Bias Reducing Estimation of Treatment Effects in the Presence of Partially Mismeasured Data

Labor market policy evaluation studies often rely on a merged database from different administrative entities, where part of the information might be archived with varying quality in different sources. Suppose that one observes inter alia a variable of dubious quality for the entire population and the same variable for a subgroup, say the treated, with good quality from an extra source. This paper introduces a bias reducing estimator of the propensity score, as a widespread tool in this area, that uses validation data in order to control for mismeasurements of the non-validation units when treatment and validation status are binary and coincide. A Monte Carlo simulation reveals that estimation of average treatment effects, based on this first-step propensity score estimation in combination with expected propensity scores, performs better with respect to bias and mean squared error compared to the case of either using or ignoring the validation data in a naive parametric propensity score model. An application to widely used German
administrative data underlines its relevance.