Auto-series fitting/forecasting: level or @pc()?

[T8SCH]

Dear all,
I realized that if I compute the fitted values in levels or using the expression of the dependent variable (in my case @pc) then values, of course expressed in the same units, actually differ! Until now I thought this option "d" was only a matter of convenience of how the output should look like...

This is more or less what I did:
eq01.fit(f=actual,d) fitted_pc
eq01.fit(f=actual) fitted_level
show fitted_pc @pc(fitted_level)

When I use option "d" I get the same numbers as from the menu (Equation\View\Actual,Fitted,Residual) so I guess I should forecast as well using the "d" option... but I actually want to understand why and what EViews8 is actually doing here?

Thanks in advance!

[EViews Gareth]

You're doing a fit, which is a static forecast (i.e. actual values of the lag of the dependent variable are used rather than forecasted).

When you do the @pc fit you're forecasting both Y and Y(-1). When you do the level fit you're only forecasting Y.

If you change your code to do a dynamic forecast rather than a fit, you'll see they match.