NaN for Theoretical Moments

Hello everyone,

After running the mod file I get bunch of NaNs for values of approximated theoretical moments in the Dynare output, even though I got IRFs as a result of the simulation. I am attaching my Dynare file and the complementary MATLAB steady state file for your reference. Both are slightly modified files from my previous post here in Dynare forum a week ago.
Model.mod (13.4 KB)
Model_steadystate.m (9.6 KB)

I thought there might be a unit root problem here but the model diagnostics did not detect any obvious problems with this mod-file. I suspect also that some variables (e.g., labor allocation for entrepreneurs and human capital stock for wage-earners) take negative values along the transition path, and this might cause all these NaNs. Not sure if my naive guess is correct though.

I will appreciate any advice here.

If you add check; you will see that model model indeed features a unit root.

Thank you, I will investigate further at this matter and will look up the similar topics (unit roots for forward-looking variables) to find the potential solution for my case.

Hi Gani,

I think you can have some results showing up (like some IRFs) but still non-stationarity issues. Think for example about the Open Economy Models (SGU 2003, JIE) with bonds but no adjustment costs and incomplete markets. These models are solvable but the Consumption difference across countries is non-stationary. In such a case, some theoretical moments will be NA.

What Johannes suggests is usually a good way to see which equation may be the one with the unit root issue. Similarly, you can also try


Also, your IRF showing up is not a sign of stationarity, they only are if you see the shock effect dying out over time and the IRF returning to zero.