Implied Correlation Index: Dispersion, Index Options, and Diversification

Table of Contents

    Suppose the implied volatility of three-month, at-the-money S&P 500 index options rises while comparable implied volatilities on the largest stocks in the index barely move. One possible explanation is that the options market is pricing more co-movement: if stocks swing together, their moves cancel out less at the index level. An implied correlation index measures how much co-movement is needed to explain index volatility relative to component volatility. That makes it useful for assessing diversification and understanding dispersion trades. [1]

    What Is Implied Correlation?

    Implied correlation is the correlation backed out of index and constituent option prices for a specified horizon and set of option observations. First, translate each option price into an implied volatility (IV): the volatility input consistent with that price under a pricing convention. Then ask how much co-movement would reconcile the index IV with the component IVs and their weights.

    What the options market is pricing

    Individual-stock volatility and co-movement both affect index volatility. With stock IVs and weights fixed, a higher index IV implies higher correlation in this model. If stock IVs or weights also move, the index IV alone cannot tell you what happened to implied correlation.

    Unlike historical return correlation, this calculation uses option prices observed today. It measures the co-movement priced into those options, which can reflect what investors pay for protection as well as their expectations for future returns. [8]

    The Two-Stock Implied Correlation Formula

    Start with a fixed-weight index containing only stocks A and B. The variance of its weighted return has two individual variance terms and one cross term. The cross term is the covariance between the two returns, multiplied by twice the product of their weights. Covariance itself is correlation multiplied by both stocks’ standard deviations. [4]

    For an option-based estimate, plug in the annualized IVs as volatility proxies. Let w be each stock’s weight and let σ be annualized IV in decimals. Subscript I denotes the index. The formula uses squared IVs because variance, rather than volatility, is the quantity that decomposes:

    Index variance with two stocks
    σI2 = (wA × σA)2 + (wB × σB)2
    + 2 × wA × wB × σA × σB × ρimp
    The first two terms give the zero-correlation baseline. The last term accounts for co-movement.

    Subtract the baseline from the index’s implied variance. Then divide by the cross term’s value at a correlation of +1. The result, ρimp, is unitless:

    Implied correlation
    ρimp = [σI2 − (wA × σA)2 − (wB × σB)2]
    ÷ [2 × wA × wB × σA × σB]
    If either stock’s weighted volatility is zero, there is no cross term to solve for.

    Worked Example: Back Out Correlation from a 20% Index IV

    Assume a hypothetical index holds 60% in Stock A and 40% in Stock B. At the same observation time, comparable at-the-money (ATM) options at a three-month horizon imply annualized volatility of 25% for A, 20% for B and 20% for the index. The same annualization convention applies to all three.

    From three volatility inputs to one implied correlation
    1. Weight each stock’s IV: 0.60 × 0.25 = 0.15 for A; 0.40 × 0.20 = 0.08 for B.
    2. Find the baseline variance: 0.152 + 0.082 = 0.0289. Its square root is 17% annualized volatility.
    3. Find the scale of co-movement: 2 × 0.15 × 0.08 = 0.0240.
    4. Solve for correlation: Index variance is 0.202 = 0.0400. The implied correlation is (0.0400 − 0.0289) ÷ 0.0240 = 0.4625.

    To check the result, add back the cross term: 0.0289 + (0.0240 × 0.4625) = 0.0400. If you display correlation on a 100-scaled index, 0.4625 reads as 46.25. That number is a correlation coefficient on a different scale, with no probability interpretation.

    A quick diagnostic is to ask what index IV the same inputs would permit at correlations of −1 and +1. The weighted IVs could subtract to 0.15 − 0.08 = 7% or add to 0.15 + 0.08 = 23%. An index IV of 24% would not fit this two-stock model; review the observations and weights before interpreting the result.

    How Cboe Calculates COR3M

    The Cboe 3-Month Implied Correlation Index (COR3M) applies the same idea with a real S&P 500 index option input and a much larger stock basket. Cboe’s public overview and calculation white paper describe these inputs and steps: [1] [2] [3]

    1. Select and weight the basket. The component side uses 50 large SPX constituents, ranked by prior-close float-adjusted market capitalization, with replacements if needed. Cboe rescales the selected stocks’ market-cap weights to sum to 100% within that basket.
    2. Match the volatility observations. Cboe uses constant three-month IVs at its 0.50-delta ATM convention, taken from the Cboe Hanweck volatility surfaces. A constant horizon is measured from each observation date; it does not follow one fixed expiration as it approaches. The correlation formula uses these IVs as inputs; the volatility surfaces themselves are fitted to market data.
    3. Solve for one correlation. From the squared IV of options on the full SPX, subtract the sum of squared-weight stock IVs, meaning each stock’s IV squared times its weight squared. Divide by twice the sum of the products of weighted IVs for each distinct stock pair.

