How Rational Are Inflation Expectations. A Vector Autoregression Decomposition of Inflation Forecasts and Their Errors

摘要

Recent successive over-predictions of inflation have renewed interest in the rationality of forecasters and what causes their forecasts to deviate from rational expectations This paper examines inflation forecast data from the Livingston Survey and the ASA/NBER Survey of Professional Forecasters over the past 30 years to determine what publicly available macroeconomic information, if any, explains the persistence of forecast errors. A reduced form VAR is used to identify potential inefficiencies and then calculate the impulse response functions and variance decompositions of forecasts errors to analyze how shocks to the other endogenous variables of the VAR affect forecast error behavior. The study finds that the majority of public information is used by forecasters efficiently and therefore, supports weak form rational expectations of inflation, however, there appears to be significant inefficiency in the use of past forecast errors and the term structure of interest rates in the forecasts of both surveys. The IRF analysis also uncovers a significant change in the structure and variance of forecast errors that occurs in the early 1980's. It is hypothesized that this structural change of inflation forecast errors is related to a change in the way the Federal Reserve has conducted monetary policy since the end of the Volcker deflation in 1983.