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...