I am working with quite complicated model. It has quite a strange feature: after a stochastic shock (even it is with AR(1)=0), it takes around 200 periods to converge to initial steady state. E.g. IRF < 0 at the beginning, then after 20 periods it becomes > 0 and after some 180 periods IRF is close to 0. I tried different calibration schemes – still the same.
Is it normal?