My question is about ordering variables in VARs. As far as I know, one of the criticisms of VARs comes from the fact that, changing the order of variables in the VAR system changes the results to be obtained.
Specifically, in my paper, there are four macro variables: Inflation (pi), Output Gap (x), Government Spending (G) and interest rates (r). My purpose is to obtain the impulse response functions from these. The question then is: which is the "best" ordering? For instance, I can order "pi=>x=>G=>r" or "x=>pi=>r=>G." I would like to examine the "r" shock and "G" shock in my paper.
Thanks in advance.
Hi Mizuki,
The concern about ordering in a VAR is specially due to the impulse response (IRF) specification.
If you're going to use a Cholesky decomposition in order to realize your IRF then you need to take care of the order. The idea is that, when you do the cholesky you'll have an lower triangular matrix or a upper triangular (it is important to check what of those is your software computing). In the lower triangular case, you will need to order with an exogeneity criterium, it is put first the most exogenous of your variables, then the second on until you put last the variable for which the all the variables have effect on it.
If you're thinking on compute an Structural VAR then you may not use Cholesky, and the matrix for your IRF has to be specified with an economic criterium, the order not matter at all if you specify good your matrix. (2002 - Monetary Policy Transmission through the ConsumptionWealth Channel - Ludvigson, Steindel, Lettau check this paper).
Best regards,
Dear Dr. Sanin-Restrepo,
Hi - many thanks for your intuitive explanation and for pointing out the paper! I use Eviews and a lower triangular matrix seems to be the key. This paper looks non-technical and accessible. I'll read the paper and understand VARs better. Thank you very much! : )
Warm Regards,
Mizuki
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments