The 36 month outcomes the post estimation results

This preview shows page 35 - 37 out of 70 pages.

the 36 month outcomes, the post-estimation results differ with respect to the outcome variable used to train the forests. Tables 3 and 4 contain the results of the estimation for the ATE (upper panel), the various GATEs that appeared to be of a priori interest (in the middle of the table), as well as summary statistics for the IATEs (lower panel). The GATEs relate to effect heterogeneity in the number of unemployment episodes in the 2 years prior to the unemployment spell analysed (12 catego- ries), the native language being a ‘Swiss’ one or a different one (2 cat.), the employability index (3 cat.), and the economic sector of the last occupation (16 cat.). As mentioned before, all effects have the interpretation of additional months of regular employment due to programme partici- pation.
35 Table 3: Average, group effects, and individualized effects: 9 months Average potential outcomes Average effects Treated Controls Expectation Stand. err. Expectation Stand. err. ATE Stand. err. p-val. in % 1.59 0.06 2.09 0.02 -0.52 0.06 0.0 Group average treatment effects (GATEs) Group Expect. Est. std. p-val. in % Group Expect. Est. std. # E spells 0 -0.52 0.06 93 Sector 1 -0.56 0.07 # E spells 1 -0.51 0.06 37 Sector 2 -0.64 0.07 # E spells 2 -0.50 0.06 93 Sector 3 -0.51 0.06 # E spells 3 -0.50 0.07 43 Sector 4 -0.41 0.09 # E spells 4 -0.50 0.07 55 Sector 5 -0.44 0.07 # E spells 5 -0.49 0.07 87 Sector 6 -0.45 0.07 # E spells 6 -0.49 0.08 39 Sector 7 -0.50 0.07 # E spells 7 -0.51 0.07 30 Sector 8 -0.50 0.06 # E spells 8 -0.48 0.08 59 Sector 9 -0.53 0.06 # E spells 9 -0.51 0.08 76 Sector 10 -0.50 0.06 # E spells 10 0.53 0.10 - Sector 11 -0.64 0.09 # E spells joint test (p-val in %) 94 Sector 12 -0.49 0.07 Swiss lang. native -0.51 0.06 68 Sector 13 -0.54 0.07 Swiss lang. not nat. -0.53 0.06 - Sector 14 -0.45 0.07 Swiss joint test 68 Sector 15 -0.46 0.07 Employability good -0.49 0.06 11 Sector 16 -0.47 0.07 Employabil. middle -0.52 0.06 52 Sector joint test (p-val in %) 0.1 Employability bad -0.54 0.07 - Employability joint test (p-val in %) 17 Individualized average treatment effects (IATEs) Mean Standard deviation Share < 0 in % Share > 0 in % Average standard error of estimate Share significant at 5% level in % -0.52 0.15 100 0 0.15 92 Note: Expect. and est. std. is defined for potential outcomes, effects, and their respective estimation errors. The p-values relate to hypothesis tests based on the (asymptotic) t- or Wald-type. For the ATE, it is the t -test that the effect is zero. For the GATEs, it is the t -test that the adjacent values are identical (therefore it is not given for the sectors that are not in any natural order) as well as the Wald test that all GATEs relating to this variables are identical. Therefore, this test has (# of groups-1) degrees of freedom. For the average treatment effect, the first four columns show the mean of the potential outcomes as well as their estimated standard errors. The remaining columns show the average effects, their estimation error, and the p-value for the hypothesis that the average effects are zero. The results imply that over nine months (3 years), these individuals work 1.6 (16.5) month if they participate in the programme and 2.1 (16.8) months if not. This negative short-run effect is statistically significant at conventional levels (despite inference being likely conservative), while the smaller 3-year effect is not.

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture