So. Luigi Zingales is in the news, as you might have heard, for inviting former Breitbart executive chairman and short-lived Trump administration Steve Bannon to debate at the University of Chicago. If you haven’t heard, my old college paper is doing an outstanding job reporting on it.
I’ve written about Zingales’s work before. The wisdom of inviting Bannon to a debate under the university’s aegis aside, Zingales is an outstanding economist, and his interest in Bannon’s views follows from his deep concern about populist movements and the kind of governments they support. Back in 2009 he sounded the alarm of how the Great Recession and the Obama administration’s response could paradoxically bolster crony capitalism while engendering rage against it. There’s a lot that’s prescient about his argument, and it’s grounded in Zingales’s experiences as an Italian living under the Trump-like figure of Silvio Berlusconi.
(This is not to say he doesn’t have a blind spot—the Chicago Maroon reported that Zingales reached out to his colleague Cathy Cohen “because he thought that she could make up for his self-admitted ignorance, as a native Italian, of racial issues in America at the debate.” Back in May of 2016, Zingales wrote an op-ed for the Tribune about the similarities between Trump and Berlusconi and conceded that “with all of his defects, Berlusconi was [not] strongly anti-immigrant,” while that was a pillar of Trump’s campaign. This goes a long way towards explaining why so many view Bannon, an architect of the most reactionary parts of Trump’s pitch, as a threat. Anyhow, Berlusconi just threatened to deport 600,000 migrants.)
But, now that I’ve cleared my throat: This post isn’t about that.
While poking around Zingales’s research, I came across a fascinating study that Zingales co-authored with Ernesto Ruben of Columbia and Paola Sapienza of Northwestern, titled simply “How stereotypes impair women’s careers in science.”
It’s a lab experiment that attempts to tease out broader social tendencies, so it pares down a decision about hiring to the bare minimum.
The researchers asked subjects to do basic arithmetic, adding up as many sets of numbers as they could in four minutes, a task for which they had evidence that men and women are equally adept at.
Then they split up the subjects into “employers” and “candidates.” Applicants had the motivation of getting more compensation if they were hired; employers had the motivation of getting more compensation for choosing the “right” candidate, based on who actually did better on a second math test after the “employment” decision was made.
Finally they gave the employers a couple different ways of choosing between candidates:
- “Cheap talk": The candidates told the employers how they thought they’d do on the second test; the employers didn’t get the results of the first test.
- “Past performance”: The employers got only the results from the first math test.
- “Decision“: Just based on appearance alone.
The results align with how people tend to think of these subjects subjectively. The most favorable situation for male candidates was when employers made the hiring decision based on appearance alone, with men being twice as likely to get the job. Female candidates got the job just 34 percent of the time, even though, with literally no other information, it should theoretically be fifty-fifty. And employers made the wrong decision—picking the low performer—45 percent of the time. When they did make the wrong decision, it went in favor of men 70 percent of the time.
Of course, the employers weren’t given anything other than their stereotypes to make the decision. How’d they do when they were given past performance data?
A lot better, but still, in the words of the researchers, “far from optimal.” Female candidates got the job 43 percent of the time, and employers only chose the low performer 20 percent of the time. When they made the wrong decision, it favored the man 64 percent of the time—far from even odds, but better than the appearance-based scenario.
Perhaps most interesting is the “cheap talk” scenario, which gave candidates the chance to put their thumb on the scale. Men took advantage of that, consciously or not: They overestimated how they would do on the second test by three correct answers. Women overestimated their performance by 0.44 correct answers—in short, they were mostly accurate.
Employers were more likely to make the right decision in the “cheap talk” scenario, choosing the low performer just 31 percent of the time, a substantial improvement over the 45 percent failure rate when they had no information about appearance.
But when they did screw up, the advantage went to the lower-performing man almost every time: 92 percent. Self-reporting gave employers some good information—there’s only so much you can do with BS—but it seems to have given men a huge advantage in the edge cases.
It mirrors the societal dialogue about the subject in interesting ways. Is there bias? Yes, and it’s especially apparent when you tease the conditions to bring it to the forefront. Should women lean in? It does appear that they would benefit from it.
And the best-case scenario, unsurprisingly, is when applicants are judged on their work product alone. But even then, unsurprisingly, the playing field still isn’t level.