Kenneth J. SINGLETON, Stanford Graduate School of Business
“Learning From Disagreement in the U.S. Treasury Bond Market”
We study the evolution of risk premiums on US Treasury bonds from the perspective of a real-time Bayesian learner BL who conditions her beliefs on measures of disagreement among professional forecasters about the future paths of bond yields. Learning about historical yields and disagreement within a dynamic term structure model leads to substantial variation in BL's subjective expected excess returns on bonds. The informativeness of disagreement is shows to be distinct from the (much weaker) forecasting power of inflation and output growth. Rather, it appears to be reflect policy uncertainty and, in particular, uncertainty about fiscal policy. BL's learning rule substantially outperforms consensus forecasts of market professionals, particularly following U.S. recessions.