This paper examines the impact of employment status on voting by formally modeling the effect of partisan government on workers' economic interests. These interests are determined by model workers' labor supply decisions, which are conditioned on their expected wage, unemployment benefits, and the probability of receiving a job offer or being laid off. Both probabilities are influenced by the party in power. The solution to the model implies that, relative to the employed, the higher the education level, the income, and the unemployment benefits of the unemployed the less likely they are to vote for the party associated with higher growth. The unemployment rate has the same impact. These hypotheses are successfully tested on NES 1972-2000 U.S. presidential elections data.To those experiencing it, unemployment is often a dramatic event, both materially and psychologically. It is natural to suppose, therefore, that voters' political choices will be affected by their employment status. Yet this simple inference may have eluded them. Students of elections, at any rate, have had trouble developing a clear and systematic understanding of how economic circumstances affect individual voting. Their difficulty is not for want of trying (for a review of this vast literature see, e.g., Lewis-Beck and Stegmaier 2000). Nonetheless, there remains general disagreement about the kind of impact personal unemployment has (e.g., Funk and García-Monet 1997;Kiewiet 1981;Mutz 1992;Schlozman and Verba 1979). Moreover, personal unemployment, whatever its role, is widely found to have a smaller impact on voting than "sociotropic" concerns with aggregate unemployment and similar indicators of national economic health (e.g., Nannestad and Paldam 1994).Findings like these are puzzling. After all, aggregate unemployment is ultimately visited on individuals. Therefore, if the public holds government responsible in some way for the level of unemployment, it is hard to see why, on average, the unemployed do not hold government partly responsible for their fate. Indeed suppose the employed and unemployed have identical levels of altruistic concern with aggregate unemployment, and the unemployed accept no more responsibility for their own situation than they would attribute to others for theirs. So long