Bonds are primed for a better 2023, but how much better?

A lot of investors might believe that bonds are probably going to snap back—they couldn’t possibly do worse than 2022. But this is by no means a safe bet. There are a number of factors in the market that suggest bonds aren’t going to snap back, or at least snap back enough to put them in positive territory.

It is easy to see why people think that bonds are due to improve. This year’s bond crash will go down in the history books as an extreme: a 24.8% plummet for long-term U.S. Treasurys through Friday, as judged by Vanguard Long-Term Treasury Index Fund ETF (VGLT). Assuming bonds end the year at their current level, their loss would be far and away the worst since 1793, according to a database maintained by Edward McQuarrie, an emeritus professor at Santa Clara University. In fact, such a loss would be nearly 4.0 standard deviations below the average annual return of the past 229 years.

Also consider: The U.S. stock market has lost 14.5% year to date, as judged by Vanguard Total Stock Market ETF (VTI). Bonds have done about 10 percentage points worse than that.

However, there are several reasons to think bonds won’t see the kind of comeback people want.

The first factor in play is what’s known as “regression to the mean.” People typically think this implies a big reversal in performance. In other words, a losing year is more likely to be followed by a winning year than another losing one, and vice versa.

But this represents a misunderstanding of what regression to the mean actually entails. It doesn’t mean that bonds will necessarily deliver a positive return next year. It instead implies that the second observation in a series—such as a year of market returns—is likely to be closer to the mean, or average, than the prior one, especially when that prior one was extreme. So, if bonds had a terrible 2022, they are likely to have a 2023 that is closer to the mean—returns could still be negative, however, just closer to the mean.

Things get even more complicated when you consider another factor—so-called trend following. This can be just as strong as regression to the mean, if not stronger, and thereby diminish its effects.

Trend following happens—in all types of assets, not just bonds—when a loss is more likely to be followed by another loss than a gain (and vice versa). In other words, the trend in play keeps going rather than reversing.

Prof. McQuarrie’s database shows how this plays out. Consider the years since 1793 in which bonds rose; they rose again the subsequent year 86% of the time. If regression to the mean were dominant, you’d expect this percentage to be higher in a year after bonds fell.

This has not been the case, however. Following declining years, bonds rose less frequently—79% of the time. This suggests that trend following in the bond market is reducing the impact of regression to the mean. Furthermore, this impact becomes increasingly strong the longer the declines last.