The Black–Scholes equation is a partial differential equation governing the value of certain financial derivatives under an idealized model of financial markets. It relates changes in a derivative’s value over time to its sensitivity to the price and volatility of an underlying asset and to the prevailing risk-free interest rate. For an underlying asset paying no dividends, its standard form is
where is the derivative’s value, is the underlying asset price, is time, is volatility, and is the continuously compounded risk-free interest rate.
The equation’s central insight is that a derivative’s risk can, under specified assumptions, be replicated through continuously adjusted holdings of the underlying asset and a risk-free account. The absence of arbitrage then determines its price without requiring an estimate of the underlying asset’s expected return. Developed through the work of Fischer Black, Myron Scholes, and Robert C. Merton and published in 1973, this framework became foundational to modern derivatives valuation and risk management.
The equation is distinct from the Black–Scholes model, which specifies the market assumptions and asset-price dynamics, and from the Black–Scholes formula, which is a particular closed-form solution for a European option. Different contractual payoffs and exercise provisions require different terminal conditions, boundary conditions, or extensions of the equation.
Historical development and conceptual significance
Before the Black–Scholes framework, option valuation already had a substantial mathematical history. Louis Bachelier’s 1900 doctoral thesis, Théorie de la spéculation, modeled price changes using a process associated with Brownian motion. His model allowed prices to become negative, whereas later work increasingly favored models for proportional price changes. During the 1960s, researchers including Paul Samuelson developed option-pricing approaches based on lognormal asset prices, and Edward O. Thorp explored practical option valuation and hedging.
A persistent difficulty was determining the appropriate discount rate for an option’s uncertain payoff. An option can be much riskier than its underlying stock, and its risk changes as the stock price and remaining maturity change. Discounting its expected payoff at the stock’s expected return, or at an arbitrarily selected risk-adjusted rate, therefore does not generally produce a consistent valuation. The problem is not simply estimating the payoff’s expected value: it is establishing how that payoff can be priced relative to other traded securities.
Black and Scholes addressed this difficulty in “The Pricing of Options and Corporate Liabilities,” published in the Journal of Political Economy in 1973. Merton’s “Theory of Rational Option Pricing,” published that year in the Bell Journal of Economics and Management Science, developed and extended the continuous-time reasoning. Their work showed how trading in the stock and a risk-free asset could eliminate the local uncertainty of an option position. Once the uncertain component had been removed, the resulting position had to earn the risk-free rate, or an arbitrage opportunity would arise.
This was a shift from pricing by a separately specified preference for risk to pricing by replication. Investors could disagree about a stock’s expected return or have different degrees of risk aversion, yet agree on the option price if they accepted the model’s assumptions and the same volatility. The reasoning connected financial valuation to stochastic calculus and to differential equations already familiar in mathematical physics. It also provided a common language for describing hedge ratios and the changing risks of nonlinear payoffs.
The Chicago Board Options Exchange began trading standardized listed call options in April 1973, giving the theory an important contemporary institutional setting. Its adoption was nevertheless a process rather than an instantaneous replacement of earlier trading methods. Traders used formulas alongside judgment about volatility, market conditions, and execution costs. In 1997, Scholes and Merton received the Nobel Memorial Prize in Economic Sciences for a new method of determining the value of derivatives. Black, who died in 1995, was not included in the award.
The historical importance of the equation lies less in the claim that markets exactly satisfy its assumptions than in its method. A complicated contingent payment could be analyzed as the outcome of a trading strategy, turning a valuation problem into a question about replication, admissible trading, and the absence of arbitrage.
Market assumptions and asset-price dynamics
The classical model contains a risky asset, usually interpreted as a stock, and a risk-free account. The account value grows according to
The stock price follows geometric Brownian motion:
where is standard Brownian motion under the real-world probability measure, is the stock’s instantaneous expected return, and is constant volatility. The model distinguishes the predictable drift term, proportional to , from the random term, proportional to the Brownian increment .
The resulting stock price is
Thus a positive initial price remains positive, and logarithmic returns over an interval of length have a normal probability distribution with mean and variance . Volatility is the standard deviation of continuously compounded returns per square root of time, not the standard deviation of the price level. For example, an annualized volatility of corresponds, within the model, to a one-year log-return standard deviation of .
