EViews.com

I am just wondering if the built-in forecasting function for VECM is wrong. I noticed that the forecasts produced from the forecast function or (button) will not produce the summary statistics for the individual equation. Can someone confirm that the function is correct? You can't produce the same statistics as the one reported in the vector error correction model.

Using a simple example below

wfopen "base.wf1"
genr logab=log(ab)
genr logrer=log(rer)
genr logus_inc=log(us_inc)
genr logwti_p=log(wti_p)

'VECM
smpl 1992M04 2008M12
var vecm1.ec(e,2) 1 2 logab logrer logus_inc logwti_p
smpl @all
vecm1.fit(e,g) _ef
smpl 1992m04 2008m12
genr diff=(d(logab)-d(logab_ef))^2
scalar SSR= @sum(diff)

SSR should produce the same in-sample sum of squared residuals of 0.395 as in the vecm1 output.

If I use the way we did it in the old version

vecm1.makemodel(mod1)
'smpl 1992m04 @last
smpl @all
mod1.solve(d=d)

The produced forecasts are very different from the summary stat provided. Can someone clarify this?