I have written a SOE model with three sectors and incomplete markets (both for external debt and public debt) that runs in the decentralised economy (I provide an external steady state function as the SS is non-linear, I a few CES aggregators). To pin down equilibrium I assume a debt elastic premium (on both foreign assets and public debt), a tax rule and that government expenditure varies a fraction of output. The competitive equilibrium is pinned down and have no collinearity problems.
Now I would like examine optimal fiscal policy, so I remove the tax rule and government expenditure. I am currently trying to implement ramsey policy with two instruments (government expenditure and distortionary taxes). I have tried to follow the example that Johannes provides for (NK_example.mod) monetary policy (using the combination of steady_state_model and initial values for the instruments). But this procedure fails for me because I cannot reduce the non-linear SS to one equation (using the steady_state_model and initval). Since the decentralised economy is highly non-linear, then this route is not viable for me.
I have insteady tried to feed SS from the decentralised case and use initval but the model struggles to find a solution as it appears that the Ramsey SS may either be for from the competitive equilibrium or the Ramsey SS does not have a unique solution.
I have also written the full problem (with multipliers) but have not beeen able to pin down the SS. This is a little bit more involved to do do in Dynare directly given that the number of equations including the constraints and FOCs is around 40. I believe that ramsey_policy finds the SS using the OLS approach proposed by SGU, which splits the problem into two (SS conditional on instrument) and then OLS approach iterating until the instrument satisfies all conditions.
What I would like to know is whether there is simpler way to compute Ramsey SS (of a model whose competitive equilibrium cannot be easily reduced to one non-linear equation conditional on instrument).
Many thanks for your help.