Financial Derivatives

Chapter 7
Greeks, Volatility and Dynamic Hedging

Graduate Program in Finance Management
Financial Derivatives • Graduate

Greeks, Volatility and Dynamic Hedging

Introduction

Derivative prices respond to several changing market inputs: the underlying price, time, volatility, interest rates, and sometimes dividends or foreign interest rates. A price alone does not reveal how exposed a position is to these inputs. The Greeks provide local measures of sensitivity and therefore connect valuation to risk management.

This chapter develops delta, gamma, theta, vega, and rho; explains delta and gamma hedging; and examines historical, implied, and local volatility. It also interprets volatility smiles, skews, surfaces, and term structures. The central practical point is that hedging is dynamic: a hedge that is effective today may become inappropriate after the underlying price, volatility, or time changes.

Learning objectives

By the end of this chapter, the reader should be able to:

Define and calculate the principal option Greeks.
Interpret the sign and scale of Greeks for calls, puts, and portfolios.
Construct and evaluate delta- and gamma-neutral hedges.
Explain discrete rebalancing, transaction costs, and hedging error.
Distinguish historical, implied, and local volatility.
Interpret volatility smiles, skews, surfaces, and term structures.
Evaluate model assumptions, governance requirements, and practical limitations.

1. The sensitivity framework

Let the value of a derivative be

V = V(S, t, σ, r, q)

where S is the underlying price, t is calendar time, σ is volatility, r is the continuously compounded risk-free rate, and q is a continuously compounded dividend yield. A small change in value can be approximated by a Taylor expansion:

∆V ≈ V_S∆S + V_t∆t + V_σ∆σ + V_r∆r + V_q∆q + ½V_SS(∆S)²

The first derivatives are local sensitivities. The second derivative in the underlying price captures curvature. This approximation is most accurate for small changes and becomes less reliable during large price movements, abrupt volatility changes, or near expiration.

1.1 Definitions and units

Greek	Definition	Interpretation	Typical units
Delta, Δ	∂V/∂S	Change in value for a one-unit underlying-price change	currency per underlying unit
Gamma, Γ	∂²V/∂S²	Change in delta for a one-unit price change	delta per underlying unit
Theta, Θ	∂V/∂t	Time passage effect under specified conventions	currency per unit time
Vega, ν	∂V/∂σ	Change in value for a one-unit volatility change	currency per volatility unit
Rho, ρ	∂V/∂r	Change in value for a one-unit rate change	currency per rate unit

Market convention often reports vega for a one-percentage-point volatility change and rho for a one-percentage-point rate change. Thus, if vega is 0.18 per volatility point, a 1 percentage-point increase in volatility changes value by approximately 0.18 currency units, not 18 currency units. Reporting conventions must always be stated.

2. Black–Scholes–Merton benchmark

For a European call and put on a dividend-paying underlying, the Black–Scholes–Merton formulas are

C = Se^(−qT)N(d₁) − Ke^(−rT)N(d₂)

P = Ke^(−rT)N(−d₂) − Se^(−qT)N(−d₁)

with

d₁ = [ln(S/K) + (r − q + ½σ²)T] ÷ (σ√T)

d₂ = d₁ − σ√T

Here, K is the strike, T is time to expiration in years, N(·) is the standard normal cumulative distribution function, and e is the base of natural logarithms. The model assumes, among other things, continuous trading, lognormal underlying prices, constant volatility and interest rates, frictionless markets, no arbitrage, and European exercise. These assumptions make the model a benchmark rather than a complete description of observed markets.

For a non-dividend-paying stock, set q = 0. For discrete dividends, a different treatment is required; simply substituting a yield may create material pricing error.

2.1 Closed-form Greeks

For the benchmark model, call and put deltas are

∆_call = e^(−qT)N(d₁)

∆_put = e^(−qT)[N(d₁) − 1]

A long call therefore has positive delta and a long put negative delta. Put–call parity implies that the difference between call and put deltas is e^(−qT).

Both calls and puts have the same gamma:

Γ = e^(−qT)φ(d₁) ÷ (Sσ√T)

where φ(·) is the standard normal probability density. Long vanilla options generally have positive gamma; short options generally have negative gamma. Gamma is largest near at-the-money strikes and close to expiration, although the precise pattern depends on the inputs.

