Asymmetric Monetary Policy Expectations

Abstract

We document some novel empirical evidence of significant time-varying skewness in the aggregate forecast distribution of the federal funds rate (FFR), i.e. asymmetric monetary policy expectations. To this end, we construct measures of the one-year ahead FFR expectations from responses to the Survey of Primary Dealers (SPD). The SPD provides a "physical" future distribution of the FFR, in contrast to measures extracted from asset prices. Importantly, this survey's unique feature allows us to explicitly compute mean and modal expectations and the discrepancy between the two measures, free of risk premia. We further show that a simple New-Keynesian model with the ZLB constraint can endogenously generate both positive and negative skewness similar to patterns in the data. The time-variation of asymmetry in the aggregate distribution highlights the importance of correctly measuring the mean when extracting FFR expectations from surveys. We argue that the FFR forecasts from the Blue Chip Survey (BCS), a popular survey measure of monetary policy expectations, track the mode more closely than the mean since 2011, when the data became publicly available. As a result, the mean measure of policy expectations extracted from the SPD implies significantly less negative term premia compared to term premia implied by BCS forecasts. The mean measure also outperforms the BCS forecasts based on the mean squared error loss, consistent with the theory of optimal forecasting.