Université de Lausanne
Ecole des HEC
Département d'économétrie et d'économie politique

Mercredi 19 décembre 2007, 17h00
Extranef, Dorigny, salle 118

Pascal LAVERGNE
(Simon Fraser University, Canada)

Efficent Smooth GMM and Dimension Reduction

Abstract
We propose a new GMM criterion for models defined by conditional moment restrictions based on local averaging that resembles a statistic used in specification testing. Our approach defines a whole class of estimators depending on whether a smoothing parameter is fixed or decreases to zero with the sample size. We show that consistency and asymptotic normality follows in both cases. At first-order, letting the smoothing parameter tend to zero yields a semiparametric efficient estimator, and we provide a two-step efficient version. We also investigate a dimension-reduction device in this context. While the resulting estimator does not attain the asynptotic efficiency bound, its small sample properties may be preferable.

Site web du séminaire (avec texte en ligne): http://www.hec.unil.ch/deep/evenements-english/e-sem-all-2007-08.htm