Do reported end dates of treatments matter for evaluation results? : an investigation based on the German integrated employment biographies sample

"The paper investigates the measurement error in end dates of further training programs in the German Integrated Employment Biographies Sample (IEBS) to gain insights on how measurement error in end dates of treatments affects evaluation results and on how to deal with this problem in future studies using the IEBS. Error-proneness of end dates in the IEBS is discussed, correction procedures are introduced and their influence on estimated employment effects is analyzed using basic descriptive analysis, propensity score snatching and a descriptive duration model. Though there is considerable measurement error in the end dates that can be corrected, the effect on evaluation results is modest, because for evaluation end dates are relevant only through indirect channels. The impact on causal effects based on matching is highest during the lock-in period and for long programs. The advice for future users of the IEBS is to decide depending on their evaluation design and goal if an explicit correction of end dates is necessary. In case an explicit correction is dispensable, it seems still important to take care that program spells are not given priority to employment spells in case of conflict." (Author's abstract, IAB-Doku) ((en))