Skip to main content
. 2019 Aug 16;134(4):1747–1791. doi: 10.1093/qje/qjz023

Table III.

Treatment Effects (Intention-to-Treat)

First year (12 months) Longer-run (24–30 months)
Outcome variable Mean No controls Post-Lasso Mean No controls Post-Lasso
(1) (2) (3) (4) (5) (6)
Panel A: Medical spending
 Total spending (dollars/month) [admin] 576.2 10.8 34.9 650.5 Inline graphic74.7 Inline graphic39.7
(48.5) (36.9) (58.5) (47.9)
[.937] [.859] [.618] [.754]
N = 3,239 N = 3,239 N = 3,152 N = 3,307 N = 3,307 N = 3,155
  Drug spending [admin] 132.0 Inline graphic8.5 Inline graphic6.1 148.8 Inline graphic25.2 Inline graphic22.2
(26.5) (12.0) (27.7) (16.4)
[.937] [.947] [.836] [.589]
N = 3,239 N = 3,239 N = 3,152 N = 3,307 N = 3,307 N = 3,155
  Office spending [admin] 69.5 Inline graphic6.1 Inline graphic2.0 74.2 Inline graphic6.6 Inline graphic4.8
(10.0) (4.4) (8.6) (5.4)
[.937] [.947] [.836] [.754]
N = 3,239 N = 3,239 N = 3,152 N = 3,307 N = 3,307 N = 3,155
  Hospital spending [admin] 313.0 22.2 24.6 353.5 Inline graphic31.7 Inline graphic20.3
(30.9) (28.1) (35.6) (31.9)
[.937] [.868] [.836] [.754]
N = 3,239 N = 3,239 N = 3,152 N = 3,307 N = 3,307 N = 3,155
 Nonzero medical spending [admin] 0.902 Inline graphic0.008 0.002 0.950 0.005 0.007
(0.011) (0.010) (0.008) (0.007)
[.937] [.947] [.836] [.754]
N = 3,239 N = 3,239 N = 3,152 N = 3,307 N = 3,307 N = 3,155
(1) (2) (3) (4) (5) (6)
Panel B: Employment and Productivity
 Job promotion [admin] 0.176 Inline graphic0.003 Inline graphic0.004 0.360 0.006 0.006
(0.013) (0.012) (0.017) (0.016)
[.952] [.922] [.996] [.996]
N = 4,146 N = 4,146 N = 4,130 N = 3,635 N = 3,635 N = 3,619
 Job terminated [admin] 0.113 Inline graphic0.013 Inline graphic0.012 0.204 0.002 0.002
(0.010) (0.009) (0.012) (0.012)
[.630] [.571] [1.000] [.998]
N = 4,834 N = 4,834 N = 4,753 N = 4,834 N = 4,834 N = 4,753
 Sick leave (days/year) [admin] 6.341 0.186 0.138 6.066 0.013 0.018
(0.230) (0.200) (0.204) (0.169)
[.816] [.880] [1.000] [.998]
N = 4,782 N = 4,782 N = 4,712 N = 4,782 N = 4,782 N = 4,712
(1) (2) (3) (4) (5) (6)
Panel B: Employment and Productivity
 Job promotion [admin] 0.176 −0.003 −0.004 0.360 0.006 0.006
(0.013) (0.012) (0.017) (0.016)
[.952] [.922] [.996] [.996]
N = 4,146 N = 4,146 N = 4,130 N = 3,635 N = 3,635 N = 3,619
 Job terminated [admin] 0.113 −0.013 −0.012 0.204 0.002 0.002
(0.010) (0.009) (0.012) (0.012)
[.630] [.571] [1.000] [.998]
N = 4,834 N = 4,834 N = 4,753 N = 4,834 N = 4,834 N = 4,753
 Sick leave (days/year) [admin] 6.341 0.186 0.138 6.066 0.013 0.018
(0.230) (0.200) (0.204) (0.169)
[.816] [.880] [1.000] [.998]
N = 4,782 N = 4,782 N = 4,712 N = 4,782 N = 4,782 N = 4,712
 Management priority on health/safety [survey] 0.790 0.057*** 0.050*** 0.784 0.028* 0.024
(0.015) (0.014) (0.016) (0.015)
[.001] [.003] [.539] [.632]
N = 3,566 N = 3,566 N = 3,514 N = 3,018 N = 3,018 N = 2,976
 Productivity index [survey/admin] 0.000 Inline graphic0.046 Inline graphic0.060 0.000 Inline graphic0.015 Inline graphic0.054
(0.061) (0.056) (0.062) (0.056)
[.450] [.283] [.805] [.328]
N = 3,309 N = 3,309 N = 3,300 N = 2,890 N = 2,890 N = 2,881
(1) (2) (3) (4) (5) (6)
Panel C: Health status and behaviors
 IL Marathon/10K/5K [admin] 0.066 0.002 Inline graphic0.005 0.052 0.006 0.001
(0.008) (0.006) (0.007) (0.006)
[.975] [.471] [.625] [.995]
N = 4,834 N = 4,834 N = 4,817 N = 4,834 N = 4,834 N = 4,817
 Campus gym visits (days/year) [admin] 5.839 Inline graphic0.062 0.401 5.047 Inline graphic0.342 0.001
(0.733) (0.360) (0.660) (0.367)
[.975] [.471] [.625] [.998]
N = 4,834 N = 4,834 N = 4,817 N = 4,834 N = 4,834 N = 4,817
 Ever screened [survey] 0.942 0.039*** 0.036*** 0.962 0.029*** 0.027***
(0.009) (0.008) (0.008) (0.007)
[.001] [.000] [.006] [.005]
N = 3,567 N = 3,567 N = 3,557 N = 3,020 N = 3,020 N = 3,010

Notes. Each estimate is from a separate regression of an outcome, specified by the row, on a treatment group indicator. Observations include the control and treatment groups. Longer-run time horizons are 24 and 30 months for survey and admin outcomes, respectively. Post-Lasso specifications control for covariates selected by Lasso to predict the outcome. Potential predictors include all available baseline variables in the same family of outcomes, strata variables, and the baseline survey variables reported in Table I, as well as two-way interactions between these predictors. Robust standard errors are reported in parentheses. */**/*** indicates significance at the 10%/5%/1% level using conventional inference. Family-wise p-values, reported in brackets, adjust for the number of outcome variables in each family. Results for all outcomes, categorized by family, are reported in Online Appendix Tables A.4a–A.4g (12-month outcomes) and Tables A.7a–A.7g (longer-run outcomes).