Lagged duration dependence in mixed proportional hazard models

We study the non-parametric identification of a mixed proportional hazard model with lagged duration dependence when data provide multiple outcomes per individual or stratum. We show that the information conveyed by the within strata variation can be exploited to non-parametrically identify lagged duration dependence in more general models than in the literature. -- lagged duration dependence ; mixed proportional hazard models ; identification ; multiple spells ; parallel data