Table 6 Treatment effects of legalization
Parametera |
Parametricb Method______ |
Polynomial Method________ |
Nonparametric Method I_____ |
Nonparametric Method II_____ |
ATET |
0.1043 |
0.2385 |
0.2635 |
0.2538 |
ATEU |
0.1002 |
0.1784 |
0.1459 |
0.1616 |
ATE |
0.1020 |
0.2069 |
0.1978 |
0.2031 |
Sorting gain |
____________0.0023 |
__________0.0316 |
________0.0657 |
________0.0507 |
aA test for essential heterogeneity in the treatment effects yielded an F-statistic (p value) of 18.19 (0.0000),
indicating self-selection arising from heterogeneous and unobserved gains for individuals in the sample (See
Heckman, Urzua and Vytlacil, 2006). bThe extent of selection bias is gauged with a comparison of the OLS and
parametric model results: selection bias = OLS-ATET= 0.0359-0.1043= -0.0684. It shows that the OLS estimate of
the average effect of legalization on earnings is downward biased, indicating a 3.6% average earnings gain relative
to the 10% average gain suggested by the ATET estimate in the parametric method. The overall bias (OLS-ATE) is
-0.0661.
□ Untreated (D=0) □ Treated (D=1)
Figure 1 Frequency of propensity score by legal status
26