// Update: Problem solved. I was misdeclaring my estimation equations. The estimation runs now!
I want to run an estimation using a model from Quint/Rabanal:
Monetary and Macroprudential Policy in an Estimated DSGE Model of the Euro Area.
First adapting the estimated model from Quint Rabanal to fit my needs, everything worked out fine and the model runs in stoch_simul with fixed parameters and single/all shocks set on on. However, trying to run the estimation does not work already in an early stage.
I already went through all forum advice I could find on the topic.
Using the initvals of Quint Rabanal did not work in the estimation model and also setting initvals such that my residuals are close to zero would not work (such as in the file attached).
The occurring error message looks as follows:
Error using print_info (line 83)
Impossible to find the steady state. Either the model doesn’t have a steady state, there are an infinity of steady
states, or the guess values are too far from the solution
Error in steady (line 104)
Error in estimationmodel (line 1438)
Error in dynare (line 268)
So, the model did not yet passed the steady and check commands. This seems to be quite a common issue for unexperienced users, however, I could not figure out a solution using the information I found so far.
Attached you’ll find the estimation model and the data: estimationmodel.zip (35.2 KB)
My Dynare Version: version 4.6-unstable-46ce6e02af4d186eaef2ac351248b7b7fe70db46
I am sure I am missing something crucial out here, but could not figure it out by my own up to now.
Many thanks and best regards,
I am glad it works. A student of mine some time ago tried to reproduce the results of the paper. He came close, but could not exactly match the IRFs of the model. Were you successful?
Dear Prof. Pfeifer,
not yet. The estimation runs entirely. However, the mode check plots do look odd. So I still have to find suitable priors and distributions, since I also estimate a slightly different set of parameters.
For sure there will be follow up questions and I am so thankful about this forum led by you. This is a really valuable source of information.
You should check your model in a version calibrated to the estimated values of Quint/Rabanal before(!) moving to estimation yourself. That is an important consistency check. If you cannot get the right result in a calibrated model, estimation will not work as well.
Dear Professor Pfeifer,
Thank you for the advice and excuse the misunderstanding. The calibrated model works perfectly! Right now, I am trying to find suitable Priors.
When you say
does this mean you can exactly replicate the IRFs in their paper?
Please excuse the late answer, Professor Pfeifer, I did not notice the reply and just stumbled over it again.
I was estimating the model using the estimated calibrated model of the MMB. In their calibrated model IRFs are quite similar to those in the paper.
Using the calibrated model, I made adaptions to estimate it with suitable data.
So I take from this that you have not verified the original paper as you are relying on the codes provided by the authors for the MMB?
This is correct. I only made some minor adaptions. Since I am relatively new to the field verifying and coding the model by myself will be an aim for the future to pursue!