Why would vgxsim and vgxpred yield different forecasts?

I have estimated a VAR(2) model and need to forecast 6 variables, which can be easily done with vgxpred. I also used vgxsim to verify that the mean path was equal to the deterministic forecast obtained by vgxpred. Why would vgxsim and vgxpred yield different forecasts (i.e.the mean paths of the simulation are different from the deterministic forecast)?

Thanks!

Answer 1

[Hang Qian]

I think vgxsim and vgxpred will produce the same forecasts as long as the presample values are specified. For a VAR(p) model, vgxpred forecasts future variables conditional on the most recent p observations, while vgxsim simulates variables conditional on p presample values. For example, If the data Y serve as both observations for vgxpred and presample values for vgxsim, they should produce similar forecasts:

load Data_VARMA22

Yf = vgxpred(Spec, 5, [], Y);

Ysim = vgxsim(Spec,5,[],Y,[],10000);

Then mean(Ysim,3) is supposed to be very close to Yf.