Variance Dispersion Trade: Cleaner Correlation Bet

Key Takeaways

• Variance dispersion trade is short an index variance swap plus long a vega-weighted basket of single-name variance swaps; net exposure collapses to realized correlation.
• An SPX desk observes implied correlation at 0.50 vs 5-year realized average of 0.35; the $100k-vega book earns ~$1.4M when realized correlation lands at 0.32.
• A common mistake: mixing variance notional and vega notional across legs, the mismatch leaves the book unintentionally directional on the vol level.
• Variance-swap dispersion has constant vega (unlike straddle-based dispersion) and eliminates pin risk at expiry, making P&L decomposition cleaner and more tractable.

What It Is

The structure is short a variance swap on an index and long variance swaps on a vega-weighted basket of constituents. Both legs are sized so that a uniform 1-point rise in implied variance across the whole surface nets to zero. What remains is an exposure to the spread between realized index variance and realized basket variance, which collapses algebraically to realized correlation.

Compared with a straddle-based dispersion trade, variance-swap dispersion has four advantages. The vega is constant across the life of the trade, not concentrated near the money. The realized leg uses daily log returns, which are robust and model-free. Pin risk near expiry disappears. And the composition of P&L is easier to decompose into correlation and vol-of-vol components.

The price for these advantages is that single-name variance swaps can be illiquid outside the top 50 or so index constituents.

The Intuition

Recall the variance identity for an index with weights w_i, single-name variances sigma_i^2, and pairwise correlations rho(i,j):

sigma_index^2 = sum(w_i^2 * sigma_i^2) + 2 * sum over i<j of w_i * w_j * sigma_i * sigma_j * rho(i,j)

If you short a variance swap on the index at strike K_index and go long a basket of variance swaps at weighted strike K_basket, your P&L at expiry is:

P&L = Notional * (K_index - RealizedVar_index) + NotionalBasket * (RealizedVar_basket - K_basket)

Because the variance identity links realized index variance to realized basket variance via realized correlation, the terminal P&L is dominated by whether realized correlation lands above or below the level implied by the strike spread at inception.

Jacquier and Slaoui derive this formally and show that the P&L has an additional volatility-of-volatility term, which explains why a variance dispersion trade does not exactly replicate a correlation swap even with identical notionals.

How It Works

Implementation is a standard vol-arb workflow:

1. Compute implied correlation from current variance swap strikes. Implied correlation is approximately:

ImpliedCorr = (K_index - sum(w_i^2 * K_i)) / (2 * sum over i<j of w_i * w_j * sqrt(K_i * K_j))

1. Compare implied correlation to realized correlation over the recent window and against seasonal averages.

2. Size notional on both legs so that total vega is matched. A common rule of thumb is to size the basket leg at roughly 80 percent of the index leg vega because not every constituent is included.

3. Transact both legs simultaneously with the dealer to avoid legging risk.

4. Monitor daily. Variance swaps mark-to-market based on realized variance so far plus the current forward variance strike for the remaining tenor.

Banks including JPMorgan and Goldman Sachs published the original client-facing primers on the structure in the early 2000s, and the technique remains a staple of multi-strategy and volatility hedge fund books.

Worked Example

A desk observes S&P 500 1-year variance strike at 225 (15 percent vol squared). The weighted basket of the top 50 constituents trades at a weighted variance strike of 400 (20 percent vol squared). Implied correlation from the relationship above resolves to roughly 0.50.

Realized 1-year correlation over the prior five years has averaged 0.35, and the current macro regime looks quiet. The desk sells $100,000 vega of 1-year SPX variance at 225 and buys $80,000 vega of 1-year single-name variance on the basket at a weighted strike of 400.

At maturity, realized correlation lands at 0.32. Approximate P&L from the correlation spread (0.50 implied minus 0.32 realized) on the adjusted vega sizing nets out to a gain of roughly $1.4 million, ignoring transaction costs and counterparty spreads.

