The RATS Software Forum

Dear Estima:

Is there any code for a finite long-restriction for a VAR? That is, the BQ restrictions are infinite long-run restrictions, but a number of papers have imposed finite long-run restrictions. For example this paper ( http://www.federalreserve.gov/pubs/ifdp ... dp832r.htm imposes finite long-run restrictions in the context of a technology shock. Papers often use these finite long-run restrictions as a robustness check on the infinite restrictions. Is there an easy way to modify the existing BQ code for some arbitrarily imposed fixed-horizon?

Thanks!

I have attached a pdf file that has another article that uses finite restrictions. See page 398. This one uses the finite LR restriction to check the robustness of using infinite LR restrictions. Thanks.

The first paper is basically doing the same thing as Faust, "The Robustness of Identified VAR Conclusions About Money", Carnegie-Rochester Conference Series on Public Policy, 49, 207-244. The program for that is http://www.estima.com/forum/viewtopic.php?f=8&t=1486

Thanks Tom,

One more question: how do I specifically adapt it to impose a finite long-run restriction. I know I first modify the code so that the BQ decomposition is in it, but after that it is unclear. It looks like the needed adjustments need to be done here:

compute [rect] step0 = %xt(irf,1)
compute [rect] step4 = %xt(irf,4)
*
compute [vect] rcond = %xrow(step4,6)
compute [vect] pcond = %xrow(step4,2)

However, I am not sure how to implement a finite long-run restriction here. Thanks in advance for any suggestions.

Just to be clear, here is specifically what I am trying to do. It is common to assume that in the long run productivity is only influenced by technology shocks. Thus, all other shocks in the VAR have zero cumulative effect on productivity; only a shock to itself--the technology shock--can permanently affect its long-run value. These restrictions, however, are imposed using infinite long-run restrictions. A finite long-run restriction would mean saying that nothing permanently affects productivity at a long horizon, say 10 years. This restrictions, thus, allows one to move beyond a business cycle horizon, but while avoiding the imposition of an infinite horizon restrictions. Thanks.

This will do the type of analysis from the Lastrapes paper. thisH is the horizon that you want to use for the restriction, while maxH is the longest horizon that you will be using. (You can do a calculation with a different value of thisH without recomputing the impulse responses). This does the BQ factor using an (inverted) finite sum of the MAR rather than the infinite sum that you get from the lag sums of the VAR.