We investigate how worker absenteeism and turnover affects firm productivity, and whether employing a mix of migrant and local workers can mitigate the costs of labour input volatility. Using detailed administrative data from a large textile firm in southern India, we develop and estimate a recursive framework in which a firm chooses its composition of resident and migrant workers to minimise expected costs while meeting a target level of output. Since output is generated through a concave production function, volatility in attendance reduces output and raises costs. Migrant workers are more consistently present and work longer hours, but they quit sooner than local workers and take longer to be replaced. We combine estimates of the attendance probabilities, exit hazard rates, and production function in a simulation to examine how hours worked, and output vary across different workforce compositions. The combination of migrant and local workers that minimizes input volatility, maximizes output per hour, and reduces labour costs involves a higher share of migrants than is currently observed at the firm. The ‘migrant premium’ is positive, suggesting that an additional migrant is more valuable to the firm at the margin. However, frictions in hiring and retaining migrants -- such as integrating them into the local community -- prevent the firm from operating at this optimal level. 

This paper examines how a worker's peer group affects individual productivity in a labour-intensive manufacturing setting. Using administrative data from a garment factory in South India, I estimate the endogenous peer effect: the impact of peers' contemporaneous productivity on a worker's own output. Identifying this effect is difficult due to selection, correlated shocks, and the reflection problem. I address these challenges by exploiting worker rotation across lines, using training test scores measured before workers interact as a proxy for baseline ability, and instrumenting peer productivity with the training scores of peers-of-peers. I find that a one percentage point increase in average peer productivity raises a worker's own productivity by 0.58 percentage points. By distinguishing between upstream and downstream peers, I show that this effect is not driven solely by the technological complementarity of the assembly line, but reflects a substantial behavioural response.

This paper asks why internal migration in India remains limited despite persistent spatial disparities in economic opportunities. Using ten years of panel data from Andhra Pradesh and Telangana, I find that migrants experience significantly greater consumption growth than non-migrants, yet report no corresponding improvement in subjective well-being. This divergence suggests that the monetary gains from migration may be offset by substantial non-monetary costs. In a qualitative analysis, I investigate the mechanisms behind this divergence in a comparable sub-sample, using semi-structured interviews and focus group discussions. The interviews suggest that non-migrants are not a homogeneous group. Those who had previously migrated reported a lower subjective well-being outside their locality, reflecting incorrect expectations regarding the non-monetary costs or the consumption premium associated with migration. Some respondents consciously chose not to migrate because they perceive that the non-monetary costs outweighed the expected economic gains, while others expressed a desire to migrate but remain constrained by factors unrelated to wage differentials. Hence, the study suggests that a purely monetary or a purely quantitative analysis may misjudge the extent to which migration can serve as a welfare-enhancing strategy.

This paper examines the relationship between human capital and the gender gap in earnings using high-quality panel data spanning 12 years from Ethiopia, India, Peru, and Vietnam. We construct latent stocks of cognitive and non-cognitive skills measured during adolescence, and investigate the relationship between these skills and subsequent earnings acquired in early adulthood. Our results suggest that women earn significantly less than men in all four countries, even after accounting for differences in carefully constructed skill endowments. Interestingly, the gender gap in earnings decreases at higher cognitive skill levels in two out of the four countries. We find that these country-level variations are driven by differences in employment status as opposed to differences in earnings among the employed, and may reflect disparities in unpaid care work. We further explore how the gender earnings gap varies in the context of the COVID-19 crisis. While earnings decreased for both men and women during this period, the pre-pandemic relationships between human capital and gender gaps persisted and were strengthened.