Different versions, different results?

[johnnyking]

Dear all,

I wrote the attached mod file and got the following error message using version 4.2.5. Nevertheless, the obtained IRFs look exactly as expected, meaning that the model itself should be correct.

[...]
Processing outputs ...done
Preprocessing completed.
Starting MATLAB/Octave computing.

Warning: Matrix is singular to working precision.
> In dr1 at 441
In resol at 147
In stoch_simul at 66
In mod2_miguel at 284
In dynare at 120
Warning: Matrix is singular to working precision.
> In dr1 at 452
In resol at 147
In stoch_simul at 66
In mod2_miguel at 284
In dynare at 120
[...]

However, it turned out that the code is not working under versions 4.2.0 (IRFs look VERY strange) and 4.3.0 (Blanchard-Kahn conditions not satisfied). Has anybody an idea what the problem could be? Has anything changed in these versions that could be the cause of this problem? I also tried random parameterizations and it didn't help. Version 4.2.5 works (reporting the singularity error), 4.2.0 and 4.3.0 don't.

[jpfeifer]

The warning in 4.2.5 already indicated that there is a problem. Dynare 4.3.1 has refined the checking for problems. There seems to be a timing issue in your model, leading to a singularity.
The endogenous variable p_fh is not present in the model at the current period.