Trading is assumed to be continuous and frictionless. Securities can be bought and sold in arbitrary fractional amounts, borrowing and lending occur at the same risk-free rate, and short selling is permitted. There are no taxes, transaction costs, or bid–ask spreads. Trading does not move market prices, and the underlying stock price has continuous paths without jumps. The basic version assumes no dividends, although a known continuous dividend yield is readily incorporated.
These conditions make exact dynamic replication possible. A hedging strategy is self-financing when purchases and sales are funded entirely from the strategy’s existing assets: no additional external cash is required after its initial establishment. Appropriate restrictions on trading strategies are also necessary to exclude pathological schemes, such as unlimited doubling strategies financed by unbounded borrowing. In the standard one-stock, one-Brownian-risk setting, the market is complete: suitably integrable contingent claims measurable with respect to the modeled information can be replicated.
The assumptions have different roles. Continuous paths permit cancellation of the single instantaneous random shock with one stock hedge. Frictionless trading permits arbitrarily frequent rebalancing without accumulating trading costs. A known volatility specifies the rate at which randomness enters prices, while a known interest rate specifies the return on locally riskless wealth. Constant coefficients make the familiar closed-form solution possible, but are not indispensable to the broader replication method.
Time-dependent deterministic interest rates and volatilities can be accommodated without abandoning the main logic. By contrast, introducing untraded stochastic volatility, sudden jumps, or trading restrictions may prevent complete replication. The equation is therefore the consequence of a particular market structure, not a universal identity satisfied by every observed derivative price.
Derivation through a self-financing hedge
Let be sufficiently smooth before maturity, with one continuous time derivative and two continuous derivatives in the stock price. Applying Itô’s lemma to the stock dynamics gives
Subscripts denote partial derivatives. The second-order term is essential: Brownian increments have quadratic variation proportional to elapsed time, conventionally expressed as . Ordinary first-order calculus would omit precisely the term that captures the effect of payoff curvature.
Consider a portfolio containing shares and a monetary amount in the risk-free account. Its value is
If the strategy is self-financing, its gains satisfy
To replicate the derivative, require . Comparing the Brownian terms in the portfolio and derivative gains gives
The required cash holding is consequently
The stock holding cancels the derivative’s instantaneous exposure to the single source of uncertainty.
Matching the remaining drift terms yields
The expected-return terms cancel, leaving
The disappearance of is the decisive result. The stock’s expected return is not needed to determine the cost of reproducing the payoff, even though it remains relevant to the stock’s actual price distribution and to an investor’s expected profits.
An equivalent argument considers a position long one derivative and short shares over an infinitesimal interval. Its stochastic component vanishes, so the locally riskless position must earn the risk-free rate. Otherwise, positions in the hedge and risk-free borrowing or lending could be combined to produce arbitrage. This intuitive argument requires care: the derivative of the marked value of a portfolio with changing holdings is not automatically its trading gain. The self-financing formulation explicitly accounts for the funds used in rebalancing and avoids treating changes in as free cash flows.
If the stock pays a known continuous dividend yield , each share generates cash at rate . Including that income in the replicating strategy gives
The coefficient reflects the distinction between the stock’s ex-dividend price appreciation and the total gain from holding it.
The equation determines how an admissible derivative value evolves, but not which contract is being priced. A call, put, or other payoff is identified by its maturity condition and any contractual boundaries. Replication also underpins the law of one price: two strategies producing identical future cash flows must have the same current value in an arbitrage-free market.
European option solutions and boundary conditions
A European call gives its holder the right, exercisable only at maturity , to buy the underlying asset for a strike price . Its terminal value is
where . A European put gives the corresponding right to sell and has terminal value
These payoff conditions select different solutions of the same pricing equation. Although each payoff has a kink at the strike, the option values are smooth for positive remaining time when and .
With constant , , and continuous dividend yield , define and
The Black–Scholes formulas are
where is the standard normal cumulative distribution function. The factors and incorporate discounting and the effect of dividends.
