Ongoing work:
Too few migrants? (with Anvita Ramachandran, Carlo Perroni, Vidya Mahambare, Amrita Dhillon and Sowmya Dhanaraj): Indian labour-intensive manufacturing firms frequently report wanting to employ more migrant workers than they can recruit, suggesting that migrant labour is particularly valuable to them. This paper investigates the sources of this migrant premium using daily attendance, employment and production records from two textile factories in Tamil Nadu over six and a half years. We estimate a structural model in which resident, in-state migrant and out-of-state migrant workers differ in their contributions to production, attendance and turnover dynamics, overtime provision, and recruitment and retention costs. Migrant regular hours are associated with substantially greater output than resident hours, but the estimated differential is sensitive to the assignment of workers across production activities, particularly in weaving. Migrants also provide a disproportionate share of overtime in response to unexpected labour shortages, although overtime replaces only a small fraction of lost hours. These findings suggest that the value of migrant labour depends not simply on worker productivity, but also on how firms organise production and respond to staffing shortages. The model quantifies the contribution of these different margins to the migrant premium under observed production arrangements.
The Cost of Competing for Government Jobs: Competition, Application Intensity, and Recruitment Design: (with Abdou Musonera Aimable Nsambina and Regine Roren): Public-sector recruitment must balance broad access with effective selection, while limiting the costs borne by applicants and government. Lowering the cost of applying can broaden participation, but it can also increase search and the number of applications that government must process and screen. We study this trade-off through reforms to Rwanda’s centralised recruitment system that moved applications online, introduced computer-based testing, and later, added psychometric screening. Applications per vacancy remained around 140 after computer-based testing, even as both applications and vacancies expanded sharply. This stable ratio masks large changes in who applied and how they searched. The female share of applications for teaching positions rose from about 12 percent to 45 percent, participation also increased among applicants with secondary education or less, and applicants frequently pursued several vacancies at once. The reforms let us study how recruitment technology changes who enters the process and how applicants are screened, but they do not tell us which combination of access and screening works best. We develop a model of application choice to compare recruitment designs that vary the costs of applying and the way applicants are screened to study these issues.