In this paper we present a continuous-time model of changes in employment status. The model implies that the binary logit model describes the probability of employment in equilibrium; it also has implications for the probability of employment out of equilibrium and for the length of spells of employment and nonemployment. Parameters of the model can be readily estimated from employment history data using the method of maximum likelihood. In an application to data from the Seattle and Denver income maintenance experiments, we find that both the proposed model and the logit model reveal significant decreases in the probability of employment of family heads under NIT programs that decrease the net wage rate and increase nonwage income. Based on the proposed model, the decrease in the employment probability results primarily from a significant increase in the average length of spells of nonemployment.