Citizens are expected to punish corrupt politicians at the polls. Lab and survey exper- iments consistently show that citizens are unlikely to vote for candidates that engage in corruption. However, observational studies and field experiments frequently conclude that corrupt politicians are only mildly punished by voters. This contradiction might be caused by some of these designs not accounting for information self-selection. An experimental design that randomly informs participants  about  corruption  and  disregards  the  fact  that  in the real-world citizens are prone to self-select information might overestimate the electoral consequences of corruption. To overcome this limitation this paper implements a Preference Incorporating Choice and Assignment (PICA) experimental design that allows us to analyze the impact of corruption on voting behavior accounting for information preferences and self-selection. Based on an online experiment conducted in Spain this design allows us to estimate the electoral consequences of information about corruption accounting for the fact that not all citizens might be exposed to such information. The PICA design increases ecological validity by explicitly modelling how citizens navigate information about malfeasance, while retaining the internal validity of fully randomized experiments. The results indicate that the electoral punishment of corruption is dependent on individuals’ information preferences and that, at the same time, not all citizens might be exposed to information about corruption scandals involving their preferred party. The potential effects of being exposed to this information are particularly large among individuals who prefer to avoid political information altogether and consume entertainment instead. This suggests that, especially in the case of inconspicuous corruption scandals, the exposure of inattentive citizens to information about those scandals would increase the electoral punishment of corruption.