White period coefficient covariance

[Pedro]

Hi,

I've read in Eviews manual that " since we wish to compute standard errors that are robust to serial correlation (Arellano (1987), White (1980)), we choose White period as the Coef covariance method". This method is not appropriate for hekeroskedasticity? Which is the equivalent method in Eviews to the fully robust covariance matrix estimator mentionated by Wooldridge (Econometric Analysis of Cross Section and Panel Data)?
Thanks in advance.

[EViews Glenn]

Fully robust to what? More specifically, which estimator.

[Pedro]

Sorry, fully robust to heteroskedasticity and autocorrelation. The thing is that Eviews provide three different robust estimators, cross-section, period and diagonal and it is not clear too me what of the three is better if I want to get a "fully" robust estimator (my panel is N = 48, T = 5 and I have used Fixed Effect Model).

Kind regards,

Pedro

[EViews Gareth]

There isn't one that is fully robust to every possible case of autocorrelation or heteroskedasticity. You need to narrow down what you're worried about, then you can correct it.

[EViews Glenn]

Gareth is correct, though from your description you probably want the White Period. "White - Period", may be thought of as a White estimator where we assume that there may be between-period correlation, where there is no correlation across cross-sections. These data would be termed clustered by cross-section in that literature and is based on Arellano (2003). Panel Data Econometrics, Section 2.3.1. p. 18. Requires asymptotics in N and fixed T.

[Pedro]

Thanks both. In fact the White Period is the estimator I had used. I infere that If my worry is heteroskedasticity I should employ White Cross Section. Then, what is the point in using White Diagonal?
Pedro

[EViews Glenn]

White diagonal is a robust to a particular form of heteroskedasticity, where individual observations have their own variances (i.e., non-constancy across both cross-sections and periods).