Correlations between asset classes are often described as if they were stable properties — historical values that can be plugged into portfolio construction models with confidence. In practice, correlations shift substantially over time, and they shift most dramatically during exactly the periods when their stability would matter most. Understanding this pattern is essential to reading portfolio diversification realistically rather than as a mathematical abstraction.
The stability illusion
Long-run average correlations between major asset classes are widely reported. US stocks and Treasury bonds have averaged a modest negative correlation over most historical periods. US stocks and international stocks have averaged a moderate positive correlation. US stocks and commodities have averaged near zero. These numbers are not wrong, but they are averages over long periods that mask substantial variation.
The variation is what matters for portfolio behaviour. A portfolio built on average correlations will behave close to expectations during normal periods but can behave dramatically differently during periods when correlations shift.
The crisis pattern
The most consistent pattern in correlation shifts is that correlations rise during periods of market stress. Assets that show modest correlation during normal times often move together strongly during crises. This is sometimes called "correlations going to one" — the observation that when markets are falling most sharply, many previously-uncorrelated assets fall together.
The 2008 financial crisis is the reference example. Over the six months from September 2008 through March 2009, the correlations between US stocks, international stocks, commodities, high-yield bonds, and some traditionally "alternative" asset classes all rose materially. Assets that provided meaningful diversification during 2005-2007 provided much less diversification during the crisis peak.
The 2020 pandemic sell-off showed a similar pattern in shorter compressed form. In late March 2020, the correlation between US stocks and Treasury bonds — traditionally the cornerstone of the 60/40 portfolio construction — briefly became positive rather than negative. Both assets sold off together as forced liquidations spread across the financial system. The pattern reversed within weeks but the temporary breakdown of the traditional stock-bond diversification was consequential for portfolios relying on it.
Why correlations rise in crises
The mechanism is well-understood. During normal periods, individual assets are driven partly by asset-specific factors and partly by common macroeconomic factors. The asset-specific factors reduce the aggregate correlation across the portfolio.
During crises, the common macroeconomic factors dominate. Every asset that would normally be affected differently by different economic conditions becomes affected primarily by the crisis-level economic condition. The asset-specific factors are overwhelmed by the common factor, and correlations rise.
A second mechanism operates through forced selling. During periods when leveraged investors face margin calls or fund redemptions, they must sell whatever assets they can, regardless of their views on those specific assets. This forced selling creates correlated declines across everything the forced sellers hold, even assets that would otherwise be uncorrelated.
The specific 2022 pattern
The 2022 experience showed a different correlation pattern from 2008. The 60/40 portfolio, which had provided diversification through most of the previous three decades, had its worst calendar year of the past several decades. Stocks and bonds both fell substantially — not because of a traditional risk-off crisis but because of a specific macro pattern (rising rates hitting both duration and equity valuations simultaneously).
The lesson: correlations shift not only during risk-off crises but also during specific macro regime changes. The stock-bond correlation depends on the underlying driver of market movements. When rates are the driver, stocks and bonds can move together. When earnings are the driver, they often move separately. Which regime prevails at any given moment is one of the more important determinants of portfolio behaviour.
The implications for portfolio construction
Three specific implications follow from the pattern of correlation shifts.
Effective diversification requires more than count. A portfolio with many holdings that correlate closely during stress provides less protection than a portfolio with fewer holdings whose correlations remain stable during stress. This is why "diversification" in retail portfolios often fails during crises — the diversification looks good on average correlation metrics but degrades during exactly the periods when it is most needed.
Alternative asset classes matter more than they seem. Assets whose correlations to traditional equity remain stable during crises (rather than rising) provide genuine diversification benefit that raw correlation numbers understate. This is one of the reasons managed futures, some hedge fund strategies, and various real assets have earned space in institutional portfolios despite modest expected returns on their own.
The 60/40 baseline is more variable than it appears. The traditional US 60/40 portfolio has provided historically strong risk-adjusted returns, but its performance in specific periods can vary substantially from long-run averages. Investors who committed to 60/40 in 2000 saw very different results in the following 5-year period than investors who committed in 2010.
The measurement problem
Correlation is typically measured over some past window — 1 year, 3 years, 5 years. The choice of measurement window matters. A one-year window captures more recent conditions but is more sensitive to any specific event in the window. A five-year window smooths out specific events but may miss regime shifts that have occurred in the interim.
For portfolio construction purposes, no single measurement window is obviously correct. Understanding that correlations shift and that recent measurements may not persist forward is a useful discipline even without knowing exactly which correlations to use.
What retail investors can practically do
Three practices help retail investors deal with the reality of shifting correlations.
Assume less diversification than the numbers suggest. Portfolios that appear diversified based on long-run average correlations often provide less diversification during specific periods. Assuming a somewhat higher effective correlation than average produces more conservative portfolio construction that is less likely to disappoint during stress.
Include exposures that show stable low correlation during crises specifically. Not all "alternative" investments provide crisis-period diversification, but some do. Understanding which asset classes have specifically shown stable low correlation during past crises helps inform portfolio construction beyond simple average correlation.
Maintain some cash. Cash correlations remain reliably at zero regardless of what other asset classes are doing. A modest cash allocation provides a specific form of diversification that survives every correlation regime. Institutional investors often maintain surprisingly large cash allocations for exactly this reason.
The rule to internalise
Correlations between asset classes are not stable properties; they are conditional relationships that shift substantially with market regime. The shifts tend to concentrate during exactly the periods when their stability would matter most for portfolio diversification. Understanding this is not an argument against diversification; it is an argument for understanding what diversification realistically provides and constructing portfolios accordingly. The retail investor who assumes long-run average correlations will hold during specific crisis periods is systematically overestimating their portfolio's crisis resilience.
Educational content only. Not investment advice.