Theta for a call is

Θ_call = −[Se^(−qT)φ(d₁)σ] ÷ [2√T] − rKe^(−rT)N(d₂) + qSe^(−qT)N(d₁)

and for a put:

Θ_put = −[Se^(−qT)φ(d₁)σ] ÷ [2√T] + rKe^(−rT)N(−d₂) − qSe^(−qT)N(−d₁)

These are derivatives with respect to calendar time when time to maturity decreases as calendar time passes. Some systems quote theta as the negative of this derivative, so sign conventions require verification.

Vega for either a call or put is

ν = Se^(−qT)φ(d₁)√T

This formula treats σ as a decimal, so a change from 20% to 21% is ∆σ = 0.01. Vega is usually positive for a long option and tends to be greatest near the money and for longer maturities.

Rho is

ρ_call = KTe^(−rT)N(d₂)

ρ_put = −KTe^(−rT)N(−d₂)

A long call normally benefits from higher rates, while a long put normally loses value, holding other inputs constant. For foreign exchange, the foreign interest rate plays a role analogous to a dividend yield, and “rho” must be specified as domestic-rate or foreign-rate sensitivity.

2.2 Interpretation beyond first-order risk

Delta is not a permanent hedge ratio. Gamma means that delta changes as the underlying moves:

∆_new ≈ ∆_old + Γ∆S

A portfolio with positive gamma gains approximately from large movements in either direction, after allowing for theta and other effects. A short-gamma portfolio tends to lose from movement and requires frequent rebalancing. The approximate daily P&L decomposition is

∆P&L ≈ ∆∆S + ½Γ(∆S)² + Θ∆t + ν∆σ + ρ∆r

Cross-effects, such as vanna (sensitivity of delta to volatility) and volga or vomma (sensitivity of vega to volatility), can be important for large moves. The decomposition is an attribution tool, not an exact identity.

3. Worked example: interpreting a position

Suppose a portfolio has delta −2,400, gamma −80 per index point, theta +1,200 per day, and vega −6,000 per volatility point. The index rises by 10 points, one day passes, and implied volatility rises by 0.5 percentage points. Ignoring rates, cross-greeks, and changes in the quoted Greeks:

∆P&L ≈ (−2,400)(10) + ½(−80)(10²) + 1,200(1) + (−6,000)(0.5)

∆P&L ≈ −24,000 − 4,000 + 1,200 − 3,000 = −29,800

The negative delta creates the largest loss, negative gamma adds a further loss, positive theta offsets part of the loss, and negative vega loses value as implied volatility rises. The estimate is local and should be compared with a full revaluation.

4. Delta and gamma hedging

4.1 Delta hedging

For a portfolio with delta ∆_P, a position h in the underlying changes portfolio delta by approximately h units per unit of underlying. A delta-neutral hedge therefore uses

h = −∆_P

If the portfolio delta is −2,400, buying 2,400 units of the underlying produces approximate zero delta. If the derivative is quoted per contract, contract multipliers must be included. A delta-neutral position is not riskless: gamma, vega, theta, jumps, basis risk, funding costs, and model risk remain.

4.2 Gamma hedging

The underlying itself has approximately zero gamma. To neutralize portfolio gamma, use another option with gamma Γ_H and hold n contracts:

n = −Γ_P ÷ Γ_H

After adding this option, the residual delta must be neutralized with the underlying:

h = −(∆_P + n∆_H)

Gamma hedging is therefore a two-step procedure. It generally introduces vega, theta, and rho exposures. A hedge option should be selected with attention to liquidity, maturity, strike, exercise style, and correlation with the risk being hedged.

4.3 A numerical gamma hedge

Consider a portfolio with delta −1,000 and gamma −50 per share. A liquid option has delta 0.40 and gamma 0.025 per share. Ignoring contract multipliers:

n = −(−50) ÷ 0.025 = 2,000

Buy 2,000 option units. The combined delta is

−1,000 + 2,000(0.40) = −200

Thus, buy 200 shares to obtain approximate delta neutrality. The resulting position is gamma-neutral but may have substantial vega and theta. If the option multiplier is 100, the required number of option contracts is 20, not 2,000.

4.4 Continuous versus discrete hedging

In an idealized continuous-trading model, a sufficiently regular claim can be replicated by continuously adjusting the hedge. In practice, rebalancing occurs at discrete times and incurs bid–ask spreads, brokerage, market impact, taxes, funding costs, and operational expenses. A short-gamma strategy tends to trade after price movements: it buys after prices rise and sells after prices fall, which can be expensive in volatile markets. A long-gamma strategy often trades in the opposite direction but pays option premium and usually suffers negative theta.

