Estimation et tests en présence d'erreurs de mesure sur les variables explicatives : vérification empirique par la méthode de simulation Monte Carlo

In this paper we present two new estimators which are robust in the presence of errors in variables. These estimators are much less erratic than their classic counterparts: The Durbin and Pal estimators. These new estimators are based upon sample moments of order greater than two. They may be viewed as special instrumental variable estimators where the instruments are obtained by taking powers of the explanatory variables. Data from Statistics Canada on consumer finances are used to evaluate the performance of our estimators. Monte Carlo simulations show that their biases are less than those of ordinary lest squares. Our estimators may be viewed as a special case of generalized method of moments (GMM). Consequently they take part in the actual trend of research in financial econometrics. Financial econometrics might benefit greatly form our new estimators in applying them to well known models, as the CAPM where the market portfolio which is essential to the empirical verification of this model is contaminated by important measurement errors. Risk premia measures which are related to this portfolio might be corrected by our estimators, this in the context of the generalized method of moments (GMM).