Busting the Myth of Treasuries as a Haven

When stocks go down, bonds go up, right? That’s the received wisdom, at least. Well, it’s not what happened when war broke out between the US and Iran last month and both stocks and Treasuries slumped. It wasn’t true last year when both stocks and Treasuries surged. And it wasn’t true when they simultaneously plunged in 2022, with Treasuries actually taking heavier losses.

Those are a lot of disappointments in a relatively short time. That also left some investors wondering if Treasuries are still the bear-market hedge they are touted to be — which prompted me to ask if they ever were. After digging into the data, I discovered a surprising answer: no.

This doesn’t mean that Treasuries don’t do what they have always done well, which is help stabilize portfolios over time. But longstanding assumptions that Treasuries are a lifeline in a bear market have primed investors for disappointment, particularly in markets hammered by war in Iran and haunted by the specters of rising inflation and snowballing federal debts and deficits.

Monthly total returns for the S&P 500 Index and long-term Treasuries had no apparent relationship in the data I reviewed, going back to 1926. Their returns were not correlated over any period, whether over one month (0.08, where a reading of 1 would indicate they moved in lockstep), one year (0.08), three years (0.1), five years (0.08) or 10 years (0.13). Even if you knew with certainty how one would perform in the future, you still couldn’t reliably guess how the other would fare. That hardly amounts to a reliable lifeline.

One reason it’s hard to anticipate how stocks and Treasuries will interact is that their relationship can change drastically. In 2020, for example, they were strongly negatively correlated; then, strongly positively correlated in 2022 and 2023 and more recently, not at all. Across rolling 12-month periods, they have moved almost perfectly together (high single-year correlation of 0.89) and almost totally in opposite directions (low one-year correlation of -0.94), and their correlation can swing wildly from year to year. The same is true when looking at longer periods, except the high and low correlations are less extreme with more time.