However, several years of pre-intervention earnings are viewed as important in determining the effect of job training programs (Angrist 1990, 1998; Ashenfelter 1978; Ashenfelter and Card 1985; and Card and Sullivan 1988). Thus, we further limit ourselves to a subset of this data in order to obtain data on earnings in 1974. Our subset, also defined using the month of assignment, includes 185 treated and 260 control observations. Since month of assignment is a pre-treatment variable, this selection does not affect the properties of the experimentally randomized data set: the treatment and control groups still have the same distribution of pre-intervention variables, so that a difference in means remains an unbiased estimate of the average treatment impact.
We present the pre-intervention characteristics of the original sample and of our subset in the first four rows of Table 1. The distribution of pre-intervention variables is very similar across the treatment and control groups for both samples (none of the differences is statistically significant), but our subset differs somewhat from Lalonde’s original sample, especially in terms of 1975 earnings. Our propensity score results will be based on our subset of the data, using two years of pre-intervention earnings. In order to render our results comparable to Lalonde’s, we replicate his analysis on our subset (both with and without the additional year of pre-intervention earnings data), and show that his basic conclusions remain unchanged. As well, in Section 5, we discuss the sensitivity of our propensity score results to dropping the additional earnings data. insurance