Research Article

Nonparametric Estimation of ATE and QTE: An Application of Fractile Graphical Analysis

Table 3

ATE estimators of the effect of training on earnings.

Nontrimmed Trim 1 Trim 2
Estimator Coef.Average Std.error Coef.Average Std. error Coef. Average Std. error

FGA ATE(a) 1572.2 1589.0 634.0 1608.6 1620.2 645.3 1480.9 1565.2 665.8
FGA ATE ( 𝑅 × 2 )(a) 1537.1 1606.4 659.0 1520.2 1634.4 659.8 1208.6 1572.4 682.3
FGA ATE(b) 1584.1 1561.9 616.7 1672.5 1625.7 637.0 1676.1 1536.1 655.5
FGA ATE ( 𝑅 × 2 )(b) 1563.2 1511.4 606.6 1604.5 1576.1 627.3 1530.2 1483.4 644.5
Hirano et al. [12] 1598.2 1612.2 631.3 1731.3 1691.9 661.0 1862.0 1589.1 682.6
PS matching
Nearest neighbor (1) 997.2 1393.8 736.4 1101.6 1389.2 721.3 869.8 1334.4 744.8
Nearest neighbor (2)156.2 1471.6 710.61262.0 1477.8 702.8 984.2 1424.2 729.9
Nearest neighbor (4)1471.8 1559.0 670.01552.7 1571.7 671.6 1346.5 1525.1 700.4
Kernel 1629.0 1639.6 631.1 1638.9 1636.3 634.3 1482.4 1590.8 657.3
Spline 1587.0 1616.6 641.4 1614.4 1613.8 638.8 1380.3 1565.6 660.9

(a) ̂ 𝛿 𝜏 , (b) ̃ 𝛿 𝜏 . Averages and standard errors are obtained using 1000 bootstrapping random samples with replacement of the original database.