The wage discrimination model developed by Becker has been criticized for predicting that competitive forces will lead to the disappearance of racial discrimination in the long run. We have reformulated the model in terms of nepotism toward white workers rather than discrimination against black workers. In this new framework, both nepotistic and taste-neutral firms are expected to survive the competitive struggle in the long run. Therefore, the new framework is consistent with long-run as well as short-run racial wage differentials.

John T. Warner, Matthew S. Goldberg, The Influence of Non-Pecuniary Factors on Labor Supply: The Case of Navy Enlisted Personnel, The Review of Economics and Statistics, Vol. 66, No. 1 (Feb., 1984), pp. 26-35

Thomas P. Frazier, Matthew S. Goldberg, Thomas R. Gulledge, Jr., Department of Defense Profit Policy and Capital Investment in the Military Aircraft Industry, The Review of Economics and Statistics, Vol. 74, No. 3 (Aug., 1992), pp. 394-403

The traditional economic analysis of is based on Gary Becker's study of taste by employers, employees, and consumers. More recent work by Kenneth Arrow (1972, 1973) has attempted to interpret intergroup wage differences in an alternative framework as a rational reaction to uncertainty in labor markets. His model of statistical discrimination demonstrates that when the screening process used to determine a worker's qualifications is costly, and prior expectations of productivity differ across race or sex groups, then wage differentials may arise between workers of identical productivity. By implicitly assuming a perfect screening process, Arrow ignores a potentially important source of wage differentials, namely the fact that the screening process might be a more reliable predictor of productivity for one group than for another.' Our paper generalizes the Arrow model in two ways. First, in contrast to Arrow, we assume that all groups have identical distributions of productivity. Secondly, the screening process used by the firm to determine an applicant's productivity is biased in the sense that: a) members of various groups may pass the test in different proportions despite their identical productivity distributions; and b) the predictive power of the test might vary across groups. Our objective is to analyze the effects of these types of biases in the screening process on the wage differentials between different population groups.