Short Term Interest Rate ARMA

[brennan6738]

I am trying to forecast the LIBOR-OIS rate using monthly data going back to 2003. The SIC seems to suggest an ARMA(1,0) or ARMA(1,1). I have tried various models, and they consistently show the rate turning negative. I think it's a possibility, but highly unlikely. Is there a way to put a lower bound or adjust the model to give me a more realistic forecast? I tried adding a constant but that didn't fix the issue. Anyone have an idea?

[startz]

estimate log(interest) as the dependent variable