Second-Order Greeks Explained

Table of Contents

    Second-order Greeks explain why an option’s sensitivities change. A delta hedge can drift even when the underlying price stays still: time passes, implied volatility changes, and yesterday’s offset no longer matches today’s exposure. Charm, vanna and vomma help separate those effects.

    This guide builds on the core option Greeks rather than repeating their introduction. It uses a hypothetical European call, with explicit time and volatility conventions, to show what these measures tell you about a hedge and where a local estimate falls short.

    What Are Second-Order Greeks?

    A first derivative measures how option value changes with one input. A second derivative measures how that first sensitivity changes. Gamma is already a second-order Greek mathematically: it measures how delta changes with the underlying price. The term is not reserved for less familiar names. [1]

    The Hedge Ratio Can Change Without a Price Move

    Charm tracks time-driven delta change. Vanna tracks volatility-driven delta change. Vomma tracks volatility-driven vega change. Each isolates one input while holding the others fixed.

    Which sensitivity changes, and why?
    Greek Sensitivity that changes Input that changes Practical question
    Gamma Delta Underlying price How does a spot move alter the hedge?
    Charm Delta Time How does the hedge drift as expiry approaches?
    Vanna Delta Implied volatility How does an IV change alter directional exposure?
    Vomma, or volga Vega Implied volatility How does an IV change alter volatility exposure?

    Charm and vanna are mixed partial derivatives: they connect price sensitivity to two different inputs. Vomma differentiates price twice with respect to volatility. For a smooth model price, vanna also measures how raw vega changes with spot; both descriptions refer to the same cross-partial when units and held-fixed inputs match.

    Charm: Declare the Time Direction

    Charm, sometimes called delta decay, measures delta’s response to time. Its sign is meaningless until you know whether “time” means calendar time passing or time remaining increasing. This article uses the calendar-time convention. [1]

    Let t be calendar time, T the fixed expiry date, and τ = T − t the time remaining. When one day passes, t rises and τ falls. With spot, volatility, interest rate and dividend yield fixed:

    Calendar-time charm
    Charmcalendar = ∂Δ / ∂t = −∂Δ / ∂τ
    Positive calendar charm means delta increases as time passes. Negative calendar charm means delta decreases. A time-to-expiry derivative has the opposite sign.

    An out-of-the-money call can lose delta as its remaining opportunity to finish in the money shrinks, even with no spot move. An in-the-money call can move toward a delta of one. Those are useful limiting intuitions, not universal sign rules at every combination of rates, dividends, volatility and moneyness. The option delta guide develops that expiration intuition.

    Charm differs from theta. Theta measures the option price’s response to passing time; charm measures delta’s response. “Delta decay” does not mean charm is always negative or that delta always moves toward zero.

    Vanna and Vomma: Two Different IV Effects

    Vanna Changes the Delta Hedge

    Vanna is the sensitivity of delta to implied volatility. If vanna is positive, a small IV increase raises the option’s signed delta, all else equal; if negative, it lowers it. For a put, “raises delta” can mean making a negative delta less negative. Vanna is a change in exposure, not a direct dollar profit estimate.

    Equivalently, vanna measures the spot sensitivity of raw vega. That equality comes from reversing the order of differentiation of a smooth price function. A platform’s vega quoted per one volatility percentage point needs a corresponding unit adjustment before the two numbers can be compared.

    Vomma Changes the Vega Estimate

    Vomma, also called volga, measures how vega changes with IV. Positive vomma means a small IV increase raises vega; negative vomma means it lowers vega. The option vega guide uses a local linear premium estimate; vomma describes part of the curvature that estimate leaves out. [1]

    A long vanilla call or put can have positive vega and negative vomma. A function can slope upward while its slope decreases. Likewise, being long an option does not guarantee positive vanna or charm. To obtain a short position’s exposure, reverse the signs of the corresponding long position’s Greeks.

    Set the Units Before Using the Numbers

    Use decimal annualized volatility in the model: 25% is σ = 0.25. A rise from 25% to 26% is one percentage point, so the decimal change is 0.01. It is not a 1% relative increase. Time is in years, using 365 calendar days per year here.

