Momentum and the Term Structure of Interest Rates

Abstract

A vast literature reports excess returns to momentum strategies across many financial asset classes. However, no study examines trading rules based on price history along individual government-bond term structures-that is, with respect to duration buckets across the curve-as opposed to across sovereign markets or individual term structures as a whole over time. Under duration-neutral and long-only constraints as well as low trading costs, this paper reports excess annualized returns of up to 120 basis points and information ratios as high as 0.79 using U.S. Treasury total return data from December 1996 through July 2013. Given a corresponding longshort strategy with no absolute duration risk, excess returns and information ratios are up to 207 basis points and 1.01, respectively. Unlike momentum strategies in some other asset classes, the excess return distributions are positively skewed, and momentum loads, if in any way, favorably on broad risk factors. Returns correlate to a degree with portfolios based on instantaneous forward term premium estimates, in turn derived from a set of Gaussian arbitrage-free affine term structure models. However, substantial variance remains unexplained, the betas are less than one, and the alphas are meaningfully positive. A caveat is that underlying behavioral explanations for momentum are lacking in the context of the U.S. Treasury market.

Another View on U.S. Treasury Term Premiums

“…Greater skepticism regarding formal tools for extracting monetary policy expectations from recent yields, and in turn increased reliance on model-free metrics, indeed could be forgiven. With respect to investors, and perhaps similar to shortcomings in the CAPM or some equity risk premium estimates, model-free anomalies along the term structure, including persistent momentum patterns both within and across government bond markets (Durham, 2013d;Asness et al, 2013), could provide alternative signals regarding required duration compensation that ATSM-based term premium estimates fail to identify. 24 Without any claim toward remote progress on estimation issues, the preceding analyses of a survey-based forward term premium proxy, simple econometric analysis of the time-series dynamics of the level of the term structure, and sample extensions of simple but fullyparameterized ATSMs raise questions regarding the very-low-term-premium consensus.…”

Betting Against Beta (and Gamma) Using Government Bonds

“…However, a low IR is neither a necessary nor sufficient condition to explain favorable absolute-risk-adjusted returns. Perhaps the most problematic result for BAB gov is that, although the ex-post loadings appear favorable, as they do for term structure momentum (Durham, 2013b), the ex-ante hedges for level, slope, and curvature exposure largely erase the gains from the strategy, using both the CRSP data on USTs and the Bloomberg data on the broader cross-section. Again, as argued previously, the fact that GATSM parameters may readily reproduce an inverse schedule between SRs and maturity does not imply abnormal BAB gov returns, but rather to the contrary that such an empirical relation is not necessarily inconsistent with the no-arbitrage condition.…”

Another View on U.S. Treasury Term Premiums

“…Greater skepticism regarding formal tools for extracting monetary policy expectations from recent yields, and in turn increased reliance on model-free metrics, indeed could be forgiven. With respect to investors, and perhaps similar to shortcomings in the CAPM or some equity risk premium estimates, model-free anomalies along the term structure, including persistent momentum patterns both within and across government bond markets (Durham, 2013d;Asness et al, 2013), could provide alternative signals regarding required duration compensation that ATSM-based term premium estimates fail to identify.…”