First‐Stage Regressions
| Dependent Variable | ||
|---|---|---|
| Three‐Year Track Dummy (1) | Unemployment Rate at Graduation (2) | |
| Predicted unemployment rate at graduation (ϕcj) | 0.010 | 0.510*** |
| (0.007) | (0.045) | |
| Pilot intensity in municipality of residence (Pcj) | 0.451*** | 0.081 |
| (0.042) | (0.145) | |
| Cohort FE | ✓ | ✓ |
| Municipality FE | ✓ | ✓ |
| Individual characteristics | ✓ | ✓ |
| Individuals | 68,241 | 68,241 |
| R2 | 0.20 | 0.89 |
| Kleibergen–Paap Wald rk F‐statistic | 67.85 | 67.85 |
Notes: As we have two endogenous regressors and two IVs, we cannot simply look at the first‐stage F‐statistics to test for weak instruments. We look instead at the Kleibergen–Paap Wald rk F‐statistic (which is cluster‐robust). Here, the eigenvalue (for 10 percent maximal IV size) is equal to 7.03. We show first‐stage estimates for the sample of individuals in the year of (predicted) graduation (68,241 individuals). Marginal effects are presented. Standard errors (in parentheses) are clustered at the municipality * cohort (j × c) level. Significance: *p < 0.10, **p < 0.05, ***p < 0.01.