Abstract

Many questions regarding labour markets require to account for unobserved worker and firm heterogeneity. Those factors can indeed explain things such as wage inequality, gross job flows or changes in sectoral employment. So far the main econometric methodology to identify firm and worker heterogeneity has been the Abowd, Kramarz and Margolis (1999) approach. This methodology is however plagued by identification bias and incidental parameter bias. Both arise from the fact that movers across firms are used to identify fixed effects. However, typically movers are a small sample, hence the small mobility bias leads to imprecise identification and to few observations (worker-firm network links) used to estimate many worker and firm fixed effects. A novel methodology that solves both problems is the one devised by Bonhomme, Lamadon and Manresa (2019). Technically, they propose a two stage-finite mixture model (worker- random effects interacted with firm fixed effects), where the first stage consists of grouping firms so that eventually the entire network is connected in the grouped set-up. Implementing those methods can be extremely valuable for researchers, but is very hard to do. Our project aims at implementing the method on the IAB datasets in a way that can be easily updated and replicated over time.