Hello，everyone. I meet a new error. the steady-state should be right.The model_diagnostics; told me nothing. I do not know how to deal with it
bayes.mod (13.9 KB)
Are you sure your exp()-substitutions are harmless?
exp((1-exp(epsilonw))*W)=(1-thetaw)*exp((1-exp(epsilonw))*Wstar)+exp((exp(epsilonw)-1)*pi)*thetaw*exp((1-exp(epsilonw))*W(-1));
has in steady state exp(-10*Wstar)=exp(-86) which is numerically zero. That seems strange.
No, now you got proper issue:
There are 12 eigenvalue(s) larger than 1 in modulus
for 13 forward-looking variable(s)
The rank condition ISN'T verified!
This is typically a timing problem.
This may be a numerical issue reflecting the very different scalings of variables. Your W has a steady state of 257.808. Applying exponential functions to it may cause issues.
Thank you professor!
Dear professor, I changed my steady state and I meet this problem again. How can I solve it~
nk.mod (10.8 KB)
One thing that you should do is replace the use of the steady_state-operator for things that cannot be computed endogenously. See
That should eliminate the most message displayed in model_diagnostics. After that, check whether additional flagged issues remain.
Dear professor. I don’t think it’s a matter of steady state operator usage nor about time. I feel like this is a steady-state problem, as the model cannot be solved in this steady-state, and I wonder why ~
What do you mean? Are there multiple steady states? More likely is a timing issue.
In fact, for the same model, the steady state solved by calibrating different parameters, can not always solve the model.
But then it’s not about the steady state but the model parameterization.
Well, for the parameter eta I calibrated, if I keep more than four decimal places, the model is not solvable, if only four digits are kept. All other conditions are equal
See my reply at