IMF GPM Model - error in estimating likelihood function

I downloaded the attached code from Doug Laxton’s website but it fails when I run it returning “error in computing likelihood for initial parameter values” and “error: BK conditions not satisfied, no stable equilibrium” despite the fact that the BK condition are satisfied (4 eigenvalues with modulus>1 for 4 forward looking variables).

I’m running Dynare 4.2.2 on Octave 3.2.4.

edu1_steadystate.m (381 Bytes)
candyn_1985Q1_2008Q1_per93.m (15.6 KB)
edu1.mod (1.72 KB)


I have the same problem, with this and other models (see my post on the same day as yours). This model works perfectly well under Dynare 4.0.4 (for example) but not under Dynare 4.2.2.

Obviously something has changed between the two versions which causes the model(s) to fail, but I am at a loss to know what it is. Let us hope that the Dynare team can point us in the right direction to a solution.


I hope so, it is awfully frustrating. (though I’m glad its not just me).


Apparently, there is a bug. In contrast to what the manual says for Dynare 4.2.2, specifying “unit_root_vars” alone is not sufficient. Try putting

into the options of the estimation command and it should run.

There is a bug in Dynare 4.2.2 which appears with some version of the GPM model. For this specific model, you should use version 4.2.1 until we release 4.2.3 (which should happen soon).