Cointegration analysis of the Fisher hypothesis: the role of the real rate and the Fisher identity

The Fisher hypothesis on the long-run relationship between inflation and nominal interest rates has recently been investigated using cointegration techniques. This paper examines the role of the expected real rate in such cointegration analyses. It is shown that, contrary to Bonham's (1991) suggestion, it is inappropriate to add a proxy for the expected real rate to hypothesized cointegrating relations involving the nominal rate and inflation. An alternative framework for analysis of the Fisher hypothesis is presented which distinguishes between the Fisher identity and long-run behavioural equations for nominal and real rates; different possible empirical outcomes are discussed.