UC Santa Barbara

This dissertation focuses on the job searching and matching in online labor markets and explores research topics including workers' job search behaviors, employers' recruitment decisions, and the role of internet job platforms in online job matching, based on the field experimental data and internal data from job boards.

In the first chapter, joint with Peter Kuhn and Kailing Shen, we study how explicit employer requests for applicants of a particular gender enter the recruitment process on a Chinese job board, focusing on two questions: First, to what extent do employers’ requests affect the gender mix of a firm’s applicant pool? Second, how ‘hard’ are employers’ stated gender requests-- are they essential requirements, soft preferences, or something in between? Using internal data from a Chinese job board, we estimate that an explicit request for men raises men’s share in the applicant pool by 14.6 percentage points, or 26.4%; requests for women raises the female applicant share by 24.6 percentage points, or 55.0%. Men (women) who apply to gender-mismatched jobs also experience a substantial call-back penalty of 24 (43) percent. Thus, explicit gender requests do shape applicant pools, and signal a substantial but not absolute preference for the requested gender.

The second chapter, based on joint work with Peter Kuhn, Taoxiong Liu and Kebin Dai, studies how workers make voluntary wage disclosure decisions in the job search process using internal data from a leading online Chinese job board, Liepin.com. We find that on average, workers' disclosure decisions are consistent with a model in which high current wages are seen as "good news" by prospective employers: Workers are more likely to disclose their wages when their wages are higher than might be expected, based on the worker's resume and where they applied. Employers' responses to workers' resumes, however, are hard to reconcile with these disclosure patterns: While employers respond positively to workers with higher-than-predicted current wages, they do so equally, regardless of whether those wages have been disclosed. This suggests that firms can infer the unobserved ability associated with a worker's current wages from other aspects of her resume and application behavior. Finally, the act of disclosing one's current wage --regardless of its level-- appears to reduce firm's interests in hiring a worker. Disclosures of low wages (which are rare) appear to be mistakes (because they reduce both application success rates and offered wages); disclosures of high wages may, however, benefit workers by filtering out unwanted low-wage job offers.

The third chapter investigates gender bias in job recommender systems. By conducting an algorithm audit in four Chinese job boards, I find that gender-specific jobs, which are only displayed to one gender, account for 9.72% of the total recommended jobs to identical male and female applicants. Gender-specific jobs differ in both the job's explicit quality and the words used in job descriptions: Compared to jobs that are only recommended to men, only-to-women jobs propose lower wages, request fewer years of working experience, are more likely to require literacy skills and administrative skills, and tend to contain words related to feminine personality, which reflect gender stereotypes in the workplace. Item-based collaborative filtering, content-based recommendation algorithms and the hiring agents' behaviors incorporated in job recommender systems are the possible drivers of the gender bias in job recommendations.

To social science researchers, the recommendation algorithms used by job boards to recommend jobs to workers are a proprietary ‘black box’. To derive insights into how these algorithms work, we conduct an algorithmic audit of four Chinese job boards, where we create fictitious applicant profiles and observe which jobs are recommended to profiles that differ only in age and gender. We then estimate the cumulative effect of pursuing these recommended jobs by applying to them in up to three rounds of successive applications. Focusing on the jobs that were recommended to just one of the two genders that applied, we find that only-to-women jobs propose lower wages, request fewer years of working experience, and are more likely to require literacy skills and administrative skills. Only-to-women (men) jobs also disproportionately contain words related to feminine (masculine) personality characteristics, as measured by three distinct approaches for identifying such characteristics. Finally, we assess the patterns in the recommendations generated by our audit study for their consistency with four processes the algorithms could be using: item-based collaborative filtering, content-based matching, matching based on recruiters’ profile views, and rules-based matching based on employers’ stated gender preferences. We find evidence suggesting that the algorithms are relying on all but the last of these processes.