In São Tomé and Príncipe, Vicente followed the random delivery of leaflets informing that vote-buying is illegal and found that this increased the number of voters who agreed they should not let their vote be bought; however, it discouraged voter turnout and increased the incumbent vote share.
São Tomé and Príncipe (ST&P) is a Portugese-speaking island nation in the Gulf of Guinea with about 193,000 citizens where oil was discovered in the late 1990s. There was a coup by the army in 2003, and an attempted coup in 2009, but otherwise elections are considered to be free. Vote-buying is a common occurrence; in a poll Vicente conducted, 64% of respondents said they had witnessed vote-buying in their area.
Map: São Tomé and Príncipe (Figure 2, pg. F364)
In 2006, 10,000 pamphlets were distributed to households emphasizing that vote-buying is illegal.
(Figure 1, pg. F362)
Canvassers also verbally told citizens that they should vote with their conscience. After the parliamentary elections but before the presidential election, a representative sample of treatment and untreated citizens were polled about their perceptions of vote-buying, corruption in general, and electoral behavior. At this time, they were given a pre-paid postcard with a written demand to publish the results of the corruption module, and told that if 50% of respondents sent the postcard, the results would be published. After the elections, the same voters were polled about their perceptions of vote-buying and their electoral behavior in presidential elections.
First, perceptions of vote-buying in a past election were not substantially different among treatment and control subjects, which Vicente interpets as meaning that subjects in the treatment condition were not merely telling the interviewers what they thought they wanted to hear (conformity bias). There were no significant differences among treatment and control groups in likelihood of sending the postcard.
Second, voters who received the pamphlet were substantially more likely to report that vote- buying did not influence the outcome of the presidential election, and that voters acted in good conscience.
Third, the campaign decreased voter turnout by 3-6 percentage points, suggesting that vote- buying may drive participation.
Fourth, the campaign diminished opposition turnout and increased incumbent turnout, both by nearly 4 percent.
This large-scale experiment suggests that discouraging vote-buying has measurable effects on elections, but perhaps not those that policy-makers might want. Decreasing voter turnout and increasing the incumbency advantage may be undesirable in many contexts. This experiment serves as a good reminder that well-intentioned interventions may have unanticipated consequences.
The extent to which ST&P generalizes to other democratic countries is debatable. Future replications might answer this question.