Recent empirical studies have found a incumbency disadvantage in many developing democracies, in marked contrast with the well-known incumbency advantage in the US and other developed democracies. We know considerably less about incumbency disadvantage than incumbency advantage. In a simple principal-agent framework, I explore the role of a prominent feature of developing democracies -corruption. When rents are constant in incumbents' tenure -a standard assumption -the conditions for incumbency disadvantage are existent but limited; however, increasing rents, possibly due to learning, a gradual build-up of rent-extraction networks or fiscal windfalls, considerably increase the possibility of incumbency disadvantage, because voters may prefer inexperienced and unconnected challengers, even if they are of lower quality. Incumbency disadvantage becomes more likely as the pace of rent increase grows, politician quality decreases, with noise in the policy outcome, and potentially even when the pool of politicians improves. It is strictly more costly than any electoral outcome with high but constant-rents. The results highlight a novel reason for control of corruption and sensitivity to its dynamics.