Sequential Learning and Economic Benefits from Dynamic Term Structure Models

This paper explores the statistical and economic importance of restrictions on the dynamics of risk compensation, from the perspective of a real-time Bayesian learner who predicts bond excess returns using a dynamic term structure model (DTSM). We propose a novel methodological framework that successfully handles sequential model search and parameter estimation over the restriction space landscape in real time, allowing investors to revise their beliefs when new information arrives, thus informing their asset allocation and maximizing their expected utility. Our setup provides the entire predictive density of returns, allowing us to revisit the evident puzzling behaviour between statistical predictability and meaningful out-of-sample economic benefits for bond investors. Empirical results reveal the importance of different sets of restrictions across market conditions and monetary policy actions. Furthermore, our results reinforce the argument of sparsity in the market price of risk specification since we find strong evidence of out-of-sample predictability only for those models that allow for level risk to be priced. Most importantly, such statistical evidence is turned into economically significant utility gains, across prediction horizons. The sequential version of the stochastic search variable selection (SSVS) scheme developed offers an important diagnostic as it monitors potential changes in the importance of different risk prices over time and provides further improvement during periods of macroeconomic uncertainty, where results are more pronounced.