I am trying to understand the scale of the impulse responses. Is it the response with respect to one standard deviation shock to the relevant forcing process?
Typically, I find irfs are of very small magnitude multiplied by 10^-3? Can I rescale these irfs by multiplying by 100 uniformly and interpret these irfss as percent deviation from the steady state?.
Yes, by default, it is the response to a one standard deviation innovation. Of course, you can rescale, if you inform people about this. But note that this rescaling only has the interpretation of the IRF to a different shock size if the model is linear or linearized (order=1). Otherwise, the size of the shock matters.
Thank you. Just now I rephrased my question. I have a nonlinear exact model. If I multiply irfs by 100, it means percent deviation from the steady state. Right?
Yes, that is correct.