The RATS Software Forum

Dear All,

I am currently working on a panel VAR with sign and zero restrictions as per Gambacorta et al. (2012) (http://www.bis.org/publ/work384.pdf). However, I have just one question: When using a Panel-VAR, are there specification issues associated with using an identical series in each of the cross-sectional (e.g. - country specific) VARs? e.g. - some kind of 'global' variable which may take the form of an index (which has the same values), which occurs in every system? I have a fear that this may create bias in the SUR estimator, but this could be unfounded.

Thanks for your time, any suggestions or comments are appreciated.

I'm not sure I fully understand. Are you saying you want to do a panel VAR where one of the endogenous variables isn't individual-specific?

Hi Tom,

Yes, you are correct - is it possible to model a common endogenous variable across all systems, if its data generating process is approximated by the other variables specific to that cross-sectional system? I am looking to compare IRFs across regions, but some data is unavailable at a region specific level, and was thinking about approximating this by a national variable. In the case of the Gambacorta et al. (2012) paper, this might be thought of as a global volatility index measuring sentiment, or something similar to that. Is this going to create a huge problem, even if it is likely that each system is able to vaguely model the DGP of the common endogenous variable?

Many thanks