National economic conditions regularly influence outcomes in U.S. presidential elections. However, beyond this simple finding, much remains unclear. How large are national economic effects? Which macroeconomic indicators? Subjective or objective measures? Retrospective or prospective? What is the role of institutions? In our analysis of the American National Election Studies, 1956–1996, we employ a National Business Index (NBI), an aggregate measure that amalgamates individual voter perceptions of the collective economy. It outperforms other national economic measures and reveals that effects have been underestimated. The assumption of strong retrospective economic voting is tested under different institutional hypotheses. Contrary to expectations, it is not found to be influenced by divided government, but it is heavily influenced by incumbency, meaning in practice whether a popularly elected president is running. Moreover, this incumbency variable highly conditions the time horizon of national economic voting. When a popularly elected president is not running, such voting is almost entirely prospective. These conditional effects go a long way toward explaining certain enduring controversies in the literature.