Shock size given an ARMA estimate

Hello everyone,

I have a little question: when we estimate a process through an ARMA model and this process is part of a DSGE model, how big should the shock be given the estimate?

Just for example, let’s suppose that I estimate the TFP through an ARMA model, and I obtain that the model is an ARMA(1,0) with persistence coefficient ar1 = 0.9 and sigma^2 = 0.205. So, the size of the shock should be:

var varepsilonv; stderr 100*sqrt(0.205);

or I’m wrong?

Thanks in advance

Yes, but why do you multiply by 100?

1 Like

In order to each unit on the y-axis represents a percentage point

If it’s a model solved at first order that is correct.