IRFs value depend on number of periods

Stochastic simulations
1

Hello, I am trying to generate IRFs of a model using 4th order perturbation with the code:
stoch_simul(order=4, irf=51, periods=0, irf_plot_threshold=0, nograph)

I notice that if I change the number of periods of the irf (irf option), or the number of periods of the simulation (periods option), irfs for some endogenous variables can change quite significantly. It is the first time this has happened to me. How can this be the case? If it can help, those endogenous variables seem to be also highly sensitive to higher order terms, as if I change the order of perturbation from 3 to 4, IRFs also change significantly.

Thanks a lot in advance

2

IRFs at higher order are generalized IRFs based on simulations. If you don’t use IRFs at the stochastic steady state instead, you need a replic that is very high to get consistent IRFs.

3

Thank you! So when computing IRFs around the stochastic steady state, we always get consistent IRFs? Is this DSGE_mod/Basu_Bundick_2017/Basu_Bundick_2017.mod at master · JohannesPfeifer/DSGE_mod · GitHub the most recent reference on how to compute IRFs around a stochastic steady state?

4

Yes, that is correct.

1 Like