With relatively few exceptions—such as a casting call seeking an actor to play Tupac—it’s been illegal for employers to hire based on race for decades. But a level of discrimination persists; some employers are less likely to call applicants with distinctively black names, for example. And after all this time, even the ethics of the issue are hardly settled: some libertarians and conservatives see civil rights laws as an infringement on economic and associational freedoms, while some on the left endorse hiring discrimination that favors underrepresented groups. An impressive new experiment—from Peter J. Kuhn and Trevor T. Osaki of the University of California, Santa Barbara, written up in a working paper released through the National Bureau of Economic Research—probes Americans’ instincts on when hiring discrimination is wrong. These judgments vary from case to case, and not always in the ways you might think.
The subjects, recruited from Amazon’s “Mechanical Turk” workforce and paid to take a roughly 15-minute survey (so, importantly, they might not be representative of the overall population), were asked to judge hypothetical examples of employment discrimination, with three variables randomly changed. The first was whether the person discriminated against was white or black. The second was the rationale for the discrimination: in economists’ terms, “taste-based” discrimination reflects simple dislike for a group, while “statistical” discrimination is the use of race as a proxy for other traits that might be hard for a potential employer to measure or access, such as productivity or a criminal record. And the third was the strength or “justifiability” of the rationale—whether the decision was more or less arbitrary, or instead rooted in solid evidence that discrimination would benefit the business.
More concretely, the example of taste-based discrimination with a weak rationale was that the business owner just doesn’t “enjoy interacting” with blacks (or whites). The stronger rationale was that the business owner has learned, through focus groups with his customers, that the customers “do not like interacting with Black [or white] people and would be hesitant about continuing to support his business if he employed them.”
The example of statistical discrimination with a weak rationale is that, in a “brief conversation,” the business owner’s neighbor told him that “he once met [another] business owner who had trouble with some Black [or white] employees.” The stronger rationale for statistical discrimination is that a “large and experienced network of local business owners” told the owner that they “have had trouble with a large share of their Black [or white] representatives” and for good measure provided “some reliable statistics from their businesses that verify these claims.”
Hiring discrimination may be illegal, but these scenarios weren’t always seen as problematic by the experiment’s subjects, who ranked each scenario they encountered on a seven-point scale ranging from “very unfair” to “neither fair nor unfair” to “very fair.” Mathematically, the researchers treated the neutral answer as a 0, with “very unfair” counted as −3 and “very fair” as +3—so that if the average response to a question was negative, the situation was seen as unfair on balance.
One intriguing finding is that the distinction between taste-based and statistical discrimination, which looms large for economists studying the phenomenon, didn’t seem to matter. What did matter was the evidence supporting the rationale. Business owners relying on personal prejudice or hearsay from a neighbor were rated poorly, a bit worse than −0.5 on average—though perhaps that’s a surprisingly weak condemnation, given that individual responses can range as low as −3. But those basing their discrimination on customer focus groups or data from other local businesses actually received slightly positive ratings, meaning that they were seen as acting more fairly than unfairly. (In the case of taste-based discrimination with a strong rationale, though, this result wasn’t statistically different from zero.)
The race of the person being discriminated against also had a large effect. Respondents saw discrimination against black workers as worse than discrimination against whites, an effect of about half a point on the seven-point scale. In questions involving anti-black discrimination based on poor rationales, the average ranking was close to −1.
The experimenters also collected some other information from the participants, such as their political ideology. Self-identified conservatives were more tolerant of discrimination in general—but unlike liberals and moderates, they did not react much differently when the discriminated-against job applicant was white versus black. As the authors put it, “Conservatives view discrimination against (fictitious and identically qualified) Black and White job applicants the same way: as more fair than unfair.”
Lastly, the authors compare their results with several different concepts of fairness, to see which ones best fit the data. The results are clearly not driven by a racial in-group bias: white liberals and moderates thought anti-black discrimination was worse than anti-white discrimination, after all.
Another option is “utilitarianism,” by which the authors are not referring to—or even factoring in—the utility of discrimination to the business (though I’d note that such a concept could explain the positive results for discrimination based on solid evidence it will benefit the operation financially). Instead, in their definition, “utilitarian welfare criteria assign higher levels of fairness to outcomes that redistribute wealth or opportunities from people with higher incomes to those with lower incomes.” Obviously, this can explain the special attention liberals and moderates pay to anti-black discrimination relative to anti-white discrimination.
And still another is “race-blind rules,” in which the reasons behind the discrimination matter but the race of the discriminated-against person does not. This is a decent fit for the data in general, and a nearly complete explanation of the results for conservatives in particular: like everyone else, conservatives reacted more positively to discrimination based on better evidence, but unlike moderates and liberals, they didn’t change their judgments based on the race of the person being passed over.
The working paper’s findings, of course, tap deeply into the current debate over race and modern conservatism. Are conservatives wrong not to make special allowances for historically oppressed groups, or are they embracing a reasonable colorblindness?
Normative questions aside, the study adds to our understanding of the different benchmarks Americans use to evaluate discrimination in hiring.
Photo: Abu Hanifah/iStock