Discrimination, Measuring

Challenges in Measuring Discrimination

Several factors make accurate measurement of discrimination difficult. Measuring discrimination requires comparing treatment across different racial or ethnic groups, but individuals usually know of only the treatment they personally give and receive. Because of the evolution of norms against discrimination and legal prohibitions, perpetrators have reasons to conceal discrimination. In response, social scientists have developed a number of methods to measure discrimination. These methods have different strengths and weaknesses; the best method depends on the outcome studied and the goals of the study.

In case law and many social science discussions, a distinction is made between two basic forms of discrimination. Differential treatment discrimination occurs when an individual is treated less favorably because of her or his race. Differential impact discrimination occurs when a group is disadvantaged by a practice that does not involve race or ethnicity directly but disadvantages members of a racial or ethnic group without a sufficiently compelling reason. In many social science discussions, institutional discrimination is often used to refer to situations similar to differential impact in law. For instance, differential impact discrimination might occur when a property owner refuses to rent apartments to single mothers without a good business justification, when this practice contributes to renting fewer apartments to members of certain racial and ethnic groups.

Recent work on counterfactual questions as the basis for causality in social science has led social science methodologists to think of the measurement of discrimination's effects as answering a counterfactual question: How different would an outcome for a target racial group have been had each person instead been a member of the dominant racial or ethnic group, but otherwise the same? The effect of discrimination on an outcome is based on the difference between the outcome that actually occurs and the outcome that would occur in the counterfactual state. In U.S. society, this question is how different would an outcome have been for a non-White individual if that person had been White? The basic problem in measuring discrimination is that we do not observe the outcomes for each non-White person if they were White but otherwise the same.

Techniques of Measurement

Because we do not directly observe both states of the counterfactual, we must instead rely on indirect techniques to measure discrimination. Four types of methods are commonly applied. They are (1) the statistical decomposition of racial or ethnic differences in an outcome characteristic like wages or earnings; (2) reports from perpetrators, either of having committed acts of discrimination or measures of prejudicial attitudes thought to predict discrimination; (3) reports from targets of having been discriminated against; and (4) field audit methods that employ mock applications sent out by investigators to measure discrimination.

...