Known results on the identification of structural duration dependence in the presence of unobserved heterogeneity depend crucially on the proportional hazards assumption. Here, I show that variation in covariates over time, combined with variation across observations, is sufficient to ensure...