I am currently working on comparing the implementation of a linear and a non-linear (non-exp type) version of the same model using Dynare. I have encountered an issue: while the linear version runs smoothly, the non-linear version raises an eigenvalue problem, preventing successful execution.
I would greatly appreciate any insights or suggestions on why this eigenvalue issue might occur despite the structural similarities between the two models. Additionally, any advice on how to address and resolve such eigenvalue issues in the non-linear model implementation would be very helpful.
@jpfeifer Thank you very much for your response. So, what you mean is that if the same model is analyzed in both linear and non-linear forms and they should theoretically produce consistent results, but it only works for the linear form and not for the non-linear form, it could indicate a potential issue with the model itself. Is that correct?
@jpfeifer Professor, I have one more question. When I express the non-linear model in an exponential (exp) form, it produces results consistent with the linear model without any eigenvalue issues. However, when I input the non-linear model in a non-exp form (as shown in the mod file attached in my original question), eigenvalue issues arise. Could you please provide some insight into what might be causing this issue?
The determinacy properties of the model are independent of whether you use an exp() substitution or not. If the non exp() model does not run, there must be a mistake somewhere.