Page 1 of 1

VAR/VECM, number of variables included.

Posted: Mon Apr 18, 2011 8:24 am
by hgaronfolo
Hi,

I am estimating VARs or VECMs depending on whether my variables are cointegrated or not. In connection with this, I was wondering how many variables I should include in a VAR/VECM model?

For instance, I have found that four macroeconomic variables are correlated with stock market capitalization, should I then include all four variables in my VAR/VECM or should I rather estimate pairwise models? (I.e. first model: market capitalization and var1, second model: market capitalization and var2 and so on)

Is it appropriate to use e.g. Akaike information criterion to judge how many variables to include when it comes to VAR/VECM models?

Any input is greatly appreciated!

Have a nice day.

/Herb

Re: VAR/VECM, number of variables included.

Posted: Wed Apr 27, 2011 11:35 pm
by dealsfe
i think, you should study about omitted variable problem