Determinants of Occupational Segregation across Race and Gender: Evidence from Sourcing, Screening, and Hiring in IT Firms

We study the determinants of the underrepresentation of women and racial minorities in tech within the hiring context. Despite the precipitous rise in the use of professional networking platforms for sourcing passive candidates, the hiring literature has largely focused on the screening interface -- who applies to what jobs; who receives an interview once they apply. In this paper, we broaden this focus by studying demand-side choices (i.e. choices of the worker) and supply-side choices (i.e. choices of the employer) in both screening and sourcing interfaces of hiring; and how these choices contribute to the underrepresentation of women and racial minorities in tech. We address this question using two novel, large-scale datasets: Applicant Tracking System data from 8 tech firms containing nearly 900k candidates, and a LinkedIn dataset containing 318 million public LinkedIn profiles. Using matched sample analyses and controlling for a rich set of job and applicant attributes found in candidate's resumes and LinkedIn profiles, we find evidence suggesting that the underrepresentation of women in technical jobs is primarily driven by supply-side choices (i.e. choices of the worker) rather than demand-side choices (i.e. choices of the employer) within the hiring interface. In contrast, the underrepresentation of Black and Hispanic workers in technical jobs is primarily driven by demand-side choices rather than supply-side choices. We find that women are less likely to apply to technical jobs, but those that do are more like to receive a callback and get hired. Passive female candidates are also more likely to be sourced on LinkedIn and invited for a phone screen. On the other hand, Black and Hispanic applicants are roughly equally likely to apply to technical jobs compared to White applicants but those that apply are less likely to receive a callback, interview, and offer. Among passive candidates that are sourced on LinkedIn, Black and Hispanic candidates are more like to accept an invitation for a phone screen and are less likely to withdraw. These results highlight how different mechanisms contribute differently to gender and racial underrepresentation in tech.