Model runs just fine but MODEL_DIAGNOSTIC message about Singularity

Perfect foresight simulations
1

Dear all,

My model runs just fine, Blanchard Conditions are verified, and DYnare finds a perfect foresight solution. However, when I run model_diagnostic with the newest version of Dynare 5.5 (I did not have that message from model diagnostic with previous versions of Dynare, the 5.0 one), I have this message:

“MODEL_DIAGNOSTICS: The presence of a singularity problem typically indicates that there is one
MODEL_DIAGNOSTICS: redundant equation entered in the model block, while another non-redundant equation
MODEL_DIAGNOSTICS: is missing. The problem often derives from Walras Law.”

Should I be worried? Or can I process with the output?

Thanks for the help.

2

Yes, you should be worried! Because the presence of redundant equations implies that some important equation is missing. Your model doesn’t represent the economic relationships that you had in mind.

3

Could you maybe share the file?

4

Thank you, currently trying to resolve the problem and add what was missing

5

I would be interested to find out why two different version give different error messages.

6

I just reran it with Dynare 5.0 and Dynare 5.5, and it gives the same message actually. One month ago it was not. At some point I used the unstable version 6.0, it might be the one. But now it gives me the same message.

Thank you

7

Adding to @MichelJuillard’s answer: your mod-file has steady state values ranging from 1e-2 to 1e7 and has about 400 equations. Any numerical singularity check will run into issues here. I would not exclude the possibility that you are not actually missing an equation but rather that the steady state values are problematic.

8

Thank you. I tried to modify the problem by taking out a redudant equation and adding another, but I end up towards the same issue.

However, when I check my SS values, they are consistent with what I find in the data for aggregate variables. My hint is that I have an issue with my 3 households and the consumption/investment of each, because again, aggregate variables of consumption, production etc are ok with what I find from national accounts and other databases. I’m gonna try to run the same model but with just one household, to see if the problem comes from here.

Plus, the model I have is not really standard macro as I’m adding many stuff from industrial ecology too, such as resource balances thanks to additive CES functions, tracing waste and material flows.

Thanks again @MichelJuillard @jpfeifer

9

Again, it may simply be numerical imprecision. You should first work on the normalization issue in Zeroporfit condition and calibration

1 Like
10

Hi,

Just to give a bit of an update. I have no collinearity issued anymore by following what was said in this thread:

  1. The main issue was indeed the fact that I had large differences in the scale of variables. I rescaled variables carefully to avoid large differences
  2. Some parameters were not dimensionless constant. I fixed that by following carefully Cantore and Levine (2011) reparametrization technique.

Many thanks