“Words Matter: Experimental Evidence from Job Applications” (with Alison Stein, Uber)
Women are underrepresented in certain jobs, particularly in STEM fields and the technology sector. It has been claimed that women in tech hold themselves to higher standards than their male counterparts when deciding whether to apply for a job. Thus, job postings that ask for “exceptional” expertise and a slew of bonus qualifications may disproportionately discourage women from applying if women believe they must meet all the listed qualifications. To investigate this hypothesis, we ran a randomized controlled trial on a sample of 60,000 potential applicants to over 600 of Uber's corporate U.S. job postings. Our treatment removed optional qualifications and softened language about the intensity of the qualifications. Job seekers meaningfully respond to language: our treatment significantly increased the total number of applications by 7 percent. Altering the language did not change the fraction of women who applied, but did close the gender skill gap. While female applicants in the control group are 6 percentage points significantly more likely to have graduate degrees than men applying for the same job, women and men in the treatment group are equally likely to have graduate degrees. Our results confirm the importance of language in the self-screening process: words matter in different ways for women and men of different educational backgrounds, and materially affect job seekers' economic outcomes.
RESEARCH IN PROGRESS
“Gender Differences in Performance Evaluations”
I use a proprietary dataset from an HR tech firm to examine gender differences in over 100,000 performance evaluations for a sample of 200 companies. Performance evaluations are an important aspect of labor markets, as they impact decisions related to worker productivity, compensation, promotion, and long-term career choices. Preliminary findings reveal that there are gender differences in the way that females and males publicly present themselves in formal evaluations. In particular, the gap between workers’ numerical self-assessment of their overall work performance and their manager’s assessment of them is more negative for female workers than for male workers, indicating that females rate themselves lower than their male counterparts, even after controlling for their manager’s beliefs. Future work will explore mechanisms for this finding, including gender differences in signaling and self-promotion.
“The Gender Earnings Gap within Firm and Job: Evidence from Online Self-Reports” (with Matthew Gibson)
Using a proprietary dataset of online self-reports, we provide evidence on the gender wage gap after controlling for detailed occupational- and firm-fixed effects, typically not permissible in publicly available data. We validate the sample by comparing it to Occupational Employment Statistics and Current Population Survey data. Estimates of the gender wage gap range from 5 to 7 percent after including detailed controls, smaller than in prior studies. This work illuminates the value of confidential, proprietary data to more accurately quantify existing gender differences in the labor market.