The Implicit Weighting of GMM Estimators

Paper recalls motivation for generalized method of moments (GMM) as well as the idea estimated generalized method of moments. Having demonstrated the danger of underrating the heteroscedasticity of disturbances and recalling Cragg's idea of improvement of estimation under heteroscedasticity, it offers a robustified version of it. Then it explains idea of implicit weighting of residuals, i. e. the idea of weighting of order statistics of squared residuals and outlines a modification of the least weighted squares resistant to heteroscedasticity. A proposal of robust version of GMM estimation concludes the paper.