The Fisher Hypothesis and the Forecastability and Persistence of Inflation

For the period 1860 to 1939, the simple correlation of the U.S. commercial paper rate with the contemporaneous inflation rate is -.17. The corresponding correlation for the period 1950 to 1979 is .71. Inflation evolved from essentially a white noise process in the pre-World War I years to a highly persistent, nonstationary ARIMA process in the post-1960 period. I argue that the appearance of an ex post Fisher effect for the first time after 1960 reflects this change in the stochastic process of inflation, rather than a change in any structural relationship between nominal rates and expectedi nflation. I find little evidence of inflation non-neutrality in data from the gold standard period.This contradicts the conclusion of a frequently cited study by Lawrence Summers, who examined the low frequency relationship between inflation and interest rates using band spectrum regression. Deriving and implementing a frequency domain version of the Theil misspecification theorem, I find that neither high frequency nor low frequency movements in gold standard inflation rates were forecastable. Thus even if nominal rates responded fully to expected inflation, one would expect to find the zero coefficient obtained by Summers