The distribution of small material resources by politicians to voters during election campaigns, often in the form of money, food, beer, or clothing, has long been common in many democracies around the world. These activities typically go by the name of vote buying. Where ballot secrecy is not legislated or enforced, buying votes can be effective because politicians can enforce vote-buying bargains. In some settings, however, vote-buying persists despite the secret ballot. In Kenya, over 40 percent of voters in some areas received money before the 2007 election; yet the ballot is formally secret, parties lack the capacity to monitor voters, and large majorities believe that powerful people cannot figure out how they voted. Vote buying in Kenya is puzzling. It is even more puzzling because Africans and Kenyans in particular are often said to vote ethnically. If African voters vote ethnically, as the conventional wisdom suggests, why must their votes also be bought?

To understand vote buying in the African context, this project develops a fuller understanding of how vote buying influences the preferences and expectations of recipients and observers. Indeed, despite the prevalence of vote-buying in many settings, there is little systematic evidence documenting its impact on support for and perceptions of politicians. This project provides such evidence while addressing largely unanswered questions: Is vote buying effective in building support for politicians? If so, with whom is vote buying effective, and why?

Answering these questions is difficult with survey data for at least two reasons. First, because of embarrassment or fear, many survey respondents are unwilling to respond honestly to survey questions about vote buying, which biases the surveys. Second, different theories of vote buying predict outcomes that are observationally equivalent in survey data.

To avoid these problems, this study conducts field experiments in Kenya. The experiments will present participants with artificial, but realistic, audio recordings discussing hypothetical political candidates, varying only the inclusion of subtle references to the politician's vote-buying behavior in the recordings heard by randomly selected treatment groups. Follow-up questions will gauge support for and perceptions of the politicians discussed in the recording, and will yield a dataset with which to analyze and measure vote buying's impact on voter support for and perceptions of politicians across a number of key dimensions. The artificiality of the experiment provides the control necessary to test competing theories. Moreover, the use of recordings allows the investigator to gain subtlety in referring to vote buying.

The intellectual merit of this project lies in its capacity to advance understanding of the workings of the critical democratic process of elections in Africa. The project illuminates in particular the widespread and puzzling phenomenon of vote buying. By contributing to our understanding of whether and why vote buying is effective, the project enhances knowledge of when and why democratic elections do--and do not--yield responsive and accountable government.

The broader implications of the project are tied to its intellectual contributions. The project can inform the design of policies and programs seeking to promote the responsiveness and accountability of governments to their citizens.

Project Report

This project studies vote buying and ethnic mobilization during African elections, with an empirical focus on Kenya, a country where between 25 to 30 percent of adults are targeted with cash handouts during election campaigns. The project addresses two theoretical puzzles: first, how cash handouts can be effective in winning votes when the secret ballot ensures that voters can accept money and vote on the basis of other criteria; and second, why cash handouts are so prevalent in a context where ethnicity is highly salient in politics and where voters are often assumed to vote almost deterministically for candidates that share their ethnic identity. The central theoretical claim of the study is that cash handouts are effective in winning votes because they convey information to voters about a candidate’s future performance. In particular, they convey information about a candidate’s personal ability to deliver resources and benefits to constituents, an ability that matters to voters in settings such as Kenya’s where the state often does not predictably provide public goods to all citizens and where access to resources often depends on personal connections to those in power. To test the informational argument, the project involved a set of novel survey experiments conducted with 1,000 Kenyans in multiple provinces. The survey experiments were designed to overcome a central challenge in the empirical study of vote buying: that for most people, vote buying is a sensitive subject and so in surveys and interviews they are likely to underreport or misrepresent their experiences in order to provide a socially acceptable response to the enumerator or interviewer. In the experiments, participants listen to radio recordings describing political candidates. Some participants are randomly assigned to receive a radio recording that contains a very short and relatively subtle reference to vote buying by the candidate being discussed. Participants are also randomly assigned to hear a recording featuring a member of their own ethnic group or a different ethnic group, or a recording in which there is no reference to the ethnic group of the candidate. As the recordings are assigned at random, differences in stated preferences for and expectations of the candidate can be attributed to the vote buying information and the ethnicity cues. As the differences in the recordings are subtle, the design minimizes response bias. I can therefore test implications of the argument without asking a single direct question about vote buying itself. The key findings of the project are as follows: First, I show that information about a candidate’s vote buying behavior increases political support for that candidate, a result that is consistent with the informational theory. I then show that this effect is strongest among the poorest participants in the study, raising concerns about the disproportionate impact of vote buying on poor voters and the strategy’s subsequent potential to undermine the political representation of the poor. Second, I show that information about vote buying is most effective when candidate and voter are members of the same ethnic group. This implies that material exchange and ethnic identity are not substitutes but rather complements. I further show that the reason for this stronger effect among coethnics is that the distribution of money during a campaign generates expectations of ethnically biased resource allocations in the future. Third, I find that an ethnic connection to a political candidate only impacts political support for a candidate when that candidate hands out money. In other words, there is no coethnic advantage in the absence of material distribution. This implies that emotional or psychological attachments to coethnics may not be the strongest drivers of ethnic voting—these drivers would matter regardless of a candidate’s distributive behavior during a campaign. Rather, the results suggest that ethnic voting is driven by cash handouts, which increase expectations of access to future resources—in other words, that ethnic voting is instrumental in nature. These latter findings have implications beyond the study of vote buying, shedding light on the drivers of ethnic voting and the reasons why ethnic identity is important in African politics.

National Science Foundation (NSF)
Division of Social and Economic Sciences (SES)
Standard Grant (Standard)
Application #
Program Officer
Brian D. Humes
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of California Los Angeles
Los Angeles
United States
Zip Code