Using data on adult male workers we first investigate the incremental effect of a year of schooling on unemployed hours, and use this calculation to explain the difference in the pro- portional effects of schooling on earnings and wages. Schooling apparently reduces unemployed hours by reducing the incidence of unemployment spells, but it does not significantly affect their duration. We next test whether unemployed hours represent real constraints on worker behavior. To do this we develop and estimate life-cycle models of labor supply for workers with and without spells of unemployment using longitudinal data. The results imply that perhaps three-quarters of the unemployed hours of male workers are part of the offer to sell labor.