How Can We Know if There is Discrimination in Hiring?

The Issue:

Groups of workers that differ by personal characteristics such as gender, race, ethnicity, sexual orientation, past criminal record, physical appearance or disability often exhibit measurable differences in wages and employment. However, it is difficult to identify and measure the extent to which differences in wages and employment are due to discrimination. This is because differences in labor market outcomes between groups may arise for a range of other reasons, such as differences in average experience or education between groups. Experimental approaches to uncover discrimination in hiring, however, offer a more robust method for uncovering some types of discrimination.

Using experiments, economists are finding evidence of pervasive discrimination in hiring for most groups that experience lower wages, as well as other groups.

The Facts:

• It is hard to measure discrimination. While data on wages is generally consistent with wage differences across groups, it is difficult to prove this is evidence of discrimination. In my review of economic research on discrimination, I note that there has been a difference of 20 percent between the earnings of black and white men in the past decade, and a persistent pay gap between men and women, currently at 22 percent. Statistical analyses of wage data attribute some portion of these differences to differences in education and experience in the labor market across black and white men, and across men and women. Even taking into account these differences, however, there seems to be a wage differential associated with race or gender, although the estimated difference due to these factors is much less than the 20 or 22 percent overall difference and it is difficult to prove that the differences attributed to race or gender are not capturing other, non-measured differences in individuals.
• As an alternative approach, there has been substantial growth in research that conducts experiments to uncover labor market discrimination. These experiments focus on hiring rather than wages or promotions. There is a range of methods that economists have used. In vignette studies, researchers simulate employers’ personnel decisions by giving participants hypothetical scenarios and asking them to select job applicants to hire or promote. Another type of experiment uses fake job applicants to apply for and interview for jobs. The fake applicants are created to differ only by the characteristic under investigation, for example, by race to study racial discrimination, but otherwise to be identical in characteristics that matter for the job.
• Experimental research generally indicates that hiring discrimination is pervasive, with discrimination in hiring for most groups that experience lower wages, as well as other groups thought to suffer from discrimination (based on non-experimental evidence). Researchers find evidence of hiring discrimination against people based on race, ethnicity, prior felony convictions, and disability. A recent study conducted on a massive scale provides compelling evidence that older workers experience age discrimination in hiring for the lower-skilled types of jobs they typically apply for later in life. Older women experience more age discrimination than older men. The study finds, for example, that the callback rate for an administrative job was 47 percent lower among older women as compared to younger women, and 30 percent lower for a sales job for older men as compared to younger men.
• Experimental research suggests men and women face hiring discrimination dictated by social norms. Despite a great deal of statistical evidence consistent with wage discrimination against women, experimental research does not indicate that women are uniformly discriminated against in hiring decisions. Instead, the evidence suggests hiring discrimination consistent with sex stereotyping of jobs. That is, men are less likely to be hired for a female-dominated job, while women are less likely to be hired for a male-dominated job.
• Early experimental research may overstate the near-uniform findings of discrimination against minorities. I recently used researchers’ data to re-examine evidence from field experiments to correct for a source of bias that that could affect the estimation of the extent of discrimination. I found results from about half of these studies of discrimination against ethnic, racial, or sexual orientation minorities are not robust when correcting for this possible bias and, in these cases, the original findings of discrimination were reversed. It seems reasonable to conclude that the overall evidence of labor market discrimination from experiments may not be as overwhelming as previously thought, although many studies continue to point to discrimination.
• Experimental research needs to make more progress to give insight into the type of discrimination in hiring. Taste discrimination is tied to prejudice, that is, employers dislike hiring a particular group. Statistical discrimination, however, is not tied to prejudice, but instead occurs when employers use stereotypes about the productivity of a group and assume it holds for a particular person. This distinction matters because the policy implications differ. If there is taste discrimination, it would be helpful to put rules in place to raise the cost of discrimination, but for statistical discrimination, it would be helpful to put rules in place that provide more information about each individual job applicant. Experimental research has tried to distinguish between taste discrimination and statistical discrimination. However, distinguishing between the two types of discrimination is a challenge, as it requires strong and likely untestable assumptions about what employers know about job applicants, and when they learn it. Regardless, either type of discrimination is illegal under U.S. law, so even if experimental studies cannot identify the type of discrimination, they are detecting discrimination that policy is attempting to reduce or eliminate.