Five of the eight AGI pilots were able to successfully embed a rigorous impact evaluation design. We also had a centralized research team that ensured standardization of the research objectives and methods as much as possible. You can access the papers from the individual pilots on our website, and you can download useful documents such as our evaluation concept notes, list of core indicators, and survey instrument in our Resource Guide.
Here are some key recommendations for further research:
Unbundle evaluation designs and provide cost-benefit information by project component. AGI evaluations weren’t able to compare the relative impact of technical training versus life skills training or measure the impacts of specific project strategies, such as mentoring or placement assistance. Similarly, we can say very little about disaggregated costs of these components.
Have a cash-only evaluation arm. Youth employment interventions of all kinds are under pressure to demonstrate that their impacts are larger than what could be achieved through giving cash directly. See, for example, this relevant blog post from Chris Blattman. As part of this agenda we also need to understand the differential impacts of cash provision on young men and women.
Determine the optimal composition, intensity, and delivery of different mixes of skills. This is particularly true for life skills training, which tends to be much more heterogeneous across contexts and is far less expensive to implement than technical or business skills. Related questions around the appropriate age to focus on different types of skills and whether training works better in sex-segregated classrooms will aid in designing the next generation of youth employment programs.
Test strategies for job placement. Progress has been made in improving the delivery of skills training and in helping youth start businesses, but much less is known about how to cost-effectively assist youth to find and retain wage jobs. Interventions—some implemented in AGI pilots—that deserve more testing include:
- Variations in the length and intensity of job placement support: Most AGI interventions included three to five months of placement support;
- Performance-based contracts for the training providers, as used in both the Liberia and Nepal AGI pilots, though these have not been tested rigorously;
- Wage subsidies, as tested among young female community college graduates in the Jordan AGI, which achieved significant short-term gains but no long-term impact;
- Partnerships with large firms to create custom training programs.
Untangle the relationships between young women’s labor and health outcomes. The AGIs in Liberia and Nepal, using a technical and vocational education and training (TVET) model, did not have significant impacts on sexual behaviors or health outcomes, while the Uganda girls' club-based approach dramatically lowered fertility and increased condom use. One distinguishing factor about the Uganda project was that it worked with younger girls, starting at age 14. Another important question to answer is whether there is an optimal age threshold or whether there are other conditions under which skills training projects can affect sexual behaviors.