Expectations, risk premia and information spanning in dynamic term structure model estimation

This article examines the nature of the empirical instability in dynamic term structure models. I show that using survey forecasts is an effective solution because it directly addresses the information imbalance at the heart of the instability: it increases the (cross-section) information on actual dynamics, bridging the gap with the large (cross-section) information on the risk-adjusted dynamics. I relate this to other information spanning problems, particularly spanning of macro factors, and discuss the desirability of anchoring models to surveys. I also show that restricting prices of risk is not effective in ensuring stable and sensible implied expectations.