According to the conventional wisdom, African voters split along ethnic lines at election time, with each group voting for its own. In Uganda, however, President Museveni's ethnic group, the Banyankole, comprise only nine percent of the population while Museveni is supported by a majority of the public. Clearly, large numbers of Ugandans must be voting for candidates outside of their ethnic group. Nevertheless, the Banyankole do seem to vote ethnically, supporting Museveni in numbers much higher than does the population as a whole. What explains this pattern of voting? Do Banyankole and non-Banyankole voters support Museveni for the same reasons?
The project hypothesizes that non-Banyankole voters, lacking ethnic cues, vote based on their assessments of Museveni's past performance in office. To explain the voting behavior of the Banyankole, the project examines five hypotheses from the literature. In two hypotheses, the Banyankole, like the non-Banyankole, vote purely based on assessments of Museveni's performance. Museveni's advantage among the Banyankole comes from the fact that his co-ethnics assess his performance more favorably, either because he has provided them with more goods or because they believe he has. In the second set of hypotheses, non-Banyankole still vote based on performance, while the Banyankole support Museveni specifically because he is of their ethnic group. Banyankole may vote for Museveni out of ethnic pride, because they believe a co-ethnic represents their best chance of future goods, or because they face social pressure to vote with their ethnic bloc.
The core of the study is a voting simulation experiment in which 750 respondents were asked to choose between candidates whose ethnicity, performance record and other characteristics have been randomly assigned. This test allows the researcher to determine if voters, whether Banyankole or not, rely on performance or ethnicity (or something else) as the basis of their vote. The voting simulation is also be used to test the social voting hypothesis: if social pressure determines voting, voters should be more likely to select co-ethnic candidates when asked to vote aloud among a group of co-ethnics than when marking a secret ballot. In the second part of the study, respondents rate their confidence that a hypothetical co-ethnic or non-co-ethnic candidate, if elected, would provide them with goods such as employment. This experiment determines whether, all things equal, voters expect more future goods from a co-ethnic candidate. Finally, the researcher creates inventories of household wealth and public service quality in 35 villages throughout Uganda. These data enable the researcher to determine whether the Banyankole have gained significantly more than other under Museveni's leadership and/or whether they overestimate what they receive. The data, compared to assessments of, predictions about, and votes for Museveni, indicate the importance of candidate performance, relative to ethnicity, in predicting vote share in the specific context of Ugandan politics.
The experimental nature of the study improves upon existing research on African voting by isolating and controlling the candidate characteristics and electoral conditions to which voters are exposed. In addition, the project will produce a unique new dataset that contains both objective and subjective indicators of public goods, in a way that allows real changes in these indicators to be linked to actual individual voting behavior. The findings, which will provide objective information about how Museveni distributes goods and how voters make their decisions, will be shared with multiple Ugandan research organizations conducting work on goods provision, and disseminated both through the public media and in direct mailings to leaders of areas in which the survey was conducted.
