Page 1 of 1

VECM forecasting

Posted: Mon May 06, 2013 2:15 am
by selin1000
Hi everybody!

I want to test the out-of-sample forecasting ability of a VECM on Eviews and am stuck. What I do is the following:
1) I divide my sample in two (I have monthly data series)
2) I estimate the VECM in my first sub-sample
3) I solve the model for my second sub-sample and manually compute the forecast errors

I would like to understand what Eviews does here exactly. Does Eviews compute 1-step ahead forecasts here, and does it employ rolling or recursive schemes?

How can I use the rolling scheme, and is it possible for me to do 2-step and 8-step ahead forecasts as well?

Thank you for your help so much!

Re: VECM forecasting

Posted: Mon May 06, 2013 7:49 am
by EViews Gareth
It does not use a rolling forecast. You'll have to write that yourself in a program. A search of the forum should bring up examples.

Re: VECM forecasting

Posted: Mon May 06, 2013 8:38 am
by selin1000
Gareth, tx a ton for your response. Is it one-step ahead forecasts? What does Eviews do? I'm a bit confused

Re: VECM forecasting

Posted: Mon May 06, 2013 8:42 am
by EViews Gareth
Depends on whether you are performing a dynamic or a static solve.

Static solve uses the fixed coefficient estimates, and the actual past values of the series. Dynamic solve uses fixed coefficients, but the forecasted/solved past values of the series.