Structural vector autoregressive models interpretation

[Laura Myles]

Hi Listers,

I am running a Structural vector autoregressive model but I am unsure on how to best interpret it as I am new to SVAR analysis.

As I cannot share my data so I am using some sample data instead.

I am unsure why the value estimated within matrix A is not the same as the value from the immediate impulse response function - I was under the impression they both captured the immediate effect of the impulse on the response.

I am not sure if/when orthogonal irf should be preferred.

As 2 lags appear to offer a good fit to the data based on -varsoc-, the output only shows me irf values for step 2, 4, 6 and 8 - does this seem appropriate?



use http://www.stata-press.com/data/r16/friedman2.dta

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

tsset time, quarterly

* I difference the data so that it is stationary
g m1Diff = d1.m1
g CompDiff = d1.consump

varsoc m1Diff CompDiff, maxlag(8) // I choose 2

qui: svar m1Diff CompDiff, lags(2) aeq(A)
matlist e(A)

irf create order1, set(var2.irf) replace step(8)
irf table irf, irf(order1) impulse(CompDiff) response(m1Diff)