Table 3.
Local linear RD regression results
| 1 | 2 | 3 | |
|---|---|---|---|
| Total | Single | Multiple | |
| Panel A: total outpatient visit | |||
| Copayment abolition | 0.633 | 0.744** | −0.816** |
| (0.479) | (0.309) | (0.253) | |
| [88.30%] | [110.47%] | [−55.77%] | |
| RD type | Fuzzy | Fuzzy | Fuzzy |
| Panel B: total healthcare costs | |||
| Copayment abolition | −0.045 | 0.475* | −0.625** |
| (0.266) | (0.261) | (0.289) | |
| [−4.40%] | [60.80%] | [−46.47%] | |
| RD type | Sharp | Sharp | Sharp |
| Panel C: healthcare costs per visit | |||
| Copayment abolition | −0.264*** | −0.099*** | 0.364*** |
| (0.018) | (0.011) | (0.035) | |
| [−23.20%] | [−9.43%] | [43.91%] | |
| RD type | Sharp | Sharp | Sharp |
| Observations | 2688 | 2688 | 2688 |
Column 1 includes all outpatient visits. Columns 2 and 3 only include one-time visits and multiple visits, respectively. Covariate variables are included in the regression. A cluster–adjusted standard error is used to account for the within-cluster correlation. We use a polynomial of order one and a triangular kernel function. A data-driven mean squared error optimal bandwidth selection is applied. *** Significant at the 1% level, ** Significant at the 5% level, * Significant at the 10% level.