Variation in standard errors in event-study design: insights from empirical studies and simulations

• Published in: Applied economics Vol. 55; no. 5; pp. 518 - 530
• Main Author: Li, Yang
• Format: Journal Article
• Language: English
• Published: London Routledge 01.01.2023; Taylor & Francis Ltd
• Subjects: Applied economics; applied economics methodology; cluster-robust standard errors; Clustering; Economic analysis; Economic theory; Event-study design; Monte Carlo simulation; Monte Carlo simulations; Recurrent; Simulation; Statistical inference; wild bootstrap
• ISSN: 0003-6846; 1466-4283
• DOI: 10.1080/00036846.2022.2091109

A growing body of empirical research in applied economics exploits event-study design to show the dynamic effects of policy intervention. This article documents an interesting pattern in these empirical applications: cluster-robust standard errors (CRSE) from event study generally exhibit a decreasing trend over the pre-event periods, while standard errors (SEs) without clustering adjustment take similar values across pre-event periods. This article explores possible explanations for this recurrent issue by documenting the pervasiveness of this pattern across empirical applications and by Monte Carlo simulations. The simulation results show that CRSE generally present a decreasing pattern in pre-event periods and more so when intra-cluster correlation is high. The SEs also exhibit a decreasing trend using wild bootstrap and a more robust statistical inference in the case of few clusters. Overall, my results suggest that within-cluster correlation contributes to explaining the pattern.