A model of the determination of the union status of workers is developed that incorporates the separate decisions of workers and potential union employers in a framework which recognizes the possibility of an excess supply of workers for existing union jobs.This theoretical framework results in an empirical problem of partial observability because information on union status is not sufficient to determine whether nonunion workers are nonunion because they do not desire union representation or because they were not hired by union employers despite a preference for union representation.The problem is solved by using data from the Quality of Employment Survey that have a unique piece of information on worker preferences which allows identification and estimation of the model.The empirical results yield some interesting insights into the process of union status determination that cannot be gained from a simple logit or probit analysis of unionization. Chief among these relate to the unioniza-tion of nonwhites and southerners.The well-known fact that nonwhites are more likely to be unionized than otherwise equivalent whites is found largelyto be due to a greater demand for union representation on the part of non-white workers. The equally well-known lower propensity to be unionized among southern workers is found to be due to a combination of a lower demand for union representation on the part of southern workers and a supply of union jobs which is more constrained relative to demand than in the North
