BVAR long-run + sign restrictions

[KOBE24]

Dear Tom,

I am estimating a Bayesian VAR where my aim is twofold

1. I want an estimate of potential output
2. I want to assess the effects of other shocks on output

Basically, I am trying to extend this paper by Luca Benati
https://ideas.repec.org/a/eee/ecolet/v1 ... 3-119.html
to allow for other shocks beyond technology.

Since I am using a Blanchard-Quah decomposition, my first try was doing it by using Bjornland - Leitemo (JME, 2009, Rats example) code.
However, I was wondering whether you have some suggestions to achieve identification of other shocks through sign restrictions, preserving the first one by using the conventional long-run one.

I hope this is not too confusing. In a nutshell: can I run a RATS code in which I combine BQ and Uhlig methodology?

Best regards,
KOBE

[TomDoan]

Your first shock would have to have enough restrictions to be fully determined, so for an m variable system, you would need m-1 zero restrictions (either impact or long run). If you have that, then you just adapt the existing Mountford and Uhlig program to compute the first shock rather than "guess" it using simulations.

[KOBE24]

Thanks for your nice answer, Tom.
I will have a look at MU code and then get back to you, once I modify it (hopefully in an appropriate way).