Dynare forums

I am looking for examples of Bayesian vector autoregression (BVAR) estimation in Dynare.

I have read the example in bvar-a-la-Sims.pdf, and am looking for something more elaborate. In particular, how do I retrieve the estimated coefficient matrix and impulse response functions? And how do I choose the hyperparameters for the prior distribution?

I am trying to replicate the BVAR's in Figure 2 of Liu, Miao, and Zha (JME, 2016): Land prices and unemployment, see here: http://www.sciencedirect.com/science/ar ... 3216300290.

Yours sincerely,
Marcus Mølbak Ingholt
Ph.D. Candidate
University of Copenhagen

Unfortunately, there seem to be not many examples of this part of the code and it is poorly documented. If you have specific questions, I can try to guide you, but I am also no expert on this part of the code.

1. The posterior information is saved in

Code: Select all

oo_.bvar.posterior

in the unstable version.
2. To generate IRFs, you need to call

Code: Select all

bvar_irf(8,'Cholesky');

where the first number is the lag number. Results are saved in

Code: Select all

oo_.bvar.irf

Dear Johannes,

Thank you for your previous reply.

I have one additional question concerning the "bvar_irf" command. Is it possible to:

• Increase the number of periods on which to compute the IRFs?
• Change the confidence interval that is used when reporting IRFs?

These options do not appear to be available in bvar-a-la-Sims.pdf or on the "bvar_irf" description website.

Yours sincerely,
Marcus Mølbak Ingholt

Thanks for pointing this out. Before the

Code: Select all

bvar_irf

command put

Code: Select all

options_.irf=40;
options_.bvar.conf_sig=0.6;


to change the IRF length to 40 and the HPDI coverage to 60 percent.

That is great. Thanks a lot!