Why the stock/bond correlation actually matters more than any single number
The entire case for a diversified portfolio rests on assets that don't all fall at the same time. For most of the last few decades, government bonds reliably rose when stocks fell — a recession scared investors out of equities and into the safety of bonds, pushing bond prices up right when you needed the cushion most. That negative correlation is the real mechanical reason a 60/40 portfolio has historically had a smoother ride than 100% equities, not just "because it's more diversified" in the abstract.
What broke in 2022
2022 was unusual precisely because that relationship flipped: rising interest rates hurt bond prices directly (existing bonds became less attractive next to newly issued, higher-yielding ones) at the same time high rates hurt stock valuations (a higher discount rate lowers the present value of future cash flows — see our DCF model for exactly why). Stocks and bonds fell together, and the rolling correlation chart above shows that shift happening in real historical data rather than as an abstract warning.
How to actually use this
Don't treat the full-period correlation number as a fixed law — treat it as a long-run average that can and does drift, especially around big shifts in interest-rate regimes. When you're building a portfolio for genuine diversification, it's worth checking whether the correlation you're relying on has held up in more recent, shorter windows, not just the multi-decade average. Gold's correlation with both stocks and bonds tends to be closer to zero across most periods — one reason it's often discussed as a diversifier when the stock/bond relationship itself is under strain.
The honest limitations
Correlation is a backward-looking statistical measure, not a guaranteed relationship — it can and does change, sometimes sharply, exactly when you need it to hold. It also only captures linear relationships and says nothing about magnitude: two assets can be highly correlated but move by very different amounts. Use this to understand historical diversification behaviour, not to assume it going forward.