Robust Estimation in Simultaneous Equations Models·

In this paper we review existing work on robust estimation for simultaneous equations models. Then we discuss three strategies for obtaining estimators with a high breakdown point, a controllable efficiency, and a reasonable computational cost: (a) robustifying Three-Stages Least Squares, (b) robustifying the Full Information Maximum Likelihood method by minimizing the determinant of a robust covariance matrix of residuals, and (c) generalizing multivariate tauestimators (Lopuhaa 1991) to these models. The latter seems the most promising approach. ·Maronna, Universidad Nacional de la Plata, Argentina; Yohai, Universidad de San Andres, Argentina and Universidad de Buenos Aires. The authors were partially supported by C.I.P.B.A. and C.O.N.I.C.E.T., respectively. Most of this work was done while the first author was a visiting professor at Universidad Carlos III de Madrid, Espafia. 1 Robust Estimation in Simultaneous Equations Models· Ricardo A. Maronna 1 and Victor J. Yohai 2 lUniversidad Nacional de La Plata, C.C. 172, 1900 La Plata, Argentina 2Universidad de San Andres, V. Dumas 284, 1644 Victoria, Argentina and Universidad de Buenos Aires •