How pervasive is hiring discrimination against blacks in the United States? Those who push for more extensive racial intervention efforts by the government claim this sort of discrimination has not left us, that civil rights and equal opportunity laws are not sufficient to eliminate it, and a combination of greater government activism as well as more extensive private initiatives. But are those claims valid?
One of the key pieces of evidence academics rely on to prove ongoing racial discrimination is the resume audit, in which resumes from black and white candidates are sent to prospective employers. The researchers measure how many of each receive calls from those employers. A meta-analysis from 2017 identified 21 different studies using this method; as summarized at the Harvard Business Review, indicated in the title, the evidence of these studies proves that “Hiring Discrimination Against Black Americans Hasn’t Declined in 25 Years” and, therefore, “affirmative action policies” and “more active intervention” are needed to end discrimination.
We’ve seen those calls grow, with increasing reports of critical race theory training in corporate America, which makes it all the more important to look critically behind the curtain at the proof that’s being used.
Consider the most recent such study: in a massive effort, a team of economists at the University of California at Berkeley and the University of Chicago claim to have demonstrated clear racial discrimination against black applicants at 23 large American firms, in which (imaginary) black applicants were less likely to receive responses to their resumes than (imaginary) white applicants, at a difference of 2.1 percentage points. But all is not as it seems.
Here’s how the researchers (and their research assistants) conducted their study. First, they identified 108 Fortune 500 companies, with 125 entry-level jobs at each, and sent 8 fictitious resumes out for each (more or less; this would total to 108,000 but they only sent 83,000 applications altogether because some identified jobs were filled during the process). For each job, they sent four “black” and four “white” resumes; they also randomly varied the age, sex, sexual orientation, gender identity, and political leaning of the imaginary applicants. They then measured the frequency with which the hiring employers contacted them, which was, on average, 24 percent.
Their results? First, they identified the gap in contact rates of 2.1 percentage points, black versus white. They also found that male and female applicants were equally likely to be contacted, although some companies, for some jobs, were more likely to contact men and others women. Applicants over age 40 were less likely to be contacted, with a differential of 0.6 percentage points.
Applications listing belonging to a LGBTQ club experienced no difference, nor did applications identifying “gender-neutral pronouns.” They found no statistically significant differences in contact rates across states, for the gaps for gender or age, but they did see greater racial contact gaps “in geographic areas with more prejudice as measured by Implicit Association Tests (IATs) and internet searches for racial epithets,” and reduced gaps at more profitable companies as well as companies with centralized hiring offices. The racial gaps were also greater in “customer facing sectors including the auto services and sales sector and certain forms of retail.”
What Is in a Name?
In all of the above descriptions, however, I’ve skipped over how “race” is indicated in the resumes. Their method is not the same as their approach towards signaling sexual orientation or politics, that is, involvement in a “Black Students Association” or an “Irish Heritage Society.”
Instead, the researchers assigned both first and last names that are either “distinctively black” or “distinctively white,” with the requirement that “more than 90 percent of individuals with that name are of a particular race.” They also “selected the most common distinctive Black and white names for those between 1974 and 1979”; for last names, they used common names with “high race-specific shares” in the 2010 U.S. Census.
What are those names? The “black” first names include: Aisha, Ebony, Keisha, Kenya, Lakeisha, Lashonda, Lawanda, Tamika, and Tomeka for women, and Antwan, Darnell, Jamal, Kareem, Marquis, Rasheed, Tremayne, and Tyrone for men.
The “white” first names included Allison, Amanda, Amy, Emily, Erin, Jennifer, Meredith, and Susan for women, and Adam, Brad, Geoffrey, Jason, Jeremy, Joshua, Nathan, Scott, and Todd for men.
The “black” last names included Alston, Bethea, Booker, Hillins, Jefferson, Muhammad, Smalls, and Washington. The “white” last names included Bauer, Becker, Burkholder, Carlson, Erickson, Hershberger, Olson, and Schultz.
In other words, with respect to first names, the first list is not simply “distinctively black”; it is distinctively black and poor. That’s not simply a matter of stereotype. A study by a University of California at Los Angeles researcher in 2017 identified this as a serious concern with this form of discrimination testing, citing a previous study that found that “[b]lacker names are associated with lower-income zip codes [and] lower levels of parental education,” and identifying that names of poorly educated black mothers are more likely to be perceived of as distinctively black.
An older study, from 2003, similarly found that in the ’80s and ’90s, naming conventions shifted and “led to a ‘ghettoization’ of distinctively Black names, namely, a distinctively Black name is now a much stronger predictor of socio-economic status” — so much so that that paper’s analysis suggests it is the correlated socio-economic status, not the name, that is behind these lower resume call-back rates.
Second, with respect to the last names, the “distinctively white” names are not just “white,” but signal ethnicity in their own way: virtually all of the “white” last names are distinctively German, with a minority specifically tied to Amish and related communities (Hostetler, Yoder) or Jewish communities (Klein, Meyer). This further muddies the issue because it makes it impossible to identify whether the black applicant was selected against, or whether positive stereotypes about Germans or Jews affected their likelihood of being selected.
On the Basis of Class
Could this study — and a variety of similar studies throughout the years — be fixed? It would be much harder to create a list of “middle-class black” and “middle-class white” names to test against each other, because these are far less likely to be distinctive. Could researchers test stereotypically “redneck” or “white trash” names and see whether Brandy, Crystal, Daisy, Billy, Clyde, Earle, or Travis (to take a few from a list that purports to be “104 Redneck Baby Names for the Bogan In All of Us“) have the same sort of reduced call-backs?
Or, since they tested for discrimination against LGBT people by mentioning membership in a club, could they not have tested paired resumes listing politically neutral clubs that signal race, such as singing in the African Methodist Episcopal choir versus a Methodist choir, or an African dance versus Irish dance club? In fact, in the year 2021, there’s an even clearer way (although admittedly with more work involved) to indicate race: we no longer include photographs on our resumes, but we have LinkedIn and other social media profiles with headshots that achieve the same result.
As it is, this study, and all such similar ones, are at least as likely to test whether applicants are rejected on the basis of social class. While it is a serious concern if lower-economic class workers struggle to find a job, they do have an option of adopting an alternate name for professional purposes, while someone cannot do likewise in regards to skin.
Job discrimination on the basis of race is certainly worth investigating, but it is incumbent on the researchers upon whose work politicians and policymakers rely to be able to correctly diagnose the severity of the problem before providing prescriptions to remedy it. Fundamentally, these resume audits simply do not prove what the researchers claim they do.
