In 2022, a standard 60/40 portfolio lost roughly 16.7%. Both equities and investment-grade bonds fell together as inflation surged and central banks tightened aggressively. It was one of the worst calendar-year outcomes for balanced portfolios in modern history (Reddy and Smith, 2025). The episode was not a freak accident. It was a regime shift that exposed a structural assumption baked into every static balanced portfolio: that stocks and bonds reliably move in opposite directions.

A recent review article by Urinson (2026) systematically examines why that assumption breaks down and what a more resilient portfolio architecture looks like. The paper draws on empirical work across G7 markets, regime-switching models, and signal-based allocation research to argue that diversification is not dead, but it has become conditional.

The correlation problem

The entire logic of 60/40 rests on the stock-bond correlation staying negative or at least weakly positive. When equities drop, bonds rally, and the portfolio absorbs the shock. This held reasonably well from the mid-1990s through 2021, a period of declining inflation and falling interest rates.

But the correlation is not a constant. Molenaar et al. (2024) document that inflation levels and real interest rates are the primary drivers of stock-bond correlation direction. When inflation shocks dominate market pricing, rising discount rates compress both equity valuations and bond prices simultaneously. The hedge stops hedging.

McMillan (2026) confirms this with evidence of switching behaviour across G7 countries: the stock-bond correlation flips between positive and negative regimes, with a renewed move toward positive correlation in recent years for many markets. A portfolio that treats the covariance structure as fixed inherits a risk it cannot see in calm periods and cannot escape during stress.

Why regimes matter

The deeper issue is that financial markets do not evolve through a single stationary process. Growth momentum, inflation pressure, policy stance, volatility, and risk appetite combine into distinct states that recur over time. A portfolio that holds the same weights through expansion, stagflation, recession, and crisis is implicitly betting that the average condition is close enough to any given condition. That bet worked for decades. It does not work when the environment turns nonlinear.

De Longis and Ellis (2022) make this point with particular clarity. Their macro regime framework shows that the return contribution of equity, credit, duration, and commodity risk premia changes materially across business cycle phases. The same allocation cannot be equally suited to all of them.

The practical consequence is straightforward. A portfolio should be treated as a conditional system. The question is not whether a balanced portfolio is diversified in the abstract. The question is whether its exposures are coherent with the regime currently governing inflation, growth, volatility, and funding conditions.

What regime-aware allocation looks like

Urinson organises the approach around four signal families that inform regime classification:

Macro and policy signals include inflation, interest rates, yield curve shape, and growth indicators. These are the slowest-moving inputs but carry the most information about which asset classes are structurally rewarded or penalised.

Valuation and fundamental signals provide context for how vulnerable or resilient an asset class is within a given macro state. A high-CAPE equity market entering a tightening cycle faces a different risk profile than a cheap one.

Sentiment and behavioural signals help identify crowding, stress, and potential turning points. Taguchi et al. (2023) show that financial text polarity indexes can produce useful rebalancing signals when embedded in a systematic process.

Technical and trend signals refine timing. They reduce the cost of entering or exiting positions too early, which matters in practice because being right about the regime but wrong about timing can still produce painful drawdowns.

No single signal family is sufficient. The allocator works with an ensemble whose purpose is to reduce classification error, improve persistence, and prevent overreaction to noise.

A minimal case study

The paper includes a deliberately simple backtest to illustrate the concept. A strategy rotates between the S&P 500 (SPY) and long-term Treasuries (TLT) based on a VIX-derived volatility regime signal. Low-volatility regime: hold equities. High-volatility regime: rotate into duration. Allocation changes with a one-day lag.

Over 2001 to 2022, this produced a terminal wealth of approximately $4,234 from a id="post-body" class="prose prose-lg max-w-none pb-16" style="font-family: 'Inter', sans-serif; line-height: 1.8; color: #374151;" data-astro-cid-xnoknld4>00 starting point, compared to roughly $879 for buy-and-hold equities and $549 for static 60/40. The maximum drawdown was about 19.5%, versus 55% for the S&P 500 during the global financial crisis. The Sharpe ratio was approximately 1.22 (Shu et al., 2024).

