research

In low-income countries, urban employment is driven by small firms that face challenges in hiring, such as costly skill verification and volatile demand. In Uganda, I document an informal solution to these problems where firms in manufacturing clusters subcontract from one another in a system that I call “labor sharing”. I conduct novel surveys of firm owners and workers in two manufacturing sectors and document detailed hiring patterns. I document that labor sharing is a widespread practice with over 80% of surveyed firms recently participating. It shares a common structure as uncompensated, reciprocal, locally-concentrated, and manager-organized. Labor sharing serves two main purposes: it effectively screens workers and provides state-contingent labor, which allows firms to flexibly expand and meet demand shocks. To estimate firm and worker valuations of labor sharing, I deploy a series of stated choice experiments, and find that firms value labor sharing at a 35% premium over other hiring channels, but workers value it similarly to other types of referrals. I estimate a partial equilibrium model of the labor market where firms face uncertainty over consumer demand and worker quality. I show that the labor sharing economy increases total welfare by 10% compared to an economy with no sharing. I use the model to show that labor sharing shapes the impact and incidence of active labor market policies, such as training and screening. Other policies, such as wage subsidies, can backfire and reduce welfare by crowding out labor sharing.

Rationing policies are often designed with equity considerations in mind, yet whether equalizing exposure delivers equal economic impacts remains underexplored. We study electricity outages in South Africa, where rationing equalizes exposure across locations. Combining high frequency outage data with geocoded transactions from over 11,000 firms from 2021 to 2023, we show that equal exposure does not imply equal impact. Average sales remain unchanged, but outages induce reallocation: below-median firms lose roughly 11 percent of average revenue, while above-median firms gain 10 percent. We provide novel evidence that consumer substitution drives these effects and that advance notice amplifies disparities.

We investigate the impact of incentives and their removal on parents' decisions to vaccinate subsequent children. We follow up with parents three years after exposure to a bracelet incentive for timely vaccination in Sierra Leone, leveraging the random assignment of clinics to implement the incentive. Since only parents with a newborn at the time of the experiment were eligible for the incentive, we exploit within-clinic variation in exposure. First, we find that parents who received an incentive for an earlier child are 5 to 11% less likely to vaccinate their subsequent child on time compared to parents in the control group. There are no effects on vaccination rates by 15 months of age, suggesting that parents delay vaccination rather than abstaining altogether. Second, parents living in communities where incentives were implemented but who were ineligible for them exhibit no changes in vaccination behavior, ruling out changes in community norms or clinic as explanations. Third, incentives that were valued as signals of vaccination completion do not result in adverse effects after their removal. Our causal forest analysis suggests that parents with higher vaccination costs are most affected by these negative effects. Taken together, our findings suggest that experience of incentives can reduce parents' intrinsic motivation. These findings have important implications: while policy makers commonly focus on moving individuals to adopt a desirable behavior, we show that incentives may significantly impact the motivation of those already performing the desired action.

In African cities, households with formal grid connections are informally reselling electricity to unconnected neighbors. This informal market serves 46% of all electrified households in urban Benin, offering lower connection costs but higher marginal prices than the utility. We quantify the welfare consequences of this informal market by randomizing informal connection costs and marginal prices in both sectors across 4,000 households. We use the experimental variation to estimate a model of electricity demand and reseller competition. Connection demand is elastic, consumption is inelastic, and formal tariff subsidies do not pass through to informal prices. We quantify the surplus the informal market generates, evaluate connection and tariff subsidies with and without resale, and ask whether offering a menu of tariffs can replicate the surplus the informal market provides.

Job seekers often face challenges in finding jobs that align with their skills and meet employer demands. To address this, we conduct a randomized controlled trial to evaluate whether artificial intelligence can assist job seekers in two key ways: 1) gaining a clearer understanding of their own skills, and 2) identifying careers that align with their skills and labor market demand. This improved understanding should enable job seekers to make more informed decisions about training programs and career paths, resulting in better labor market outcomes — such as securing jobs more quickly, staying employed longer, and achieving greater job satisfaction.