Re: [Gretl-users] different IRFs using VAR and VECM

On Thu, 13 Jun 2013, Artur Tarassow wrote: > Hi all, > > I am a bit confused about the following. I estimated following two > models which give exactly the same results. The first one uses the > VECM command and the second one the VAR command. For both I compute > the IRF(1,1) but they look different, which should not be the case, > or? Or do I miss something here ... I don't have time to look into this with very much attention, sorry, but it seems to me that the trouble is that your're "faking" a VAR in differences plus exogenous (mode no. 2) when in fact your "exogenous" variable is a linear combination of lagged Ys; as a consequence, the companion matrix is not caluculated correctly. ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------