Estimating the Effect of Training on Employment and Unemployment Durations: Evidence From Experimental Data

Using data from a social experiment, we estimate the impact of training on the duration of employment and unemployment spells for AFDC recipients. Although an experimental design eliminates the need to construct a comparison group for this analysis, simple comparisons between the average durations or the transition rates of treatments' and controls' employment and unemployment spells lead to biased estimates of the effects of training. We present and implement several econometric approaches that demonstrate the importance of and correct for these biases. For the training program studied in the paper, we find that it raised employment rates because employment durations increased. In contrast, training did not lead to shorter unemployment spells.