This paper investigates the impact of referral-based hiring systems on workforce diversity and output through an experiment in a manufacturing firm in India and a lab-in-field extension. At the firm, I directed a disproportionate number of referrals towards workers from minority lower caste groups to assess how this affects worker composition, retention, productivity, and team cohesion. I find that increasing referral opportunities for lower caste workers significantly raises their representation and improves retention. Treatment teams experience improved output, driven by tasks requiring high levels of coordination. In the lab-in-field extension, I show that referral-based hiring does not depress cohesion (and thus output). These findings suggest that referral-based hiring programs can promote diversity without short term cohesion or productivity costs. These results have broader implications for designing labor market interventions in stratified societies.

Land is power. Governments and revolutionaries have understood this for centuries, but the causal impacts of land concentration are notoriously difficult to study. We study how differences in village land concentration stemming from the granting of feudal titles hundreds of years ago affect service delivery and labor markets in the present day. A fertile literature evaluates the effects of land tenure systems on agricultural productivity and downstream economic outcomes. However, most of this literature focuses on colonial and post-colonial land tenure policies, and evaluates a narrow set of agricultural and policy outcomes. We exploit variation in pre-colonial land tenure systems at a vastly more granular level than is seen in the literature to evaluate the impacts not only on service delivery but also labor markets. We implement a regression discontinuity along feudal borders that no longer correspond with modern administrative boundaries. Large discontinuities in land concentration persist across these boundaries. These differences are associated with 7% lower agricultural wages for women, but not men who are more able to travel and seek outside options. The main government scheme meant to provide an outside employment option is less well implemented in these areas with 71% fewer person-days offered during peak agricultural months and no difference the rest of the year. This work stresses the long lasting effects of land inequality and suggests that ensuring workers have more outside options can help reduce persistent rural inequities. 

We study store credit, a deferred payment system offered by small businesses to customers across the developing world. We collected data from local shops in an urban Indian settlement, randomly offering subsidies for stores to provide either store credit, a price discount, or a business-as-usual control. Store credit increased businesses’ market share by encouraging more visits and higher spending. Even after subsidies ended, stores continued to extend credit to treated customers. Customers who received credit during the experiment were equally likely to repay as stores’ standard credit customers and shifted some spending from non-credit stores to credit-offering ones. We find suggestive evidence that credit helps customers smooth consumption, and increase consumption expenditure overall. Our results underscore the role of small businesses as local lenders and explain the prevalence of store credit for consumption smoothing and market access in developing countries. We suggest that customers in these settings have an unmet demand for credit but struggle to demonstrate creditworthiness, leading stores to under-experiment with lending. These findings indicate potential for increased credit access by subsidizing business experiments and reducing lending default risks.

We use admissions lotteries to estimate the long-run effects of a residential middle school system targeted at historically disadvantaged communities in India. The school system boosts years of schooling, test scores, college attendance, senior secondary school graduation, and labor force participation. However, we also find that treated students have substantially smaller and mroe caste-homogenous networks. While treated students have better observable employability, their labor market outcomes are weakly worse, likely due to slower job arrival rates as a consequence of more homogenous networks. Our findings illustrate the impacts of delivering high-quality educational infrastructure at historically disadvantaged groups, while also highlighting the importance of measuring long-term and non-test score outcomes in evaluating the effectiveness of education programs.