Forecasting with a BVAR

[EviewsUser1]

I estimated a BVAR with Minnesota priors, then I created a model, then I solved it to get a one-step ahead forecast for a single time period (2014:Q3) which is one-quarter beyond the end of my sample.
The forecasts are all zero except for one variable (the first variable in teh variable list has a non-zero forecast). What did I do wrong?

[EViews Gareth]

We'd need to see the workfile/VAR/Model.

[EviewsUser1]

let me know if you need to see more:

var var_test.bvar(prior=lit) 1 5 lctgh lcsh lfneh lfnph lfrh lfnsh lxmh lxsh lmmh lmsh lgfh lgsh vh_y
var_test.makemodel(bvarmod)
smpl 2014:3 2014:3
solve bvarmod
stop

[EViews Gareth]

We'll need the workfile.

[EviewsUser1]

workfile is attached

[EViews Gareth]

Is your copy of EViews 8 up to date?

The BVAR that's in your workfile hasn't estimated properly (open it and look at the coefficients). Yet if I re-estimate it, it works fine.

[EviewsUser1]

yes, I can get it to estimate the BVAR, but then when I solve the model all of the coefficients go back to zero

[EViews Gareth]

What is the build date of your copy of EViews? (Help->About EViews)

[EviewsUser1]

Jun 19 2013

[EViews Gareth]

You should update to the latest version.

[EviewsUser1]

That worked! Thanks.