Voters are allegedly uninformed, fickle, and sensitive to irrelevant events, or they are stubborn, tribal, and hyper-partisan. As a result of evidence along these lines, many scholars argue that democracy does not function as it should, and some have gone so far as to argue that we should abandon democracy altogether. At the same time, studies on elections and policy outcomes tend to find that electoral selection and incentives work reasonably well. In this paper, we offer a reconciliation of these two literatures. Even if individual voters leave something to be desired, electorates can still select the best candidates and incentivize elected officials to do a good job. We theoretically compare electoral accountability in a world with a single, rational, representative voter to one with many voters who exhibit the limitations documented in the literature, and we ask when the aggregation of imperfect voters can still produce desirable outcomes. We also empirically assess the extent to which voters change their votes or change their turnout decisions in response to the qualities of candidates, and we use the results to explore how different partitionings of voters into electorates affect incentives for incumbent effort.