On mean reversion in real interest rates: An application of threshold cointegration

Using data from Germany, Japan, UK, and the U.S., we explore possible threshold cointegration in nominal short- and long-run interest rates with corresponding inflation rates. Traditional cointegration implies perfect mean reversion in real rates and hence confirms the Fisher hypothesis. Threshold cointegration accounts for the possibility that this mean reversion is active only conditional on certain threshold values in the observed variables. We investigate whether findings of such effects can be exploited for interest rate prediction.