Working Paper 36
Evaluating the outcomes of labour market programs (Part B)
Abstract
Part B of the paper introduces two econometric techniques to deal with selection bias when selection to treatment is based on characteristics unobservable to the researcher. The first technique is called the Instrumental Variables (IV) method. The IV method is the standard econometric treatment of the problem due to the correlation between an explanatory variable and the error term. In the case, the correlation between the treatment status and the error term of the outcome equation is generated by the selection bias. Angrist, Imbens and Rubin (1996) proved that the IV estimate of the treatment effect has an unambiguous interpretation under the potential outcomes framework. The second method is due to Heckman (1979) which saw the problem as a specification error. The error can be eliminated by adding an extra independent variable in the outcome equation. This extra variable captures the impact of the treatment status on the conditional mean of the error term in the outcome equation, whose construction requires an estimation of a structural model of treatment status selection.
Download Evaluating the outcomes of labour market programs (Part B) [PDF 80kb]
Contact Apprenticeships info 1800 210 210 or Training Queensland 1300 369 935
Other languages