Published online by Cambridge University Press: 26 February 2010
We pursue a semiparametric approach to examining core implications of the Fisher hypothesis, namely cointegration linking nominal interest rates and inflation, and homogeneity of the potential equilibrium relation. The sample is an unbalanced panel and comprises monthly time series from more than 100 economies. The time period of at most 45 years is subdivided into three regimes according to dominating monetary policies. To exploit the cross-sectional dimension for inference on parameter homogeneity, we apply mean group estimation of functional coefficients that allow the conditioning of key model parameters on economic states. The evidence in favor of cointegration is weakened over states of negative real interest rates that are likely to coincide with scenarios of high inflation. The ex post real interest rate is mostly diagnosed as unstable. The Fisher hypothesis is particularly confirmed for states characterized by large positive interest rate adjustments during the inflation-targeting regime.