By LARRY SWEDROE
Government bonds represent about 30 percent of overall market capitalisations across asset classes. Guido Baltussen, Martin Martens and Olaf Penninga contribute to the literature with their June 2020 study Predicting Bond Returns: 70 Years of International Evidence.
They examined the predictability of international government bond market returns over the period 1950 through May 2019. Their data sample covered ten-year government bonds across six major markets (U.S., U.K., Germany, Japan, Canada and Australia) and two secular rate cycles — the period 1950-1980 (a secular rise) and the post-1980 period (a secular decline) — providing tests of robustness. As a further test of robustness, they also examined nine other European markets (Austria, Belgium, France, Italy, Netherlands, Norway, Spain, Sweden and Switzerland). The variables they employed were:
— The steepness of the yield curve (i.e., yield spread)—10-year yield minus the cash rate.
— Past bond returns (i.e., bond trend)—the sign of the past 12 months bond return.
— Past equity returns—the past 12 months equity return in excess over cash.
— Past commodity returns—the past 12 months commodity index return.
— Combinations of the above four.
Here is a summary of their findings:
— The average bond market return was -0.3 percent over the first sub-period and 3.9 percent over the second.
— There is consistent and ubiquitous evidence for bond market predictability, with economically strong and generally statistically significant Sharpe ratios for each of the four variables.
— Their equal-weighted model combination produced a Sharpe ratio (SR) of 0.73 over the post-1980 period and 1.09 over the pre-1980 period. Over the full period, the SR was 0.87 for the equal-weighted combination.
— The combined SR was much higher than that of each variable individually, as correlations between the variable returns were close to zero (averaging to 6%).
— The results were robust over both the pre-1980 (mainly rising rates) and post-1980 (mainly falling rates) periods.
— Performance was substantial per decade, with decade SR varying between 0.33 (2000-2009) to 1.15 (1970-1979). Performance was also substantial for the last decade—SR of 0.58.
— Bond return predictability held in a further out-of-sample test across nine additional developed government bond markets.
— Predictability is robust across recessionary and expansionary periods, high and low inflation periods, and equity bull and bear markets.
— Predictability is not due to structural bond risk (only a small part of the performance can be explained by the average bond beta) but to bond market timing (higher bond beta in rising markets than in falling markets), with bigger performance in times of larger market moves—the strategy had on average a significantly positive bond beta of 0.22 in months bond returns were positive, and a significantly negative bond beta of –0.13 in months bond returns were negative.
Their findings led the authors to conclude that their results suggest that “predictability has decreased over time, albeit remaining significant and economically large.” They then provided evidence and insights into the economic channels of bond return. They showed:
— The yield spread and past equity returns predict GDP growth, and bond trend and past commodity returns predict inflation—they are forward-looking indicators for the macro economy.
— Since contemporaneously bond prices react negatively to both GDP growth and inflation, the predictive ability of these variables for bonds is partly driven by their predictive ability for GDP growth and inflation.
The authors then explored practical applications of their documented bond return predictability. Using bond market-timing strategies, they performed asset allocation tests and found:
— Investors benefit, with improvements in the Sharpe ratio of the ex-post mean-variance portfolio rising from 0.34 for a traditional equity-bond portfolio to 0.66 with inclusion of a simple bond market-timing portfolio.
— The results are exploitable in practice after accounting for transaction costs—the full-sample net SR was 0.71. Net SRs were 0.91 and 0.58, respectively, for the January 1950-September 1981 and October 1981–May 2019 sub-periods.
— There is strong evidence of bond market predictability on a subsample of highly liquid (nine European) 10-year bond futures.
Summary
Baltussen, Martens and Penninga found strong and robust evidence of government bond return predictability across the major developed bond markets, with the economic profits being sizeable (strong SRs). They also found little evidence of decay of bond market predictability. The predictability was robust over time periods and market episodes, including prolonged periods of rising or falling rates.
In addition, the authors concluded, “the predictability relates to changes in future macroeconomic states (i.e., growth and inflation), originates from market timing, is exploitable after transaction costs, and adds substantial value for the bond market or multi-asset investor.”