Working Papers

To Follow the Crowd? Benefits and Costs of Migrant Networks [Draft] [JMP] 

Abstract: Globally, migrant workers often cluster with hometown peers in the same sectors and locations. This paper quantifies two countervailing effects of migrant networks: learning benefits and labor market congestion costs, using data on millions of migrant workers from a food delivery platform in China. First, I provide direct evidence of knowledge spillovers. New workers learn from their peers’ past delivery experiences, decreasing search time by 10%. Using quasi-random variation in workers' location choices induced by entry bonuses, I show that having one hometown peer nearby increases new workers' productivity and earnings by 2%. Second, I quantify the congestion cost of clustering due to correlated shocks. Since migrant workers tend to work longer during adverse hometown shocks, clustering results in hometown workers competing for deliveries when their labor supply surges simultaneously, decreasing real wages by 10%. Third, I build an equilibrium model to quantify the trade-off between the learning benefits and congestion costs. Counterfactuals show that providing insurance for hometown shocks doubles clustering levels and increases productivity.

Firm Visibility and Quality Upgrading: Evidence from Randomized Location Allocations [Draft]

Abstract: How much do firms benefit from having greater visibility to new customers? I study the impact of consumer visibility on firm growth and market dynamics using a unique experiment in a porcelain market, where booth locations were randomly allocated across porcelain makers each season. Firms assigned to a more visible location (close to areas with high consumer traffic) have 20% higher sales. Small and medium-sized firms benefit the most and are most likely to upgrade product quality in response to higher sales. Informed by the empirical results, I develop a model with consumer search frictions and firm capital constraints on quality investments. I compare random booth allocation to a system where firms bid for booth locations. When consumers have sufficient heterogeneous preferences, random allocation can maintain a less concentrated market and increase long-run aggregate demand by enabling small firms to accumulate capital for quality upgrading. These results highlight the benefits of visibility for high-potential yet capital-constrained firms. 

Commitment at Scale: Evidence from 13 Million Social Media Users [Draft Available Upon Request]

(with Peter Hickman and Ruru Houng)

Abstract: Do people want to limit their social media use? We document demand for a soft commitment device through an experiment with 13 million active social media users. In the intervention, users were prompted to set a usage alarm. We estimate a 4.8% take-up rate among a broad population. Using a model with present bias and additional variation, we explore four reasons for the low take-up rate. First, around half of users do not perceive self-control issues. Second, our results indicate the alarm has a limited impact on people's actual usage. Third, take-up significantly responds to intervention timing. Finally, our large-scale experiment allows us to study sample selection issues. In a second experiment, we focus only on participants who voluntarily answer a very simple two-click survey and find the average take-up rate almost doubles, rising to 8%. 

Selected Work In Progress

Civil Wars and Strategic Subsidies [Draft Coming Soon]

(with Jaya Wen and Cheryl Wu)

Quality Incentives and Upgrading in Uganda's Coffee Supply Chain [Draft Coming Soon]

(with Jie Bai, Lauren Falcao Bergquist, Ameet Morjaria and Russell Morton)

Make Services Tradable: The Effects of Delivery Technology on Firm Locations [Draft Available Upon Request]