This project is read-only.

Panel VAR

Topics: VAR, SVAR, FAVAR
Feb 2, 2015 at 9:46 AM
Hello,

I refer to the tutorial about panel VARs found here: https://www.codeplex.com/Download?ProjectName=iristoolbox&DownloadId=852590

To be honest, I'm still not completely sure what the difference between panel VARs and large standard VARs is.

If I follow the tutorial, it looks like a panel VAR is like the three individual VARs with the restriction that the companion matrices are the same e.g.

Y_{i,t} = alpha_i + A*Y_{i,t-1} + epsilon_{i,t}, i=1,2,3,

where alpha_i is the fixed effect and A is the same between the different groups. Now I'm wondering why
i) in the example of the tutorial the fixed effects are exactly the same for all three countries
ii) is this really a "panel VAR"?
A panel VAR is defined on page 7 of https://ideas.repec.org/p/cpr/ceprdp/9380.html (equation 2). The companion matrix A_i are allowed to be different between the groups and y_{i,t} of group i depends on the lagged variables of all countries and not only of country i. But for me, it looks like I could write that model of equation 2 as a large standard VAR??

Thanks for your help!

Regards,
Sven
Feb 2, 2015 at 12:12 PM
Hi Sven

Thanks. First off, there are many types of panel VARs, or panel models in general, not just one. The panel VAR in IRIS is just one example of them, albeit the simplest possible one (but I never claimed to be utterly sophisticated on the VAR front -- this is not my primary area of expertise and interest): a so-called fixed effect model. In fixed effect VAR models, all countries/groups share the same transition matrix polynomial and the covariance matrix, and can have differing constant terms. When estimating the panel VAR, you can also weight the individual countries/groups so that the estimates of the transition matrix and the covariance matrix can reflect the importance/size/etc. of individual countries/groups.

And you are right, all fixed-effect regressions can be thought of as one large model with the little tweak that constant terms are allowed to differ across countries/groups.

Best,
Jaromir
Marked as answer by jaromirbenes on 2/2/2015 at 4:12 AM
Feb 4, 2015 at 10:39 PM
Hello Jaromir,

thanks for clarification and thanks for making all the great work available to others!

Regards,
Sven