One of the canonical causal claims in political science links individuals’ evaluations of the national economy with their votes. Yet there are reasons to expect that these economic perceptions are endogenous to vote choice, meaning that existing cross-sectional models cannot provide a valid test of the causal retrospective voting claim. Using an instrumental variables approach, we assess the effect of sociotropic evaluations on the decision to vote for the incumbent president or his party’s candidate in eight recent U.S. presidential elections. In contrast with prior work, our results reveal that while there is a correlation between sociotropic evaluations and vote choice, individuals’ subjective evaluations only exert a causal effect on votes when there is not an incumbent president on the ballot. These results suggest that, when incumbents are on the ballot, individuals’ economic perceptions are particularly clouded by appraisals of the incumbent and thus do not operate as an exogenous influence on votes.