Unobservable factors and panel data sets: an investigation in the labour market

This paper investigates the effects of unobservable factors that, as is well-known, contaminate two of the variables most used in labour market research, namely the stock of unemployment and the stock of vacancies. Using a matching function framework, we compare different panel data estimators using a number of appropriate Hausman tests robust to deviations from the classical errors assumptions. The relevance of the choice of the model specification is underlined. It is shown to what extent conclusions lacking a rigorous statistical analysis may be misleading