Fig 6. Difference-in-differences (DID) estimates for different weighting schemes and samples and their 95% confidence intervals.
Confidence intervals are calculated by clustered robust standard errors clustered by municipality. All DID estimates are estimates for βτ in Eq (1). The unit of “benefit” is 1,000 points (10,000 JPY or around 100 USD in standard areas). “OLS” is a normal ordinary least squares regression and “WLS” is a weighted least squares regression in which the weight is the number of elderly individuals for the outcomes of Q and C and the number of certified individuals for the outcome of B. The trimmed sample includes the municipalities in the baseline sample whose outcome values are within the minimum and maximum of outcome values of the treated municipalities in all three pre-disaster years. The Fukushima sample contains the evacuation and non-evacuation municipalities in Fukushima prefecture.