Mixed proportional hazard models are commonly used to estimate duration dependence and unobserved heterogeneity in unemployment exit rates. Some strong assumptions are made in this framework, i.e. that the various influences on the individual unemployment exit rate are separable. The model we use in this paper allows for both the individual duration dependence pattern and the inflow composition to exhibit cyclical variations, thereby relaxing two of the three separability assumptions. The aim of this paper is to assess the validity of the third separability assumption, namely that the duration dependence pattern is the same for all individuals.