VECM IRF confidence bands with "we" variables

[ylijohtaja]

Hi all,

I have been trying to modify the VECM impulse response confidence bands code so that it would take account of the weak exogeneity of a number of variables. Unfortunately, it seems that my programming does not work properly. My modeling yields responses that appear to be either based on a non-restricted alpha vector or on a VAR model (i.e. all the alphas zero) - I cannot get IRF's that would correspond to a VECM with weakly exogenous variables. I wonder what is wrong and whether there is a code to derive the proper IRF's for a VECM with weakly exogenous variables? The code is attached.

Best,
E

[TomDoan]

It would help to have the data as well, but the use of the weakly exogenous variable breaks the symmetrical structure that is relied upon for doing standard MC integration. You would need to use the Gibbs sampler as in the MONTESUR.RPF example. You'll probably find that easier to use anyway.

[ylijohtaja]

Thanks Tom!

So the easiest way would probably be to modify the "Montesur" code to cater for the long-run dynamics, i.e. to specify the cointegrating vector and to include the "ect" line in the estimated model for the adjusting variables - is this correct?

E