Zero and sign restrictions in SVAR

Apr 3, 2015 at 12:56 AM
I am following the paper Transmission of the ECB's Balance Sheet Policies (Boeckx, Dossche, Peersman, 2014). I would like to impose zero and sign restriction on the contemporaneous matrix of shocks. The command 'S(1,3,1) == 0 ', isn´t working. I would like to ask, if it is possible to do this with IRIS and how.
Thank you very much.

Apr 3, 2015 at 12:05 PM
Hi Matej

The sign restriction algorithm in IRIS is based on random factorization of covariance matrix using Householder transformation. This means that you cannot really specify exact equality restrictions, such as zero restrictions, because the probability of a randomly drawn number being exactly numerically equal to zero (or to some other fixed number) is zero. You would have to be very lucky to draw such a number at least once in a huge sample... You can think of it this way: try to test how many draw it takes until rand()==0.

A workaround is though relatively easy: you simple to specify a narrow range around your fixed number (zero in your case), e.g. testing abs(S(1,3,1))<=0.1.

Marked as answer by jaromirbenes on 4/3/2015 at 5:06 AM