Estimating the Effect of COVID-19 Interventions Using a Regression Discontinuity Design

Abstract

Identifying the causal effect of non-pharmaceutical interventions for preventing the spread of COVID-19 is difficult because of the many variables that affect political leaders’ public-health decisions. In this thesis, I use a regression discontinuity design to assess these causal effects in Andalucia, a region in Spain, using the regional government’s policy of assigning municipalities to more or less stringent lockdowns according to sharp thresholds in local case numbers. After reviewing the literature on regression discontinuity, especially about the optimal bandwidth selection, I run a simulation study comparing different methods. Using the Imbens-Kalyanaraman optimal bandwidth and a local linear regression, I do not find a significant effect of the policies on cases 10 to 14 days after their implementation. However, these results are not conclusive because of the relatively small sample size used and large variance in case rates across municipalities.