Correlation-based diversification metrics can be misleading during liquidity crises because correlations are not stable and become entangled with marketwide liquidity shocks. Academic and market research shows that co-movement between assets typically increases in extreme stress, eroding the benefits that historical correlation estimates imply.
Why correlations rise during liquidity crises
Empirical work by Francis Longin HEC Montréal and Bruno Solnik HEC Paris documented that correlations across equity markets increase substantially in extreme negative returns, a phenomenon sometimes called correlation breakdown. Robert Engle New York University Stern School of Business developed dynamic conditional correlation models that demonstrate correlations are time-varying and rise in turbulent periods. Yakov Amihud New York University Stern School of Business linked illiquidity to expected returns and showed how trading frictions amplify price moves. Together these findings indicate that rising correlations during crises are often driven by common liquidity channels: margin calls, forced deleveraging, concentrated market-making inventories, and sudden withdrawal of liquidity providers. These forces make seemingly independent assets move together not because fundamentals converged but because liquidity evaporated.
Consequences for portfolio construction and risk management
Relying solely on historical correlation matrices can understate tail risk and lead to portfolios that perform worse than expected when liquidity dries up. Practitioners who treat correlation as a fixed input may face amplified losses, concentrated exposures, and cascade effects like fire sales and widening bid-ask spreads. Human and territorial factors matter: smaller or emerging markets with thinner trading depth often experience stronger co-movement and slower recovery, while market microstructure and cultural trading norms shape how liquidity providers respond in stress.
Mitigating this requires combining time-varying correlations with liquidity-aware measures and stress testing. Dynamic models such as the DCC family from Robert Engle New York University Stern School of Business, liquidity proxies inspired by Yakov Amihud New York University Stern School of Business, and scenario-based exercises that impose funding and market liquidity shocks provide a more resilient framework. No single metric cures the problem, but integrating liquidity into correlation estimates and running reverse stress tests helps reveal vulnerabilities that static diversification metrics conceal.
In short, correlation-based diversification remains a useful concept, but during liquidity crises it must be augmented by liquidity-adjusted risk metrics, dynamic correlation modeling, and explicit scenario analysis to avoid misleading conclusions about protection and exposure.