Extended Data Fig. 3. Treatment effects: negative binomial regression.
In each panel, the horizontal axis indicates dates in 2020, and the vertical axis represents the coefficient estimates of the treatment variable. The vertical turquoise line corresponds to the survey date. The thick black line indicates the point estimates of the treatment variable’s coefficients, and the shaded grey area presents the 95% confidence intervals. Cluster-robust standard errors are utilized where clusters are a pair of matched municipalities and a municipality. a, April 6. b, April 10. c, April 16. d, Irregular results of treatment effects by negative binomial regression. For other survey dates, we cannot estimate negative binomial regression parameters.