A risk manager must choose a rebalancing rule, such as time-based, delta-band, volatility-triggered, or liquidity-aware rebalancing. The choice is an optimization problem rather than a purely mathematical decision.

5. Volatility: three meanings

5.1 Historical volatility

Historical volatility estimates variability from observed returns. For logarithmic returns

r_t = ln(S_t ÷ S_t−1)

the sample standard deviation is

s = √[∑(r_t − r̄)² ÷ (n − 1)]

If there are m observations per year, annualized historical volatility is

σ_H = s√m

This annualization assumes the return increments are sufficiently comparable and that the sampling convention is appropriate. It does not imply that volatility is constant, normally distributed, or forecastable. Measurement choices—close-to-close versus intraday data, look-back window, treatment of corporate actions, and outlier handling—can materially alter the estimate.

5.2 Implied volatility

Implied volatility is the volatility input that makes a pricing model equal an observed market price. For a call:

C_market = C_model(S, K, T, r, q, σ_imp)

It is an inverse-pricing statistic, not a direct forecast and not necessarily the expected future standard deviation. It incorporates risk premia, supply and demand, jumps, liquidity, model conventions, and the possibility that the benchmark model is misspecified. Implied volatility should be compared only after aligning strike, maturity, settlement, interest-rate curve, dividends, and contract specifications.

5.3 Local volatility

A local-volatility model makes instantaneous volatility a deterministic function of time and underlying level:

dS_t = (r − q)S_tdt + σ_loc(S_t,t)S_tdW_t

Under suitable regularity and arbitrage-free option prices, the Dupire framework links the local-volatility function to the observed call-price surface:

σ_loc²(K,T) = [∂C/∂T + (r−q)K(∂C/∂K) + qC] ÷ [½K²(∂²C/∂K²)]

This expression assumes a sufficiently smooth, arbitrage-free call surface and consistent rates and dividends. In practice, numerical differentiation amplifies market noise, so smoothing and interpolation are essential. Local volatility exactly fits a sufficiently rich vanilla surface in theory, but it may produce unrealistic future smile dynamics and poor pricing for some path-dependent products.

6. The volatility surface

An option’s volatility is not one universal number. Market quotes vary by strike and maturity. A volatility surface maps implied volatility across these two dimensions.

6.1 Smile and skew

A volatility smile occurs when out-of-the-money calls and puts have higher implied volatilities than at-the-money options. A skew or smirk describes asymmetric variation by strike. Equity-index markets commonly show higher implied volatility for downside puts, reflecting crash insurance demand, leverage effects, and risk aversion. Foreign-exchange markets may display different, sometimes more symmetric, patterns.

A surface is often represented using delta rather than strike because delta-based quoting conventions are liquid and intuitive in some markets. However, delta itself depends on the volatility and pricing convention, so the transformation must be handled consistently.

6.2 Term structure

The term structure describes implied volatility across maturities. It may slope upward, downward, or contain humps associated with events such as earnings, elections, central-bank meetings, or scheduled announcements. A longer maturity does not necessarily mean higher volatility. Event risk can be concentrated in a short maturity even when long-run volatility is stable.

6.3 Arbitrage and interpolation

A usable surface should avoid obvious static arbitrage. For fixed maturity, call prices should generally decrease with strike and be convex in strike:

∂C/∂K ≤ 0, ∂²C/∂K² ≥ 0

Across maturity, discounted call prices should satisfy appropriate monotonicity conditions under the chosen rate and dividend conventions. Interpolating implied volatilities directly can violate these restrictions. Professional systems therefore often interpolate prices or carefully parameterize a surface, then test butterfly and calendar arbitrage.

7. Applied case: an equity-index options book

Illustrative case. A bank is short a six-month portfolio of out-of-the-money index puts sold to institutional clients. The book has negative delta, negative gamma, and negative vega. Recent market data show a steep downside volatility skew and a short-dated volatility spike before a major political announcement.

Problem. A simple hedge using the index would neutralize current delta but would not protect against a large fall, a further steepening of downside skew, or an increase in implied volatility.

Analysis. The risk team first calculates sensitivities by strike and maturity, then aggregates delta, gamma, vega, theta, and rho. Scenario analysis shifts the index by ±5%, ±10%, and ±20%; shifts the volatility surface by parallel, skew, and term-structure shocks; and includes a jump scenario through the announcement. Full revaluation is used because the Taylor approximation becomes unreliable for large shocks.

