I've read several paper that all compare different cumulative IRF of the same VAR equation for statistically significant difference. The IRF they use are simply the sum of the coefficients of the VMA representation resulting from Cholesky decomposition. I wonder how one can construct tests for IRF. The authors mention that they use Newey-West standard errors. Could it be that they don't just take the sum of the coefficients but run two different simulations, one with and one without a shock? I still don't know how to apply NW to it and how to derive significance from that though. Here you can find one of the paper (p.16).

Thanks for your hints!


When I worked with IRFs I bootstrapped confidence intervals around them to use as a significance measure. I used the methodology described here in:

H. Lutkepohl & Jurgen Wolters A. Benkwitz. Comparison of bootstrap con dence intervals for impulse responses of german monetary systems. Macroeconomic Dynamics, 5:81-100, 2001.

( I made a pretty thorough description of the derivation of impulse response functions and the bootstrap confidence intervals around them, it's not published or anything but I think my master thesis might be useful to you, and relevant to your question. https://dl.dropboxusercontent.com/u/16744194/Thesis_finished.pdf )

  • $\begingroup$ Let me know if you find it useful, I would like to know! $\endgroup$ – fredrikhs Jan 8 '14 at 22:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.