Office of Research Analyzing Effects of Job-Trainings Suffering Dropouts with an Optimal Multiple Matching

We analyze the effects on employment probabilities of job-trainings for unemployed Korean women during January 1999 to March 2000, where the job-trainings suffered substantial dropout problems. Three groups are pairwise compared: the control group with no treatment , the dropout group with a partial treatment, and the treatment group with the full treatment. To control for covariates, three propensity-score matching schemes are used: sequential pair matching, optimal pair matching, and optimal multiple matching with a variable number of controls. We find that dropouts occur mostly because a job was found or the subject self-evaluates the treatment as bad, and that the job-trainings lowered the employment probabilities, because the post-training unemployment duration (a " negative " benefit of the job-trainings) was not shortened enough to compensate for the training duration (a cost). The contribution of this paper is twofold: showing a way to deal with dropout problems taking advantage of information contents in dropouts and presenting an extremely simple algorithm to implement the optimal multiple matching.