UC Berkeley

Development aid comprises a significant share of government budgets in many resource-poor countries and is a significant source of funding for health and social service delivery. Yet little is known about the strategies that political actors in these countries use to determine the geographic allocation of aid resources or how such allocation patterns affect the political behavior and attitudes of citizens. Poor aid reporting and a lack of transparency in budget processes have made it extremely difficult to track the sub-national allocation of aid projects. As a result, much of what we know about its political impact comes from longitudinal cross-national studies that mask large in-country variation in allocation and that tell us little about in-country political decision-making. This dissertation explores these issues, using newly available, geo-coded aid data to conduct a statistical analysis of sub-national aid allocation patterns, public opinion, and electoral outcomes in Malawi.

The first paper uses a two-part model strategy to estimate separately the probability of an area being selected to receive an aid project in a given year and the determinants of aid funding levels once an area has been selected to receive aid. It finds that in aggregate, aid allocation in Malawi exhibits little association with local need. Rather, to the extent that it is a significant factor in the study models, need tends to be negatively associated with both aid project placement and funding levels. In the models, the proportion of residents that share the President's ethnicity has no influence on the probability of an area being selected to receive aid and is negatively associated with the amount of aid funding received. Instead, areas with high proportions of smaller, non-aligned ethnic groups have higher probabilities of being selected to receive aid projects and receive disproportionately high levels of funding once selected. There is tenuous evidence that past electoral support for the incumbent party increases the probability of an area being selected to receive social services-related aid projects. However, support for the incumbent party has either no influence on the amount of aid dollars an area receives, or exhibits a slightly negative association. These results suggest that the area selection process in aid allocation decisions might be more prone to political targeting than the determination of aid funding levels and that in this targeting, political leaders might be using aid resources primarily to build cross-ethnic coalitions with non-aligned ethnic groups.

When I examine the electoral effectiveness of aid allocation, i.e., whether it mobilizes citizens to vote or induces them to support the ruling party, I find that higher aid levels are associated with increased incumbent vote share and higher voter turnout. The positive impact of aid on turnout is strongest in areas that have been electorally competitive in the past. This suggests that in Malawi, aid allocation has the potential to entrench incumbent political parties.

In the third essay, I report the results of a multinomial logistic regression model that I developed to estimate the association between aid levels and citizens' perceptions of corruption in their local leaders. I find respondents in districts receiving relatively high amounts of aid dollars no more likely to view local leaders as corrupt than those in lower aid districts. Instead, there is a tendency for districts that have more project activity to perceive low local corruption even though there is strong evidence that actual corruption, in the form of bribe solicitations, is positively associated with aid levels. The relationship between corruption perceptions and aid varies over the study period, becoming more positive over time. The negative association between aid and perceptions of corruption is less pronounced for those who share the President's ethnicity and those with strong ethnic attachments.

The results of these three studies suggest that citizens in Malawi value aid projects and may see them as a sign of fairness and competence in government. Governments that provide aid resources are rewarded with votes and may be viewed relatively favorably by citizens.

These studies add nuance to our understanding of distributive politics in sub-Saharan African democracies by highlighting a case in which political leaders seemed to have used resources not only to shore up support in narrow core constituencies based on shared ethnicity but also to win over opposition voters and ethnic groups with weak partisan attachments. Such a case may be relevant to other competitive, open, electoral democracies in sub-Saharan Africa. The study highlights the value of having detailed, project-level, location-specific data on aid projects in order to allow this kind of research.