Research Summary
Research Summary
An Unlimited Moments GMM Estimator
Description
A short time series relative to the number of moment conditions in a GMM framework yields an inconsistent estimator. To circumvent this problem, researchers generally restrict the number of moment conditions to some fraction of the length of the time series. I develop an estimator that allows for consistent estimation with unlimited moment conditions. It works as follows. GMM estimates are obtained from all size-k subsets of the moment conditions, where k denotes the maximum number of moment conditions that can be used while still yielding a consistent GMM estimator. A weighted average of these subset estimates forms the final estimate. This estimator can be thought of as a generalized U-statistic. I characterize the properties of the estimator and develop its distribution theory. I show that it dominates the standard GMM estimator in terms of both efficiency and robustness.