摘要：This paper investigates the performances of GMM estimates using kernel methods with and without prewhitening and the VARHAC method in a representative agent exchange economy. A Monte Carlo study is conducted to evaluate the issues of estimating the spectral density functions, e.g., parametric vs. nonparametric, data-based bandwidth selection, and prewhitening procedures. The Monte Carlo results show that kernel methods with prewhitening procedure outperform others in terms of statistical inferences. The deviations from true parameter values, however, are larger for kernel methods with prewhitening procedure. Therefore, there exists efficiency/bias trade-off when choosing HAC covariance estimation method.
