In low-income communities, pressure to share income with others may disincentivize work, distorting labor supply. We document that across countries, social groups that undertake more interpersonal transfers work fewer hours. Using a field experiment, we enable piece-rate factory workers in Côte d’Ivoire to shield income using blocked savings accounts over 3-9 months. Workers may only deposit earnings increases, relative to baseline, mitigating income effects on labor supply. We vary whether the offered account is private or known to the worker’s network, altering the likelihood of transfer requests against saved income. When accounts are private, take-up is substantively higher (60% vs. 14%). Offering private accounts sharply increases labor supply—raising work attendance by 10% and earnings by 11%. Outgoing transfers do not decline, indicating no loss in redistribution. Our estimates imply a 9-14% social tax rate. The welfare benefits of informal redistribution may come at a cost, depressing labor supply and productivity.

Human actions can alter the regional climate, particularly via land use. We analyze the Great Plains Shelterbelt, a large-scale forestation program in the 1930s across the US Midwest. This program led to a decades-long increase in precipitation and decrease in temperature. Changes extended to downwind unforested areas up to 200km away---enabling us to study climate adaptation and its economic consequences. Where growing conditions improved, farmers expanded corn acreage, switched to more water-intensive production, and experienced less crop failure. In a period of intense farm consolidation and mechanization, this led to the survival of more small farms and a reduction in the use of farm machines. This paper highlights the endogeneity of climate to land use change, and the potential for tree planting to regionally mitigate climate change impacts with lasting consequences for agricultural development.

Commuting is notoriously challenging in lower-income country cities, limiting firms’ ability to attract workers from different neighborhoods and reducing the economic benefits of urban agglomeration. This paper examines whether decreasing the non-financial cost of commuting—by commuting with peers rather than alone—reduces this friction and increases labor supply. We conduct two field experiments in urban Côte d’Ivoire. In the first experiment, job seekers are 16pp more likely to accept a formal full-time factory job if their peers also receive a job offer, and 15pp more likely to remain in that job four months later—but only if they will be employed in the same shift (rather than different shifts). These effects are driven by workers with long commute times, who can commute to work together. Consistent with this channel, workers’ own attendance and turnover are predicted by the attendance and quits of co-commuting peers. In a second field experiment with a different firm, we again randomize whether a worker’s peers are offered a job and whether they could commute together. We also randomize job location—inducing exogenous variation in commute time. We replicate the finding of complementarities in labor supply, but only in the case of long assigned commute times. These findings indicate that commuting together serves as an important job amenity with large impacts on labor supply. Our results provide a novel explanation for key features of urban labor markets, including firms’ widespread use of referrals for hiring and persistent gaps in employment across social groups. 

In this paper, we present an ongoing large field experiment in collaboration with the Tax Authority in Paraguay and the Inter-American Development Bank. The experiment randomizes tax audit notices (treatment) to firms connected through a large network determined by inter-firm VAT transactions. While the ultimate goal is to optimize tax audit policy, the short-term goal is to estimate causal effects of tax audit notices on firm behavior. Of particular interest is to understand spillovers, that is, the response of firms that are not treated but are connected to firms that are treated. We argue that current popular approaches to experimenting on networks are limited by the reality of inter-firm networks, such as their size, high interconnectivity and heavy-tailed degree distributions. Our approach to experimentation leverages subtle sub-structures in the network and allows the application of Fisherian-style permutation tests of causal effects. These testing procedures can be computationally efficient and finite-sample valid, qualities that are important for testing for treatment effects from complex policy interventions in a robust way.

We introduce a novel framework for inferring the private marginal abatement cost to comply with tightened environmental regulations before they are enacted. We apply our approach to the multiperiod structure of the US National Ambient Air Quality Standard (NAAQS) for ozone, which saw considerable changes in 1997. The revamped standard employs a three-year running average, designed to diminish the influence of unpredictable meteorological factors that could push ambient air quality beyond the acceptable range in a given year. By modulating emissions in the current year, anomalies from preceding years can be offset, ensuring the 3-year average stays below the designated threshold. Following the 1997 modification, we detect a pronounced negative counterresponse: Ozone disturbances caused by temperature fluctuations in earlier years are counteracted in the present year. This adjustment is facilitated by curbing NOx and VOC emissions from sources like industrial boilers and petroleum plants. The magnitude of the observed correlation suggests that firms don't deem it excessively burdensome to modify emissions. Lastly, we derive the Bellman equation from minimizing the anticipated environmental compliance cost, which strikes a balance between abatement cost and the repercussions of falling into non-attainment that were identified in previous studies. By simulating the counterresponse across various marginal abatement cost scenarios, we select the one that aligns the inferred counterresponse with the actual one observed in our dataset.