The Hidden Assumption Inside a 60/40 Portfolio

Direct answer: Stocks and bonds can fall together because both prices respond to inflation, expected interest rates, and the compensation investors demand for bearing risk. The negative relationship familiar to many balanced-portfolio investors is therefore conditional, not contractual. In the lab's frozen sample, the rolling SPY–TLT correlation changed sign. That observation demonstrates that diversification protection can depend on the market regime; it does not identify one cause, forecast the next regime, or prove that the positive relationship will persist.

The allocation decision hidden inside the shorthand#

A 60/40 portfolio appears to divide capital between two asset classes. Its protection, however, depends on a second allocation that is easy to miss: the allocation of risk across economic shocks.

When weak growth is the dominant concern and inflation is contained, falling equity expectations can coincide with expectations of monetary easing. Stock prices may decline while government-bond yields fall and bond prices rise. The two sleeves respond differently, and the bond allocation can offset part of the equity loss.

When inflation or term premia dominate, the same pairing can behave differently. Higher expected rates reduce the present value of distant cash flows. That pressure can lower equity valuations and nominal bond prices at the same time. The portfolio still owns two asset classes, but both can load on the same discount-rate shock.

The hidden assumption is not simply that bonds are safer than stocks. It is that the shock that hurts equities will continue to help nominal government bonds.

Three mechanisms, not one forecast#

The distinction becomes clearer when the possible shock is named before the asset response.

These are scenario mechanisms, not predictions. The Bank for International Settlements reports that the United States equity–government-bond correlation switched sign in mid-2021 and connects the change to the growing importance of inflation news for expected monetary policy. The Federal Reserve's June 2022 Monetary Policy Report records a period in which Treasury yields rose amid sustained inflation and expected tightening while broad equity prices fell. Longer history also resists a permanent-rule interpretation: European Central Bank analysis documents generally positive stock–bond correlations from the late 1960s through the late 1990s and predominantly negative correlations during the 2000s.

That institutional record makes a mechanism plausible. It does not prove that any one mechanism caused the lab's measured reversal.

Measured in the lab#

Measured evidence. The lab uses a frozen price snapshot from 2 January 2015 through 6 July 2026. A 126-trading-day rolling correlation between SPY and TLT averaged −0.35 during 2015–2019, reached +0.37 during 2022–2023, and remained positive at +0.31 at the end of the sample. The supporting diversification decision brief documents the window choice, decision consequence, and failure modes.

A second test asks a different question. It holds individual asset volatilities fixed while moving the full estimated correlation matrix toward perfect positive correlation. At the observed correlation structure, the studied portfolio had 11.6% annualized volatility. At the perfect-correlation limit, volatility rose to 15.8%—roughly four percentage points of diversification removed.

The allocation in that stress is precisely:

The familiar 60/40 label describes the broad investor question. It does not describe the lab's exact test portfolio, which is a 55/30/15 balanced multi-asset allocation.

Established context. Central-bank and international-institution research documents environments in which inflation news, expected policy, and term premia can expose equities and nominal bonds to a common channel. That context supplies mechanisms worth testing against the measured history.

Interpretive implication. Diversification should be evaluated across named shocks and several correlation states, not accepted from an asset-class label or one estimated covariance matrix. A defensive sleeve is protection against a specified risk, not universal portfolio insurance.

What changes in practice#

A more defensible portfolio review asks five questions:

1. Which shock is each sleeve expected to absorb?
2. Which pairwise relationship supplies most of the current diversification benefit?
3. What happens when that relationship weakens, reaches zero, or changes sign?
4. Do duration, liquidity, volatility, and rebalancing assumptions remain realistic under the same scenario?
5. What evidence would cause the allocation thesis to be revised?

This does not require a confident regime forecast. It requires acknowledging that a historical relationship is an input with uncertainty, then testing the portfolio when that input changes.

Limitations and failure modes#

• The SPY–TLT result comes from one frozen historical sample and one 126-day rolling-window specification.
• SPY and TLT are liquid proxies, not complete representations of global equities or every bond exposure.
• The historical sign change measures co-movement; it does not identify inflation, monetary policy, fiscal policy, liquidity, or geopolitics as the cause.
• The correlation stress is deliberately stylized. Real stress can change asset volatilities, liquidity, spreads, and correlations together.
• The tested portfolio is one fixed allocation with monthly rebalancing. Other weights, instruments, horizons, and implementation costs require their own tests.
• Neither the observed reversal nor the counterfactual stress predicts the timing, duration, or magnitude of a future regime.

The bounded conclusion is still useful: diversification protection was regime-dependent in the observed record, and a decision process should be able to survive that relationship changing again.

Continue into the evidence#

Examine the stock–bond regime dossier. It separates the measured relationship from the institutional context, follows the consequences into portfolio construction and model calibration, and states what the evidence cannot establish.