Exploitable Government Bond Return Predictability?

Are government bond returns exploitably predictable? In their June 2020 paper entitled “Predicting Bond Returns: 70 Years of International Evidence”, Guido Baltussen, Martin Martens and Olaf Penninga examine predictability of international 10-year government bond returns with emphasis on two subsamples, January 1950 through September 1981 (mostly rising interest rates) and October 1981 through May 2019 (mostly falling rates). They consider five predictive variables, each transformed into a binary signal:

1. Yield spread – 10-year government bond yield minus the cash rate, standardized relative to historical values.
2. Bond trend – sign of past 12-month 10-year government bond return.
3. Past equity return – past 12-month equity index return in excess of cash return, standardized relative to historical values.
4. Past commodities return – past 12-month commodity index excess return, standardized relative to historical values.
5. Combination – equal-weighted combination of signals 1 through 4.

They use a spliced 10-year government bond sample, using excess return on a representative bond index before inception of associated futures and futures returns thereafter. Using monthly returns for 10-year government bond indexes/futures and cash rates for Australia, Canada, Germany, Japan, UK and U.S. during January 1950 (except October 1961 for Japan) through May 2019 (7,497 monthly returns), they find that: (more…)