When Your Hedge Correlation Breaks
A hedge built on a historical correlation is only as good as that correlation staying stable — and correlations between a position and its hedge can shift or invert exactly when the hedge is needed most.
Prerequisites: Choosing The Hedge Instrument
A desk hedges a gold-miner stock by shorting gold futures, because for years the two have moved together closely — the miner's revenue is basically leveraged gold exposure. Then a labor strike hits the mine specifically, or gold rallies on a flight-to-safety bid while mining-sector equities sell off broadly on unrelated growth fears. The correlation the hedge was built on simply stops holding, and the "hedged" position now loses money on both legs at once. This isn't a bug in the hedge — correlations are estimated from history, and history is not a promise about the future.
Why correlations shift
Correlation between two assets isn't a fixed physical constant; it's a statistical property of a particular period, and it changes as the underlying drivers of each asset's price change. A stock and its sector ETF might correlate closely in normal times because both respond to the same broad market moves, but diverge sharply when something company-specific dominates — an earnings miss, a lawsuit, a regulatory action. Correlations also tend to shift during market stress: in a genuine crisis, many previously uncorrelated or negatively correlated assets can suddenly move together (correlations rise toward 1), while pairs that were reliably correlated in calm markets can decouple.
Spotting it before it costs you
Rolling correlation — computed over a moving window, say the trailing 60 or 90 trading days — is the standard tool for watching a hedge relationship for signs of drift. A rolling correlation that has quietly fallen from 0.85 to 0.4 over the last month is a signal the hedge ratio built on the old relationship no longer reflects reality, even if nothing dramatic has happened yet on any single day.
| Signal | What it suggests |
|---|---|
| Rolling correlation trending down steadily | Structural drift — the two assets' drivers are diverging |
| Correlation sharply negative for one day, then recovers | Likely a one-off event, not a broken relationship |
| Correlation spikes toward 1 in a broad selloff | Normal crisis behavior, not specific to this pair |
The scatter below shows what a stable hedge relationship looks like at different correlation levels — drag the correlation down toward zero to see the points spread out and the fit line lose predictive power, which is exactly what a rolling-correlation chart shows happening to a real hedge pair when the relationship drifts.
What this means in practice
A hedge should be monitored on an ongoing basis, not set once and trusted indefinitely. Desks that run cross-asset or proxy hedges typically track the rolling correlation (or a regression-based hedge ratio) alongside the position and re-size or replace the hedge when it drifts materially, rather than waiting for a bad day to discover the relationship had already broken down weeks earlier.
A hedge ratio estimated from historical correlation can become stale as the relationship between the position and the hedge instrument drifts. Tracking rolling correlation is how a desk catches that drift before, rather than after, the hedge fails on a day it was needed.
The worst time for a correlation to break is exactly the tail event the hedge was meant to protect against — idiosyncratic shocks that hit one leg and not the other are common precisely during the sharp, fast-moving events where a hedge matters most.
Further reading
- Ang, Asset Management (ch. 14)