Author Description

In this report, we used the data from the 1998 Adult Education and Training Survey (AETS) to estimate the impact of participating in adult education and training on the employment and earnings of Canadians. We applied methods that assume "selection on observables", including both standard regression-based methods and propensity score matching methods. We also applied methods based on instruments or exclusion restrictions, including standard instrumental variables estimation and the well-known Heckman bivariate normal selection estimator. These methods aim to deal with "selection on unobservables". We find that none of the methods we examined produced plausible estimates of the impact of adult education and training, although the methods that assume selection on observables produced more reasonable estimates than those that assume an instrument or exclusion restriction. Based on the results of our analysis, we suggest improvements to the AETS that would make it a better tool for estimating the labour market impacts of adult education and training.