Rank conditions


I am doing a deterministic simulation of a New Keynesian model. When setting the inflation response parameter in the Taylor rule to be less than one, I should have indeterminacy. But in my simulation, even though the rank condition is not verified, I still get a result. My question is that how Dynare chooses eigenvalue in the indeterminacy case? Can I choose eigenvalue manually by myself? Following is the simulation message:

There are 16 eigenvalue(s) larger than 1 in modulus
for 17 forward-looking variable(s)

The rank conditions ISN’T verified!


variable m period 1
1- err = 0.22499
err_f = 0.0075637
Time of this iteration : 0.125
variable m period 2
2- err = 3.7746e-017
err_f = 2.7756e-017
Time of this iteration : 0.078

    Total time of simulation        : 0.281

    Convergence achieved.



It looks like you are doing a deterministic (i.e. perfect foresight simulation). Dynare does not select an eigenvalue, it only looks for a path which verifies the initial and terminal conditions. There may be several of these, and the Newton-type solver has selected one depending on the initial conditions.

Note that, in your case, Dynare will refuse to do a rational-expectations (stochastic) simulation.

There is currently no interface to let the user choose between various solutions in the indeterminacy case, but that feature may be added in the future.