Bounding a matching estimator: the case of a Norwegian training program

In this paper we evaluate a Norwegian vocational training rehabilitation pr ogram by comparing employment outcomes of trainees and nonparticipants usin g nonexperimental data. A matching estimator is used to calculate the train ing effect for different subgroups of the sample. We demonstrate how boundi ng the matching estimator can be used to evaluate the intrinsic uncertainty of estimated training effects due to selection on unobserved individual ch aracteristics. After adjustment for observed selection into training programs we find that the overall training effect is around six percentage points. This is mainl y due to a high and significant effect for individuals with a low probabili ty of program participation. After calculating upper and lower bounds on th e test-statistics used to test the hypothesis of no training effect we find that the overall effect is sensitive to unobserved selection. However, the result that the training effect is positive for individuals who are less l ikely to participate in a training program is not sensitive to selection bi as. These individuals also have the lowest employent probabilities, which i ndicates potential harmful cream skimming in the Norwegian vocational rehab ilitation sector.