Skip to content

Stock-Bond Correlation and Duration Risk Allocation: summary and key takeaways

The paper sorts weekly stock-bond correlations, built from intraday futures data, into deciles and tests whether they predict Treasury bond returns. Low correlation weeks were followed by stronger 10-year bond returns, and high correlation weeks by weaker ones. The authors argue investors should hold more duration risk when recent correlation is low.

Xinyi, L.; Hua, F. · 2016 · The Journal of Portfolio ManagementSummary updated 20 November 2026
Published2016Journal article
Asset focusStocks+BondsStock-bond correlation
TypePortfolio MgmtStock-Bond Correlation
Reading time4 minthis summary

Stock-Bond Correlation and Duration Risk Allocation at a glance

The question
Does the recent level of the stock-bond correlation forecast future Treasury bond returns and interest rate moves?
Dataset & period
S&P 500 futures and 10-year Treasury futures at 10-minute frequency, one correlation estimate per week, from Jan/11/1985 to Sep/06/2013, 1,496 weekly observations. Main decile results use the sub-sample starting in 1997.
Method
Each week the authors compute a realized stock-bond correlation from intraday futures prices. They rank it against its own recent history over several window lengths and convert the decile rank into a bond weight that is higher when the correlation is low. Significance is tested with a bootstrap, and a regression checks the strategy against yield momentum, curve spread, VIX and equity factors.

Key findings of Stock-Bond Correlation and Duration Risk Allocation

  • In the 1997 to 2013 sub-sample, the highest correlation decile underperformed the lowest decile by 32% (annualized) in the first week after ranking.
  • The simple correlation-sorted 10-year bond strategy had excess return of 3.9%, volatility of 5.9% and a Sharpe ratio of 0.66, which the paper reports as statistically significant.
  • After controlling for yield momentum, curve spread, VIX and equity returns, the 10-year strategy kept a weekly alpha of 0.066%, which the paper converts to 3.4% a year.
  • The 10-year bond's equity beta in the lowest and highest deciles was -0.14 and -0.05, so higher expected bond returns coincided with lower measured equity risk in the paper's results.
  • Following the lowest decile, the 10-year minus 1-year curve spread steepened from 1.04% to 1.65% on average over a year, so the strategy works against yield curve carry.
  • Average high-frequency correlation was 0.35 before 1997 and -0.30 after 1997.

Robustness, as reported in the paper

  • The paper says the 10-year result holds across different sampling windows and quantile choices.
  • High-frequency correlations had consistently more predictive power than low-frequency correlations for the 10-year bond.
  • For the 1-year bond, most of the excess return is explained by the yield momentum, curve spread and VIX factors, and alpha shrinks.
  • Results were weaker when only short windows or the past week alone were used, which the paper attributes to noise.

Caveats and limits

  • Many supporting exhibits are said to be available online and are not in the paper text, so those checks cannot be reviewed here.
  • The explanations (under-reaction by markets or policymakers, and to the safe-haven status of bonds) are proposals, not tested causal results.
  • Transaction costs and implementation constraints are not reported in the text (inferred).
  • The core sample ends in September 2013, and the later signal charts in the paper are shown as images only (inferred).

Items marked (inferred) are our reading of the design, not statements by the authors.

Why it matters

The paper links a measurable market signal to the timing of duration exposure. It is relevant to investors with bond-like liabilities, who face higher surplus risk when the stock-bond correlation is low. It also suggests the signal could act as an overlay on a stock and bond portfolio, though the authors' evidence is limited to their sample.

Questions about this paper

What does a low stock-bond correlation predict?

The paper finds it forecasts falling 10-year yields over the coming weeks and falling short-term yields over the next year. A high correlation points the other way.

Who should care most about the result?

The authors say investors with long-term bond-like liabilities should take more duration risk when recent correlation is low, since low correlation raises surplus risk and signals higher liability values.

Why might the effect exist?

The authors propose that markets or policymakers react slowly to the economic conditions the correlation implies, and that markets initially under-react to the long-term bond's safe-haven status.

Where can I read the original paper?

The original is at its publisher: https://doi.org/10.3905/jpm.2016.42.2.056. We link to the paper and do not host the PDF.

Follow new quant research

Download MacroFinance market data as CSV or Excel, or pull it by API, and test the ideas in these papers yourself. New summaries are added as we review them.

Create a free account
Summaries are written by MacroFinance from the published papers, in our own words, for information only. They are not investment advice and may omit detail. Papers are not hosted here: each page links to the original at its publisher.