## Impulse Response functions

Questions and discussions on Vector Autoregressions

### Impulse Response functions

Dear Rats Users,

If I run a VAR model where all variables are expresed in log (not differenced), then all my impulse-response function are expresed in logs too? Te results are not expresed in percentage variation?

SG
Welcome from Venezuela.
sguerra

Posts: 6
Joined: Mon Dec 04, 2006 11:12 am
Location: Venezuela

If the variables are in log form (whether differenced or not), the responses are also in logs. If, for instance, you are looking at the response of log y, then a response of .05 means (roughly) a 5% upward change in y. If the variable is differenced logs, then you what you are seeing are the responses of the growth rates. If you want to convert that into the change in the underlying variable, you need to accumulate the responses, so if you get responses of .01, .01, .005; the third step response of log y itself is .01+.01+.005=.025.

You can do this last calculation after the IMPULSE instruction by using ACCUMULATE or you can put in an identity which does this for you, such as

equation(coeffs=||1.0,1.0||,identity) logyid logy
# logy{1} dlogy
TomDoan

Posts: 2720
Joined: Wed Nov 01, 2006 5:36 pm