Primary Outcomes (explanation)
We believe the training program will directly impact peoples’ entrepreneurial readiness and confidence and encourage them to engage in entrepreneurial activities (outcomes #1 and 2 above). We then want to test whether their increased entrepreneurial confidence engender more hopeful and optimistic personal outlooks, promote attitudes towards regional peace, and/or decrease inclinations towards radicalization. In this way we can see primary outcomes #1-2 as “first stage” variables and outcomes #3-5 as “second stage” variables. Since we think the efficacy of the training program in generating a first stage may vary with observables such as training location, gender, and level of preparation (e.g., college major), we will analyze treatment effects on our second stage variables (#3-5) in two ways: first in a standard reduced form intent-to-treat framework (regressing the outcome variables on a dummy for treatment), and second, in an IV framework where the endogenous regressor is outcome #1, the dependent variables are outcomes #3-5, and the instruments are a dummy for being in the treatment group interacted with variables which affect the first stage such as training location, gender, and preparation.
We believe the first endline will be too soon after the intervention to impact wages and new business creation, but can test whether people are using the tools they learned to engage in entrepreneurial activities (outcome #2).
We move outcome #6 (self employment and income/profits) from a secondary to a primary outcome since enough time has elapsed for the intervention to potentially impact this outcome.
We correspondingly move outcome #1 (entrepreneurial readiness/knowledge/confidence) from a primary to a secondary outcome since enough time has elapsed for entrepreneurial readiness to translate into actual entrepreneurial activity (outcome #2), which is a more important and more cleanly-measured outcome
We will again use the same two analysis frameworks from before: (1) the OLS ITT framework and (2) the IV-style framework that analyses the treatment effects for outcomes #3 onwards only among those for whom we see a first stage on outcome #2.
We add self-efficacy as a primary outcome because we saw effects on that measure in the second endline among those for whom we have a first stage on outcome #2 and want to test if these effects persist.
We again will adopt both analysis frameworks outlined for the second endline, paying particular attention to the (2) IV-style framework that focuses analysis for outcomes #3 onwards only among those for whom we have a first stage on outcome #2. We pay particular attention to framework (2) since the first stage on outcome #2 was much larger in certain subsamples than others.
In addition, because in the second endline we saw a much higher first stage on outcome #2 in certain training locations, we will conduct heterogeneity analysis and descriptive analysis to try to understand what factors drive the differences across locations. This analysis relates to the literature on entrepreneurship training which contains many examples of effective trainings and many examples of ineffective trainings; we will see if we can use our within-sample variation in effectiveness to shed light on potential reasons for this heterogeneity in effectiveness.