When trying to replicate (with some additions) “Gauging the effects of fiscal stimulus packages in the Euro Area” by Coenen et al (2012), and after obtaining the steady state, I receive the following error in Dynare 4.5.1:
"There are 19 eigenvalue(s) larger than 1 in modulus
for 18 forward-looking variable(s)
The rank condition ISN’T verified!"
I know it might have something to do with the timing of the predetermined variables (capital services, public capital, government debt (bonds)), but for the life of me can’t figure it out. Might there be something else amiss?
I attach the .mod file below.
SAFiscal.mod (31.0 KB)
Did your replication suceed before changing anything? If yes, what did you change/add?
Thank you for the reply.
I have some reference code for the article in question, but I went back to the original model it is based on (“The New Area Wide Model of the Euro Area”, Christoffel et al (2008)) as the reference code included some features that I could not identify/replicate in the analytical solution.
I included (as in the Coenen paper) heterogeneous consumers (non-Ricardian, Ricardian) and government capital, but also added back in the capacity utilisation and private capital services feature in the original model. I also added an equation for the target rate of consumer inflation (a South African feature is that the implicit target was not constant over the last decade and a half). Finally, I added an equation for domestic marginal cost (the original code did not have a separate equation but incorporated it in the Phillips curve relationships).
I adjusted the code somewhat to reflect the analytical solutions in the Christoffel (2008) paper.
Any help would be appreciated.
Any thoughts on why the error? For reference, I include the reference/replication codes for Coenen et al (2012) - note that the code only works on earlier versions of Dynare (4.3).
I also attach my steady state files should you want to run the model.
Coenen et al (2012).zip (15.0 KB)
SAFiscal.zip (12.9 KB)
Any thoughts? Could it be a parameterization problem?
It almost always is a timing problem, not a parameter issue. The problem is that the model is really big, making it extremely hard to debug. Usually, I would recommend to use the \LaTeX-capabilities of Dynare to get a more readable version of the mod-file for checking.