SEMINAR: Penalised Continuous-Updating GMM Estimators: Shrinking the No-moment Problem Seminar

Event details

Speaker: Dennis Kristensen (UCL)

Abstract:  We propose a class of penalised estimators as a solution to the no--moment problem of the Continuously Updated Estimator (CUE). We analyse the finite-sample and asymptotic properties of the CUE and its penalised version in a general nonlinear setting. We show that adding a penalty term to the objective function of the CUE help reducing its finite--sample variability. We also analyse the higher--order properties of the penalised version which provides guidelines for how to choose the penalty. Our preferred penalised CUE, which we call the quasi-likelihood GMM (QL-GMM) estimator, is obtained by adding the log-determinant of the optimal weighting matrix to the usual GMM objective function that defines the CUE. This penalty is justified asymptotically since the QL-GMM objective function is the large-sample log-likelihood of the sample moments. The penalised estimators we propose are easy to implement and perform well in simulations, restoring the finite sample moments of the CUE, at a small price in terms of slightly bigger biases compared to the CUE in some settings.

Host: Francesco Bravo (York)