Firm Data Analysis in Linked Employer-Employees Datasets

In linked employer-employee databases, when the employees are sampled first and then the employers data are attached, the set of firms that enter the sample is biased towards the large dimension. In this paper I discuss two strategies to impute inflating factors to the employers’ records, and to obtain a sample representative of the entire population of firms. A comparison between the two strategies is made with the help of Monte Carlo simulations. As an application, I build a rotating panel of Italian firms from a linked employer-employee longitudinal database of administrative source for the years 1986 to 1998, and compare some stylised facts derived on it with the existing knowledge on Italian firms’ size distribution.