Near VAR with Long-Run Restrictions

[macro_man]

Estima:

Is there a straightforward way to modify the montesur.prg so that the shocks are identified using the BQ long-run restrictions?

Thanks.

[TomDoan]

Unfortunately, no. Long-run restrictions are conveniently handled only in the situation where you have a full VAR with a just-identified structural model. With that combination, the long-run restrictions don't put any restrictions on the lag coefficients. Any departure from that (either an overidentified structural model or a restricted VAR) puts further, very ugly, restrictions on the lag coefficients.

[JBG]

Dear Tom,

Is it possible to estimate a Bayesian VAR with block exogeneity (ie a NEAR-VAR with priors) with RATS?

Thank you in advance.
JBG

[TomDoan]

The easiest way to do that is with an asymmetrical prior which makes the standard errors on the excluded variables almost equal to zero. The final example on page UG-255 (RATS v8 manual) is an example, though you can safely replace the .01's with even smaller numbers like 1.e-6.

[JBG]

Thank you very much for your answer.

Actually, that is what I have done, using this matrix:

"declare rect priormat(neqn,neqn)
input priormat
1.0 0.00001 0.00001 0.00001 0.00001 0.00001
0.5 1.0 0.5 0.00001 0.00001 0.00001
0.5 0.5 1.0 0.00001 0.00001 0.00001
0.5 0.5 0.5 1.0 0.5 0.5
0.5 0.5 0.5 0.5 1.0 0.5
0.5 0.5 0.5 0.5 0.5 1.0"

PS If there is a more complicated way, which allows to improve the precision (by reducing the number of parameters) of the estimated part of the mixed estimation technique, please let me know.

[TomDoan]

A more complicated method would give you almost identical results. With the ultra-tight prior, those added parameters really aren't affecting the precision of the others.