Robust Estimation: An Example

The use of Ordinary Least Squares and its generalizations is widespread among economists, although it should be common knowledge that a single outlying observation can cause this technique to produce arbitrary estimates and hence incorrect t-values. In this paper a robust alternative, an example of a General M estimator, is discussed. Not only has this estimator a breakdown point of 50%, it also yields consistent estimates and, as we show by means of a simulation experiment, is more efficient than Rousseeuw’s Least Median of Squares estimator. We also propose a specific correction factor which improves both the resampling algorithm and the projection algorithm for computing the Minimum Volume Ellipsoid estimator. We are much indebted to Teun Kloek (Erasmus University Rotterdam) for guiding us into the field of robust estimation. André Lucas (Erasmus University Rotterdam) provided valuable comments. Robert Waldmann (European University Institute) increased our understanding. The usual disclaimer applies. © T he A ut ho r(s ). Eu ro pe an U ni ve rs ity In st itu te . D ig iti se d ve rs io n pr od uc ed b y th e EU I L ib ra ry in 2 02 0. A va ila bl e O pe n Ac ce ss o n C ad m us , E ur op ea n U ni ve rs ity In st itu te R es ea rc h R ep os ito ry .