Implied correlation

Definition

Implied correlation is the average pairwise correlation between index members that the options market is pricing — extracted from the arithmetic link between index implied volatility and single-stock implied volatilities. Index variance is member variances plus correlation terms; with index IV and member IVs observable, the implied correlation solves the equation. It is the options market’s live estimate of “how much do stocks move together,” and the documented finding (Driessen et al) is that it is persistently priced ABOVE realized correlation — a correlation risk premium.

How it works / structure

• The arithmetic: index variance = Σ wᵢ²σᵢ² + Σᵢ≠ⱼ wᵢwⱼσᵢσⱼρ̄; index IV vs the weighted member-IV basket implies ρ̄ (Cboe’s COR indices publish standardized versions). Index IV BELOW the member average is diversification being priced; the gap’s size is the correlation estimate.
• The premium finding (Driessen-Maenhout-Vilkov): implied correlation exceeds subsequently realized correlation on average — buyers of index protection pay for correlation insurance (crashes are when correlations converge — risk-correlation-exposure); selling that insurance (strategy-dispersion economics) earns the premium and holds the crash risk, the standard insurance-shaped return profile.
• The regime signal (engine-relevant): implied correlation spikes toward 1 in stress (macro dominates everything — regime-volatility companion series) and falls in calm, dispersion-rich markets (earnings seasons, single-name catalysts); its LEVEL calibrates how much diversification the market currently believes in (port-diversification-math priced live).
• Reading pairs: index IV low + member IVs high = low implied correlation (single-name event risk without macro fear); both high = macro regime; the decomposition separates “which kind of volatile.”

When it applies

Dispersion-trade evaluation (the premium IS this series’ gap to realized); portfolio-hedge selection (high implied correlation makes index hedges expensive relative to single-name hedges and vice versa); regime classification (a correlation input alongside vol level and term structure); diversification honesty (when the market prices correlation at 0.8, calm-period portfolio math claiming 0.3 is arguing with the price of insurance).

Risk profile & failure modes

• Estimate, not observable: implied correlation inherits every IV measurement noise doubled (index and members); short-window readings are jumpy — level and trend, not tick precision.
• Premium ≠ free: the documented gap compensates crash convexity; harvesting it concentrates loss in correlation-spike events exactly like every insurance premium in this KB (indicator-realized-vs-implied-vol family).
• Composition drift: mega-cap concentration changes the index-member arithmetic (a few names dominating weights makes “average pairwise correlation” less meaningful) — documented interpretive caveat in concentrated-index eras.
• Signal crowding: correlation-regime reads are widely followed; the series informs risk posture, not standalone entries.

Evidence & limits

The extraction arithmetic is standard; Cboe publishes methodology-documented indices; Driessen et al (2009) is the peer-reviewed premium evidence, with successor literature debating magnitude and time variation. The premium’s insurance interpretation is the standard account; alternative explanations (demand pressure, segmentation) are studied — labeled ongoing.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Implied correlation above its 90th percentile coincides with realized index vol above its 80th percentile within a month in most episodes (stress-coincidence check)” — falsified by the paired history.
• “The implied-minus-realized correlation gap remains positive over rolling years (premium persistence)” — falsified by the long-window measurement.

Cross-references

• The trade built on it: strategy-dispersion
• The inputs: opt-implied-volatility; the premium family: indicator-realized-vs-implied-vol
• The risk it prices: risk-correlation-exposure, port-diversification-math
• The regime layer: regime-volatility