Stock-Bond Diversification Offers Thinner Protection as Correlation Holds Positive
The rolling 36-month correlation between global equities and government bonds stood at about 0.59 at the end of September 2026. Reuters
The estimate covers a three-year window on global benchmarks. It keeps stock-bond co-movement in positive territory.
On February 18, 2026, the International Monetary Fund published analysis titled "Stock-Bond Diversification Offers Less Protection From Market Selloffs". IMF
Historically, when stocks fell, investors sought safety in bonds, and bonds rallied, cushioning losses and stabilizing portfolios. That description formed the starting point for the Fund's analysis.
Analysts noted in September 2026 that a positive correlation between stocks and bonds emerged after the COVID-19 pandemic-induced inflation shock. Reuters
The stock-bond correlation tends to turn positive when inflation runs high, usually around 2.7%, except in recessions. Reuters
The broader context here is familiar to multi-asset desks. Positive stock-bond correlation changes joint distribution assumptions. Duration stops offsetting equity beta in drawdowns. Portfolio variance rises for a given equity weight, and ex ante diversification ratios overstate realized risk reduction. Volatility targeting becomes more procyclical. Rebalancing adds equity exposure into weakness without the usual offset from duration gains. Cross-asset hedges priced on negative rate-equity beta decay faster.
In my view, the persistence matters more than the level. That print on a 36-month window is slow moving. It reflects sustained common exposure to nominal shocks rather than a brief risk-on episode. That has consequences for hedging, for liability-relative positioning, and for stress calibration. When inflation beta dominates both legs, term premia and equity risk premia compress and expand together. Convexity offers less portfolio-level protection. Correlation assumptions anchored in a prior disinflationary regime will misstate tail dependence and joint drawdown severity.
Looking at what this means for risk practice, the issue is measurement, not just allocation. Backward-looking covariance matrices will carry the inflation regime forward. Forward-looking scenarios need separate demand-shock and supply-shock paths, with term premia and equity multiples responding with different signs. That distinction determines whether duration functions as ballast or as amplifier. For risk committees, the practical question is how much weight to place on conditional correlation. Unconditional averages blend regimes. Conditional estimates tied to inflation and growth states give a cleaner read on diversification when it is needed most.
