Does the early bird always get the worm? First round advantages and second round victories in Latin America

Runoff systems allow for a reversion of the first-round result: the most voted candidate in the first round may end up losing the election in the second. But do voters take advantage of this opportunity? Or does winning the first round increase the probability of winning the second? We investigate this question with data from presidential elections since 1945, as well as subnational elections in Latin America. Using a regression discontinuity design, we find that being the most voted candidate in the first round has a substantial positive effect on the probability of winning the second round in mayoral races – especially in Brazil –, but in presidential and gubernatorial elections the effect is negative, though not statistically significant at conventional levels. The positive effect in municipal races is much stronger when the top-two placed candidates are ideologically close – and thus harder to distinguish for voters – but weakens considerably and becomes insignificant when the election is polarized. We attribute these differences to the disparate informational environment prevailing in local vs. higher-level races.