    The result is a single effective correlation. A pair with larger weights or IVs has more influence on the calculation. The numerator still starts with an option on the full S&P 500, so the result describes the relationship under Cboe’s top-50 approximation rather than an equal average of all pairwise correlations in the S&P 500. The two-stock calculation above illustrates the algebra; even with access to stock option quotes, it does not reproduce COR3M.

    Implied Correlation vs. Realized Correlation, DSPX and VIX

    Start by checking the horizon and units. COR3M, the Cboe S&P 500 Dispersion Index (DSPX) and VIX answer different questions, even though each uses S&P 500 option information. [1] [5] [11]

    Compare the inputs and units before comparing the readings
    Measure What it measures
    COR3M Three-month option-implied co-movement, quoted on a correlation index scale.
    Realized correlation Unitless co-movement computed from observed returns. To compare with an implied reading, measure the returns over its subsequent horizon and match the basket, weights and return conventions as closely as possible.
    DSPX Thirty-calendar-day implied dispersion in volatility points. The methodology subtracts the S&P 500 implied variance from a weighted average of selected constituent implied variances, floors the difference at zero, and takes the square root.
    VIX Thirty-day implied S&P 500 volatility in volatility points, derived from SPX options. Individual-stock option IVs do not enter its calculation. VIX options are separate contracts.

    In a January 26, 2026 market note, Cboe described the one-month correlation index (COR1M) as near a low of 8.33 and reported DSPX at 34.8 volatility points. A low correlation reading can coexist with substantial dispersion when individual stocks are expected to make sizable moves. The two numbers are on different scales. [10]

    What Drives Dispersion Trade Profit and Loss?

    A long dispersion position typically buys options on constituent stocks and sells options on the index. One structure uses straddles, a call and put at the same strike and expiry on each underlier. If only implied correlation rises, the model says index IV rises relative to the stock IVs, which can hurt the short index options. But the actual position makes or loses money on option prices and hedges. Correlation is one driver of that result. [6] [7]

    Lower realized correlation can still produce a loss

    To isolate the key relationship, return to the 60/40 basket and replace straddles with three idealized variance swap payoffs. The index leg is short with $1,000,000 per 1.0000 unit of decimal annualized variance. The A and B legs are long with $360,000 and $160,000 per unit, obtained by multiplying the index notional by 0.602 and 0.402. The strikes are set to the squared example IVs: 0.0400 for the index, 0.0625 for A and 0.0400 for B. This is a hypothetical choice, not an observable swap quote.

    Assume fixed weights, consistent annualized variance and covariance from the same return observations, and no financing or trading costs. Because of the squared-weight sizing, the single-stock variance terms cancel against those inside the index term. What remains is the entry cross term of 0.0111 from the worked example minus the realized cross term, all multiplied by the index notional. The realized term is 2 × 0.60 × 0.40 × realized vol A × realized vol B × realized correlation.

    Gross payoff on the same hypothetical variance positions under two outcomes
    Realized inputs Realized cross term Gross payoff
    Realized volatilities match the IV inputs: 25% and 20%; correlation 0.20 0.0240 × 0.20 = 0.0048 +$6,300
    Volatilities rise: 35% and 28%; correlation 0.30 0.04704 × 0.30 = 0.014112 − $3,012

    In the first case, (0.0111 − 0.0048) × $1,000,000 = $6,300. In the second, (0.0111 − 0.014112) × $1,000,000 = −$3,012. Both realized correlations are below the entry estimate of 0.4625. In the second case, the larger product of the two stocks’ realized volatilities pushes realized covariance above what the trade priced at entry. The short index leg loses more than the stock legs gain.

    What changes with listed options?

    • Volatility and sizing. Vega measures estimated option-price sensitivity to implied volatility. Matching the long legs’ dollar vega to the short leg at entry only offsets the specified volatility move at that point. The exposures change, and individual-stock and index IVs need not shift by the same amount.
    • Path and timing. Delta hedges reduce first-order directional exposure when placed. Gamma changes delta as prices move, gaps can leave a hedge stale, and time decay affects every leg. A stock’s earnings release or a difference in settlement times can make one leg cover risks another leg misses.
    • Weighting and costs. A partial basket or different option sizes leaves extra exposure to uncovered names and factors. Bid-ask spreads, fees, hedge trades and margin can change both the net result and the ability to hold the position through stress.

