The RATS Software Forum

Dear all,
I am estimating a model of monetary policy and I apply the Gali (1992) method (combinaison of short and long run restrictions).
I followed the method and I've adapted it to my case by integrating the three matrix and mcgraphirf procedure, but the results of the IRFs are strange and not logic.
The data and adapted programm are attached (setup and varirf).
Please, see if the procedures are right and help me to find and to resolve the problem of IRFs.
Thank you.

Ben

These just identify the model so you don't need the added constraint and your lr + sr_id greatly overidentifies the shocks. Is there a reason that you need the r6 constraint in your model? You might want to look at the Bjornland and Leitemo example which does a model identified with short and long run constraints without the added complication of the "A" model constraints that Gali uses.


input lr
. 0 . . 0
. . 0 . .
. . . 0 .
. . . . 0
. . . . .
input sr
. . . . 0
. . . 0 .
. . . . 0
. . 0 . .
. . . 0 .

Hello,
Tom thank you for your reply.

I have already used the method of Bjornland and Leitemo, the results are inconclusive, so I want to compare it with that of Gali (1992) in my paper.
I eliminated the sr_id (Gali (1992) and I kept both SR and LR for a justidentified estimtion.
I have adapted the program and I got the IRFs for inflation, but others do not.
Can you see what is and where is the problem?

New prog attached

Regards

1. If you just identify the model with SR and LR restrictions, that's Bjornland-Leitemo, not Gali.

2. You have source svarsetup.src not source svarsetupnew.src

3. On svarsetupnew.src your LR-SR model isn't correctly identified. You have 1-1-2-2-4 zeros in the five columns between the matrices. You need to have 0-1-2-3-4 in some order.

4. Your IMPULSE instruction is doing Cholesky factors not the structural shocks.

5. You have some series with strong seasonality. Just 4 lags isn't going to handle that particularly well. You need at least five.