Published on: 2026-09-07
Updated on: 2026-09-07
Assets often become more correlated during market crises because the forces driving prices become more concentrated. In calm markets, equities, bonds, commodities and currencies can respond to different fundamentals. Under stress, common macroeconomic shocks, funding pressure, deleveraging and shifts in risk appetite can dominate those differences, pushing realised cross-asset correlations higher and weakening diversification.
Correlation measures how closely the returns of two assets move together. A reading near +1 indicates strong movement in the same direction, near 0 suggests little consistent linear relationship, and near -1 indicates that returns tend to move in opposite directions.
Those relationships are conditional rather than permanent. In calm periods, a company may trade mainly on earnings, a bond on rates and credit quality, a commodity on supply and demand, and a currency on relative monetary policy. During stress, a dominant macroeconomic, funding or risk factor can overwhelm those individual drivers and cause several markets to react to the same information.
Realised cross-asset correlation can therefore rise abruptly even when long-run historical correlation remains moderate. Equities, high-yield credit, emerging-market assets and cyclical commodities are especially exposed because they often share sensitivity to growth, financial conditions and risk appetite. One of the main channels through which that common shock spreads is liquidity.
Market stress often creates an urgent demand for cash. Investors may face redemptions, margin calls, collateral requirements or funding obligations that have little to do with the fundamental quality of the assets they own.
When cash becomes the priority, highly liquid equities, government securities or other quality holdings may be sold simply because they are easier to convert into cash. Assets with little direct economic connection can then decline together because they are being sold for the same liquidity-related reason.
This can also explain why defensive assets sometimes weaken early in a crisis. Their long-run characteristics may remain intact, but an immediate need for liquidity can dominate short-term pricing. Correlation spikes can therefore reflect funding pressure as much as changing fundamentals.
Liquidity stress becomes more powerful when leverage turns discretionary selling into forced risk reduction.
When prices fall and volatility rises, leveraged portfolios may need to cut positions to remain within margin, funding or risk limits.
The process can reinforce itself. Lower prices raise measured portfolio risk, higher risk triggers position cuts, and those cuts create additional selling. Value-at-Risk limits, volatility-targeting strategies and margin requirements can all push exposure lower at the same time.
If many funds react to the same volatility shock, a bond, equity position and commodity exposure can all be sold even when their underlying fundamentals differ. The resulting correlation is partly mechanical, driven by common portfolio constraints rather than a new economic link between the assets.
This is why crisis correlation cannot be understood through macroeconomic narratives alone. Funding structure and portfolio rules can materially alter realised co-movement during stress.
A full-period correlation number can hide the relationship that matters most during a severe drawdown. Two assets may have only moderate average correlation across gains, losses and quiet trading, yet become much more positively correlated when markets fall sharply.
This asymmetry is often described as downside correlation or, in more extreme cases, tail dependence. CFA Institute research on international equity markets has documented significantly higher correlations in bear markets and argued that risk estimates should account for correlation behaviour in the tails rather than relying only on average covariance relationships.
The same idea appears in forward-looking markets. S&P Global research on option-implied correlation shows that implied correlations are higher at lower strike levels, meaning equity option markets price stronger expected co-movement under more adverse scenarios. A single average correlation estimate can therefore understate the dependency structure that markets expect in a downside event.
That helps explain why diversification can look robust over a long sample yet provide less protection during the observations that cause the largest losses.
The phrase “correlations go to one in a crisis” is a useful warning, not a literal rule. Correlations can rise sharply during stress, but the direction and size of the change depend on what causes the stress. CFA Institute research supports higher correlations in bear markets, not the claim that every asset pair mechanically converges to +1.
| Type of market stress | Typical correlation effect |
|---|---|
| Growth or recession shock | Equities, high-yield credit and cyclical assets may move more closely together, while high-quality government bonds may diversify the decline |
| Inflation shock | Equities and long-duration bonds can weaken together, making stock-bond correlation more positive |
| Liquidity shock | Correlations can rise across a broad range of assets as investors raise cash |
| Interest-rate shock | Rate-sensitive equities and long-duration bonds may decline together |
| Credit shock | Equities, high-yield bonds and financial assets can become more tightly linked |
Safe-haven behaviour is therefore conditional. Government bonds can rally during a growth-driven equity selloff if weaker activity is expected to bring lower rates, but they may fall alongside stocks when inflation or term-premium concerns drive yields higher. Gold can also benefit from defensive demand in one crisis and weaken in another if cash needs or rising real yields dominate.
The key question is which risk factor is driving the crisis and whether an asset is positively or negatively exposed to it.
The stock-bond relationship provides a clear example of correlation changing with the macro regime. IMF research found that for roughly two decades before the pandemic-era inflation shock, rising market volatility generally coincided with falling stock prices and rising bond prices. In the subsequent inflationary regime, stocks and bonds became more likely to sell off together during sharp stress episodes, weakening the traditional hedge.
MSCI has documented the same regime dependence. In one comparison, an equity drawdown of 34% coincided with a 5% bond gain, leaving a hypothetical 60/40 portfolio down 18%. Under a high-inflation selloff, equities fell 26% and bonds 14%, leaving the same mix down 21%.
Diversification did not permanently break. The macro regime changed. When falling growth and disinflation dominate, bonds can offset equity weakness. When inflation and rising yields are the source of stress, the same two asset classes can become positively correlated.
Rising correlation increases portfolio risk because covariance depends on both individual asset volatility and how closely assets move together. If several holdings become more positively correlated, losses are less likely to offset one another and total portfolio volatility can rise.
This creates a practical form of diversification failure. A portfolio may contain dozens of securities and still be concentrated in the same underlying exposures, such as growth, interest rates, credit conditions or liquidity. The number of holdings can therefore overstate the number of genuinely independent risks.
Calm-period correlation matrices can make that problem harder to see. A model calibrated mainly on normal conditions may underestimate losses when volatility and correlations rise simultaneously. The IMF has specifically argued that stress tests should incorporate correlation-breakdown scenarios because historical relationships can understate risk when traditional hedges weaken.
Scenario testing should therefore examine not only asset-price shocks but also changes in the relationships between assets. Stressed or downside correlations can reveal how a portfolio behaves when positions that normally diversify one another begin responding to the same factor.
Elevated correlations do not necessarily remain high once stress fades. As liquidity improves, forced deleveraging slows and uncertainty declines, individual fundamentals can regain influence. Companies can separate according to balance-sheet strength, credit markets can differentiate between issuers, and commodities can return to their own supply and demand drivers.
It is useful to distinguish structural correlation from stress-driven correlation. Structural correlation reflects lasting economic links, while stress correlation can be temporary and regime dependent. This helps explain why a long-run average may describe neither calm markets nor crisis conditions particularly well.
Correlations rise during many market crises because prices stop being driven primarily by separate asset-specific fundamentals and start responding to common macroeconomic, funding and risk factors. But the pattern depends on the shock: recession, inflation and liquidity crises can produce very different stock-bond and safe-haven relationships. Portfolio diversification should therefore be judged by underlying risk exposures and by stressed or downside correlations, not simply by the number of holdings or long-run historical averages.
