UC Berkeley

Can personalized mentorship by experienced workers improve young job seekers’ labor market trajectories? To answer this question, in the first chapter of this dissertation I study “Meet Your Future”, a mentorship program we designed and randomized which assisted a subset of 1,112 vocational students during their school-to- work transitions in urban Uganda, where youth unemployment is high. The program improved participants’ labor market outcomes. Relative to the control, mentored students were 27% more likely to work three months after graduation; after one year, they earned 18% more. Call transcripts from mentorship sessions and survey data reveal that mentorship primarily improved outcomes through information about entry level jobs and labor market dynamics, and not through job referrals, information about specific vacancies, or through building search capital. Consistent with this finding, mentored students revise downward their overly optimistic beliefs about starting wages and revise upward beliefs about the returns to experience. As a result, they lower their reservation wages and turn down fewer job offers. The results emphasizes the role of distorted beliefs among job seekers in prolonging youth unemployment and proposes a cost effective and scalable policy with an estimated internal rate of return of 300%. In the second chapter of this dissertation, I move to investigate whether hiring processes themselves can disadvantage women and consequently explain part of the gender wage gap and the occupational segregation documented in many labor markets across the world. Specifically, I look at referrals, a significant factor in hiring decisions and one of the primary ways to land a job. With my coauthor, we conduct a correspondence experiment to examine how referrals by firm employees may perpetuate occupational gender segregation among Uganda’s skilled workers. We start by presenting pairs of gender-differing profiles of potential candidates to workers in a wide range of industries, and ask who they would refer to their firm for an internship we subsidize. We randomize the gender of the high-experience profile to elicit discriminatory preferences while mitigating endogeneity in network formation. To validate the findings in this anonymous setting we subsequently offer participants the possibility to use their networks as referral choice sets, the lifelike setting. We further randomize the disclosure of the referral source’s name to the employer. We document three facts. First, discrimination in referrals exists against both genders and is correlated with subjects’ gender and the gender dominance of their sector; however, discrimination against the non-stereotypical gender is more prevalent in male-dominated sectors. Second, the intrinsic preferences of employees are a significant driver of their discrimination in referrals, which, in general, do not simply reflect passthrough from employers’ preferences. Thirdly, when the referral is private, subjects in male dominated sectors are more likely to refer women, indicating that beliefs regarding employer preferences are a significant driver of pro-male bias in these sectors. In the last chapter I investigate gender disparities in the effect of COVID-19 on the labor market outcomes of skilled Ugandan workers. Leveraging a high-frequency panel dataset, my coauthors and I find that the lockdowns imposed in Uganda reduced employment by 69% for women and by 45% for men, generating a previously nonexistent gender gap of 20 p.p. Eighteen months after the onset of the pandemic, the gap persisted: while men quickly recovered their pre-pandemic career trajectories, 10% of the previously employed women remained jobless and another 35% remained occasionally employed. Additionally, the lockdowns shifted female workers from wage-employment to self-employment, relocated them into agriculture and other unskilled sectors misaligned with their skill sets, and widened the gender pay gap. Pre-pandemic sorting of women into economic sectors subject to the strongest restrictions and childcare responsibilities induced by schools’ prolonged closure only explain up to 65% of the employment gap.