Quant Memo
Core

Hedges That Go Stale

A hedge put on to offset a specific risk can lose its effectiveness over time as the underlying position or market relationship changes, silently leaving a portfolio exposed even though it still looks hedged on paper.

A hedge is only as good as the relationship it relies on, and that relationship can drift after the hedge is put on. A classic example is a stock position hedged with an options position sized to a particular delta: as the underlying price moves and time passes, the option's actual delta changes (this is exactly what gamma measures), so a hedge that was neutral on day one can become badly under- or over-hedged weeks later if nobody rebalances it. The same problem shows up with pairs and index hedges — a stock hedged against a sector ETF because the two were historically 90% correlated can find that correlation quietly drop to 60% after a change in the company's business, leaving the position exposed to risk the hedge was supposed to remove, even though the trade blotter still shows a "hedged" position.

The practical fix is scheduled review rather than "hedge and forget": re-checking hedge ratios and the underlying correlation or delta assumption on a regular cadence, and treating any hedge older than that review window with suspicion until it's re-verified.

Staleness tends to creep in fastest around events that change the underlying relationship itself — an earnings surprise that shifts a stock's sensitivity to its sector, a merger announcement, or simply enough elapsed time for gamma to accumulate on an options hedge — so many desks also trigger an unscheduled hedge review whenever such an event occurs, rather than waiting for the next calendar checkpoint.

A hedge's effectiveness depends on a relationship (a delta, a correlation, a beta) that can drift after the hedge is placed, so hedges must be periodically re-checked rather than assumed to still be doing their job.

Related concepts

Further reading

  • Taleb, Dynamic Hedging (1997)
ShareTwitterLinkedIn