    In a historical U.S. options study, Driessen, Maenhout and Vilkov found that spreads and margin weakened the apparent opportunity to profit from priced correlation risk. The study is evidence that execution matters, not an estimate of a current strategy’s return. [8] [9]

    How to Use Implied Correlation in a Diversification Check

    1. Identify what changed. Compare the index and component IVs first. Ask whether both jumped, only the index jumped, or only the component basket jumped. If you compare readings through time, hold the tenor and delta convention constant.
    2. Check which risks are shared. Cboe’s backtested three-month implied correlation series rose alongside VIX during the March 2020 selloff. A broad equity shock can raise expected co-movement at the same time that volatility rises. That is a reason to stress-test exposure to common factors rather than just count the number of holdings. [6]
    3. Apply the scenario to your own weights. If you already have a correlation assumption for your holdings, the portfolio variance calculator shows what that assumption does to portfolio risk. It does not infer correlation from option quotes. Run a higher-volatility scenario as well as a higher-correlation one.

    Common Mistakes

    • Combining unmatched IVs. Use the same snapshot and horizon, comparable strike or delta conventions, a consistent annualization basis, and the correct underlier and basket. Quotes from different times, very wide spreads, and contracts covering different settlement windows can create a misleading residual.
    • Subtracting IVs directly. Square the volatilities and use squared weights in the baseline. The difference between index volatility and average stock volatility is not a correlation.
    • Clamping an implausible estimate. A two-stock result outside −1 to +1 means your assumptions or inputs do not fit the model. Check weights and quotes before using it; forcing it back into range hides the error.

    Limitations of an Implied Correlation Index

    One number cannot describe every pair or every market state

    A top-50 U.S. equity index proxy does not tell you how a specific stock pair will behave, how your bonds and equities will move together, or how stocks will co-move in the next selloff. The ATM reading also leaves out strike-by-strike differences in how the market prices large moves. [1] [2]

    Prices can embed a risk premium for co-movement. A difference between implied and later realized correlation can reflect risk pricing, a change in market conditions, or a mismatch in how you measure the two. Treat the index as one input to a risk check. It does not say whether the market will rise or fall, or whether a particular dispersion trade is cheap enough after costs.

    Frequently Asked Questions

    Yes. If the example’s index IV rises to 22% while its stock IVs and weights stay fixed, the implied correlation rises from 0.4625 to (0.222 − 0.0289) ÷ 0.0240 = 0.8125. The index options are pricing more co-movement relative to those stocks under the model.

    No universal threshold applies across baskets, maturities or delta conventions. Compare a reading with the history of the same series and compare the contemporaneous stock IVs too. High correlation with low component volatilities can imply less index volatility than modest correlation with very high component volatilities.

    No. VIX reflects a 30-day horizon and uses option prices across strikes. COR3M uses a three-month ATM delta-relative IV. Even VIX3M matches the horizon only, so substituting it would change the method.

    They describe different risks. The equity measure here uses stock and index options to infer stock return co-movement. Implied default correlation describes dependence among credit defaults under a credit model and can be calibrated to credit tranche prices.

    Sources

    1. Cboe: Implied Correlation overview.
    2. Cboe: Implied Correlation Index White Paper, pp. 4–8.
    3. Cboe Hanweck: Options Analytics Content Overview, v1.5.
    4. Harry Markowitz: Portfolio Selection, pp. 80–81 (1952).
    5. S&P Dow Jones Indices: Cboe S&P 500 Dispersion Index Methodology (January 2026).
    6. Cboe: Implied Correlation fact sheet (2021).
    7. Jacquier and Slaoui: Variance dispersion and correlation swaps (2010).
    8. Driessen, Maenhout and Vilkov: The Price of Correlation Risk: Evidence from Equity Options (2009).
    9. OCC: Characteristics and Risks of Standardized Options, chapters VIII–X.
    10. Cboe: Traders Defensively Positioned as Equities Retrace from Record Highs (January 26, 2026).
    11. Cboe: Index Insights, October 2026 (November 3, 2026).
    Educational Use

    The examples are hypothetical and do not constitute a recommendation. Written options can generate large losses and require additional margin. Read the OCC options disclosure document for details on standardized U.S. options.