This is great sir! However, I will also suggest that in the SVAR package that the SVECM model should be included directly just as in the JMulTi. Also, the VAR and the VECM impulse response should be allowed to be set to one standard error and any other form that the cholesky ordering. 

Thanks. 

On Fri, Jan 24, 2020, 8:12 PM Sven Schreiber <svetosch@gmx.net> wrote:
Am 24.01.2020 um 19:31 schrieb anzervas@yahoo.com:
Dear all (especially Sven and Riccardo),

A recent strand in SVAR literature uses heteroskedasticity to identify the structural shocks - those interested in the topic may read mainly Rigobon (Review of Economics and Statistics 2003), Lanne and Lutkepohl (Journal of Money Credit and Banking 2008) and / or  Bacchiocchi and Fanelli (Oxford Bulletin of Economics and Statistics 2015). If one has 2 variance - covariance matrices, there is no need to add zero or other restrictions to identify the structural matrices A or B.

Prompted by your request I have just discovered (two minutes ago) the R package "svars" which is described as doing "Data-Driven Identification of SVAR Models" and looks very interesting. They explicitly mention Rigobon 2003. Given that one of the authors is Helmut Herwartz my prior expectation is that the package should be well suited for your problem.

Any suggestions / corrections are welcome. This also seems be a good functionality to add to the SVAR package (in addition to add AB model functionality to VECMs, for completeness purposes).

I'm actually quite happy that somebody else already implemented it for open-source software ;-)  Isn't division of labor great?

At the Berlin gretl conference Helmut Lütkepohl (no, there are not that many Helmuts in Germany, just a mild coincidence) of course already suggested in his keynote to add such things to gretl. And of course it would be nice to have. But first of all I'm a bit skeptical at least in the macro context: We are having so many problems to get the first moments of the data right, and now we want to base everything on changes of the second moments? Or as somebody said: With T=100 it just takes one or two observations to move from a Gaussian to a fat-tailed and/or heteroskedastic distribution. So the whole thing sounds very tempting (less assumptions needed), but I would always worry about robustness of those animals in practice.

Secondly there are always tradeoffs in terms of time and effort. Right now Jack and I are working on something else in SVAR. (Not sure if I'm already allowed to say what.)

About AB & SVECM: I'm open to hear demands and arguments, but I don't think we'll do it just "for completeness". The fundamental question is always: What is the use case, where is the need?

cheers

sven

_______________________________________________
Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/