Structural Var Ordering of Variables

[Frederic Mueller]

Hello,

I am conducting a Var / Svar analysis using Impulse Response Functions (IRF). Hereby, I have the following 6 variables:

- output
- labour
- capital
- R&D investment 1
- R&D investment 2
- Price

Using the lower Choleski Decomposition defined by the following two matrices:

matrix A = (1,0,0,0,0,0\.,1,0,0,0,0\.,.,1,0,0,0\.,.,.,1,0,0\. ,.,.,.,1,0\.,.,.,.,.,1)
matrix B = (.,0,0,0,0,0\0,.,0,0,0,0\0,0,.,0,0,0\0,0,0,.,0,0\0 ,0,0,0,.,0\0,0,0,0,0,.)

I computed the following Svar model:

*************************(1) *************************
svar x y z a b c, small lags(1/3) aeq(A) beq(B)

irf create order1, step(10) set(myirf1)

I am now looking at the specific IRF with impulse(a) and response(x)

Now, according to my understanding the picture of the IRF should change when I swap the ordering of the variables to for example:

*************************(2) *************************
svar a b c y z x, small lags(1/3) aeq(A) beq(B)
irf create order2, step(10) set(myirf1)

I am now looking again at the specific IRF with impulse(a) and response(x) however my result from the IRF in the first example is identical.
After seeing this I have tried several combinations of ordering but the IRF's are always the same. Surprisingly when I do the same analysis with OIRF's the ordering seems to matter and I receive different plots. I do not really understand the result as my textbook says that ordering is deterministically important.

Thanks in advance.

[Frederic Mueller]

I was able to solve my problem myself. So if one uses the Svar command the applicable IRF's can only be received through the sirf command while if one has an underlying Var model the irf command is valid.