We analyze the relationship between electoral systems and corruption in a large sample of contemporary democratic nations. Whereas previous studies have shown that closed-list proportional representation is associated with greater (perceived) corruption than open-list PR, we demonstrate that this relationship fails to hold once district magnitude is considered. The theory underlying our study draws on work on "the personal vote" that suggests that the incentives to amass resources - and perhaps even to do so illegally - increase with district magnitude in open-list settings but decrease in closed-list contexts. Extending this insight, we show that political corruption gets more (less) severe as district magnitude increases under open-list PR (closed-list PR) systems. In addition, once district magnitude exceeds a certain threshold - our estimates are that this is as low as 15 - corruption is greater in open list than in closed list settings. Only at unusually small district magnitudes (those below 15) is closed-list PR associated with more corruption, as conventionally held. Our results hold for alternate measures of corruption, for different specifications of the country cases to be included, for different measures of district magnitude, and regardless of whether the political system is presidential or parliamentary and the number of effective parties. Using an alternative, objective measure of political corruption, we also present corroborating evidence across Italian electoral districts. Within the open-list environment characterizing Italy for the eleven legislative periods that spanned the second half of the twentieth century, larger districts are more susceptible to suspected political corruption than smaller ones. This finding is consistent with the underlying model that predicts that larger electoral districts are associated with higher levels of corruption in open-list settings. It is robust to the inclusion of various important control variables, including per capita wealth and the efficiency of the judiciary.