Uninformed votes: information effects in presidential elections

Theory: Recent scholarship has emphasized the potential importance of cues, information shortcuts, and statistical aggregation processes in allowing relatively uninformed citizens to act, individually or collectively, as if they were fully informed. Hypotheses: Uninformed voters successfully use cues and information shortcuts to behave as if they were fully informed. Failing that, individual deviations from fully informed voting cancel out in a mass electorate, producing the same aggregate election outcome as if voters were fully informed. Methods: Hypothetical 'fully informed' vote choices are imputed to individual voters using the observed relationship between political information and vote choices for voters with similar social and demographic characteristics, estimated by probit analysis of data from National Election Study surveys conducted after the six most recent United States presidential elections. Results: Both hypotheses are clearly disconfirmed. At the individual level, the average deviation of actual vote probabilities from hypothetical 'fully informed' vote probabilities was about ten percentage points. In the electorate as a whole, these deviations were significantly diluted by aggregation, but by no means eliminated: incumbent presidents did almost five percentage points better, and Democratic candidates did almost two percentage points better, than they would have if voters had in fact been 'fully informed.'