Deviation size of response variables (percentage) to a shock

Hi there,
I have a question regarding the Dynare output. If I run a log-linearized model (linearized by hand - with the model instruction model(linear)), how do I interpret the sizes of the response variables to the shock? It should be log-deviations from steady-state, e.g. approximatively percentage growth rates. If I shock the model with for example: var e; stderr 0.01; and I get a series for a variable like 0, 0.004, 0.001… how big are these rates in percentage deviations? Do I have to multiply them by 100 to get percentage points?? In this case is it 0.004 percent or 0.4 percent?
Thank you for your help.
Greetz Björn

yes, everything is relative to the st. deviation of the shock. so in your example, 1% shock leads to 0.4% change in the variable.

Just to add:

Do I interpret the shock size which is 0.01 to be 1% of the s.d. of the shock?


By default, the IRFs are level deviations from steady state.

The behavior described by otb is only present if you put the “relative_irf” option.