GIRF in a TVAR

[giangp]

Dear Tom,

Is there an easy way to extend the codes for TVAR so that one can get GIRF for each of the regimes specified in a model?

Thank you.

[TomDoan]

I'm not sure what you mean by "extend"? If you want the responses in a single regime, you can just estimate the VAR using a SMPL option and do standard impulse responses.

[giangp]

The problem is that under a VAR with threshold variable, standard IRFs are valid only in regions close enough to the threshold due to nonlinearity of the model. In the general case, some authors often use GIRFs based on bootstrap methods for each regime.

So my question is how to get this kind of GIRF within the TVAR procedure.

[TomDoan]

I'm still confused about your terminology of using GIRF's "for each regime". The GIRF's for a TVAR model have to be done by simulating the entire model, switching between regimes as needed. There are several examples of this in the forum. One is http://www.estima.com/forum/viewtopic.php?f=30&t=948. If you search for GIRF, you'll find others.

[giangp]

Dear Tom,

Sorry, I meant "conditioning on each regime" but the best terminology would express something like "conditioning on entire past history, the state of the economy and the size of the shock".

Thanks for the suggestion. Is it possible to calculate confidence bands for the GIRF using the quantiles of empirical distributions obtained from Monte Carlo simulations?

As an example, see Afonso, A., J. Baxa and M. Slavik (2011), “Fiscal Developments and Financial Stress: A Threshold VAR Analysis”, ECB Working Paper Series 1319, European Central Bank.

Thank you in advance.

[TomDoan]

See Balke RESTAT 2000