I’ve been trying to solve a two-country model with EZ preferences and long-run risks in dynare++. One step is to set 1/IES = risk aversion so that EZ collapse to CRRA and see what happens, and suddenly the two identical countries are asymmetric! To illustrate what’s going on I’ve cooked up the following example:
h.mod and f.mod are exactly the same set of equations except that h.mod uses home country notation and f.mod uses foreign country notation. Not surprisingly, the two files give identical results.
But once I combine the two files into one, even though the two countries have no relation with each other in any way, suddenly the results are not symmetric anymore. I think the same thing happens if I use dynare as well. If I delete the long-run processes, the problem gets resolved somehow. Any idea why this is happening?
Thanks a lot!
code.zip (3.8 KB)
Could it be that you are relying on simulations and the random shocks are not the same anymore across countries?
Hi Professor Pfeifer,
Thanks for your reply! Yes, what you are suggesting is absolutely possible. But (1) the asymmetry seems severe and doesn’t go away if I change the periods or number of simulations; (2) what you are suggesting should disappear if I delete the long-run processes, since the remaining random shocks should still be different, but if I do delete the “z” processes, suddenly the two countries are symmetric again. Any thoughts?
Thanks a lot!
Solved. I compared the policy functions for the two countries and they are symmetric. I then re-simulated the moments with dynare_simul and got roughly symmetric results if I average over a large number of simulations.
Glad it worked. The issue with long-run shocks can be that they introduce persistent deviations from the steady state that may take a long time to average out.