Latest Posts

Categories

Tags

Twitter Feed

Can You Compare Jennifer and Tawanda?

April 16, 2024 laurawattenberg 6 Comments

Can You Compare Jennifer and Tawanda?

April 16, 2024 LauraWattenberg 6 Comments

A major economic study found that employers respond differently to resumes with “Black” and “White” names. But which names, exactly? A closer look raises questions.

Names send messages, and we can’t help receiving them. We form impressions of age, gender, race, class, culture, even personality just from people’s names. That makes name-based prejudice a constant risk.

A 2022 study from the National Bureau of Economic Research measured the effects of name-based prejudice in hiring. The researchers sent out over 83,000 job applications with fictitious resumes, each bearing a full name chosen to signal race and gender. The results seemed clear: a significant disadvantage for Black names in the job market. In a highly publicized new follow-up, the authors have identified the companies with the biggest racial disparities.

I was immediately curious. Names are the very heart of this study, the variable the economists used to distinguish between conditions. Which first names did they use, and how did they choose them?

THE NAMES

The article clearly lays out the way the authors identified the “Whitest” and “Blackest” names among a sample of people born in a particular time period. It doesn’t indicate that they controlled for other ways the names might vary.

When you set up an experiment, you take care to balance or avoid any outside factors that might skew the results. In short, you ensure you’re making a fair, apples to apples comparison. In the resume study, the researchers took steps like randomizing the assignment of names to resumes. But were the Black and White names themselves fair comparison sets? Did they differ in any way other than race?

Judge for yourself. Here is the full list of female names they used.

BLACKWHITE
AishaAllison
EbonyAmanda
KeishaAmy
KenyaAnne
LakeishaCarrie
LakeshaEmily
LakishaErin
LashondaHeather
LatashaJennifer
LatishaJill
LatonyaJulie
LatoyaKristen
LawandaLaurie
PatriceLori
TamekaMeredith
TamikaMisty
TanishaRebecca
TawandaSarah
TomekaSusan

By the time I hit the third spelling of Lakesha, I started to suspect that the lists might not be comparable at all. For future researchers hoping to do science with names, here’s a roundup of some of the biggest confounding variables I would worry about in this study.

POPULARITY AND FAMILIARITY

Familiar, popular names are seen as likeable and approachable. A 2008 study found that the commonness of a name strongly predicted whether subjects liked a job candidate and would want to hire them. Notably, common vs. unique name was found to be a bigger factor than the perception of a candidate’s race.

This graph shows the average historical popularity of the Black and White female names in the resume study. The White names are 25 times as common.

AGE CUES

The two curves in that graph peak at the same time period. The difference is, in the Black name graph the peak is all there is. Names like Lakesha and Tawanda disappeared from use years ago.

That leaves the Black names strongly date-stamped. While a job applicant named Allison could be any age, a Tawanda is reliably in her 40s or 50s. And older-sounding names are known to be a disadvantage to job applicants.

SOCIOECONOMIC CUES

Affluent, highly educated parents tend to choose more traditional names and spellings. People perceive such socioeconomic cues in names, and they act on them. One study found that when a child’s name suggests low parental education, teachers treat the child as having lower potential, even compared to their own siblings with more conventional names.

The White lists in this study are studded with age-old classics like Anne and Sarah. In contrast, none of the names on the Black female list even existed in American baby name stats in the 1900s. The Black names also feature more prefixes and suffixes and creative spellings, markers that correlate with lower income and education. The result is two lists which differ profoundly in socioeconomic signals, above and beyond race.

NAME STRUCTURE

The Black and White names diverge on concrete measures of length, syllables and stress patterns. These basic factors can be surprisingly consequential. For example, almost half of the White male names are very short or monosyllabic, including nicknames (e.g. Greg, Jay), but none of the Black names fit that description. Simple names and nicknames are considered most approachable and reliable. That’s why politicians and salespeople tend to go by them.

Meanwhile, three quarters of the Black male names are stressed on the second syllable, vs. none of the White male names. In fact, two-syllable names with that stress pattern are hardly ever given to White American boys. (I’ve written in the past about how the pattern is used to signal “otherness” in racist urban legends.) Such names are likely to be less familiar and comfortable to white hiring managers. That’s a bias that’s intertwined with race, but not the same as race itself. For instance, second-syllable stress is also common in names from many other cultures.

ASSESSING THE RESULTS

The researchers clearly demonstrated name-based prejudice in hiring decisions. But they weren’t interested in names per se; they only used them as representations of race. Given the broad incomparability of their name lists, I’d hesitate to draw any conclusions about the specific role of race in their results.

I reached out to the authors of the study, who kindly responded to my concerns. Here is their explanation of how they controlled for confounding variables:

“We find no evidence whatsoever that employer contact rates vary within race and sex categories despite the potentially sizable differences in socioeconomic background conveyed by these names. It therefore seems unlikely that the differences we find reflect factors other than employer perceptions of race or sex.”

In case that’s not clear, here’s how I read it:

  • The names within a given list, e.g. Black Female, could, potentially, be extremely diverse on dimensions other than race and sex that would affect response rates.
  • Response rates within each list did not vary dramatically.
  • Therefore race and sex were the factors that produced different results between lists.

