Files
Abstract
Lancaster and Nickell (1980) have argued that in proportional hazard models the effects of time dependence and of unobserved sample heterogeneity cannot be distinguished. We show that both effects can be identified if the model allows for observed regressor variables in hazard function.