Heteroskedasticity and Autocorrelation Corrections
Many time series studies, including in particular those estimated by generalized method of moments, involve disturbances that are serially correlated and, possibly, conditionally heteroskedastic. The serial correlation and heteroskedasticity often are of unknown form. Corrections for serial correlation and heteroskedasticity are required for inference and efficient estimation. This article surveys procedures to implement such corrections.
KeywordsAutocovariances Bartlett kernel Generalized method of moments Heteroskedasticity Heteroskedasticity and autocorrelation consistent (HAC) covariance matrix estimation Kernel weights Long-run variance Moving average processes Newey–West estimator Quadratic spectral kernel Serial correlation Spectral density estimation Truncated estimators Vector autoregressions
I thank Steven Durlauf, Masayuki Hirukawa and Tim Vogelsgang for helpful comments, and the National Science Foundation for financial support.
- den Haan, W.J., and A.T. Levin. 1997. A practitioner’s guide to robust covariance matrix estimation. In Handbook of statistics: Robust inference, ed. G. Maddala and C. Rao, Vol. 15. New York: Elsevier.Google Scholar
- Hamilton, J. 1994. Time series analysis. Princeton: Princeton University Press.Google Scholar
- Priestley, M.B. 1981. Spectral Analysis and Time Series. New York: Academic Press.Google Scholar