Restrictions on the risk-pricing in dynamic term structure models (DTSMs) can unleash the power of no-arbitrage by creating a tighter link between cross-sectional and time-series variation of interest rates. This paper presents a new econometric framework for estimation of affine Gaussian DTSMs under restrictions on risk prices, which addresses the issues of a large model space and of model uncertainty using a Bayesian approach. A simulation study demonstrates the good performance of the proposed method, both for model choice and for inference about the objects of interest. I obtain novel results for the U.S. Treasury yield curve. The data strongly favor tight restrictions on risk pricing: only level risk is priced, and only changes in the slope affect term premia. Incorporating the restrictions into an otherwise standard model substantially alters its conclusions. Interest rate persistence is significantly higher than in a maximally-flexible model, hence expectations of future short rates are more variable, and the role for term premia is somewhat diminished. Hence, restrictions on risk prices help resolve the puzzle of implausibly stable short-rate expectations which has plagued this literature. Restricted models attribute a larger share of the secular decline in long-term interest rates over the last twenty years to the expectations component, consistent with survey evidence on expectations of future interest rates and inflation.
no-arbitrage prices of risk Bayesian model selection term premium