Analysis to tease out the effect of one variable?

[yanivd]

Hello all!

I have a model in which, let's say, I have 3 endog. variables: X, Y and Z.
I estimate the VAR model etc. and calculate the betas, Cov. matrix and IRFs and all.
Now, the first question that is interesting to me is what is the effect of X on Y, for example... standard VAR practice.
But, the second perhaps more exciting question is: what is the effect of actually HAVING a Z on the X and Y relationship.

So, my question to you:
Given the IRFs (or FEVD), is there a way to try to estimate the counter-factual effect of X on Y (without the existence of Z)? Is it "legal" to maybe zero coefficients and the cov. matrix to emulate a situation in which Z's effect is zero? (while using the IRF estimations for the full model...)


Thanks!