impulse response functions with dummies

[justin_m]

Dear all

I want to run a VAR-X in which each endogenous variable is regressed on the lags of endogenous variables of the system and on those same regressors multiplied by a dummy that is one for t larger than a certain period and 0 otherwise.

My question is whether:

1) is it more appropriate to run separate VARs for the two sample periods or include the dummy?

2) in case I include the dummy, I would still like to compute impulse response functions with the dummy switched on and off. Is this possible? For instance, plot the difference of the two impulse response functions (with dummy switched on and off)? If so, how?

Kind Regards,

Justin

[dakila]

If your sample is large enough then it is better to run separate VAR.
For question 2 try the TVSVAR add-in.