Problem with estimating NOEM models with a UIP condition

[bigbigben]

Hi, guys:

I have trouble with estimating a simple two country model with a UIP condition. The model works perfectly if I have a perfect international risk sharing condition, however it runs into indeterminacy so often if I switch to a UIP condition, which means an "incomplete financial market" by Chari et al (2004) or Heathcote and Perri (2002). I guess the problem is that a UIP condition introduce an eigenvalue of 1, while exchange rate becomes a forward looking variable if defined in this way. If I do simulation, I can simple set qz_criterium=.99 or something (It might be cheating), but I don't know how to do it in the estimation procedure.

There seem to be quite a few papers estimating open models with incomplete market, and I would appreciate if anyone can share me the tricks working around the indeterminacy issue.

Thanks in advance.

[SébastienVillemot]

It’s hard to give a precise answer to your problem without more details, however here is what I am aware of.

It is a well-known fact that NOEM models face an indeterminacy problem in the absence of some trick. Typically, the net foreign assets will be indetermined at the steady state. The easy solution for this issue is to add a small cost premium on the bond returns. This cost increases as the NFA deviates from its long run level. This long run level is not an endogenous feature of the model, but an exogenous value imposed by the modeler. Typically this cost premium will end up in the UIP condition.

See the description of the GIMF model (http://www.douglaslaxton.org) for an example.

Hope this helps,