    Option value below is quoted in USD per underlying unit, comparable to a per-share equity-option quote. Delta is the price derivative in dollars per dollar, interpreted as a share-equivalent exposure per underlying unit. A contract multiplier converts that exposure to shares or dollar sensitivities. OCC specifies 100 shares for standard equity contracts, while adjusted contracts can differ. Our fictional European contract explicitly assumes a 100-unit multiplier. [3]

    Raw derivatives and the units used in this article
    Measure Raw derivative Useful rescaling
    Vega ∂V / ∂σ Multiply by 0.01 for USD per unit per IV percentage point
    Vanna ∂Δ / ∂σ Multiply by 0.01 for delta change per IV percentage point
    Vomma ∂2V / ∂σ2 Multiply by 0.0001 for USD per unit per IV percentage point squared
    Calendar charm ∂Δ / ∂t, with t in years Divide by 365 for delta change per calendar day

    Vomma needs two factors of 0.01: one for changing raw vega to vega per percentage point, and one for changing the volatility input to percentage points. This rescaled vomma also estimates the change in displayed per-point vega caused by a one-point IV move. Check a data provider’s documentation before assuming its labels use these conventions.

    Second-Order Greek Formulas Under Black-Scholes-Merton

    The following formulas use a European option, positive time remaining and volatility, constant continuously compounded interest rate r, continuous dividend yield q, and lognormal diffusion without jumps. S is spot and K is strike. Volatility σ is a model input; bumping it compares model scenarios. These are not exact formulas for every American option or a guarantee of market-price behavior. The starting price, delta and vega equations come from the dividend-aware Black-Scholes model; the second derivatives below follow by differentiation. [2]

    Let N be the standard normal cumulative distribution function and φ its density. Define the model’s two standardized terms:

    Inputs to the Greek formulas
    d1 = [ln(S/K) + (r − q + σ2/2)τ] / (σ√τ)
    d2 = d1 − σ√τ
    Time remaining τ is in years; r, q and σ are decimals.

    Raw vega is S e−qτφ(d1)√τ. Differentiating delta with respect to volatility, and then vega with respect to volatility, gives:

    Vanna and vomma, before percentage-point scaling
    Vannaraw = −e−qτφ(d1)d2 / σ
    Vommaraw = Vegaraw × d1d2 / σ
    Vanna’s sign follows −d2. Vomma’s sign follows d1d2, because raw vanilla vega and σ are positive.

    For the call’s charm, first define A as the rate at which d1 changes when time remaining increases. Applying the product rule to call delta e−qτN(d1), then reversing the time direction, gives:

    Calendar-time charm of a European call
    A = [(r − q + σ2/2)τ − ln(S/K)] / (2στ3/2)
    Charmcalendar = q e−qτN(d1) − e−qτφ(d1)A
    The result is delta change per year of calendar-time passage. Divide by 365 for the local per-day coefficient.

    These expressions also explain why simple sign slogans fail. For example, when S = K and r = q, d1 is positive and d2 negative, making long-option vomma negative. A positive vega alone cannot establish the sign of its derivative.

    Example: A Hedge Drifts With Spot Unchanged

    Assume a hypothetical European call on a dividend-paying underlying: S = $100, K = $105, 60 calendar days remaining, IV = 25%, r = 4% and q = 1%. Rates and dividend yield use the continuous conventions above. These are invented teaching inputs, not a live security or quote. Hold spot, strike, rate and yield fixed throughout.

    Using τ = 60/365 gives d1 ≈ −0.382019 and d2 ≈ −0.483380. Initial price is $2.267661 per unit, delta is 0.350647, and vega is $0.150119 per unit per IV percentage point. The local coefficients are:

    • Calendar charm: −0.650465 per year / 365 ≈ −0.00178210 delta per day.
    • Vanna: 0.715903 × 0.01 ≈ +0.00715903 delta per IV percentage point.
    • Vomma: 11.088387 × 0.0001 ≈ +$0.00110884 per unit per IV percentage point squared.

    Now reprice the same call at IV of 24% and 26%, and separately let one calendar day pass with IV unchanged. The combined case changes both time and IV. Each row is a fresh model calculation, rather than a linear Greek prediction.

    Recomputed exposures with the underlying held at $100 (hypothetical)
    Days remaining / IV Price (USD per unit) Delta Vega (USD per unit per IV point)
    60 / 24% 2.118119 0.343250 0.148943
    60 / 25%: baseline 2.267661 0.350647 0.150119
    60 / 26% 2.418314 0.357584 0.151167
    59 / 25% 2.233656 0.348846 0.148587
    59 / 26% 2.382782 0.355828 0.149647

    One day alone reduces delta by about 0.00180056, close to the charm estimate of 0.00178210. A one-point IV increase alone raises delta by about 0.00693686, versus vanna’s 0.00715903 estimate. Differences arise because the sensitivities themselves change along the finite move. Calculations use unrounded values.

