Search and Signaling on an Online Labor Market

This dissertation consists of three papers on search and signaling on a large online labor market. The abstract of each chapter is as follows:Online resume and job post sites like LinkedIn and Monster.com has made it increasingly easy for employers to search for and invite workers who would not otherwise have applied to their job post. I leverage the A/B test of a new resume screening tool on a large online freelancing platform to identify which types of workers and employers are most likely to benefit from lower screening costs in the invitation channel. The screening tool increased the number of workers contacted by 12% but there was no increase in overall hiring. Treated employers substituted away from screening and hiring workers in their existing applicant pool to sending invitations. They hired workers with higher platform reputation at higher cost. Moreover, invitations are more concentrated among a minority of workers than applications. Thus the screening tool likely led to an increase in inequality for workers on the platform. Treated employers also had better job outcomes - at least half of this effect is due to employers avoiding hiring less well matched workers from their applicant pool. Benefits accrue primarily to employers looking for expert freelancers and willing to pay higher prices.How do employers respond to hiring tools? We examine the hiring decision as a process comprised of a series of decisions. We claim hiring indicators act as a “minimal cue” by elevating a job applicant to being noticed early in the hiring process, but giving way to the information an employer privately gathers later in the hiring process. Regression discontinuity analyses of over 1.5 million job applications by freelancers for over 150,000 short-term jobs on an online market for contract labor demonstrate support for our contention. Dramatically, being algorithmically recommended increases a job applicant’s unconditional likelihood of being hired over applicants of observationally similar quality by approximately 50%. 70% of this effect can be attributed to the increase in likelihood of recommended job applicants being viewed. Furthermore, conditional on being interviewed, the recommendation has no influence on the hiring decision. Underscoring the idea of it being a ’minimal cue’, the effect of a recommendation is stronger for low value jobs than high value ones.The rapid growth of online information on workers, like LinkedIn’s profiles or Monster.com’s resume database, has dramatically lowered the cost for employers to directly reach out to (headhunt) workers. The promise is that by providing employers with an additional channel of sourcing matches, it will increase the probability of filling vacancies and the average quality of hires. I construct a search theoretic model of hiring that explicitly models both the headhunting channel and the application channel. I study equilibrium outcomes when workers can optimally respond by deciding whether or not to apply. I show that while lower headhunting costs improve the average quality of hires, it can actually decrease the probability of filling vacancies in equilibrium. This is due to workers optimally choosing to apply with a lower probability