The RATS Software Forum

Hi,

I do not currently have RATS loaded up but have a problem that I hope RATS would be able to solve.

I am currently trying to create an SVAR that has block exogeneity (a foreign block that is determined only by its own variables, and a domestic block that is determined by its own history and also the foreign block). The implication here is that there are zero coefficients for the foreign block variables with regard to domestic variables. On a number of econometric package these lagged coefficient restrictions cannot be carried out and retain the full functionality of their respective VAR processes. For example, in EViews I can certainly estimate a system that is block exogenous but I lose the ability to do things like estimate contemporaneous restrictions, impulse response functions and variance decomposition, not to mention the diagnostics.

I see from various articles that RATS is quite often used with SVARs and I am wondering whether RATS can handle lagged coefficient restrictions and still provide the various features I noted above.

Thanks

Andrew

You can estimate a structural near-VAR using the two-step procedure of estimating the lag model, then applying CVMODEL or some other technique for estimating the covariance matrix model. These will give consistent estimates. The main question is under what situations you can easily get maximum likelihood estimates. Unlike the case with a full VAR, restrictions on the covariance matrix now have an effect on the maximum likelihood estimates for the lag coefficients. Thus, there is no easy way to get maximum likelihood estimates for the model if the covariance model is overidentified—you would have to estimate the lag coefficients and covariance model jointly. However, if you have a just-identified structural model, then a two-step approach of estimating the lag model (using iterated SUR to get maximum likelihood), then estimating the structural model in a second step will give maximum likelihood for the full model.

Tom,

would it be possible to get an example of a just-identified Near-SVAR model?

Thanks in advance

We don't have one out of the literature. There's an example of a near-VAR at:

http://www.estima.com/forum/viewtopic.php?f=4&t=522

That does a Cholesky factor rather than an SVAR, but the hard part is estimating the near-VAR, not the SVAR.