Dispersion Trading: Index Volatility, Stock Volatility, and Correlation

For educational purposes only; not investment advice.

Direct answer

Dispersion trading takes opposite volatility positions in an equity index and a weighted basket of its constituent stocks. A conventional long-dispersion position sells index volatility and buys constituent volatility; the reverse is short dispersion. The purpose is to isolate part of the gap between index variance and single-stock variance, which is strongly influenced by how closely constituents move together.

It is not a simple comparison of quoted implied volatilities and it is not riskless arbitrage. A working book may contain dozens of options plus repeated Delta hedges, while retaining exposure to volatility levels, correlation, skew, jumps, changing index weights, liquidity, settlement, and model choices.

Variance and implied correlation

If index return is approximated by weighted constituent returns, its variance is

σ_index² ≈ Σ(w_i²σ_i²) + 2Σ(i<j)(w_iw_jσ_iσ_jρ_ij)

The first term is weighted single-stock variance; the second contains pairwise covariance. Holding weights and stock volatilities fixed, greater correlation generally raises index variance. Index-option and constituent-option implied volatilities can therefore be used to estimate an average implied correlation, but that estimate compresses a full correlation matrix into one model-dependent number.

Cboe describes long dispersion as selling an at-the-money SPX straddle and buying weighted at-the-money straddles on constituents. Professional implementations may instead use option strips or variance exposures. Matching share weights or contract counts is insufficient: multipliers, notionals, Vega, Gamma, maturities, and variance sensitivity differ. Initial neutrality also drifts as prices, Greeks, and index composition change.

Correlation changes without stock volatility changing

Consider ten equal-weight stocks, each with annualized volatility of 30%, and assume one common pairwise correlation. The variance formula simplifies to

σ_index² = 0.1×0.30² + 0.9×0.30²×ρ.

At correlation 0.20, index volatility is about 15.9%. At correlation 0.80, it is about 27.2%. No constituent volatility changed; stronger common movement alone raised index volatility substantially.

Now suppose a long-dispersion book buys stock variance and sells index variance because market prices imply correlation near 0.58, while the manager expects 0.35. If several companies jump 7%, 9%, and 6% after unrelated earnings news while their directions offset in the index, the stock legs may outperform. Profit is still not assured: the stock options may have been purchased above realized variance, index downside skew may become more expensive, or hedge and execution costs may consume the gap.

In a stress event, constituents can fall together and correlation can approach one. The short index-volatility leg may then lose rapidly, and long stock Gamma may not offset it because weights, surfaces, trading times, and liquidity do not match perfectly.

Risk checklist

Convert every leg to a consistent variance, Vega, Gamma, multiplier, maturity, and currency basis.
Use current index weights and model additions, deletions, rebalances, splits, mergers, and special dividends.
Compare full volatility surfaces, not only at-the-money volatility; index put skew can carry a large crash-protection premium.
Maintain an event calendar for earnings, litigation, mergers, halts, and delistings that can dominate one stock.
Stress correlation at 0.20, 0.90, and near one, plus single-stock and marketwide jumps.
Separate P/L from index variance, constituent variance, covariance, Delta hedging, surface changes, and costs.
Account for dozens of bid-ask spreads, market impact, commissions, borrow, financing, and failed simultaneous execution.
Reconcile cash-settled index options with physically settled stock options, including last trading time and exercise rules.
Treat a reduced constituent basket as tracking error, not a full replication.

Common misconceptions

“Dispersion is a pure correlation bet.” Correlation is central, but volatility level, skew, jumps, weights, and costs remain.
“Stock index weights determine option quantities.” Option sensitivities and contract terms must also be normalized.
“Low implied correlation must rise.” Historical averages do not determine future regimes or executable value.
“Long dispersion is protected in a crash.” Correlations and index skew can rise sharply together.
“Initial Vega neutrality lasts.” Spot, time, volatility, and constituent weights continually change the hedge.
“A profitable book proves the correlation forecast.” P/L attribution may show Gamma, Vega, or execution was the actual source.