Evaluation of the Vocational Training Grant Fund in Namibia: Baseline Report

Publisher: Washington, DC: Mathematica Policy Research
Aug 28, 2015
Authors
Evan Borkum, Arif Mamun, Linus Marco, and Malik Khan Mubeen

Key Findings:

  • The baseline treatment and control samples were very similar in demographic characteristics.
  • However, there were some large and statistically significant differences between the treatment and control groups in the baseline indicators most closely related to the follow-up outcomes, including training, employment, and earnings outcomes.
  • Because the baseline survey often was conducted several months after the start of VTGF training, the baseline indicators might have been affected by the training. Therefore, the treatment-control differences that we identified do not invalidate the design, and the baseline similarity in demographic characteristics gives us confidence that the design is valid. 

Mathematica is conducting an impact evaluation of the Vocational Training Grant Fund (VTGF) in Namibia, which was funded by Millennium Challenge Corporation and provided scholarships to Namibian youth to attend vocational training. To assess the validity of the random assignment design that we are using to estimate impacts, we examined baseline survey data collected from the applicants who were randomly assigned. These data suggest that the baseline treatment and control samples were very similar in demographic characteristics, increasing our confidence that there were no underlying differences between the two groups. In contrast, there were large and statistically significant differences in the baseline indicators most closely related to the follow-up outcomes. In particular, the treatment group was significantly more likely to have been enrolled in vocational training in the five years before the baseline survey, was significantly less likely to be employed at the time of the survey, and had significantly lower average earnings. However, these differences likely do not reflect underlying treatment-control differences that could threaten the validity of the design, but rather reflect a delay in the timing of the baseline survey relative to training. Specifically, because in most cases the survey was conducted several months after the start of VTGF training, many baseline indicators in the treatment group would have been affected by enrollment in VTGF training. Overall, the implications of the observed treatment-control differences are that the impact evaluation design still is likely to be valid.