Bayesian estimation with unit roots

Hello, Prof. Pfeifer. I have a supplementary question concerning the application of Bayesian estimation in Dynare for models that include unit root variables. I’m seeking to align the model with stationary empirical data, yet my model also includes variables like nominal exchange rates and price levels, which can have unit roots. I understand that one approach is to manually re-express the model to exclude these variables, but this can be labor-intensive. Therefore, I’m interested in knowing if it’s feasible to estimate a model in Dynare using Bayesian methods when some of the variables have unit roots. Could you provide some guidance on this?


You can see this topic in Dynare forum. I think it can be helpful.

observable variabels in a DSGE model.

Yes, that is feasible. Typically you just need to set the diffuse_filter-option.

Thank you, Eisa. Let me check it.

That’s really valuable information. I’ve already derived the complete equilibrium system using stationary variables. Still, it’s always beneficial to minimize unnecessary work. Thank you!