Identification problem in simulated data from DSGE

Dear all,

Noobie here with Dynare. I’m trying to estimate a very simple DSGE model using simulated data from the same model (just to make sure that the model works properly). Therefore, I expect the posterior of the parameters to converge to their true (calibrated) values.

The estimation and identification works for all but one parameter (“bet”, the discount factor). When I include this one, the estimation crashes for all the optimizers. I have tried with different prior distributions for this parameter as well but without results.

Any suggestions on what I’m missing?

model_inv.mod (4.0 KB)

Attached is .mod file.

Thanks a lot in advance.


Why are you using such a strange prior for beta? It allows for beta>1 which is not allowed by the model due to stability problems. Also, starting at 0.8 if the true value is 0.992 is crazy. The simulated data will strongly reject a 20 percent steady state interest rate, but your prior thinks that an interest rate bigger than 10 percent is the most likely value.

Dear Johannes,

Thank you for you answer. I see that now. However, I have tried with several specifications for that prior, for instance bet,0.99,0.5,0.99999,normal_pdf,0.99,0.01;

When estimated alone, and even with some combination of parameters, there is no problem in the estimation. But when I try to estimate all of them, it crashes. Even when I estimate only alpha (capital share) and beta it crashes as well.

What do you suggest to check?

Again, thanks for your time and patience.

model_inv.mod (4.1 KB)


Dear Johannes,

I couldn’t solve this with different priors, solvers, and definitions of the model. Simulated data provides all the observables, and still beta can’t be identified. Any suggestions?

Thank you very much in advance

I would say: calibrate beta.