Regime-aware, tail-risk-sensitive portfolio construction

---

This article examines a dynamic approach to portfolio construction and regime-aware investing where diversification could fail during market structural shifts. It draws on six source articles: When Diversification Fails by Sébastien Page and Robert A. Panariello; Diversification during Hard Times by Najah Attig  and Oumar Sy; Asset Allocation Models and Market Volatility by Eric Jacquier and Alan J. Marcus; International Asset Allocation with Regime Shifts by Andrew Ang and Geert Bekaert; A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle by James Hamilton; and Why Static Portfolios Fail When Risk Regimes Change by Bruno Luiz Buriozzi.

---

The last decade has challenged one of the fundamental assumptions underlying portfolio construction: that historical relationships between asset classes provide a reliable guide for future allocation decisions.

During the COVID-19 liquidity market shock of 2020, diversification failed because liquidity disappeared, investors sold risky assets simultaneously and correlations rose as investors sought cash. Two years later, diversification failed again, but for an entirely different reason. Inflation, aggressive monetary tightening and rising interest rates caused stocks and bonds to decline together, producing one of the worst years on record for the traditional portfolios. Despite similar outcomes, these failures were driven by fundamentally different factors. They were not the result of poor security selection or inadequate diversification, rather of relying on static assumptions about risk in a market environment where risk evolves over time.

Diversification is conditional and time-dependent. Regimes are not temporary deviations around a single average state; they represent structurally different macroeconomic fundamentals and environments with respect to expected returns, volatilities, correlations, liquidity conditions, factor exposures and downside risks, and portfolio weights must change with them. Regime shifts and left-tail correlation rise are closely intertwined, and a regime-aware, tail-risk-sensitive framework can identify when the assumptions behind the current portfolio become unreliable in order to adjust exposures before static diversification leads to self-enforced deleveraging.

Risk changes through regimes

Financial markets are characterized by structural shifts and tend to fluctuate between low-volatility and panic-driven high-volatility states that fundamentally alter the relationships that matter most to portfolio construction—stock-bond correlation, equity-credit correlation, cross-country correlation, industry correlation, liquidity premia and volatility. While traditional strategic allocation frameworks often assume that long-run averages are sufficient guides for future risk, true portfolio risk is often determined by extraordinary conditions when these relationships become unstable.

James Hamilton’s pioneering work introduced the concept that economic time series transition between distinct regimes rather than evolving smoothly. Instead of assuming a single stable data-generating process, regime-switching models recognize that economies alternate between expansion and contraction, with each state exhibiting different statistical characteristics. This insight extends to portfolio management because the relationship among asset returns also changes as macroeconomic conditions evolve.

Rather than viewing risk as a constant, we should recognize that volatility, correlations, expected returns and liquidity are conditional on the prevailing economic regime.

This means that portfolio risk should be viewed as state-dependent. A portfolio that appears diversified in one regime may become concentrated in another if its assets become exposed to the same macro factor. Each regime requires different defensive positioning, illustrating why portfolios optimized for one environment may perform poorly in another. Regime-aware investing therefore begins with identifying the prevailing regime and recognizing that the investment opportunity set itself changes over time.

Rather than viewing risk as a constant, we should recognize that volatility, correlations, expected returns and liquidity are conditional on the prevailing economic regime.

Correlations rise during periods of market stress

A change in regime goes beyond market volatility. Correlations between risky assets typically increase when new risks emerge (i.e., the COVID-19 pandemic) and during periods of market stress as systematic risk overwhelms idiosyncratic factors, reducing the effectiveness of diversification precisely when investors need it most. Left-tail correlations are consistently higher than those observed under normal market conditions, indicating that portfolios often contain substantially more downside risk than suggested by historical averages. The asymmetric relationship between the left-tail (downside) and right-tail (upside) correlations can adversely affect the benefits of portfolio diversification.

Jacquier and Marcus explain an important mechanism behind this correlation breakdown. When market-wide factor volatility rises, systematic risk becomes more important relative to idiosyncratic risk. Assets that normally differ due to sector, country or security-specific characteristics begin to move together because a common factor dominates their return variation. A large portion of the variation in correlation structures can then be attributed to variation in market volatility.

Diversification often appears effective when measured across the entire return distribution, but the investor’s real vulnerability lies in the left tail.

