This dissertation sheds light on why accountability for corruption fails and corrupt politicians survive democratic elections. This study relies on a multi-method approach that brings together quantitative and qualitative data from Brazil as well as a broader set of countries. Challenging prior works on this topic, which focus on the low salience of corruption in voters’ minds, this dissertation shows that even when corruption is a salient issue to voters, responses to corruption are not more clearly observed. Rather than working just like accountability for other components of government performance, accountability for corruption follows a particular dynamic, one that is distinct from accountability for other electoral issues. Accountability for corruption fails because the issue of corruption itself tends to be perceived with cynicism, an attitude characterized by the belief that all politicians and political parties are equally incompetent in dealing with corruption. When voters perceive corruption to be a constant among candidate options, they are likely to overlook this aspect of government performance and base their vote on different issues. This perceived lack of differentiation among available alternatives is a function of actual levels of corruption. When corruption is widespread, more politicians are likely to be implicated in it, which in turn fosters the perception that politicians are indistinguishable when it comes to either fighting corruption or refraining from it. This explanation predicts that electoral accountability for corruption will, ironically, be weakest where it is needed most. Political corruption perpetuates through a self-fulfilling prophecy. While democratic elections and voters may not represent an effective solution to the problem, independent and credible institutions with the legitimate power to enforce efficient punishment for corrupt behavior may change voters’ poor expectations regarding the prospects of corruption in politics and break the vicious cycle of corruption.