Employees choose hours of training partly on an unobserved ability that also
drives promotion, so training is endogenous. A randomly assigned training
subsidy shifts training; in this simulation it has no direct effect on
promotion, so the true exclusion-restriction violation is zero. Training
pays off more under a supportive manager. The same file is shipped as
inst/extdata/promotion.dta for the Stata command.
Format
A data frame with 4,000 rows and 6 variables:
- promoted
Promoted (0/1); the outcome.
- training
Hours of training (standardized); the endogenous treatment.
- subsidy
Training subsidy (standardized); the instrument.
- tenure
Tenure (standardized); a control.
- manager
Supportive manager (0/1); a control and moderator.
- team
Team identifier, 200 teams of 20; for clustered standard errors.
Details
Data-generating process (data-raw/example_data.do, seed
20260613): training = 0.7 subsidy + 0.3 tenure + 0.2 manager + a,
promoted = 1[-0.3 + 0.5 training + 0.4 tenure + 0.3 manager +
0.4 training x manager + e > 0], with a ~ N(0, 1) the unobserved
ability and e = 0.5 a + sqrt(0.75) u, u ~ N(0, 1).
Examples
data(promotion)
summary(promotion)
#> promoted training subsidy tenure
#> Min. :0.0000 Min. :-4.66581 Min. :-3.656443 Min. :-3.446013
#> 1st Qu.:0.0000 1st Qu.:-0.75280 1st Qu.:-0.669994 1st Qu.:-0.645404
#> Median :0.0000 Median : 0.05736 Median : 0.009475 Median : 0.005696
#> Mean :0.4692 Mean : 0.09299 Mean : 0.004036 Mean : 0.010388
#> 3rd Qu.:1.0000 3rd Qu.: 0.94379 3rd Qu.: 0.692029 3rd Qu.: 0.675735
#> Max. :1.0000 Max. : 4.02559 Max. : 3.649563 Max. : 3.932362
#> manager team
#> Min. :0.000 Min. : 1.00
#> 1st Qu.:0.000 1st Qu.: 50.75
#> Median :1.000 Median :100.50
#> Mean :0.507 Mean :100.50
#> 3rd Qu.:1.000 3rd Qu.:150.25
#> Max. :1.000 Max. :200.00