Table 4.
Treatment effects of left-behind experience in each grade.
| Grade | Treatment Effect | Nearest Neighbor Matching (k = 1) |
Nearest Neighbor Matching (k = 4) |
Nearest-Neighbor Matching with Caliper (k = 4, caliper = 0.004) |
Caliper Matching (caliper = 0.004) |
Spline Matching |
|---|---|---|---|---|---|---|
| 2014 (n = 3958) |
ATT | 3.667 *** (1.187) |
3.671 *** (1.139) |
3.674 *** (1.143) |
3.840 *** (1.124) |
3.731 *** (1.081) |
| ATU | 7.112 *** (1.703) |
7.114 *** (1.673) |
7.215 *** (1.693) |
5.978 *** (1.436) |
6.188 *** (1.409) |
|
| ATE | 7.025 *** (1.671) |
7.028 *** (1.645) |
7.126 *** (1.663) |
5.924 *** (1.417) |
6.126 *** (1.390) |
|
| 2015 (n = 3919) |
ATT | 3.523 *** (1.101) |
3.623 *** (1.089) |
3.623 *** (1.090) |
3.800 *** (1.064) |
3.837 *** (1.047) |
| ATU | 2.676 (1.644) |
3.139 ** (1.518) |
3.200 ** (1.549) |
2.493 *** (1.375) |
2.404 * (1.425) |
|
| ATE | 2.695 * (1.618) |
3.149 ** (1.496) |
3.209 ** (1.526) |
2.522 *** (1.359) |
2.436 * (1.407) |
|
| 2016 (n = 4170) |
ATT | 4.482 *** (1.072) |
4.129 *** (1.045) |
4.129 *** (1.043) |
4.163 *** (0.979) |
4.254 *** (0.948) |
| ATU | 6.940 ** (2.723) |
7.839 *** (2.558) |
7.774 *** (2.655) |
5.471 ** (2.177) |
6.730 ** (2.386) |
|
| ATE | 6.890 ** (2.675) |
7.764 *** (2.516) |
7.700 *** (2.609) |
5.444 ** (2.141) |
6.680 ** (2.347) |
|
| 2017 (n = 3757) |
ATT | 4.636 *** (1.013) |
4.910 *** (0.992) |
4.910 *** (0.990) |
4.722 *** (0.980) |
4.700 *** (0.982) |
| ATU | 7.696 *** (1.481) |
5.639 *** (1.366) |
6.560 *** (1.434) |
5.990 *** (1.339) |
7.272 *** (1.471) |
|
| ATE | 7.600 *** (1.448) |
5.616 *** (1.340) |
6.508 *** (1.405) |
5.950 *** (1.314) |
7.192 *** (1.441) |
Note: *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors in parentheses are estimated by bootstrap method with resampling 500 times. Each treatment observation is matched to control observations with replacement when performing nearest neighbor matching.