The problem is, the individual lists aren’t actually diverse on other dimensions. The Black Female list in particular is almost comically uniform. And the differences between groups systematically dwarf any within-group diversity.

Are factors like name popularity and socioeconomic cues correlated with race? Sure. Could race be the sole factor driving the results? That’s certainly possible. We can’t know for sure, though, because the economists—who were meticulous in crafting the economic aspects of their study—didn’t control for any of the obvious disparities in their first name selections. And again, names were the entire experimental manipulation.

Names are a valuable tool in a social scientist’s toolkit, but not a simple one. They’re rich, multi-dimensional signals, and many of those dimensions are interrelated and culturally potent. If you’re going to base a research study on names, start by taking the names seriously.

LauraWattenberg
LauraWattenberg

Namerology founder and "Baby Name Wizard" author Laura Wattenberg is a globally recognized name expert, known for her scientific approach to understanding name trends and culture.

All posts

6 Comments

  • TheOtherHungarian
    TheOtherHungarian April 17, 2024 at 2:53 am

    Thank you! The minute I saw the articles about that study start to circulate, I was wondering, what names did they use, and did they do their due diligence to make sure that race was the only variable? I’m disappointed to see that not only did they fail to do so, they failed to understand or even acknowledge their mistake.

    Given the names on the “white” list, what list of “black” names would you have paired with it? (The other way around might be interesting, too, but for a different reason: it would maybe demonstrate how out-of-touch the researchers were with actual name demographics.)

    • LizA April 17, 2024 at 1:12 pm

      That is exactly what this great post made me wonder! What could have been good comparisons, Laura?

    • Namerology
      Namerology April 17, 2024 at 9:01 pm

      I’ve been thinking about that! There simply aren’t any distinctly Black names that can compare with the historical popularity and familiarity of a classic like Anne or Sarah. (In fact, names like those are a scarce commodity, period.) So instead of trying to make the Black names match the White names, I would have changed the White names to match the Black names.

      At the very least, when the economists were tallying name frequencies by race they should have noticed the overwhelming discrepancy in total usage and capped the allowable frequency of the White names to be within the same range as their Black selections. Also eliminate nicknames, because the familiarity & popularity of a name like Greg is based on the more common Gregory. And getting rid of the nicknames would have the extra benefit of narrowing the simplicity/friendliness gap. Those two basic steps would have gone a long way toward reducing confounds. I don’t have access to the data they used, but I suspect that names like Christen, Kristine and Krystina would be a decent comparison set for Lakisha, Lakeisha and Lakesha.

      • holey
        holey April 17, 2024 at 11:18 pm

        Agreed. I think it would be a lot more logical to compare made-up names that we associate with the black American community to made-up names that we associate with the white American community. And make sure that the names being compared have a similar popularity charting (in terms of slope, peak level AND peak decade).

        Probably you’d want to come up with something pretty date-stamped, made up of popular name components, and – well – probably mentally associated with poor rural areas with a majority white population. Maybe check the West Virginia name stats for a particular time period and see which names were disproportionately popular there! Something like Jana, Sherilyn, Marnie, Misti, Loriann? Lakesha vs. Loriann seems a lot more fair than Lakesha vs. Emily!

        I want to apologize for potentially being offensive about states like West Virginia. Since we’re dealing in stereotypes, my suggestions are meant to reflect and cater to those stereotypes.

  • HungarianNameGeek
    HungarianNameGeek April 19, 2024 at 6:47 pm

    I looked up the male names in the PDF.

    Per NameGrapher, the lowest-ranking “white” male names are Geoffrey, which peaked at 192nd in the 1970s, Neil, which peaked at 157th in the 1930s, and Brad and Brendan, which peaked at 123rd in the 1970s and 1990s, respectively.
    The _highest_-ranking “black” male names are Leroy, which peaked at 53rd in the 1920s, Maurice, which peaked at 101st in the 1910s, and Reginald, which peaked at 131st in the 1960s.

    In other words, there is no comparing the two lists on the “familiarity” axis: even the lowest-ranking “white” names rank higher than all but two of the “black” names — and those two exceptions are way out of range of the intended mid-1970s demographic, sounding more like great-grandpas than job applicants. (The “white” list skews the other way, with names like Jacob [the top name among current middleschoolers], Joshua and Nathan [peak popularity in the 2000s, Joshua at number 3], and Jason and Matthew [peaked at 3rd, in the 1970s and 80s, respectively].)

    So yeah, the study unquestionably showed that names affect hiring — people with familiar names are more likely to be hired.

  • holey
    holey April 21, 2024 at 11:22 am

    I notice that the study also assigned different surnames based on race. In one way that’s a good idea. However, as with the first names, the surnames they picked aren’t totally comparable. On average, the “white” surnames are more common. Plus, one of the “black” surnames is Muhammad, and if that one invites discrimination it could certainly be due to factors other than race! How could the researchers have decided to leave that in there?

    It was funny to me that having a Chief Diversity Officer was correlated with an increased difference in callbacks. Realistically, though, if there is a cause and effect it’s probably the other way around: the company is aware of a discrimination problem and therefore has a CDO.

  • Leave a Reply

    I accept the Privacy Policy

    ×