Skip to main content

Table 9 Sensitivity analysis, calibrated confounders for kernel matching estimate of ATET, men

From: The effect of vocational rehabilitation on the employment outcomes of disability insurance beneficiaries: new evidence from Canada

 

Outcome effect

Selection effect

ATET

SE

Leaving disability rolls

No confounder

--

--

0.065

0.073

Neutral confounder

1.008

1.051

0.065

0.071

Confounders

    

High school

1.703

3.508

0.065

0.081

College

3.194

2.213

0.065

0.090

University

1.566

1.325

0.065

0.083

Married

0.601

0.862

0.065

0.074

Mental disorders

0.750

0.812

0.065

0.084

Musculoskeletal disorders

0.660

0.972

0.065

0.081

Zero earnings 1-year prior to application

0.641

1.322

0.065

0.081

Zero earnings 2-years prior to application

0.713

1.515

0.065

0.074

Gainful employment

No confounder

--

--

0.074

0.081

Neutral confounder

1.011

1.028

0.074

0.084

Confounders

    

High school

1.981

3.640

0.074

0.071

College

2.990

2.175

0.074

0.083

University

1.247

1.278

0.074

0.072

Married

0.560

0.868

0.074

0.078

Mental disorders

0.862

0.805

0.074

0.082

Musculoskeletal disorders

0.599

0.994

0.074

0.083

Zero earnings 1-year prior to application

0.554

1.304

0.074

0.071

Zero earnings 2-years prior to application

0.647

1.509

0.074

0.077

Substantial gainful employment

No confounder

--

--

0.051

0.075

Neutral confounder

1.017

 

0.051

0.069

Confounders

    

High school

1.852

3.634

0.051

0.069

College

2.946

2.196

0.051

0.065

University

1.352

1.298

0.051

0.087

Married

0.569

0.872

0.051

0.062

Mental disorders

0.085

0.797

0.051

0.073

Musculoskeletal disorders

0.639

0.974

0.051

0.069

Zero earnings 1-year prior to application

0.510

1.300

0.051

0.087

Zero earnings 2-years prior to application

0.593

1.532

0.051

0.070

  1. Notes: ATET denotes average treatment effect on the treated; SE denotes standard error. The outcome and selection effects are odds ratios. The rows in the tables denote the confounding variable, the first two rows in each panel of estimates show the ATET with no cofounding factor and with a neutral confounder. The kernel and local linear matching estimates were obtained with the Epanechnikov kernel, with bandwidths computed using cross-validation. All standard errors are obtained by bootstrapping with 1000 replications.