Working Papers
"The Costs of Affirmative Action: Evidence from a Male-Favoring Quota" (Job Market Paper)
SEA2024, AEA2025 Poster session, PacDev2025, MWIEDC2025, CEA2025, WEAI2025, SOLE2025, AYEW2025, NEUDC2025
Working Papers
"The Costs of Affirmative Action: Evidence from a Male-Favoring Quota" (Job Market Paper)
SEA2024, AEA2025 Poster session, PacDev2025, MWIEDC2025, CEA2025, WEAI2025, SOLE2025, AYEW2025, NEUDC2025
Hiring quotas can reshape who enters the workforce and how organizations perform. Such policies are often intended to redress disparities arising from past discrimination. I study a gender quota that instead preserves the representation of a historically dominant group: men in China's civil service. After merit-based hiring boosted women’s share of new recruits, one-third of county tax bureaus introduced one-to-one gender quotas mandating equal hires by sex. Using a new dataset I built through web-scraping and digitizing position-level civil service exam records and county tax revenues, I show that counties with more recent female hires were more likely to adopt quotas, especially in tax-collection roles. Exploiting variation in the timing and intensity of adoption across counties, I use a matched staggered difference-in-differences and event-study framework to show that quotas decrease female representation, lower candidate quality, and reduce worker productivity: tax revenue falls by 4.7%, or about $4 million per county per year. Evidence on mechanisms suggests these effects stem from reduced personnel ability and gendered institutional practices. These findings show that “reverse” affirmative action can be costly when institutions that serve advantaged groups persist.
This paper studies how online job boards affect gender disparities in employment. While men often benefit from broader referral networks, women may rely more on transparent information provided by job boards. Exploiting geographic and temporal variation in the rollout of one of China’s largest online job boards, we construct an expansion timeline using historical screenshots from the Internet Archive and link it to restricted data from the monthly Urban Household Survey. We find that the job board’s introduction reduces the gender employment gap by increasing women’s employment, with no significant effect on men. The effects are strongest among less educated and married women but vanish for those with young children. A placebo test using a real estate platform rules out confounding from general internet access. Evidence points to two main channels: improved access to job information and greater mobility to find jobs.
"Timing Motherhood: A Field Experiment on Job Prospects" (with Myongjin Kim, Daniel Nedelescu, and Leilei Shen)
Young women face a dilemma over when to enter the labor market: before or after having children. This study implements a correspondence experiment in China’s accounting sector, randomizing marital status and number of children on fictitious resumes to signal the likelihood of maternity leave under the national family planning policy. We submitted 12 applicant profiles to 985 entry-level accounting job postings and analyzed employer callbacks to assess how marital and parental status shape hiring decisions. The results show that marriage lowers women’s callback rates by about 3 percentage points, driven by discrimination against married women without children. The penalty disappears once women have reached the maximum number of children permitted by policy. A simple hiring cost model suggests that employers’ concerns focus on anticipated maternity leave costs rather than productivity losses among working mothers. Notably, men face no comparable marriage and childbearing penalty.
"Extreme Heat and Public Services: Evidence from Vaccinations in Pakistan" (with Karrar Hussain, Muhammad Yasir Khan, and Gabriel Tourek)
Extreme temperatures pose risks for public services where delivery mechanisms depend on physical effort. This paper studies how temperature affects access to health services in the setting of field vaccinations in Pakistan, a nation plagued by extreme heat. We find that occurrences of extreme heat reduce completed vaccinations in a day by 8% translating to reductions in timely vaccination of newborns for recommended doses by 5.4 to 7.4%. Vaccinator and citizen effort appear to both contribute to these patterns. These findings highlight the susceptibility of key inputs to public health to the rising prevalence of extreme temperatures.
Publications
"Credit Cards, Risk Coping, and Stock Market Participation in Urban Households" (with Lihe Xu and Qing He). 2019. Journal of Financial Research (in Chinese), 465(3): 149-167