Hi. I’ve successfully implemented ramsey_policy in Version 4.1 for a relatively small (21 endogenous variables, at least 2 just for reporting) nonlinear model with 2 policy variables. I wonder if there is some way to recover the actual minimum discounted loss that is computed, since this would be important for comparing different (nested) models. Thanks in advance for some answer.


This feature is not implemented in Dynare 4.1, but it is now implemented in the unstable (development) version of Dynare.

If you can’t wait for Dynare 4.2, use the snapshot (see the website): the ramsey_policy there now displays the value of planner objective function under Ramsey policy (it also stores it in oo_.planner_objective_value).


Thank you Sébastien for your reply, and sorry for my delay. I did get the loss using the snapshot. But I seem to detect some anomalies in the output, although maybe there is none and I’m confused. But I’d like to be sure. In particular, in the smaller model I am working with now there is a purely quadratic loss function (no linear terms) and there are six nonlinear equations (four of them forward looking) plus six AR(1) shock processes that are written out as six more equations. But ramsey seems to confuse the role of these variables. In particular, 1) whereas in the model equations only two of these variables appear lagged, the resulting POLICY AND TRANSITION FUNCTIONS (including the economic policy rules) respond to all the lagged values of the shock variables (as well as the six shocks themselves); 2) two of the six shock variables respond to a shock variable different from themselves; and 3) two of the shock variables respond to a lagrange multiplier. Am I right in thinking there is something wrong here? Or is it simply that there is always more than one way of writing out the correct policy functions? When I solve directly with my own Matlab code my policy variables only respond to the contemporaneous shock variables (i.e. the shocks), two of the six lagged shock variables (the ones that appear lagged in the model equations), the three model variables that appear lagged, and the six lagrange multipliers. Thank you in advance for your intuition or opinion.

Hi. I found a mistake in my code and now the output seems to be OK. Thanks anyway.
gescude :smiley: