Abstract
As the Fisher Effect is either rejected or accepted without a real consensus in empirical studies, it is interesting to test for a unit root in a local-to-unity framework. Moreover, given the inflation expectation behaviour before and after 1979, we shall let a shift occur at a significant time break of our sample, so as to deal with potential non stationarity in the real interest rates series, instead of using Markov switching regimes models. The main innovation is to rely on structural breaks in the deterministic part while combining this method with an efficient unit root test. Empirical results reject a stochastic trend for the US short-term real interest rate from 1951 to 2000. This is consistent with an ex ante real rate constant over time.