svar identification question

[Alex_Co]

Dear Dynare Community,

I am currently going through the DH10.mod example in order to learn the MS-SBVAR estimation package in Dynare. I have a question concerning the exclusion restrictions from the SvarExclusionInterface on WikiDynare (http://www.dynare.org/DynareWiki/Markov ... gInterface):

How should I interpret these lines:
svar_identification;
exclusion lag 0;
equation 1, y, pi;
equation 2, pi, r;
exclusion lag 1;
equation 1, y, pi;
equation 2, pi, r;
end;

Thanks for your help

Alex

[Alex_Co]

Hi,

I think I found out the source of my confusion. In the DH10.mod example, there is a identification restriction on the coefficient in front of KCFSI_t because the Dynare MSVAR is defined with a diagonal matrix in front of the structural shocks. Since in Davig and Hakkio (2010), there is the A(s_t) matrix in front of the Sigma matrix, I have to premultiply the whole equation by A^-1. That ends up giving me the same kind of restriction. Is my interpretation right?

I have three more questions:
(1) In Davig and Hakkio (2010), some parameters are state-independent. Any ways to do this in Dynare?
(2) Is there a way to get IRF conditional on the regime?
(3) The IRF I get from the example are quite different from the ones in the paper (see attached mod file). How come?

Thanks a lot

Alex

[StephaneLhuissier]

I have three more questions:
(1) In Davig and Hakkio (2010), some parameters are state-independent. Any ways to do this in Dynare?
(2) Is there a way to get IRF conditional on the regime?
(3) The IRF I get from the example are quite different from the ones in the paper (see attached mod file). How come?

(1) no you cannot, except if it is for an entire equation.
(2) By recovering A(st) and F(st). I am wondering if Dynare can do it. It was a bug.
(3) As explained in the example, there are few differences with their original paper.