These numbers deserve context. The strategy is intentionally extreme: it goes all-in on one asset at a time. A production portfolio would use graded transitions, broader opportunity sets, and explicit risk budgets. The backtest also covers a specific historical period that included two severe equity drawdowns (2008 and 2020) where the regime switch paid off handsomely. Different periods with different regime dynamics would produce different magnitudes.

Still, the core result holds: a regime-sensitive rule that detects stress transitions early enough can materially change the distribution of portfolio returns, even in a minimal two-asset setting.

The limits are real

Regime identification is noisy. Macro data arrive with lags and revisions. Signals conflict with each other. Online classification in real time looks nothing like the clean ex-post labeling of a historical study. Whipsaw is a persistent threat when transitions are fast and market narratives reverse abruptly.

Shu et al. (2024) are explicit about these constraints. Their evaluation applies 10 basis points of one-way transaction costs and warns against relying on in-sample state labeling for model selection. These are not theoretical concerns. They are the difference between a backtest that looks impressive and a live strategy that survives.

There is also a behavioural dimension. Regime-aware allocation often demands decisions that feel wrong in the moment. Reducing equity exposure after a euphoric rally, or adding risk when fear is widespread, requires a level of procedural discipline that most investors underestimate. A rules-based framework helps convert those moments from emotional dilemmas into process decisions.

What this means for systematic investing

The argument in this paper is not that 60/40 is useless. It is that 60/40 is a special case: a portfolio optimised for a particular macro regime that happened to persist for an unusually long time. When that regime ends, as it did in 2022, the portfolio's defensive properties can vanish precisely when they are needed most.

The alternative is not prediction. It is conditional design. Build a portfolio that adjusts its exposures based on observable macro and market state, within disciplined boundaries, with explicit governance around when and how changes occur. That architecture does not eliminate losses. It reduces the probability of catastrophic losses in environments where static allocation is structurally exposed.

For anyone running systematic strategies, the question is no longer whether to incorporate regime awareness. It is how to do it without introducing more noise than signal.

References:

de Longis, A. and Ellis, D. (2022) 'Tactical Asset Allocation, Risk Premia, and the Business Cycle: A Macro Regime Approach', The Journal of Portfolio Management, 49(4), pp. 103-126. doi: 10.3905/jpm.2022.1.456.

McMillan, D. G. (2026) 'Stock-bond return correlation: Understanding the changing behaviour', Journal of International Financial Markets, Institutions and Money, 106, 102242. doi: 10.1016/j.intfin.2025.102242.

Molenaar, R., Senechal, E., Swinkels, L. and Wang, Z. (2024) 'Empirical Evidence on the Stock-Bond Correlation', Financial Analysts Journal, 80(3), pp. 17-36. doi: 10.1080/0015198x.2024.2317333.

Reddy, S. and Smith, D. (2025) 60/40 strategy regains strength. State Street Investment Management. Available at: https://www.ssga.com/us/en/institutional/insights/mind-on-the-market-03-october-2025 (Accessed: 12 April 2026).

Shu, Y., Yu, C. and Mulvey, J. M. (2024) 'Downside risk reduction using regime-switching signals: a statistical jump model approach', Journal of Asset Management, 25, pp. 493-507. doi: 10.1057/s41260-024-00376-x.

Taguchi, R., Sakaji, H., Izumi, K. and Murayama, Y. (2023) 'Constructing Sentiment Signal-Based Asset Allocation Method with Causality Information', New Generation Computing, 41(4), pp. 777-794. doi: 10.1007/s00354-023-00231-4.

Urinson, M. (2026) 'The End of 60/40: Building Resilient Portfolios with Regime and Signal Intelligence', Universal Library of Business and Economics, 3(1), pp. 102-106. doi: 10.70315/uloap.ulbec.2026.0301013.

Disclaimer: Educational and informational purposes only. Past performance does not guarantee future results. Not investment advice.