Incumbency Effects in Brazilian Mayoral Elections: A Regression Discontinuity Design

I use a regression discontinuity design to study incumbency effects in Brazilian mayoral elections. For mayors elected in 1996 I find no evidence of an incumbency effect on the probability of being elected in 2000. For the 2000-2004 electoral cycle I also find no effect except for races where the mayor elected in 2000 belonged to the a party in the center-right coalition and the runner-up belonged to a party in the center-left coalition. In these races I find an incumbency disadvantage. For mayors elected in 2004 I find a strong incumbency advantage in the 2008 election across all races. I also show some novel incumbency effects. Winning a mayoral election does not have a positive effect on the future prospects of a politician’s career at the state, national or local level. Losing a mayoral election increases the probability of a politician switching parties.