We propose a two-stage instrumental variable estimator that is consistent if there is a selective compliance in the treatment group of a randomized experiment and the outcome variable is a censored duration The estimator assumes full compliance in the control group We use the estimator to...