ordering variables in VAR yields different IRF

[peter_h]

HI all,

i have this problem that I'm runnig VAR with four variables and when I change the order of input of endogenous variables, my impulse responses change and yields into conflicting results, e.g. response of A to B was significant in order of variables A B C D, but insignificant when I changed the order let's say to A C D B. Any idea? what is the problem? thnx

[rajamohsin]

Hi there
Ordering of variables is important for Impulse Response Function specially when using Cholesky method. It happens due to the assumption of zero lower triangle matrix and so the ordering is set according to theory. For referrence see
Introductory Econometrics for Finance by Chris Brooks, page 299

[ajkoziol]

A better reference is Lutkepohl (1991) p53-56 Google books allows you to read this section in preview.