Testing for Discrimination the Case for A National Report Card

Article excerpt

Despite the fact that minorities have made substantial economic and social progress over the past 30 years, significant disadvantages based on race persist within the United States and serve as markers of continuing policy failures. A body of empirical and anecdotal evidence indicates that discrimination based on race and ethnicity has yet to be eliminated by the nation's civil rights laws. For example:

* The hourly earnings of black men are 65 percent those of white men;

* Black men pay more than $1000 more for the same new car as white men;

* Deep disparities persist in the receipt of state and local contracts for all minority groups;

* Schools and neighborhoods are becoming more not less segregated as we approach the 21st century.

Why Testing Is Needed

While these statistics suggest the persistence of racial and ethnic discrimination, they do not, in and of themselves, help us gauge its extent with any accuracy. In other words, disparity of results does not prove discrimination. The absence of understandable and compelling information about the extent of discrimination in our society contributes to sharp differences in the way groups interpret patterns of inequality and the obligation of government to alter them. So while 60 percent of whites think conditions for blacks have improved during the past few years, only 35 percent of blacks share those views. Indeed, it is probably fair to say that beneath many of the current controversies about race and ethnicity in the United States lurk fundamental differences of perception about the empirical reality: To what extent are racial and ethnic minorities subject to discrimination?

It is not surprising that there is so little social consensus over the contribution of discrimination to social inequality. As Peter Siegelman notes, blatant Jim Crow discrimination is largely a thing of the past and the so-called have-a-nice-day discrimination--to the extent that it exists--is harder to detect, measure, and ultimately counteract. At the same time, progress toward integration paradoxically may mask an overall decline in discrimination, as the noted scholar Orlando Patterson argues. That is, increased interaction between members of differing racial or ethnic groups may lead to greater friction and more perceived acts of discrimination--despite the fact that the broader trend may be toward less discrimination and fewer discriminators. Thus, while "have-a-nice-day" discrimination may lead to premature claims that we have achieved a color-blind society, conflicts associated with progress towards integration may generate exaggerated claims of victimization. Both types of distortion, along with the misguided policies that flow from them, can be corrected by more accurate and widely understandable measures of discrimination.

Evidence of discrimination has come from several sources, including analysis of aggregate employment, housing, and other data sets. While the statistical techniques employed in these analyses have much to offer, they fail to provide the clear, direct measures and narrative power offered by paired testing. In a paired test, two individuals are matched for all relevant characteristics other than the one that is expected to lead to discrimination. The testers apply for a job, an apartment or some other good and the outcomes and treatment they receive are closely monitored.

Paired testing is an excellent vehicle for understanding and measuring actual discrimination (understood here simply as the practice of treating people

differently because of their membership in a protected group). First, testing provides a feasible method for directly observing discriminatory treatment by comparing two people equally qualified for the transaction in question, who differ significantly only in their group membership. In technical terms, paired testing design minimizes "omitted variable bias"--the possibility that differences in outcome are caused by variables that the researcher cannot observe. …