0
$\begingroup$

I came across this text that seems to estimate a non-recursive SVAR in Eviews, but the restrictions are kinda different from what I expected.

In their recursive SVAR, they have the following results in Eviews which are the restrictions I expected given the model.

enter image description here

The results above has the following restrictions for matrix A and B. So we know the coefficients of the contemporaneous variables.

enter image description here

In the non-recursive SVAR, they have the following results

enter image description here

But I expected the following retrictions for matrices A and B instead.

enter image description here

As we can see, their results show the restrictions below. How can we then know the coefficients of the contemporaneous variables? I know e = (A^-1)Bu will match the model in the non-recursive example above, but I don't understand why A is an identity matrix and B is not diagonal. And why my expectation in the red matrixes above is wrong. I actually want B to be diagonal or identity matrix in the non-recursive SVAR, not matrix A. Thanks for any help.

enter image description here

$\endgroup$
5
  • $\begingroup$ Does your confusion have something to do with the fact that the ordering of the coefficients in EViews is columnwise in the A matrix and indices with zero are skipped? for example, the columns of a lower-triangular matrix would be (1) c1, c2, c3; (2) 0, c4, c5; (3) c6. $\endgroup$
    – BrsG
    Jul 10, 2021 at 8:21
  • $\begingroup$ @BrsG My confusion is that if A matrix is diagonal, then what are the contemporaneous coefficients? When I read books on SVAR, matrix A (containing the contemporaneous restrictions) is never diagonal, yeah? Like in Ae = Bu, where u orthogonal structural shock, how can A be diagonal? I guess, then there are no contemporaneous effects in the model? Thanks! $\endgroup$ Jul 11, 2021 at 10:59
  • $\begingroup$ Indeed, when A is diagonal, this corresponds to the special case with no contemporaneous interaction between the variables. In $Ae=Bu$ case, you can impose restrictions on $A$, $B$, or $S=A^{-1}B$, each meaning different things. $\endgroup$
    – BrsG
    Jul 11, 2021 at 15:29
  • $\begingroup$ @BrsG. Thanks for the reply. Lemme ask though for a hint. When you impose restrictions directly on S, what is your goal? Only correct identification of the shocks? $\endgroup$ Jul 18, 2021 at 6:09
  • $\begingroup$ $S$ gives you the long-term restrictions. For example, by setting a zero in the appropriate position you can impose that shocks have no impact on a certain variable at an infinite horizon. $\endgroup$
    – BrsG
    Jul 19, 2021 at 8:32

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.