I have a the nonlinear NK model and the linearized one.
The question is: are the second moments (st.dev) obtained from the stochastic simulation of both models the same?
I get some small difference (+/- 5 %) in the linearized NK model. Are they supposed to be 100 % the same ?
That depends on what you compare. If you compare theoretical moments at order 1 from the logarithm of the levels in the nonlinear model to the percentage deviations in the linearized one, they should be exactly the same.