    Translate Delta Drift Into Hedge Shares

    For 10 long contracts with 100 units each, initial share-equivalent delta is 10 × 100 × 0.350646780 ≈ 350.65 shares. An idealized stock hedge shorts that many shares. Fractional shares here describe equivalent exposure; an actual hedge has execution and position constraints.

    With spot fixed, one day of passage alone reduces the option position’s delta by about 1.80 shares, so restoring model delta neutrality would require buying back about 1.80 of the short shares. The one-point IV rise alone increases delta by about 6.94 shares, requiring an additional short of about 6.94 shares.

    For small joint changes, including a possible spot move, the local delta estimate is:

    Local change in delta
    Change in delta ≈ Gamma × Change in S
    + Vannaraw × Change in σ
    + Charmcalendar × Change in t
    Decimal volatility and years; rate, dividend yield and strike held fixed. Higher-order terms are omitted.

    For one day and a one-point IV rise together, with spot unchanged, the estimate is 1,000 × (−0.00178210 + 0.00715903) ≈ +5.38 shares. Fresh recomputation instead gives 1,000 × (0.355828212 − 0.350646780) ≈ +5.18 shares. About 5.18 extra short shares would restore model delta neutrality. The hedge adjustment is not a profit figure; financing, borrow, transaction costs and other risks are outside this calculation. See delta hedging for the broader hedge mechanics.

    Use Vomma to Refine a Vega Estimate

    Return to the original 60-day call and change only IV from 25% to 26%. Vega alone estimates a premium gain of $0.150119 per unit. Adding volatility curvature gives:

    Local IV-only premium change
    Change in V ≈ Vegaper point × p + ½ × Vommaper point squared × p2
    ≈ $0.150118845 + ½ × $0.001108839 = $0.150673264
    Here p = +1 IV percentage point. The result is USD per underlying unit, before any contract multiplier.

    Exact model repricing gives a $0.150652791 gain per unit. Vomma improves this particular estimate, but it does not make the truncated formula exact. It also predicts displayed vega rising by about $0.00110884; the recomputed rise is about $0.00104814 per unit per IV point.

    Check the Meaning With Smaller Bumps

    A derivative describes a limiting small change. To estimate vanna numerically, calculate delta at IV just above and below the same baseline, then divide the difference by the full volatility gap. For example, a symmetric ±1-percentage-point bump uses 26% and 24%, a decimal gap of 0.02:

    Central finite-difference estimate of raw vanna
    [Δ(0.26) − Δ(0.24)] / (0.26 − 0.24)
    ≈ (0.357583639 − 0.343249602) / 0.02 ≈ 0.716702
    The baseline remains 25% IV and 60 days. This is a derivative estimate, not the one-sided delta change from 25% to 26%.

    For vomma, use the same central difference on raw vega. For calendar charm, subtract delta at an earlier calendar time from delta at a later calendar time: the later point has fewer days remaining. Repeating those calculations with smaller symmetric bumps gives:

    Smaller bumps approach the local derivatives (hypothetical)
    IV bump each side Vanna, raw Vomma, raw Time bump each side Calendar charm per day
    1 percentage point 0.716702 11.119517 1 day −0.001782322
    0.1 percentage point 0.715911 11.088697 0.1 day −0.001782098
    0.01 percentage point 0.715903 11.088390 0.01 day −0.001782096
    Analytic derivative 0.715903 11.088387 Analytic derivative −0.001782096

    The convergence checks the formulas and clarifies their local meaning. Making a numerical bump arbitrarily tiny is not always better: subtracting nearly equal computer values eventually introduces rounding error. For material scenario changes, recompute prices and exposures rather than extrapolating one set of Greeks indefinitely.

    Charm vs. Vanna vs. Vomma in Practice

    Managing Directional Exposure

    • Gamma: spot-driven delta change.
    • Charm: time-driven delta change.
    • Vanna: IV-driven delta change.
    • Use the combined exposures to assess how a current delta hedge may drift.

    Managing Volatility Exposure

    • Vega: first-order premium response to IV.
    • Vomma: how that response changes with IV.
    • Vanna: how raw vega changes with spot.
    • Reprice joint scenarios when spot, IV and time all change.

