IRP Discussion Papers

To measure economic discrimination, one must be able to measure the productive capacity of people, the opportunity structure they face, and outcomes in the form of income and earnings. The first two are difficult to measure but crucial to any assertion that discrimination exists. To illustrate the uses and limits of statistical analysis to overcome these difficulties, two case studies of ethnic discrimination in American history--Irish-Americans in 1900 and Japanese-Americans in 1940--are examined. A persistent problem in using statistical models of discrimination is distinguishing between an opportunity structure that is imposed on a worker from one that reflects the choices or preferences of that worker. For example, although a low wage relative to others with the same productive capacity may be the result of discrimination, it may also reflect the choice by the worker to work at a part-time job at more convenient hours. Such a choice may explain why some married women fail to hold jobs that fully use their productive capacity. Measuring discrimination against women--who tend to have dual careers--is especially difficult. Their wages may be low because they have spent less time in the paid labor market, but it is also plausible that the reason they have devoted more time to housework and less to market work is that they have faced restricted opportunities in the marketplace. The causation can run either way. Statistical models of economic discrimination measure associations but do not establish causation. Therefore, evidence based on statistical analysis is necessary but not sufficient for establishing the presence of economic discrimination. It must be supported by historical and institutional evidence and interpreted within a theoretical framework. The Uses and Limits of Statistical Analysis in Measuring Economic Discrimination This paper begins with a brief discussion of the basic concepts of economic discrimination. In the second part, two case studies of ethnic discrimination from American history illustrate the uses and limits of statistical analysis. (The detailed evidence of discrimination against these groups is presented in Appendix A.) The next part deals with current analyses of economic discrimination, and a summary concludes that evidence based on statistical analysis is necessary but not sufficient for establishing the presence of economic discrimination. It must be supported by historical and institutional evidence and interpreted within a theoretical framework.