Hi,

I have been evaluating long run welfare under two monetary policy regimes by using the ‘burn’ option in Dynare++ to discard the first 100 values from each simulation, in order to randomise the initial conditions. The model has to be solved to 2nd order as policy affects risk premia, and the advantage of Dynare++ is that it makes it feasible to simulate the model to 2nd order a large number of times.

Having discovered that policy B dominates policy A over the long run, I now want to see if this conclusion is overturned if I account for the transition from regime A to regime B. If I understand correctly, this could be done in Dynare++ using the ‘initval’ option. Specifically, my idea is to run a simulation of regime B in which the intial values of endogenous variable are set equal to the stochastic fix point of regime A, and where burn = 0 so that the transition periods are not disregarded in the calculation of welfare. Will this work in Dynare++?

I have also looked into whether this could be done with Dynare. If I understand correctly there are two options: (1) use initval as described above, making sure additionally that the ‘steady’ command is commented out; or (2) model the transition mid-simulation by creating an indicator variable that becomes 1, rather than 0, once regime B is adopted, by mixing stochastic and deterministic shocks using the ‘varexo_det’, ‘periods’ and ‘forecast’ commands as described in the user guide. I have been able to get (2) working, but only in models solved to 1st order. Once I set order = 2 in these models, Dynare (v 4.3.1) no longer solves them, saying 'Undefined function or variable “gx” '. But the same models are solved fine to 2nd order if the deterministic shocks are removed. Can Dynare solve models with both stochastic and deterministic shocks to 2nd order?

Thanks in advance,

mch