Erreurs de mesure sur les variables économiques et financières

Abstract

The ordinary least squares (OLS) estimator is biased when there are errors in variables. This article discusses two new robust estimators in the presence of measurement errors. These new estimators are based on the higher moments of the explanatory variables. These estimators can be viewed as special types of instrumental variables estimators, in which the instruments are built by taking the explanatory variables up to specific powers. The performance of these new estimators is evaluated by performing Monte Carlo simulations which show that the biases are lower than those associated with the ordinary least squares estimator. The estimators developed in this article are consistent with the estimator of the generalized method of moments (GMM). The field of financial econometrics could greatly benefit from our new estimators by applying them to well-known models such as the CAPM. In this model, the market portfolio, which is the basis for the empirical validity of the model, is measured with errors. The measures of the risk premium associated with it could be corrected using our robust instruments with the GMM method. Finally, a financial application is presented where we apply these estimators to estimate the market beta, which is used in the formula of the average cost of capital (WACC). In this new application, the parameter estimate increases significantly, which is consistent with the empirical facts. (English) [ABSTRACT FROM AUTHOR]