Stereotypes, Imperfect-Information Theories, and Statistical Discrimination in Labor Markets.

Abstract

This dissertation comprises three Essays analyzing a firm's use of stereotypical information in making individual employment decisions. Essay I first compares the distribution effects of recognizing versus ignoring group information. It extends the basic one-job, two-wage model to a two-job, three-wage model and to differentially valid testing models. Essay I then models the efficiency implications of statistical discrimination. Although stereotypical information costlessly increases the accuracy of employers' productivity predictions, it does not necessarily increase total output over a world that ignores group information. One source of inefficiency is through labor supply allocations--lowering one group's wage may discourage its able workers more than raising another group's wage encourages its able workers. Second, firms may choose to capture the greater expected productivity of one group and forego productive complementarities between groups. Third, using stereotypical information may decrease long-run productivity through decreased human-capital investments. Essay II examines the legal treatment of statistical discrimination. One goal is to compare society's revulsion toward invidious discrimination with its more ambivalent feelings toward statistical discrimination. Essay II finds that the Fair Employment Laws prohibit any use of race as a productivity predictor, although the bona-fide-occupational-qualification exception allows, in exceptional circumstances, employers to use sex and age as imperfect proxies for productivity. Essay II then uses an equal protection analysis to conclude that statistical discrimination against any group is distasteful unless justified by significant efficiency gains, and becomes more unpalatable if it hurts immutable and pervasively burdened groups and if the proportion of misclassified persons is large. Essay III makes endogenous the firm's choice of productivity predictors by examining alternatives to statistical discrimination. Firms may use stereotypical information as a cheap screen even though individual determinations are possible. The Essay inquiries under what conditions a firm forced to make individual determinations will still want to use group information to reduce the residual uncertainty. The Essay then presents a decision-theoretic model to show that firms may neglect alternate information sources because testing prior beliefs about a stereotype is too costly, ex ante.