Shutdown Methodology

[PTillmann-436]

Dear Tom,

Is there a code on counterfactual impulse responses such as in this paper (section 4)

http://www3.nd.edu/~esims1/bachmann_sim ... august.pdf

or in Sims and Zha (Macroeconomic Dynamics, 2006)?

They construct counterfactual shocks in order to set the reponse of one variable to an identified shock to zero, that is, they show down a certain transmission channel.

Thanks a lot for your help.

Peter

[TomDoan]

Unlike the historical decomposition which zeros out shocks, their analysis shuts down responses. That requires sequentially solving for a shock which counteracts the effects of other shocks. An example is provided at viewtopic.php?p=19222#p19222.

[PTillmann-436]

Fantastic, thanks a lot, Tom

[PTillmann-436]

Tom,

Do you have an idea about how to do it with sign restrictions? There, every accepted draw has to be neutralized by the shutdown shock.

Thanks again
Peter

[TomDoan]

Do you have a reference on that? The paper cited above doesn't do anything with sign restrictions.

[PTillmann-436]

Benati constructs counterfactuals based on sign restrictions:

http://users.ugent.be/~gpeersma/gert_fi ... er2014.pdf

Thanks for your help!

[TomDoan]

Whatever he's doing (and unfortunately it's rather vague), it's quite different from the shutdown methodology described above. That's impulse response calculation (this sequence of shocks produces this sequence of responses). He's describing calculations that make alterations to the shocks to make sure the interest rate doesn't go negative, so is valid only at a specific set of initial conditions.

[RomainR]

Dear Tom,
I would like to run a "shutdown" SVAR (in the spirit of Sims&Zha and Bachman&Sims JME 2012) but applied to technology shocks, rather than to government spending shocks. Accordingly, the identification scheme is a long-run one and is based on Blanchard and Quah (AER 89). In Bachman&Sims instead, fiscal shocks are identified using an identification closely related to Blanchard-Perotti QJE 2002. Do you think, it's possible to adapt the code shutdown.rpf for a long-run Blanchard-Quah identification?
Changing the following line is needed (I think):
line impulse(model=varmodel,steps=nstep,results=irfs,factor=%decomp(%sigma),noprint)
and replacing it by something like
compute bqfactor=%bqfactor(%sigma,%varlagsums)
impulse(model=pvarmodel,steps=nstep,results=irfs,factor=bqfactor,noprint)

But after that, I'm not sure what I have to modify (and I'm not sure at all that the Bachman&Sims shutdown analysis can be applied to long-run identification???).
Thanks.
Romain