This dissertation seeks to examine why democratic systems and electoral competition can sometimes fail to secure clean government in the interest of the electorate. The question of why voters support corrupt politicians, despite disapproving of corruption itself, is of critical importance if it is to be believed that corruption has a detrimental effect on development. The core argument of this dissertation is that party system features that improve accountability by shaping the efficacy of elections as tools to select and control politicians, play a vital and overlooked role in conditioning the scope for corruption. I conceive of governmental corruption as a classical principal-agent model, in which voters‘ relationships with their representatives are mediated by the extent to which party systems enable the electorate to select non-malfeasant politicians who seek to curb corruption and to hold accountable those who do not. This thesis purports that party systems which reduce agency problems confronting voters, by making available information regarding the quality of their incumbents and potential challengers and structurating effective, choices at the polls, decrease the latitude for governmental corruption. This thesis probes this argument through a controlled comparative analysis of corruption in 91 contemporary democracies and three nested-design case studies. The large-N analysis and the case studies of Panama, India, and Mexico offer broad support for these hypotheses.