Testing for a nonlinear Fisher relationship

Empirical evidence regarding the Fisher effect is mixed. One reason may be a nonlinear adjustment process in the real interest rate. The nonlinear unit root test proposed by Sollis, Leybourne, and Newbold (Journal of Money, Credit, and Banking 34: 686-700, 2002) is used to test for stationarity of the U.S. real interest rate over the 1934:01-2011:02 period and selected subperiods. The unit root null in the real rate of interest can be rejected over the full sample, evidence of a Fisher effect. Weaker evidence of a Fisher relation is found in the subsample for 1934:01-1959:12 for which the unit root null can be rejected for one measure of the real interest rate. However, there is no indication of a Fisher effect for subperiods starting in 1960 or later. A conjecture is that temporal aggregation of the interest rate data may explain the different results, but the findings are also consistent with other explanations.