Research Summary
Research Summary
Optimal Heteroskedasticity Autocorrelation Consistent Covariance Estimators for GMM Weighting Matrices
Description
This paper considers the optimal bias-variance tradeoff for estimators of the long run covariance matrix used to generate GMM weighting matrices in time series contexts. Minimum MSE HAC estimators do not yield minimum MSE GMM estimators. Instead, achieving minimum MSE GMM estimates requires greater HAC bandwidths than those giving the minimum MSE long run covariance estimates. I derive a closed-form solution for the optimal bandwidth and introduce a data-dependent method for selecting its value.