IRF comparison - parameters values to be used

Estimation ML/Bayesian estimation
1

Hello everyone,

I want to compare the effects of R&D(or another specification) within a DSGE model estimated using the Bayesian method by analyzing the Impulse Response Functions (IRFs). I have then two models: one with R&D and one without. Both models have estimates for structural parameters and shock parameters. To explore the response of variables after a TFP shock in both models,I wanna plot IRFs of each variable for both model in the same figure. Then, my question :

  1. Should I simply compare these bayesian IRFs by the using it in estimation command? In this case, I would use the posterior means of the estimated parameters, which are different between the two models, especially for the deep parameters and the TFP shock parameter.
  2. Or should I apply the same size of the TFP shock to both models (for example, a 0.1 deviation from the steady state), while keeping the other estimates different, and then run stoch_simul in a separate MATLAB file for each model?
  3. Alternatively, should I take both models and use the same set of parameters (for instance, using the estimated parameters from the R&D model in both models, with the only difference being the R&D equation in the second model), and then run stoch_simul for each model?

Thanks in advance.

2

In that type of situation, you may want to use bayesian_irf, potentially with relative_irf. The relative_irf option would make sure that size of the shock that you consider is comparable across models. The size of the estimated shock standard deviation would still allow comparing the average size of shocks, while the relative_irf is informative about the transmission of equally sized shocks.

Note that bayesian_irf will not evaluate the IRFs at a parameter vector like the posterior mean but rather will compute the distribution of IRFs across the posterior. As such, it is a summary measure capturing the full distribution of parameters that will differ across models.

3

Thanks Pr., It’s always a pleasure to hear from you.