retrieving decision rule coefficients symbolically

[mauriceroche]

Hi Guys,
Is it possible to retrieve decision rule coefficients symbolically (ie in terms of parameters and steady states) without solving the model? If so how?

I would like to use the 2nd order appproximation decision rule coefficients (from Euler conditions) in a different program (RATS) so that I can perform more analysis than just calculating unconditional moments.

If it is not possible to retrieve decision rule coefficients symbolically I could write some matlab code to perform what I need done during simulations. In this case in which dynare files would I need to call this routine/file?

Maurice

[StephaneAdjemian]

Hi Maurice,

Hi Guys,
Is it possible to retrieve decision rule coefficients symbolically (ie in terms of parameters and steady states) without solving the model? If so how?

Unfortunateley, dynare is unable to provide a symbolic expression of the decision rules (with dynare v4 we only have a symbolical expression of the approximated model, not of the solution). To our knowledge there is no way to analitically compute the generalized schur decomposition needed to solve the model (first or second approximation). Even if in some circonstances we are able to obtain the policy rules by hand (the canonical RBC model for instance)...

I would like to use the 2nd order appproximation decision rule coefficients (from Euler conditions) in a different program (RATS) so that I can perform more analysis than just calculating unconditional moments.

If it is not possible to retrieve decision rule coefficients symbolically I could write some matlab code to perform what I need done during simulations. In this case in which dynare files would I need to call this routine/file?

Yes, you should use matlab. You can use matlab codes in your mod file (even loops around stoch_simul). The solution is saved in the structure oo_.dr (fields ghx and ghu).

Best, St

[mauriceroche]

St

[OndraKamenik]

Maurice,

you can also use dynare++ to solve the model (any order of approximation). The solved decision rule is then dumped to a MAT file and can be simulated/evaluated for custom shocks and initial values with dynare_simul.m.

Ondra K.