Question on Basu Bundick (2017)

Stochastic simulations
1

Hi Johannes,

I am trying to learn the Basu Bundick model. I have checked their Dynare code and their preference volatility shock generates non-zero impulse responses in the level of preference (a_{t}). From my understanding, this should not be possible as their specification of the shock process does not have a Jensen’s inequality effect. Is it possible that this is due to the impulse response being calculated using 3rd order approximation simulation?

Best,
Tian

2

That sounds strange. Can you provide files to replicate the issue?

3

Hi Johannes,

This is the example I used to learn the model. The stochastic simulation shows that the volatility shock generates non-zero responses of variable a.

Best regards,
Tian

Basu_Bundick_2017.mod (13.2 KB)
Basu_Bundick_2017_editM_steadystate.m (4.1 KB)

4

You are using simulation-based ergodic GIRFs. You will need a really large replic in the order of several ten thousands for a to stay flat. That’s why most papers use GIRFs at the stochastic steady state.