Table 3.
Determinants of share of vaccinated 16-17-year-olds in treated and neighboring municipalities.
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| Treatment | 0.129*** (0.018) | 0.094*** (0.014) | 0.109*** (0.026) | 0.073 (0.028) |
| Neighbor indicators | No | No | Yes | Yes |
| Share foreign-born | No | −0.531*** (0.114) | No | −0.418 (0.273) |
| Share high education | No | 0.215** (0.084) | No | 0.150 (0.181) |
| COVID-19 deaths | No | 6.375 (5.195) | No | −4.999 (17.053) |
| Constant | 0.722*** (0.013) | 0.774*** (0.027) | 0.761*** (0.034) | 0.843*** (0.087) |
| R2 | 0.782 | 0.900 | 0.934 | 0.976 |
| Moran's I (residuals) | 0.093 | −0.191 | −0.198 | −0.409** |
| Conley SE (treatment) | 0.019*** | 0.010*** | 0.013*** | 0.013*** |
Notes: The dependent variable is the share of 16–17-year-olds vaccinated in week 49 in the 16 included municipalities. Ordinary least squares regressions controlling for Treatment (pre-booked appointments), Neighbor indicators (one dummy variable for each treated municipality, indicating its neighbors), as well as the control variables Share foreign-born, Share high education, and COVID-19 deaths. Moran's I for spatial residual autocorrelation and Conley standard errors accounting for spatial autocorrelation were computed assuming a maximum distance for spatial autocorrelation of 65 km (the minimum distance from which all area centroids shared at least one neighbor) and a Bartlett kernel using the acreg package for Stata. Other distance choices and a uniform kernel led to similar results, but standard errors could not be computed in Model 4 using a uniform kernel. *p < 0.1, **p < 0.05, ***p < 0.01.