The Changing Role of Expectations in US Monetary Policy: A New Look Using the Livingston Survey

Abstract

Using a Bayesian structural vector autoregression (TVP-SVAR) with time-varying parameters and volatility we investigate monetary policy in the United States, in particular its interaction with the formation of inflation expectations and the linkages between monetary policy, inflation expectations and the behaviour of CPI inflation. We use Livingston Survey data for expected inflation, measured at a bi-annual frequency, actual inflation, unemployment and a nominal interest rate to estimate the VAR and show the significant changes that have occurred in the responses of these variables to monetary policy shocks or to shocks to expected and actual inflation. In so doing, we generalize the analysis undertaken by Leduc, Sill and Stark (2007) to allow for a more nuanced and detailed look at questions such as the impact of different chairmanship regimes at the Federal Reserve Board, the role of good policy versus good luck, and second round inflation effects. While some of the questions asked have a relatively long history, the methods used to undertake our investigations are very new, and the time-varying structure allows us to offer a more detailed picture. In using these methods we also undertake a substantial technical discussion to unearth the appropriateness of the TVP-SVAR models hitherto estimated in the literature, in particular the role of the choice of priors in determining the outcome of the estimations. As we discuss in the paper, this is an important issue which has remained rather hidden in the discussions surrounding the estimation of TVP-SVARs, yet may have a substantially important role to play in determining the results obtained.