The Jobs benchmark (LaLonde, 1986; composition of Shalit et al., 2017)
combines a randomised experiment (the National Supported Work programme)
with observational controls. The treatment is participation in a job
training programme and the outcome is employment status. The e column
flags units belonging to the randomised subsample, which enables
evaluation on real data via the experimental subset.
jobs_trainA data frame with 2,891 rows and 20 columns:
Covariates (standardised continuous and binary).
Binary treatment indicator: job training (0/1).
Binary outcome: employment (0/1).
Indicator of membership in the randomised experimental subsample (0/1).
LaLonde, R. J. (1986). Evaluating the econometric evaluations of training programs with experimental data. The American Economic Review, 76(4), 604-620.
Shalit, U., Johansson, F. D., & Sontag, D. (2017). Estimating individual treatment effect: generalization bounds and algorithms. Proceedings of the 34th International Conference on Machine Learning.
data(jobs_train)
head(jobs_train)
#> x0 x1 x2 x3 x4 x5 x6 x7 x8
#> 1 -0.6142821 1.4647266 0 0 1 0 2.39325000 2.7461965 -0.65331121
#> 2 -0.8024627 0.1018350 0 0 1 0 0.10988519 0.4982713 -0.78528431
#> 3 -0.8965531 -0.2388879 1 0 1 1 -0.08521244 -0.1480969 -0.84731167
#> 4 -0.8965531 -0.2388879 0 0 0 1 0.40558116 0.3255935 -0.84731167
#> 5 0.1384405 -1.6017795 1 0 1 1 -0.72253135 -0.2127338 -0.01984034
#> 6 0.1384405 0.4425579 0 0 1 0 2.16202319 1.9130996 -0.01984034
#> x9 x10 x11 x12 x13 x14 x15 x16 t y e
#> 1 -0.6569127 1.62753104 2.46233663 2.93724407 0 0 2.84390896 0 0 1 0
#> 2 -0.7434072 -0.02250229 -0.17719290 0.08253676 0 0 0.03842226 0 0 1 0
#> 3 -0.7816058 -0.36134842 -0.28622144 -0.30361510 0 0 -0.19130379 0 0 1 0
#> 4 -0.7816058 -0.36134842 0.02301955 -0.03962994 0 0 0.17310822 0 0 1 1
#> 5 -0.1560193 -1.42208413 -0.51456318 -0.33155179 0 0 -0.77922690 0 0 1 0
#> 6 -0.1560193 0.34580872 2.08071790 1.60539169 0 0 1.92402928 0 0 1 0