In the midst of a national reckoning on race, many companies have reaffirmed their commitment to racial equity and justice. But as business school professors, we're used to hearing that large, prosperous companies are making efforts to diversify their workforces and yet have found it difficult to make progress. What is holding them back?
To find out, we needed to learn more about the hiring practices of these companies, the kind with active diversity and inclusion offices and with social media feeds declaring "Black Lives Matter." What we found was alarming.
These firms said they were actively seeking diverse candidates, yet they still discriminated against people of color and women. Our results suggest they did not even realize that they were biased in their evaluations of potential new hires.
For this research, we examined 72 employers who recruit graduating seniors at the University of Pennsylvania for full-time positions. Among these were Fortune 500 companies, tech firms, big banks, consulting firms, hedge funds and nonprofit organizations.
We gave each employer 40 randomly generated resumes to review. They knew that the resumes were not from actual students but also knew that their ratings of these resumes (on a 10-point scale) would be used by our machine learning algorithm to match them with real candidates. This gave them an incentive to rate the resumes carefully and honestly.
Since the resume components were randomized, employers saw resumes with various GPAs, academic majors, work experiences and skills. We also randomly assigned names indicative of gender and race, allowing us to see if recruiters responded differently to resumes with the same qualifications but different demographic characteristics.
What we found surprised us. While 90% of the employers reported explicitly factoring into their hiring the goal of increasing racial and gender diversity, we saw no preference for female or minority candidates in their ratings. In fact, employers hiring in STEM fields penalized resumes with minority or female names. The effect was big: These candidates were penalized by the equivalent of 0.25 GPA points, based solely on the name at the top of the resume. That meant such a candidate needed a 4.0 GPA to get the same rating as a white male with a 3.75.
Our study suggested that implicit, or unconscious, bias was responsible for this disconnect between intentions and outcomes. Research shows that when decision makers are fatigued or are moving fast, their decisions are more likely to be based on cognitive shortcuts that can introduce bias. Indeed, in our study, discrimination against minorities and women was more pronounced toward the end of each set of resumes the employer viewed, when less time was spent evaluating each one.
Implicit bias, combined with perceived pushes for diversity, can help explain other troubling patterns in the employers' ratings. Employers across all fields gave minority and women applicants less credit for having had a prestigious internship the prior summer. Working for a top employer (such as Goldman Sachs, McKinsey or Google), rather than having a less competitive internship, boosted ratings of white male resumes as much as an extra 0.5 GPA points. Women and minorities got roughly 33% less credit for the same internship. Employers also rated minority and female candidates as significantly less likely to accept a job offer than white male candidates.
It seems that employers who think the firms in their industry are actively seeking diverse candidates assume that a female or minority candidate who snagged a prestigious internship had a leg up, leading them to give minority and female candidates less credit for such internships. Employers may similarly assume that these candidates are being aggressively recruited by other diversity-seeking firms, leading them to think these candidates would be less likely to accept a job, and potentially causing them to extend fewer job offers to these candidates.
The problem with this logic, of course, is that the firms in our study — the kind giving out prestigious internships and going after top candidates — did not rate minorities and women more favorably, and employers recruiting for STEM jobs rated them worse than white males with the same qualifications.
To make headway on diversity, employers must first recognize that they and their competitors may not be as diversity-seeking as they think. From there, firms must take steps to mitigate implicit bias. Attempts to do this in the past, however, have been fraught: Standard forms of diversity training do not appear to work; covering up names does not eliminate race and gender cues elsewhere on a resume; and artificial intelligence recruiting tools can reproduce historic biases.
But technological solutions are improving. And researchers are developing new ways to tackle bias. In our research method, for example, we could measure employer preferences for candidate characteristics (such as GPA and work experience) independently from race and gender. This enabled us to suppress race and gender bias when we matched employers to more than 200 real Penn seniors, allowing us to send the firms more diverse candidates than they would have likely selected on their own. Employers could integrate our method or other similar solutions into their resume screening process.
Bias in resume screening is just one part of the problem. Eliminating bias in employment will require rethinking every aspect of the hiring and promotion process. Our research shows that good intentions and pro-diversity goals aren't enough.
Judd B. Kessler is an associate professor of business economics and public policy at the Wharton School of the University of Pennsylvania. Corinne Low is an assistant professor of business economics and public policy at the Wharton School. This piece was written for the Los Angeles Times.