DMCA
Asymptotic properties of a robust variance matrix estimator for panel data when T is large. Journal of Econometrics. forthcoming (2005)
| Citations: | 55 - 5 self |
BibTeX
@MISC{Hansen05asymptoticproperties,
author = {Christian Hansen},
title = {Asymptotic properties of a robust variance matrix estimator for panel data when T is large. Journal of Econometrics. forthcoming},
year = {2005}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract. In this paper, I consider the asymptotic properties of a robust covariance ma-trix estimator which is commonly advocated for use in panel data. The estimator is a generalization of the conventional heteroskedasticity consistent covariance matrix estimator for panel data which allows arbitrary correlation within each individual. Under the usual panel asymptotics where the cross-section dimension, n, grows large with the time series dimension, T, fixed, this estimator has good properties while allowing an essentially uncon-strained time series pattern of correlation. However, many panel data sets are characterized by a non-negligible time dimension. I extend the usual analysis by examining cases where n and T go to infinity jointly, considering both non-mixing and mixing cases, and show that conventional t and F tests based on the robust covariance matrix estimator are consistent. In addition, when T → ∞ with n fixed and other regularity conditions are satisfied, I show that the usual t and F statistics can be used for inference despite the fact that the robust covariance matrix estimator is not consistent but converges in distribution to a limiting random variable. I also explore the properties of a direct test of the assumptions underlying simpler covariance matrix estimators analogous to that of White (1980). The properties of the robust covariance matrix estimator and tests based upon are it examined in a short Monte Carlo study.
Keyphrases
panel data asymptotic property robust variance matrix estimator robust covariance matrix estimator direct test uncon-strained time series pattern time series dimension mixing case non-negligible time dimension simpler covariance matrix estimator short monte carlo study usual panel asymptotics many panel data set good property robust covariance ma-trix estimator usual analysis arbitrary correlation regularity condition cross-section dimension limiting random
