A question about IRF

Dear all,

I am not sure if IRF’s show the level deviation of variables from their steady state or percentage deviation from their steady state.

Thank you.





But doesn’t this depend on how you write the model? If you write the model in terms of logs (say production fct: y=alpha(a+n) + (1-alpha)*k, )
don’t you get output in terms of log-deviations, which is approximately equal to % deviations?

Of course it is in the level of the variables used in the model. If it happens that the variables in the model are the log of the original economic variables, your interpretation is correct.



If the variables are in logs then then the deviations are in percent form. That is fine. When you say percent form do you mean fraction form or multiplied by 100.

I would assume fraction form. However I am working with the Smets and Wouters code and the deviations seem to be in percent form multiplied by a factor of 100. This is a bit perplexing. Could you clarify.