Approach. The team buys downside put options to reduce negative gamma and vega in the most dangerous region, uses futures to manage residual delta, and selects maturities that cover the event. It imposes limits on net short gamma, vega concentration, and model-dependent local-volatility exposure. Rebalancing bands are widened during illiquid periods to reduce market impact, while the jump scenario is monitored separately because continuous diffusion hedges cannot replicate discontinuous price changes.

Outcome. The hedge costs premium and creates some positive theta exposure only if other options are sold, but it materially reduces losses under downside and volatility-spike scenarios. The desk documents liquidity assumptions, model versions, valuation adjustments, and independent price verification.

Learning points. Delta neutrality is only one dimension of risk. Hedge design must reflect the shape of the volatility surface, event timing, liquidity, transaction costs, and non-linear scenario losses. A hedge that appears optimal under a single model may be fragile under alternative volatility dynamics.

8. Limitations, governance, and technology

Greeks are local derivatives, not guarantees. They can change sharply near barriers, expiration, corporate actions, and discontinuities. Model risk arises from incorrect distributions, volatility dynamics, rates, dividends, correlations, or exercise assumptions. Market-data risk arises from stale quotes, sparse strikes, inconsistent settlement times, and unstable interpolation.

Good practice includes independent market-data checks, documented conventions, sensitivity validation by finite differences, back-testing of hedge performance, stress testing, and limits expressed in both Greeks and economic loss. Finite-difference validation should use sufficiently small but numerically stable bumps; too-large bumps measure nonlinearity, while too-small bumps may be dominated by numerical error.

Regulatory and governance frameworks generally require prudent valuation, model validation, audit trails, valuation adjustments, and controls over risk-taking authority. Ethical practice requires transparent reporting of hedge effectiveness and avoidance of presenting model-based precision as certainty. AI and machine-learning methods can assist with volatility-surface cleaning, anomaly detection, and forecasting, but they do not eliminate arbitrage constraints, explainability requirements, regime shifts, or accountability for decisions. Any AI-generated estimate should be benchmarked against established models and subjected to out-of-sample validation.

Best-practice checklist

Define all Greek conventions, units, quote bases, and contract multipliers.
Revalue portfolios after large moves rather than relying solely on Greeks.
Monitor surface shape, not only at-the-money volatility.
Include transaction costs, liquidity, funding, and jump risk.
Use both historical and forward-looking scenario analysis.
Validate interpolation and test static arbitrage conditions.
Separate market risk, model risk, and operational risk in reporting.
Review hedge performance and recalibrate assumptions periodically.

Chapter Summary

The Greeks measure local derivative sensitivities. Delta measures first-order underlying exposure; gamma measures curvature and changes in delta; theta measures time passage; vega measures volatility exposure; and rho measures interest-rate sensitivity. The Black–Scholes–Merton formulas provide useful benchmark expressions, but their assumptions must be stated and their outputs interpreted carefully.

Delta hedging offsets current underlying exposure, whereas gamma hedging uses an additional option to reduce curvature before residual delta is hedged with the underlying. Discrete rebalancing, transaction costs, liquidity, jumps, and model error mean that hedging is never perfectly continuous or riskless.

Historical volatility summarizes past returns, implied volatility is extracted from market prices, and local volatility is a state-dependent function calibrated to an option surface. Smiles, skews, term structures, and surfaces contain information about market pricing and risk premia. Robust practice combines Greeks, full revaluation, scenario analysis, arbitrage-aware surface construction, and strong model governance.

Key Terms

Greek: A derivative-price sensitivity to a market or contractual input.

Delta: First derivative of value with respect to the underlying price.

Gamma: Second derivative of value with respect to the underlying price.

Theta: Sensitivity of value to the passage of calendar time.

Vega: Sensitivity of value to volatility.

Rho: Sensitivity of value to an interest rate.

Delta hedge: A position designed to offset first-order underlying exposure.

Gamma hedge: A position designed to offset curvature exposure.

Historical volatility: Volatility estimated from observed returns.

Implied volatility: Volatility inferred by inverting a pricing model against a market price.

Local volatility: A deterministic volatility function of underlying level and time.

Volatility smile: Strike-dependent pattern in implied volatility.

Volatility skew: Asymmetric strike dependence of implied volatility.

Volatility surface: Implied volatility as a function of strike and maturity.

Term structure: Dependence of volatility or rates on maturity.

