How clientelism works: Evidence from the Barinas special election.

Do politicians target the benefits of social programs to party loyalists or to swing voters? Traditional tests of this question are clouded by an identification problem caused by the simultaneity of politician and voter choices to participate in the exchange of assistance for votes. I use the holding of an unanticipated repeat gubernatorial election in the Venezuelan state of Barinas in 2022 as a natural experiment to identify the effects of elections on the distribution of government assistance. I estimate that the holding of the election led to an increase in the probability of voters in Barinas receiving food packages in comparison with the control group of voters in the state of Apure. I also find that moderate opposition and third-party voters received larger increases in food benefits. These results are consistent with the predictions of the spatial model of distributive politics, according to which elections lead governments to direct more benefits to swing voters instead of core supporters. The findings illustrate why investigation of cross sectional correlations is insufficient to test the implications of theories of voting if it is not accompanied by a clear identification strategy to help isolate the source of the underlying shocks. • I use an unanticipated repeat vote in Venezuela as a natural political experiment. • The election led to an increase in the probability of voters receiving food packages. • Third-party voters received more benefits, consistent with spatial models. • Traditional tests using cross sectional correlations lack an identification strategy. • This is the first time a repeat election is used to test theories of politics. [ABSTRACT FROM AUTHOR]