Given the robust economy of the late 1990s--when the NJCS was performed--it is particularly important to think about the effectiveness of JC during downturns, when the economy provides less favorable local labor market conditions.
This appendix describes the construction of the LUR variables and some of the issues arising in matching these variables to the restricted-use NJCS data.
The restricted-use NJCS data contains the respondent's zip code of residence at the time of the 48-mo follow-up interview, between December 1998 and May 2000 (see Schochet 2001).
The ZTCA was then merged with the zip code reported at the 48-mo interview in the NJCS data.
The original NJCS sample consists of all individuals who completed a 48-mo survey since the compliance-adjusted difference-in-means estimator reported in the NJCS only requires information at the 48th month.
The NJCS was sponsored by the Department of Labor following a mandate from Congress to assess the effectiveness and social value of the JC program.
The original NJCS estimates reported below are based on the 48-mo interview for which the effective response rate was 79.
All our estimates throughout the article employ NJCS sampling weights.
We point out that Schochet, McConnell, and Burghardt (2003) reexamined the survey-based NJCS results using administrative earnings records data, finding smaller and less statistically significant effects of JC on participants.
85 hr of experience per week might appear small, it translates to 385 hr over the duration of the NJCS (208 wk) or about one-fifth of a full-time work year.
Given that there are other posttreatment variables available in the NJCS data, we address this concern by estimating similar effects to those in Table 2 with a number of such other variables.