Eddy Zou

We study the role of middle managers in shaping police performance, focusing on territorial divisions within the London Metropolitan Police Service. We assemble a novel dataset that reconstructs the careers of Borough Commanders from 2005-2019 and map them into division-by-month outcomes constructed from police records and police strength data. Using an AKM-style framework and bias-corrected fixed effects, we find that middle managers account for about 12% of the variance in (log) clearances per worker, about one-fifth as large as permanent differences across divisions. High-performing managers raise clearances across offense types, strengthen proactive policing as reflected in substantial increases in drug detections, and generate more police-issued sanctions and court convictions. These improvements do not stem from reductions in underlying crime and are achieved by reducing back-office staff, reflecting more efficient use of personnel. The results show that internal organization via middle-management quality is a key input into police effectiveness in resource-constrained environments.

How does judicial bias arise and persist in the legal system? We trace the lifecycle of judicial bias in India's criminal courts, leveraging the quasi-random assignment of judges to courts and cases. We first document substantial variation in bias: within the same court, assigning a same-religion defendant from a judge at the 25th to 75th percentile of bias increases acquittal probability by 7.5 percentage points, 45% of the mean acquittal rate. We then examine how different triggers across a judge's professional career affect their bias. Exposure to Hindu-Muslim communal riots during a judge's first five years of service leads them to increase same-religion acquittal rates by 17.5%. Yet, judges who work alongside colleagues from different religions during this critical period show no such effect. Next, social interactions between judges reinforce judicial bias. On-the-job exposure to biased colleagues leads judges to increase same-religion acquittal rates by 3.2 percentage points, with effects that persist for over a year. These findings suggest that judicial bias is neither innate nor inevitable, but rather shaped by experiences and relationships, and that thoughtful institutional design can prevent and mitigate discriminatory practices across various professions.

Heatwaves, intensified by climate change, hit the poorest the hardest. Many are exposed to dangerous temperatures through outdoor work or limited access to adaptive resources. In 2024, 37 cities in India surpassed 45°C (113°F), and around 40,000 heat stroke cases were reported. How can social protection systems evolve to address the growing losses caused by extreme heat? We evaluate an innovative intervention in India that offers automatic daily wage payments to low-income workers when temperatures exceed a predetermined threshold. Developed by the Self Employed Women's Association (SEWA) — a union representing over 3 million informal workers — the scheme is the world's first parametric heat insurance product targeting earnings loss. We use a randomized encouragement design, incentivizing SEWA officers to promote enrollment in 2,821 treatment villages, while 2,821 control villages receive no targeted outreach.

How does the presence of international trade in used durable goods, such as vehicles, affect the efficacy and relative welfare effects of unilateral policies to decarbonize transportation? This project combines a micro-founded structural model with novel data from multiple countries, to understand the effectiveness, potential economic costs, and environmental impacts, of restricting trade in used vehicles.