Static arbitrage: Riskless profit implied by inconsistent prices across strikes or maturities.

Dynamic hedging: Repeated adjustment of positions as risk factors change.

Vanna: Sensitivity of delta to volatility, or of vega to the underlying, under consistent definitions.

Volga: Sensitivity of vega to volatility.

Review Questions

Short Answer Questions

What does gamma measure, and why is it important for delta hedging?
Distinguish historical volatility from implied volatility.
Why does a long vanilla option usually have positive vega?
What is the purpose of a volatility surface?
State two limitations of continuous delta hedging in practice.

Long Answer Questions

Derive and interpret the first- and second-order P&L approximation for an option portfolio.
Explain the construction of a gamma-neutral and delta-neutral hedge, including contract multipliers.
Compare the assumptions and practical uses of Black–Scholes–Merton and local-volatility models.
Discuss the economic causes and risk-management implications of volatility smiles and skews.
Design a governance framework for an options desk using Greeks, scenarios, and volatility-surface controls.

Analytical / Application Questions

A portfolio has delta −800 and gamma −20. A hedge option has delta 0.35 and gamma 0.01. Determine the number of hedge-option units required for gamma neutrality and the underlying position required for subsequent delta neutrality.
A short-gamma portfolio is delta-neutral at the start of a day. Explain qualitatively why a large underlying move can generate a loss even when the end-of-day delta is rebalanced to zero.
An index’s downside implied volatility rises sharply while at-the-money volatility is unchanged. Identify the likely surface movement, explain two possible economic causes, and propose two hedging responses.

Multiple Choice Questions with Answers

Question 1

Which Greek measures the change in an option’s delta for a change in the underlying price?

A. Theta

B. Gamma

C. Vega

D. Rho

Correct Answer: B. Gamma

Question 2

A portfolio has delta −600. Ignoring other exposures, what underlying position makes it approximately delta-neutral?

A. Sell 600 units

B. Buy 600 units

C. Buy 60 units

D. Sell 60 units

Correct Answer: B. Buy 600 units

Question 3

For a European option in the Black–Scholes–Merton model, which exposure is generally positive for a long vanilla option?

A. Vega

B. Short-option gamma

C. Put rho for every option

D. Negative delta for every option

Correct Answer: A. Vega

Question 4

If a portfolio has gamma −40 and a hedge option has gamma 0.02 per unit, how many hedge-option units are needed for gamma neutrality?

A. 20 long units

B. 200 long units

C. 2,000 long units

D. 2,000 short units

Correct Answer: C. 2,000 long units

Question 5

Implied volatility is best described as:

A. The average of all past daily returns

B. The volatility input that matches a model price to an observed market price

C. The risk-free interest rate expressed as a percentage

D. A volatility estimate that is independent of the pricing model

Correct Answer: B. The volatility input that matches a model price to an observed market price

Question 6

A volatility skew refers primarily to:

A. Variation in implied volatility across strikes that is asymmetric

B. Variation in interest rates across currencies

C. The time decay of an option

D. The number of shares in a hedge

Correct Answer: A. Variation in implied volatility across strikes that is asymmetric

Question 7

Which condition is consistent with no butterfly arbitrage for call prices at a fixed maturity?

A. ∂²C/∂K² < 0

B. ∂²C/∂K² = −1

C. ∂²C/∂K² ≥ 0

D. ∂²C/∂K² must equal the option delta

Correct Answer: C. ∂²C/∂K² ≥ 0

Question 8

A local-volatility model specifies volatility primarily as a function of:

A. Only the risk-free rate

B. Underlying level and time

C. Only the option premium

D. The portfolio’s accounting profit

Correct Answer: B. Underlying level and time

Question 9

A one-percentage-point increase in volatility is represented when volatility is stored as a decimal by:

A. ∆σ = 1.00

B. ∆σ = 0.10

C. ∆σ = 0.01

D. ∆σ = 0.001

Correct Answer: C. ∆σ = 0.01

Question 10

Why does a delta-neutral portfolio remain exposed to market movement?

A. Delta neutrality eliminates gamma automatically

B. Gamma, volatility, theta, rates, jumps, and model risk may remain

C. Delta neutrality guarantees a fixed option price

D. A delta hedge removes transaction costs

Correct Answer: B. Gamma, volatility, theta, rates, jumps, and model risk may remain

Financial Derivatives
Chapter 7 of 15

Chapter 7 Greeks, Volatility and Dynamic Hedging