Consider a setting where a treatment that starts at some point during a spell (e.g. in unemployment) may impact on the hazard rate of the spell duration, and where the impact may be heterogeneous across subjects. We provide Monte Carlo evidence on the feasibility of estimating the distribution...