L'estimation de modeles de regression lineaire autoregressifs avec erreurs residuelles autocorrelees et erreurs sur les variables

The very unfavourable effects of errors in the variables on the estimation of autoregressive linear regression models with autocorrelated disturbances, are first recalled. Then, the properties of estimators using shifted exogenous variables as instruments are analyzed, for large samples as well as for samples of smaller sizes. This latter case is also examined through Monte Carlo studies and heuristic solutions are proposed to overcome the problems encountered in these experiments. In conclusion, other possible approaches that could prove more efficient in certain cases, are suggested.