Can interest rate moves – and therefore bond returns – be accurately predicted? This question has intrigued finance academics and practitioners for decades, as government bonds represent one of the most important asset classes for investors. The size of the global government bonds market represents about 30% of overall market capitalizations of all investable asset classes.

Bond market returns fluctuate substantially over time, mainly due to changes in bond yield levels. It is therefore very tempting to try to find a way to predict bond moves systematically. Over the past decades, many studies have therefore examined this issue. But the evidence has so far remained limited, with existing studies facing three major challenges.

First, their sample is usually rather narrow, typically to the post-1980 period for only the US bond market, a period in which rates generally declined. However, many investors are currently more worried about a potential rise in interest rates. Second, existing studies typically use different methods to examine predictability on relatively small samples, raising datamining concerns. Thirdly, most studies use predictive regressions to test statistical significance, casting doubt about the significance of the results of many studies.

In a new research paper,¹ we look at the predictability of international government bond market returns comprehensively, based on a broad sample – spanning all major developed market bonds since 1950 – and using an approach that is easy to apply for investors. This is not via a typical academic approach of regressions, but via investable real-time trading strategies. Our research period thus includes the large rise in interest rates in the 1960s and 1970s.

We find consistent and ubiquitous evidence of bond market predictability, with economically strong and generally statistically significant Sharpe ratios for the simulated investment strategies. Value, momentum, economic growth and inflation measures strongly predict where interest rates will move to. For example, a strategy based on these themes, or ‘styles’, leads a Sharpe ratio of 0.87 since 1950.

Our findings hold for all sovereign developed bond markets, and hold for every decade since the start of our sample (i.e. 1950). This period includes 30 years of out-of-sample data on international bond markets and an out-of-sample set of nine additional countries. Further, the results hold independent of economic conditions, including during prolonged periods of rising or falling rates.

Finally, the results are exploitable after transaction costs and can add substantial value for the bond or multi-asset investor. In short, whether interest rates move up or down can be robustly predicted. From a practitioner perspective, our findings imply potential exploitable opportunities for the active duration management in government bond portfolios.

Footnote

¹Baltussen, G., Martens, M., Penninga, O., 2020, ‘Predicting Bond Returns: 70 years of International Evidence’, working paper.