If instead a credit shock had driven realized correlation to 0.75, both legs would have settled with losses totaling several million dollars. The asymmetric payoff profile is a core characteristic of the trade and why dealers charge a premium over the fair strike derived from realized history.

Common Mistakes

1. Ignoring the volatility-of-volatility term. The AQR, GSAM, and Jacquier research all emphasize that variance dispersion P&L includes a volga component. Assuming the trade is a pure correlation bet overstates the precision of the hedge.

2. Skipping the single-name liquidity check. Variance swaps on top index names are quoted daily by multiple dealers. For mid-cap constituents the market is thin, bid-ask spreads can exceed a full variance point, and quotes widen dramatically in stress.

3. Mismatching notional conventions. Variance swaps are quoted either in variance notional or vega notional. Mixing the two across legs is a common desk error that leaves the book unintentionally directional on vol.

4. Failing to plan the unwind. Early termination of a variance swap basket involves negotiating new forward strikes with the dealer. The trader may have to accept unfavorable unwind levels, especially if only one desk was on the trade originally.

5. Treating "correlation converges" as a rule. There are long stretches, sometimes more than a year, in which realized correlation stays above implied. The edge exists on average but not unconditionally, and undercapitalized books fail during those stretches.

Frequently Asked Questions

Q: What is a variance dispersion trade in simple terms? The trade shorts an index variance swap and buys variance swaps on a basket of its constituents, sized so that a parallel shift in all vol levels nets to zero. What remains is a bet that realized correlation among stocks will be lower than the correlation implied by the spread between index and basket variance strikes.

Q: How does the variance dispersion trade affect investment decisions? It provides cleaner correlation exposure than straddle-based dispersion because vega is constant throughout the trade life. P&L decomposition into realized-correlation and vol-of-vol components is also more tractable, making risk monitoring more precise.

Q: What is a real-world example of the variance dispersion trade? SPX 1-year variance strike at 225 (15% vol), basket at 400 (20% vol), implied correlation ~0.50. Realized correlation at maturity: 0.32. On $100k/$80k vega sizing, the spread of 0.18 correlation points generates approximately $1.4M profit.

Q: How can practitioners manage liquidity risk in variance dispersion? Single-name variance swaps are liquid on the top 50 S&P constituents but illiquid on mid-cap names. Limit the basket to the most liquid subset and reduce notional accordingly. In stress, bid-ask spreads on single-name variance swaps can exceed a full variance point.

Q: How is variance dispersion different from a standard dispersion trade using straddles? Straddle dispersion has vega concentrated near the money and rising gamma near expiry, requiring active rehedging. Variance-swap dispersion has constant vega throughout the trade life, no pin risk at expiry, and a model-free realized-variance leg that uses only daily log returns. The tradeoff is dependence on OTC variance swap liquidity.

Sources

1. Jacquier, A. and Slaoui, S. Variance Dispersion and Correlation Swaps. Imperial College London. https://www.ma.imperial.ac.uk/~ajacquie/index_files/Jacquier,%20Slaoui%20-%20Dispersion.pdf
2. Bossu, S., Strasser, E., Guichard, R. Just What You Need to Know About Variance Swaps. JPMorgan Equity Derivatives. http://docs.sbossu.com/bossu-strasser-guichard-varswap.pdf
3. Risk.net. Variance Swaps: Pricing and Dispersion Trades. https://www.risk.net/sites/default/files/import_unmanaged/db.riskwaters.com/data/asiarisk/articles/dec05/variance.pdf
4. Foresi, S. and Vesval, A. Equity Correlation Trading. Goldman Sachs Asset Management / NYU Stern. https://web-docs.stern.nyu.edu/salomon/docs/derivatives/GSAM%20-%20NYU%20conference%20042106%20-%20Correlation%20trading.pdf

Disclaimer

This article is educational content only and is not financial advice. Nothing here is a recommendation to buy, sell, or hold any security. Consult a licensed advisor before making investment decisions.