This has direct implications for risk modelling. A covariance matrix[1] estimated from full-sample or trailing historical data can understate portfolio risk if it does not adjust for the current volatility regime. Volatility is not only an input into risk; it is also a signal about how diversification itself may change, A regime-aware model should expect correlations among risk assets such as stocks and bonds to rise when market volatility increases are driven more by broad macroeconomic or liquidity shocks, such as inflation and interest rates, than by business cycles and risk appetites. Page and Panariello show that full-sample correlations are misleading and should not be used in risk models without stress-testing correlation assumptions and adding tools such as downside risk measures and scenario analysis. Full sample correlation is an average of extremes, and conditional correlation reveals how, during crises, diversification across risk assets almost completely disappears. Diversification often appears effective when measured across the entire return distribution, but the investor’s real vulnerability lies in the left tail. During selloffs, correlations across risk assets tend to rise; during rallies, correlations may fall. This is the opposite of what investors want. A portfolio can appear efficient on average while being exposed to severe loss when risk assets move together.

A regime-aware portfolio construction framework

Expected returns, volatility and correlations vary across regimes, and portfolio weights should depend on the current regime. It is worth noting that identifying the current regime or its characteristics is no easy task. Portfolio construction, therefore, becomes a dynamic process in which asset allocation adapts as the probabilities of different economic regimes change. Such an approach requires continuously reassessing whether the assumptions underlying the strategic allocation remain valid. This implies that diversification should follow the dominant source of return variation. For example, if the shock is regional, country diversification may matter most. If the shock is sector-specific, industry diversification may matter most. If the shock is inflation or rates, duration exposure may stop diversifying equities. If the shock is liquidity, even assets with different fundamentals may sell off together. The portfolio construction process must map assets to their underlying drivers rather than assume that labels such as “equity,” “bond,” “credit,” “real estate” or “alternative” are sufficient. For example, Merton (1974) defined a corporate bond as a combination of a risk-free bond and a short put position on the company’s assets; as the probability of default increases, the risk characteristics of debt approach those of unlevered equity. Hedge funds are typically short volatility and liquidity risk, which is comparable to writing an option on the equity index; this justifies the jump in left-tail equity beta during financial crises. Private assets are exposed to many of the same factors that drive stock and bond returns.

In crisis regimes, when volatility and correlations typically rise,
the priority becomes capital preservation.

In stable regimes, when volatility is low and correlations are behaving as expected, there is less need for defensive or regime-driven adjustments, and portfolios can remain closer to their strategic allocations. Diversification across asset classes, regions, industries and factors is likely to be more reliable, and security selection may matter more. In crisis regimes, when volatility and correlations typically rise, the priority becomes capital preservation.

A wide variety of portfolio optimization methodologies directly address the non-normal left-tail risk, with the most flexible being full-scale optimization. Looking beyond diversification, to manage portfolio risk, investors should leverage tail-aware analytics prior to making trading and portfolio construction decisions and calibrate their risk tolerance accordingly. Tail risk hedging with equity put options or proxies and explicit downside protection could become increasingly important for assets with non-linear payoffs. Managed-volatility overlay strategies that scale down risk assets when volatility is high and offset tail-risk correlation spikes, defensive momentum strategies with risk factors that embed short positions, and selective use of derivatives that reshape the left tail all provide better left-tail protection than traditional diversification.

A practical framework for regime-aware portfolio construction can thus include the following:

1. Identify the prevailing macroeconomic regime by monitoring indicators of growth, inflation, interest rates, liquidity, monetary policy and financial conditions
2. Determine which risk factors dominate market behavior under the current regime
3. Update forward-looking assumptions for expected returns, volatility, correlations and downside dependence using conditional rather than historical estimates
4. Optimize the portfolio using these regime-specific inputs, incorporating scenario analysis and stress testing to evaluate resilience to adverse outcomes
5. Monitor regime-transition indicators continuously and adjust portfolio exposures as evidence suggests that a new regime is emerging

This framework moves portfolio construction toward a more dynamic, adaptive process that recognizes the structural changes of markets over time.

Conclusion

Portfolio construction must evolve from a historical analysis to an adaptive, forward-looking, dynamic process. This process recognizes that markets shift between states, volatility contributes to correlations, left-tail dependence overwhelms long-term averages and diversification, and the best method integrates the underlying driver of the crisis and asset returns into decision analytics.

Regime-aware, tail-risk-sensitive portfolio construction strengthens the principles of diversification or optimization by acknowledging that the investment environment is structurally shifting and dynamic, and portfolios should adapt as the nature of risk changes, moving the discussion from reactive risk management to proactive portfolio construction.

---

Written by

Sanaz Danielle Fotoohi, CFA, MBA, AFM, is a volunteer member of CFA Society Toronto’s Strategic Content Committee and Editorial Committee, and a contributor and regular features editor for The Analyst.

[1] A table that measures the variance of multiple variables, i.e., how the returns of unique assets move in relation to each other.