The RATS Software Forum

Hi Tom,

In estimating a VAR, do you know how to impose some restrictions on the coefficients of each equation in the VAR?
For example, for a small open economy, I would like to have the foreign variables influencing output, interest rates but not consumption etc...and in this case, the foreign variables are included as exogenous variables.

Many thanks!

MC

Hmm, thought I posted a reply to this, but it seems not to have gone through. Trying again....

I think the most common approach is to simply define the model as a near-VAR, omitting certain exogenous variables from certain equations.

Alternatively, you could estimate the full system using SUR and then use RESTRICT to test the restrictions (and RESTRICT(REPLACE) with SUR(CREATE) to generate the restricted results).

Hi,

Thanks for your reply. I will try to use a near-var structure to estimate the model. another question is that i want to do a structural identification on the shocks (i.e. SVAR). Will there be any issues involved, particularly with regard to the error band? Can I still use the @MCVarDodraws and @MCgraphirf to generate the impulse and the error bands?

Or should I use follow the example of montesur.prg? If this is the case, then how to alter the structure of the code to take into account of alternative factorization rather than the cholesky?

Thanks.

MC

Hi Tom and everyone,

I have tried to alter the montesur.prg to take into account of alternative factorization, and incorporated the resulting changes into the format of MCVARDoDraws.src and MCgraphirf.src. I am not sure whether I did it correctly.....Do you mind giving me some advice?

In particular, the error band given by the option center=input, impulses = impulses, and the option stddev is quite narrow. Is it a problem with the coding? (For your information, the file MCSURgraphirf.src made the following changes to the file MCGraphirf.src:
***************************************************************************************************************************
compute frac=%wfractiles(work,%%weights,request)
*
* Choose the central value
*
if center==1
compute resp(j)(k)=%avg(work.*%%weights)/%avg(%%weights)
else
if center==2
compute resp(j)(k)=frac(npercent+1)
else
compute resp(j)(k)=impulses(yshow(i),j)(k)

if %defined(stddev) {
compute sigma=sqrt((%normsqr(work.*%%weights)/(%sum(%%weights)^2))-(resp(j)(k)^2/draws))
******************************************************************************************************************************


Thank you so much.

MC