This project is read-only.

SVAR

Topics: VAR, SVAR, FAVAR
Feb 29, 2016 at 4:59 AM
Edited Feb 29, 2016 at 4:59 AM
Hi Jeromir,

I am trying to do a SVAR model with short run restrictions using IRIS tool box. My model has foreign interest rate (i*), gdp(y), Inf (p), domestic interest rate (i) and real exchange rate (q) in the given order.

Deviating from the normal cholesky decomposition, I wish to have simultaneous contemporaneous effects between domestic interest rate (i) and real exchange rate (q) so there will be one non-zero element in the upper triangle. In order to get the minimum number of restrictions, I will assume domestic interest rate only respond to y, p and q and not for i* contemporaneously. Could you explain how I could achieve this identification.

Should I be using 'householder' as the method and test a string that specify this particular identification?
e.g. [s,sd,B,COUNT] = SVAR(v,vd,'method=', 'householder','test =', test_string);
And how should I specify the string to impose zero restrictions? I tried something like
test_string = ['S(1,2,1)== 0',......] but the program just crash.

I would really appreciate if you could give some advice on this.

Thanking you in advance.
Mar 8, 2016 at 5:43 PM
I would also be interested in this question.

How can I impose my arbitrary (not sign and not Cholesky zero) restrictions on the contemporaneous matrix?

Thanks!
Mar 8, 2016 at 5:47 PM
Relating to the above: is it possible to impose less restrictions in the contemporaneous matrix than what is needed for full identification? When I am only interested in identifying one particular shock and do not care about the rest, then maybe it is better to avoid making more restrictions by a full Cholesky (lower triangular matrix). Can this be done? This relates to the above question whether we can set the restrictions by hand.

Thanks!