Computation of autocorrelations

[ErikG]

Hi,

Thanks for a great job maintaining and developing Eviews, which we use on a daily basis.

I was reading the help file for autocorrelations (AC) and stumbled across this section:

Note that the autocorrelations estimated by EViews, which follow those in the literature, differ slightly from the theoretical definition of the autocorrelation:

[EQ 11.25]

The difference arises since, for computational simplicity, EViews employs the same overall sample mean [Ybar] as the mean of both Y_t and Y_t-k. While both formulations are consistent estimators, the EViews formulation biases the result toward zero in finite samples.

In these days of superfast computers, perhaps you should do away with computational simplicity and go for the exact formulation instead?

Regards,
Erik,
NIER Sweden