Why Strategy Correlations Blow Out in Stress

August 2007 is the standard example: dozens of quant equity funds running strategies that had shown near-zero correlation to each other for years suddenly all lost money on the same days, in the same direction, at the same time. None of the strategies had changed. What changed was that one or more large funds needed to delever fast, started selling the same crowded positions those strategies shared, and every fund holding similar positions got hit by the same selling pressure regardless of how "different" its strategy looked on paper.

This is the general pattern behind why strategy correlations blow out in stress: calm-market correlation measures the relationship between strategies' investment theses, but crisis-period correlation is often driven by something else entirely — who else holds the same positions and whether they're being forced to sell.

The forced-deleveraging mechanism

Most strategies that blow out together in stress share a specific chain of events: a shock causes losses somewhere, that loss triggers margin calls or risk-limit breaches, the affected fund sells its most liquid positions first to raise cash, and if many funds hold similar positions (because many funds independently discovered the same statistically cheap trades), that selling pressure hits all of them simultaneously. The correlation that shows up isn't between the strategies' logic — it's between their investor bases and their liquidity profiles.

Worked example

A firm runs two sleeves it estimates at 0.10 correlation in normal markets: a statistical-arbitrage equity sleeve and a merger-arbitrage sleeve. During a sharp, unrelated credit market shock, several large multi-strategy funds facing margin calls sell their most liquid holdings across all books at once, including stat-arb longs and merger-arb targets that happened to be widely held across the industry. Over that one week, the firm's two sleeves — whose underlying theses had nothing to do with each other — both lose money simultaneously, realizing a correlation of 0.6 for that week alone, a number the calm-period estimate gave no warning of.

What this means in practice

Because this kind of correlation spike is driven by crowding and forced selling rather than by strategy logic, the fix isn't better strategy diversification alone — it's checking how liquid each sleeve's positions are and how widely held they are across the industry. A sleeve trading illiquid, uncrowded names is less exposed to this mechanism than one trading liquid, widely-held names, even if the two look equally "diversifying" on a calm-period correlation matrix.