Dispersion trading

Definition

Dispersion trading sells index volatility and buys single-name volatility on the index’s members — long the names moving individually, short them moving together. Its P&L is a direct position on correlation: if members disperse (single-name moves cancel at the index level), the long single-name vol pays while the short index vol stays quiet; if everything moves together, the short index leg loses most. It is the trade built to harvest the documented correlation risk premium (opt-implied-correlation).

How it works / structure

• The classic structure (engine-parameterizable): short index straddles/variance vs long a weighted basket of member straddles (strategy-straddle building blocks), vega-weighted so the net position is approximately pure correlation (greek-vega balanced across legs); delta-hedged throughout (mgmt-delta-hedging — the P&L should come from realized vol spreads, not direction).
• The premium being harvested (Driessen et al): implied correlation persistently above realized — index options are expensive relative to member options because index protection embeds correlation insurance; the dispersion seller collects that gap on average and pays it back in correlation-spike events.
• Parameter surface: member subset (full replication vs liquid-name proxy — tracking error vs cost), tenor (near-dated earnings-rich windows maximize single-name dispersion), weighting (vega- vs theta-neutral variants), and the entry gate (implied correlation percentile — selling correlation when it’s already low has documented poor asymmetry).
• Simplified retail form: long a few single-name event straddles vs short an index-proxy straddle in matched tenor — the same exposure at rougher granularity and higher relative friction.

When it applies

Earnings seasons (scheduled single-name catalysts with muted index effect — dispersion’s documented seasonal habitat); high implied-correlation entry points (the premium is largest when macro fear has lifted index vol above the members’ aggregate); as a portfolio diversifier for short-vol books (its loss regime differs from plain short vol in calm-grind markets).

Risk profile & failure modes

• Correlation-spike loss (the defining tail): macro shocks converge correlations toward 1 — the short index leg loses faster than the long member legs gain; this is the insurance-premium payback, arriving exactly in crises (risk-correlation-exposure).
• Leg-management burden: many-legged delta-hedged books have real operational and friction costs — documented as the gap between paper and realized dispersion returns; the trade is capacity- and cost-sensitive.
• Member-selection basis: proxy baskets miss the names that actually disperse; full replication pays spread on illiquid members — the tracking/cost trade-off is structural.
• Crowding cycles: the premium compresses when dispersion capital is abundant (documented institutional cycles) — entry-gate percentiles keep the harvest honest.

Evidence & limits

Driessen et al (2009) and successors document the correlation premium the trade collects; practitioner literature documents the structure. Public performance data for pure dispersion is thin (institutionally traded) — the KB treats magnitude expectations as regime-dependent and friction-sensitive rather than citing a headline return.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Entered above the 80th implied-correlation percentile, vega-balanced dispersion is profitable in most quarters of the replay, with losses concentrated in correlation-spike months (premium-shape check)” — falsified by the replay distribution.
• “Earnings-season dispersion (member straddles over earnings vs index short) outperforms off-season dispersion per unit vega (seasonal-habitat check)” — falsified by the seasonal comparison.

Cross-references

• The priced quantity: opt-implied-correlation; the premium family: indicator-realized-vs-implied-vol
• The building blocks: strategy-straddle, mgmt-delta-hedging, greek-vega
• The tail it holds: risk-correlation-exposure