Never Breaking the Wall The old assimilation paradigm confronted to recent econometric techniques ∗

In this paper we analyze the process of integration into the European societies of the cohorts of immigrants that reached Western Europe before the mid-1990s using a longitudinal dataset. The relevant question of the paper concerns the assimilation of immigrants to natives in terms of earnings. However, because not all the population is employed, as recognized in most of the existing literature, in the empirical analysis we model the sample selection mechanism, without sacrificing the panel nature of our dataset, as instead always happens in the literature. Using a OLS we would conclude that the assimilation process is strong and alive, but, as going to the FE and to an attrition bias FE consistent estimator, results are dramatically changed and we finally reject the assimilation hypothesis in earning. Where the assimilation has some importance is instead in the employment mechanism. Amongst the various robustness checks, of particular relevance is the cure for possible endogeneity between earning and length of stay in the host country, which requires the generalization to instrumental variables of available estimators