In this paper, we have provided some support for several hypotheses about the determinants of which governors get reelected. The benefit from being a member of a particular party varies from state to state and from year to year. Personal characteristics such as age are also important. The logits give some support to the importance of coalition formation; reelection is easier in states with low voter turnout and in farm states. The paper is most concerned with the connection between the economic performance and the electoral success of incumbent candidates for governor, and we find support for a model of electoral acountability, in which governors are powerful in state governments and state governments have the ability to differentially tax fixed factors relative to neighboring states. This paper raises some important issues regarding the measurement of variables in political economy, which have wide applicability to other studies in the economics of politics. Peltzman (1988) finds that the difference between the growth rate in state personal income and the national growth rate over a one to four year period prior to the election does not affect gubernatorial electoral outcomes. Concurrently, we find that the current year's growth rate in state personal income and its difference from the national growth rate are not significantly related to electoral success but that the average deviation from predicted state personal income during the governor's tenure in office is significantly related to the odds of getting reelected. That is, the data reject simplistic views of voter behavior and support a sophisticated model of voter behavior. Similarly, Peltzman (1988) has greater success using more sophisticated, cumulative measures of national economic performance. This suggests that great care must sometimes be taken to create variables that are capable of giving our economic models a fair hearing. Voting studies often find that only current period effects matter, which is attributed either to voter myopia or — and this is the more popular explanation nowadays — to the confinement of effects by rational expectations to current period, ‘white noise’ effects. The specifications in these voter studies ignore the cumulative information generated in political economy. Both our results and Peltzman's (1988) results suggest that variables that reflect this cumulative information may be more successful than variables that are based only on the current year.