On the use of robust regression in econometrics

The use of robust regression estimators has gained popularity among applied econometricians. The main argument invoked to justify the use of the robust estimators is that they provide efficiency gains in the presence of outliers or non-normal errors. Unfortunately, most practitioners seem to be unaware of the fact that heteroskedastic and skewed errors can dramatically affect the properties of these estimators. In this paper we reconsider the interpretation of the specific robust estimator that has become popular in applied econometrics, and conclude that its use in this context cannot be generally recommended.