The RATS Software Forum

Hi,

I'm trying to implement the Loría, Armando and Salgado (2010). "New Evidence on the Monetary Approach of Exchange Rate Determination in Mexico 1994–2007: A Cointegrated Svar Model"

This is the URL:
http://www.sciencedirect.com/science/ar ... 0609000771

The paper uses cointegrated SVAR model, does anyone have sample code?
Thank you

That's behind a pay wall, but in general there's nothing all that special about doing a VECM SVAR as you can (as with regular VAR's) do that in two steps---estimate the VECM, take the residual covariance matrix and apply CVMODEL to do the structural contemporaneous model.

TomDoan wrote:That's behind a pay wall, but in general there's nothing all that special about doing a VECM SVAR as you can (as with regular VAR's) do that in two steps---estimate the VECM, take the residual covariance matrix and apply CVMODEL to do the structural contemporaneous model.

Dear Tom,
Thank you very much for reply

Is this method applicable for the Cointegrated SVAR-GARCH Model?

No. Any type of SVAR-GARCH model needs to be estimated by full sample maximum likelihood---by the nature of it being "GARCH", there's no fixed covariance matrix to model separately.

TomDoan wrote:No. Any type of SVAR-GARCH model needs to be estimated by full sample maximum likelihood---by the nature of it being "GARCH", there's no fixed covariance matrix to model separately.

That means there is no way to estimate Cointegrated SVAR-GARCH Model? if not could you explain with more detail.
Thank you!

Does the paper you cited do that? There's no really solid definition of what "SVAR" means in the context of a GARCH model.

No. Actually the authors estimated just cointegrated svar model. But, there is an article (below) in which authors used cointegrated VAR-GARCH model. Hence, is it possible to use a similar methodology for estimate cointegerated SVAR-GARCH? I would be grateful if you could possibly guide me.

http://econpapers.repec.org/paper/corlouvco/1997080.htm

The mhhp799.rpf example (from the Martin, Hurn and Harris book) does an SVAR-GARCH model. It's rather complicated and not really a simple combination of the two ideas. It also (in this case) doesn't really fit well; it does much worse than a simple CC-VAR-GARCH. If you're interested in learning more, there's a detailed description of the MHH example in the GARCH/Volatility e-course.