Predicting Future Yields and Risk Premia: The Blue-Dot Affine Model

Abstract

The authors present a new affine model that can predict future yields and risk premia in the monetary conditions of the past decade more convincingly than current state-of-the-art statistical models. Despite making use of very different sources of information, it produces remarkably similar changes in risk premia as the most popular statistical return-predicting factors. However, it predicts very different—and, they argue, more believable—levels for risk premia and expectations. The model is extremely parsimonious, is financially motivated, fits market yields accurately with very few interpretable parameters, and naturally recovers important qualitative features of the joint ℙ and ℚ dynamics of yields. TOPICS: Analysis of individual factors/risk premia, factor-based models, statistical methods Key Findings • A new affine model of the term structure is shown to give more plausible estimates of risk premia and expectations than the current state-of-the-art yield curve statistical models. • The model uses information from the Fed expectations of the future federal funds rate. • The model is financially justifiable, very parsimonious, and fits observed market yields very well.