    To analyze a position, first confirm the model, units and time sign. Multiply each per-unit sensitivity by the signed contract count and actual multiplier, then sum across legs. Keep different expirations and volatility inputs identifiable: one aggregate vanna or vomma number does not specify which part of a volatility surface moves. A core-Greeks calculator does not automatically report these additional derivatives.

    Common Mistakes

    • Reading charm’s sign without its convention. A negative calendar charm is the same local effect as a positive derivative with respect to time remaining. Check the definition before planning a hedge adjustment.
    • Mixing decimal IV and percentage points. A one-point change is 0.01 in the formulas. Using 1 with a raw derivative creates a 100-fold error; second-derivative scaling needs both factors.
    • Confusing delta units with dollars. A vanna-driven delta change becomes share-equivalent exposure after the multiplier. It is not itself an option premium gain.
    • Assuming every long-option second-order Greek is positive. Long/short ownership and the sign of a particular derivative are separate questions. Compute each exposure.
    • Treating a local Greek as a large-move forecast. Time, spot and IV changes interact. A five-day or ten-point scenario requires fresh calculations, especially near expiry.

    Limitations of Second-Order Greeks

    A Model Sensitivity Is Conditional

    The examples isolate smooth European-option derivatives. A live option’s IV, spot and other inputs can change together, while exercise features, jumps and trading costs affect hedge results. Delta neutrality at one moment does not eliminate risk.

    Model and contract choice matter. Standard U.S. equity options are American style, so early exercise and discrete dividends can require a different valuation approach. Continuous yield is a simplifying assumption. The European formulas should not be presented as exact listed-stock-option values. [3]

    Near expiry, the relevant scale shrinks. One day is a small fraction of a two-month option’s life but the entire remaining life of a one-day option. Delta can change sharply around the strike, and the payoff is kinked at expiry. Do not assume every second-order Greek diverges everywhere, or carry a smooth pre-expiry approximation through the terminal kink.

    Surface behavior and execution remain separate. Holding one option’s IV fixed defines a model partial derivative; it does not predict the market’s joint spot-and-volatility move. Volatility skew and term structure can change unevenly, and continuous hedging is an idealization. Discrete rebalancing leaves error. [2]

    These measures do not establish a market-flow forecast. The Greek of a hypothetical option tells you nothing by itself about dealers’ net inventory, when they rebalance or whether their trades will move the market. Those conclusions require additional positioning and execution evidence.

    Frequently Asked Questions

    They may differentiate with respect to different time variables. Calendar time increases as expiry approaches, while time remaining decreases. For otherwise identical inputs and units, those derivatives have opposite signs. Also check whether the number is annualized, daily, per unit or per contract before comparing it.

    Under the same Black-Scholes-Merton inputs, a European call and put with the same underlying, strike and expiry have identical vanna and vomma. Their price difference under put-call parity has no volatility dependence. Calendar charm is different when dividend yield q is nonzero: call charm minus put charm equals q e−qτ per year. They are equal when q = 0. These equalities are model results, not universal claims for American options.

    No. Vega is the premium’s sensitivity to volatility. Vanna is delta’s sensitivity to volatility, or equivalently raw vega’s sensitivity to spot in a smooth model. Their outputs have different units. A position can have little net vega yet still have meaningful vanna because its individual legs respond differently.

    Yes. Positive vega means the option value slopes upward with volatility. Negative vomma means that positive slope becomes smaller as volatility rises locally. In the European model, vomma’s sign follows d1d2. A long option can therefore have positive vega and negative vomma at the same time.

    Sources and Calculation Notes

    1. Options Industry Council: May Office Hours FAQs, June 2026, “Second-Order Greeks” and “Credit Spreads and Net Greek Exposure.” Accessed November 14, 2026.
    2. Martin Haugh, Columbia University: The Black-Scholes Model, 2016, sections 1.2–4. The article’s second-order expressions are derived from the dividend-aware price/delta/vega equations. Worked values were checked using analytic derivatives and smaller central finite differences; all inputs are hypothetical.
    3. OCC: Equity Options Product Specifications, “Unit of Trade,” “Premium Quotations” and “Exercise Style.” Accessed November 14, 2026.
    Disclaimer

    This article is for educational and informational purposes only and does not constitute investment advice. Options involve risk and are not suitable for all investors. Model sensitivities and hypothetical hedge calculations do not guarantee trading results. Review your contract specifications, model assumptions and the applicable options risk disclosure before trading.