Research Summary
Research Summary
Heteroskedasticity Autocorrelation Consistent Covariance Matrix Estimation with Wavelets
Description
I propose a new HAC estimator based on the wavelet representation of the spectral density. Whereas kernel-based HAC estimators [e.g. Newey West (1987) Andrews (1991)] have a fixed bandwidth, a wavelet estimator has bandwidths that vary across wavelet resolution levels, allowing for targeted smoothing of estimator components. As a result, the wavelet HAC estimator achieves near-parametric convergence rates and has superior efficiency compared to kernel-based estimators. A Monte Carlo study confirms and then builds on the analytical results, showing that tests using the wavelet HAC estimator have dominating size control and power functions under both standard and fixed-b asymptotics.