Authors	van den Berg, G., Caliendo, M. and Uhlendorff, A.
Year	2009
Reference	Working Paper, Bonn.
Keywords	Program Evaluation, Matching, Multivariate Duration Models, Monte Carlo Simulation
JEL-Classification	C41, C21
Presentations	2009 ESEM in Barcelona
Download	Available soon.
Abstract	In the absence of instrumental variables, two popular strategies for the estimation of treatment effects exist: matching methods and multivariate duration models. Both approaches are based on different assumptions. While matching methods rely on the conditional independence assumption, but generally do not impose any parametric assumptions for the estimation of the treatment effect, multivariate duration models allow for the selection on unobservables, but imply parametric assumptions. This paper analyzes and compares the performance of both approaches in an environment with dynamic assignment into treatment.