摘要:In neoclassical models, workers are classified a priori into discrimination groups. We develop a probabilistic model of wage discrimination in which workers need not be classified a priori. Our model is a generalization of the standard framework, whereas Becker's model is an extreme case. A second implication is that the traditional approach to measuring discrimination (the Oaxaca–Blinder approach) must be modified to take into account this probabilistic framework.