Analyzes the relationship between Job Corps performance measures and center-level impact estimates from the National Job Corps Study (NJCS) conducted in 1993.

"This study explores associations between performance measures used by Job Corps and estimates of impacts for individual Job Corps centers. The analysis relies on data gathered in 1994-1996 in the course of Mathematica’s National Job Corps Study (NJCS), a rigorous random assignment evaluation of the program, and Job Corps’ contemporaneous administrative data. The NJCS found that Job Corps centers with higher scores on the aggregate performance measure used by the program did not show higher impacts on student outcomes. Extending the analysis in this study, the authors find that even after controlling for student demographic and socioeconomic characteristics, neither the aggregate center performance measure nor its program achievement, placement and quality components can be reliably associated with impacts on student outcomes" (OPDR Abstract). (Abstractor: Author)

Full publication title: Analysis of Associations between Contemporaneous Job Corps Performance Measures and Impact Estimates from the National Job Corps Study


Major Findings & Recommendations

- "Students in high-performing centers are significantly different from students in low-performing centers; though the characteristics that are more common among students in high-performing centers are generally associated with better outcomes, there are some exceptions. Differences in the characteristics of students served by centers in different performance terciles are relatively small in magnitude" (p.53). - "Regression-adjusting for characteristics changes center performance rankings, but not dramatically. Regression-adjusted and unadjusted performance measures are positively correlated, although there are differences" (p.53). - "Regression-adjusted performance measures are no better than unadjusted performance measures at distinguishing between centers with larger impacts and those with smaller impacts. The correlations between impacts and performance measures are generally weak and insignificant. Similar results apply using the ETA-652 data and the more detailed NJCS baseline survey data" (p.53). - "Findings hold for overall measures of performance as well as components of center performance and different program years; that is, the relationship between impacts and different performance measure components is also generally weak" (p.53). - "Among the subgroups analyzed, there are not particular groups of centers for which performance measures track impacts" (p.53) - "Outcomes for treatment group participants as measured using the NJCS follow-up survey data—which are conceptually similar to performance measures—are positively correlated with impacts" (p.53). (Abstractor: Author)