Search Results

Title: Correction of Estimation Bias in Evaluating the Labor Market Outcomes of Youth Participating in School-Based Learning Programs in the United States
Resulting in 1 citation.
1. Yu, Wei-Ting
Correction of Estimation Bias in Evaluating the Labor Market Outcomes of Youth Participating in School-Based Learning Programs in the United States
Ph.D. Dissertation, The Pennsylvania State University, 2005. DAI-A 68/12, Jun 2008
Cohort(s): NLSY97
Publisher: ProQuest Dissertations & Theses (PQDT)
Keyword(s): Apprenticeships; College Enrollment; College Graduates; Colleges; Labor Economics; Labor Market Outcomes; Program Participation/Evaluation; Schooling; Seasonality; Wage Differentials

Permission to reprint the abstract has not been received from the publisher.

Evaluations of training programs, among others, have struggled with the prevention of estimation bias, especially when examining programs using non-experimental data. Alternative econometric approaches have been developed to correct for this estimation bias. This bias is mostly due to discrepancies in observed and unobserved characteristics between the participants and the counterfactual. The purpose of this study is to explore the unbiased estimates derived from alternative corrections and compare variation in predictive effectiveness among different corrections from evaluations of school-based learning (SBL) programs.

This study uses data from the National Longitudinal Survey of Youth 97 (NLSY 97), a non-experimental database, to evaluate the effects of U.S. youths' school-based learning (SBL) program participation on early labor market outcomes. Estimates from ordinary least squares regression, the linear regression with probabilistic matching or instrumental proxy, are compared to those obtained from bootstrapping regression analysis. Four outcome variables, including employment in college, employment, total worked hours and hourly wage rate, are used to gauge the early labor market outcomes of youth from the NLSY97.

Findings, at an alpha level of .05 when the first type of instrumental variable (IV) correction method is adopted, reveal that SBL program participants are significantly less likely than non-participants to enroll in college, and that SBL program participants have a lower probability of enrolling in college than do non-participants.

In comparison with college enrollment, findings from the analysis of employment and total worked hours outcomes on SBL program participation are not statistically significant with selection bias corrections. Rather, only the estimates derived from the Tobit regression for the correction of the censored data show that SBL program participants have a lower number of total worked hours than do non- participants. Due to evidence which shows that youths' wage differential is small in the early labor market, the findings from Heckman's two-step correction with an alpha level of .10 show that SBL program participants are more likely to have higher hourly wages than non-participants.

In addition, in looking at the seven specific types of SBL program participation, the significant likelihood of enrolling in college or being employed, all of the estimates derived from Heckman's two-step correction show no significance.

For the total worked hours outcome, the correction of the censored data using the Tobit regression shows that internship/apprenticeship program participants have lower total worked hours than non-participants, with an alpha level of .001.

Due to the small wage rate differential for youth in the early labor market, a significance level of .10 is used for this outcome. Heckman's two-step correction reveals that SBL program participants are more likely than non-participants to have higher hourly wage rates. In view of the seven types of SBL program participation, the findings from Heckman's correction reveal that internship or apprenticeship program participants are more likely than non-participants for all SBL programs to have a higher hourly wage rate.

In summary, this study shares corrected estimates from alternative approaches. Based on criteria from the 200-time and 500-time bootstrapping, the best selection bias estimation among these four corrections uses Heckman's two-step correction.

Bibliography Citation
Yu, Wei-Ting. Correction of Estimation Bias in Evaluating the Labor Market Outcomes of Youth Participating in School-Based Learning Programs in the United States. Ph.D. Dissertation, The Pennsylvania State University, 2005. DAI-A 68/12, Jun 2008.