The purpose of this paper is to investigate the loss of efficiency of the simultaneous error components model's estimators in the case of incomplete panels. The static and the dynamic cases were analysed, when from a panel data base those individuals are dropped for which the observations are not complete.