New Hope for the Fisher Effect? A Re-Examination Using Threshold Cointegration

This paper reassesses the long-run relation between nominal interest rates and inflation using German data. It shows that the empirical rejection of the strict Fisher effect in previous studies, i.e., the finding of interest rates not fully adjusting to changes in inflation, can be attributed to the particular time series behavior of inflation and interest rates which cannot be accounted for by standard non-stationary models. It is argued that the stochastic process governing the bivariate system of inflation and interest rates depends on the level of the variables and should be modeled as a threshold cointegration (TC) model. Contrary to the unit root hypothesis this model can be given an economic interpretation in terms of the opportunistic approach to disinflation. The full Fisher effect, even in its tax-adjusted form, cannot be rejected when a threshold cointegration model is estimated. The TC model not only explains the downward bias of the coefficient estimates, but also the sample and country sensitivity observed in previous studies. The TC model may prove useful in testing other long-run relations such as uncovered interest rate parity or purchasing power parity.