How do monetary policy expectations and term premia respond to news? This paper provides new answers to this question by means of a dynamic term structure model (DTSM) in which risk prices are restricted. This leads to more precise and more reliable estimates of expectations and term premium components. I provide a new econometric framework for DTSM estimation that allows the researcher to select plausible constraints from a large set of restrictions, to correctly quantify statistical uncertainty, and to incorporate model uncertainty in the inference about risk pricing. The main empirical result is that under the restrictions favored by the data the expectations component, and not the term premium, accounts for the majority of high-frequency movements of long-term interest rates and for essentially all of their procyclical response to macroeconomic news. At both high and low frequencies, term premia are more stable than implied by a DTSM with unconstrained risk prices. The apparent disconnect between long-term rates and policy rates that has puzzled macroeconomists for some time is resolved by appropriately restricting the risk adjustment in models for bond pricing.
