, Mali, Malta, Mauritania, Mauritius, Mexico, Mongolia, Morocco, Myanmar, Namibia, Nepal, Netherlands Antilles, New Caledonia, Nicaragua, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Rwanda, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, Samoa, Sao Tome and Principe, Senegal, Seychelles, Sierra Leone, Somalia, Sri Lanka, Sudan, Swaziland, Syrian Arab Republic, Tanzania, Thailand, Togo, Tonga, Tunisia, Turkey, Uganda, Uruguay, Vietnam, Yemen, Zambia, Zimbabwe. Corresponding author Winston Moore can be contacted at: winston.moore@cavehill.uwi.edu To purchase reprints of this article please reprints@emeraldinsight.com Or visit our web site for further details: