event. Arguably legal status would have been more difficult to attain with the additional checks
and safeguards that were put in place. That is not to say that the tightening on legal status would
have necessarily had a significant adverse effect on foreign farm workers; if anything, these
workers are more likely to have migrated illegally across the US border with Mexico. The
magnitude of the after 1986 dummy suggests that legal status has been difficult to acquire since
the last major legalization in 1986 (the SAWs program). The farm work weeks effect indicates
that more weeks of farm work reduce the likelihood of having legal status.
Table 4 presents the parametric model wage results for the treated and untreated worker
groups. Wage results for the nonparametric methods (polynomial, nonparametric I,
nonparametric II) are reported in Table 5.5 All parameter estimates have the expected direction
of influence on the wage results, and the parameters are statistically significant at the 1% level of
significance; the exceptions are the foreign farm work experience and age variables in the treated
and non-treated groups, respectively. For both groups, the magnitude and statistical significance
of the piece rate and after 2001 estimates suggest dominant influences on farm wages relative to
the other variables of the model.
Table 6 reports the estimated treatment effects of legalization which are all positive. The
average treatment effect (ATE) reflects the expected gain for a random foreign farm worker who
became legalized, the average treatment effect on the treated (ATET) indicates the return to
those workers who became legalized, and the average treatment effect on the untreated (ATEU)
indicates the potential return for those who were not legalized. The order of magnitude of the
estimates generated by each method indicates positive sorting on the gains associated with
legalization (ATET>ATE>ATEU), wherein those foreign workers who were most likely to
5 The three nonparametric estimators are different alternatives to assuming normality of the disturbances (Heckman,
Urzua, and Vytlacil, 2006b).
13