A Simple Procedure for the Evaluation of Treatment Effects on Duration Variables

Often, a treatment and the outcome of interest are characterized by the moment they occur, and these moments are realizations of stochastic processes with dependent unobserved determinants. We develop a simple and intuitive method for inference on the treatment effect. The method can be implemented as a graphical procedure or as a straightforward parameter test in an auxiliary univariate single-spell duration model. The method exploits information on the timing of the treatment relative to the outcome that is discarded in binary treatment analyses