This article examines the impact of clientelism on voting behavior in Benin through a unique field experiment conducted during the 2001 presidential elections. The experiment involved randomly selecting villages to be exposed to either "purely" clientelistic or "purely" public policy platforms, with real candidates competing in the elections. The study aims to test hypotheses about the determinants of voters' demand for public goods and the effectiveness of clientelism in different contexts. The article highlights several key questions addressed by the experiment, including the effect of ethnic affiliation on voting behavior, the comparative advantage of different types of messages, and the responsiveness of female and younger voters to clientelism. Clientelism is defined as transactions between politicians and citizens where material favors are exchanged for political support, and it is characterized by narrow corporatist and local interests. The experimental design involves selecting eight noncompetitive districts, four each dominated by incumbents and opposition candidates, and partitioning each district into three subgroups: clientelism treatment, public policy treatment, and control. The experiment compares voting behavior in these groups to understand the relative effectiveness of clientelism versus public policy messages. The empirical results show that clientelism has a positive effect on voting behavior, while public policy messages have a negative effect, particularly in northern districts and for regional candidates. Women are more responsive to public policy messages and are less likely to support northern candidates. The study also finds that past voting behavior is a strong predictor of current voting behavior, but ethnic identity does not entirely determine voting behavior. The article concludes by discussing the implications of the findings for understanding electoral politics in Benin and the broader literature on clientelism and voting behavior. It suggests that more realistic models should explicitly include factors such as incumbency and the scope of competing candidates.This article examines the impact of clientelism on voting behavior in Benin through a unique field experiment conducted during the 2001 presidential elections. The experiment involved randomly selecting villages to be exposed to either "purely" clientelistic or "purely" public policy platforms, with real candidates competing in the elections. The study aims to test hypotheses about the determinants of voters' demand for public goods and the effectiveness of clientelism in different contexts. The article highlights several key questions addressed by the experiment, including the effect of ethnic affiliation on voting behavior, the comparative advantage of different types of messages, and the responsiveness of female and younger voters to clientelism. Clientelism is defined as transactions between politicians and citizens where material favors are exchanged for political support, and it is characterized by narrow corporatist and local interests. The experimental design involves selecting eight noncompetitive districts, four each dominated by incumbents and opposition candidates, and partitioning each district into three subgroups: clientelism treatment, public policy treatment, and control. The experiment compares voting behavior in these groups to understand the relative effectiveness of clientelism versus public policy messages. The empirical results show that clientelism has a positive effect on voting behavior, while public policy messages have a negative effect, particularly in northern districts and for regional candidates. Women are more responsive to public policy messages and are less likely to support northern candidates. The study also finds that past voting behavior is a strong predictor of current voting behavior, but ethnic identity does not entirely determine voting behavior. The article concludes by discussing the implications of the findings for understanding electoral politics in Benin and the broader literature on clientelism and voting behavior. It suggests that more realistic models should explicitly include factors such as incumbency and the scope of competing candidates.