Dispersion Trading Signals from Sentiment Divergence

Dispersion trading profits from the difference between implied volatility of an index and the volatility implied by its component stocks. When index IV is high relative to component IVs, the index has embedded "tail risk" premium. Identifying periods when dispersion is mispriced requires understanding sentiment divergence: when do sentiment and expectations diverge between index and constituents?

The Dispersion Trade

A classic dispersion trade:

• Sell index volatility (e.g., short index straddle)
• Buy component volatility (long straddles on constituent stocks)
• Delta-hedge the position

If realized index volatility is less than implied, and realized component volatility exceeds component-implied levels, the trade profits. Conversely, if dispersion contracts (index IV falls relative to components), the trade loses.

Why Dispersion Exists

Index volatility reflects systematic (market-wide) volatility plus hedging demand for tail risks. Individual stock volatility reflects idiosyncratic risks plus systematic risk. In calm markets, dispersion is low; systematic vol dominates. In crisis, dispersion spikes (everyone's selling, correlated decline).

Additional reasons for dispersion variation:

• Correlation changes: as correlation increases, index vol rises (components move together)
• Hedging demand: investors buy index puts for protection, raising index vol
• Event risk: some companies have upcoming catalysts (earnings, M&A) raising their IVs

Sentiment Divergence as Predictor

When market sentiment diverges—some components very pessimistic, others optimistic—dispersion typically widens. When sentiment is uniform (all pessimistic or all optimistic), dispersion narrows.

Machine learning can predict dispersion changes by analyzing sentiment signals:

• Options flow: are investors buying more puts on index or components?
• Analyst sentiment: consensus revisions on individual stocks
• News sentiment: analyze news and social media tone for index vs stocks
• Realized correlations: recent price correlations between components

Feature Engineering for Dispersion Prediction

Features capturing sentiment divergence:

• IV skew divergence: is index skew more negative than component skew?
• Put-call ratio divergence: more put buying in index or components?
• Sentiment score variance: standard deviation of sentiment scores across components
• IV term structure divergence: shape difference between index and average component term structure
• Correlation of component IV with component returns: divergence from index

Predicting Dispersion Cycles

Dispersion follows cycles: periods of wide dispersion often precede narrow periods and vice versa. Temporal models (ARIMA on dispersion, or neural networks on vol surfaces) can forecast these cycles.

By predicting dispersion will widen, traders establish short-dispersion positions (sell index, buy components). By predicting dispersion will narrow, they establish long-dispersion positions (buy index, sell components).

Multi-Factor Modeling

Advanced models decompose dispersion into factors:

• Systematic volatility component
• Idiosyncratic volatility component
• Correlation component

Each component has different drivers and predictability. Models that separately predict each component often outperform single-variable models.

Practical Implementation

Dispersion trading requires:

• Access to index and component option data
• Ability to execute hedged positions (short index, long components)
• Careful hedging and rebalancing to maintain delta neutrality
• Sophisticated risk management (dispersion can move against you quickly)

Conclusion

Machine learning can predict dispersion changes by analyzing sentiment divergence between index and constituents. These predictions enable profitable dispersion trading, a classic derivatives strategy enhanced by predictive modeling.