Hi,

I want to estimate my dsge model using IRFMating and I have a problem in calibrating the variance of exogenous shocks in dynare.

My problem is about calibration of the variance of shocks in a dsge model in dynare in such a way that it replicates empirical impulse response functions(e.g. from a var model).

Imagine that our exogenous shock in our empirical model is:

a_t=0.5+0.3a_{t-1}+\epsilon _t where var(\epsilon_t)=0.25

(We have a SVAR model which the above equation is one of them)

Now, we want to have the same shock to a_t in the model in dynare, but due to other normalization assumptions, the steady-state is a different number than in the empirical model (var model). That is, 0.5 in the above equation is say 0.1.

what should the shock equation and variance in the model in dynare look like?

Shall it be like:

a_t=0.1+0.3a_{t-1}+\epsilon _t where var(\epsilon_t)=0.25

or we should manipulate the multiplier 0.3 and the variance of 0.25.

You need to make the IRFs comparable. If there is a different level, try expressing everything in percentage deviations from the mean.

Thanks. How should I express them in percentage?

would you please explain it about the example equations I mentioned?

What I am saying is: usually the exogenous processes are in logs.

Thank you for your response.

You mean that if my shock in the VAR model is:

log(a_t)=0.5+0.3log(a_{t-1})+\epsilon_t where var(\epsilon_t)=0.25

the shock in the dynare model should be:

log(a_t)=0.1+0.3log(a_{t-1})+\epsilon_t where var(\epsilon_t)=0.25

That is, just the contant term is change to the SS of the model. and we should not proportionally change other numbers in the equation?

Yes, because now the shock has the same size, i.e. 25% of the steady state.

I got it now. Thank you so much.