This paper develops the theory of statistical discrimination in the form of unequal employment criteria and interviews. Workers differ by imperfectly observed 'quit rate.' Profit maximization leads firms to set stricter employment criteria or interview fewer workers from the group with a greater proportion of high-quit-rate workers. The theory can explain unemployment differences between groups and the total exclusion of a group from a labor market. Although statistical discrimination constitutes economic discrimination according to a criterion of harm to a group, it may or may not be efficient because of the presence of externalities. Copyright 1998 by Economics Department of the University of Pennsylvania and the Osaka University Institute of Social and Economic Research Association.