Dispersion Trading

An index moves less violently than the stocks inside it. That's not luck, it's Diversification: on any given day some stocks rise while others fall, and the offsetting moves cancel out, so the basket is calmer than its parts. Dispersion trading turns that everyday fact into a tradeable bet, specifically, a bet on how much the stocks move together. It's really a trade on correlation wearing a volatility costume.

The setup is a pair of positions: volatility on the index against volatility on its individual components. If you sell index volatility and buy single-stock volatility, you're "long dispersion", you win when the stocks scatter (low correlation) and lose when they move in lockstep (high correlation). Flip both legs and you're short dispersion, betting the opposite.

Why an index is calmer than its stocks

The math behind the whole trade is the variance of a portfolio. For a basket, the index variance is built from every stock's variance and every pair's covariance:

$$\sigma_{\text{index}}^2 = \sum_i w_i^2 \sigma_i^2 + \sum_{i \ne j} w_i w_j\, \rho_{ij}\,\sigma_i \sigma_j .$$

The symbols: $w_i$ is stock $i$'s weight in the index, $\sigma_i$ its volatility, and $\rho_{ij}$ the correlation between stocks $i$ and $j$. The first sum is each stock's own wobble; the second is the cross-terms, and they depend on correlation. If every $\rho_{ij} = 1$ (everything moves together), the index vol equals the weighted-average stock vol. Any correlation below 1 shrinks that second sum, so the index vol comes out lower. The bigger the gap between average single-stock vol and index vol, the lower the correlation the market is pricing.

The shortcut: implied correlation

Traders collapse that whole formula into one number. Assuming roughly equal weights and similar vols, it simplifies to a clean rule of thumb:

$$\sigma_{\text{index}} \approx \sqrt{\rho}\;\; \sigma_{\text{avg}},$$

where $\rho$ is the average pairwise correlation and $\sigma_{\text{avg}}$ is the typical single-stock vol. Rearranged, the correlation the options market is implying is

$$\rho_{\text{implied}} \approx \left(\frac{\sigma_{\text{index}}}{\sigma_{\text{avg}}}\right)^2 .$$

This is the price you're trading against. If you think realized correlation will come in below the implied number, you go long dispersion; if above, short.

Worked example

An index's options imply a volatility of 18%. The average implied vol across its component stocks is 30%. What correlation is the market pricing?

$$\rho_{\text{implied}} \approx \left(\frac{0.18}{0.30}\right)^2 = 0.6^2 = 0.36 .$$

So the market is baking in an average pairwise correlation of about 0.36. Now suppose your research says these stocks have been drifting apart, earnings-driven, sector-rotating, and you expect realized correlation nearer 0.25. You go long dispersion: sell index straddles (short index vol) and buy straddles on the components (long single-stock vol), sized so the two legs roughly cancel on volatility level and leave you exposed only to correlation.

If correlation indeed realizes at 0.25, the components will have moved more, relative to the index, than the 0.36 you traded against. Your long single-stock volatility gamma earns more than your short index volatility loses, and you collect the difference. If instead a macro shock hits and everything sells off together, correlation snaps toward 1, the index moves as much as its stocks, and the trade loses, badly.

Why the edge exists, and why it's dangerous

Long dispersion tends to be a positive-carry trade in calm times: index options are usually a touch expensive relative to single-name options because investors buy index puts for portfolio insurance, bidding up index vol (and implied correlation) above what tends to realize. Harvesting that is closely related to the variance risk premium and to volatility arbitrage generally.

But the risk is specific and vicious: correlations spike in crashes. In a panic, Diversification evaporates, everything falls together, so implied correlation jumps toward 1 exactly when you're short it. The long-dispersion trade that dripped in profit for months can give it all back in a single bad week, which is why it's often described as short a rare, correlated disaster.

What makes it hard to run

• Lots of moving parts. You're trading options on the index and on dozens of single names, each with its own spread and liquidity. Costs pile up fast, and thin single-name options can be hard to fill.
• Keeping it correlation-pure. The two legs must be balanced so you're left with correlation exposure and not an accidental bet on the level of volatility, or on a handful of names. Rebalancing as vols drift is constant work.
• Weights and composition drift. Index membership and weights change; a clean replication today is stale after a rebalance.
• Single-stock jumps. An earnings surprise or takeover in one name moves its vol without touching correlation, adding noise the model doesn't want.