For a non-dividend-paying stock, a European call tends to zero as the stock price tends to zero. At very high stock prices, it approaches : the right to pay the strike later is nearly equivalent to owning the stock while retaining the strike’s present value. For a put, the limiting value as approaches zero is , while its value approaches zero as becomes arbitrarily large. Such boundary behavior is important for analytical uniqueness and for numerical implementations.
Subtracting the formulas gives put–call parity:
This relationship does not require the full Black–Scholes diffusion assumption. It follows from comparing portfolios with the same maturity payoff, provided the European options share a strike and maturity and the stock’s dividend payments can be handled as assumed. It is therefore a broader no-arbitrage restriction against which model prices and market quotations can be checked.
For illustration, take a non-dividend-paying stock with , , year, , and . Then , , and the call value is approximately . The put value is approximately . Their difference, about , equals , as required by parity.
The familiar formula is thus one solution, not the equation itself. Other terminal payoffs produce other prices. A cash-or-nothing digital call paying one unit if , for example, has value . Its discontinuous payoff changes both the valuation expression and the behavior of its hedge near maturity.
Risk-neutral valuation and the connection to diffusion equations
The same prices can be obtained through risk-neutral valuation. Under the model’s pricing probability measure, usually denoted , a non-dividend-paying stock follows
With continuous dividend yield , the price drift becomes . This change does not assert that real investors expect the stock to earn the risk-free rate. It changes the probability weights used to represent arbitrage-free prices, while the real-world drift remains relevant for forecasting and investment analysis.
For a European terminal payoff , the price is
Under suitable regularity and integrability conditions, the Feynman–Kac formula establishes the connection between this conditional expectation and the Black–Scholes differential equation. The equation and the expectation are therefore two representations of the same valuation problem. One emphasizes local trading and changes in value; the other emphasizes discounted future payoffs under a pricing measure.
In the no-dividend model, the discounted stock price is a martingale under : its conditional expected future discounted value equals its current discounted value. More generally, when dividends are present, it is the appropriately discounted cumulative-gain process that has the relevant martingale property. In the standard complete model the equivalent pricing measure is unique, corresponding to the uniqueness of replication-based prices.
The normal probabilities in the European call formula have precise interpretations. Under , is the probability that the call finishes in the money. The quantity is not that same exercise probability. It arises from weighting the stock component of the payoff and can be interpreted as an exercise probability under a stock-based change of numeraire. Confusing the two obscures why separate normal terms appear in the formula.
The equation also has a direct relationship with the heat equation. For the no-dividend case, set , , and write . Then
Writing removes the discounting term. A further shift of the spatial coordinate removes the first derivative, leaving a standard diffusion equation. Gaussian averaging of the transformed payoff then produces the familiar normal-distribution formulas.
This correspondence explains several mathematical features of option values. Diffusion smooths terminal kinks for positive time before maturity, while sharp features reappear as maturity approaches. It also explains why techniques developed for parabolic differential equations can be used for derivative valuation. The financial problem runs backward from a known terminal payoff, but after reversing time it has the structure of a forward diffusion problem.
Sensitivities, hedging, and the balance between time and curvature
The derivatives of an option’s value with respect to market inputs are known as option Greeks. They translate the pricing equation into quantities used to describe and manage risk. The most direct is delta,
which is the number of underlying shares in the replicating portfolio for one unit of the derivative. For a European call with continuous dividend yield,
while a European put has delta . A short call is therefore locally hedged by holding a positive number of shares, not by shorting more shares.
Gamma measures how rapidly delta changes:
For both a European call and put,
where is the standard normal density. Positive gamma represents convexity: delta increases as the stock rises. Near the strike and close to expiry, gamma can become large, requiring substantial changes in the hedge after comparatively small price movements. Far from the strike, an option near expiry is more likely to behave either like its exercise payoff or like a nearly worthless claim.
Theta is the derivative with respect to calendar time,
The Black–Scholes equation can be written as
This identity expresses the balance between time effects, convexity, stock exposure, dividends, and financing. For many long-option positions, maintaining positive convexity comes with negative time decay. Nevertheless, theta is not universally negative: its sign depends on the contract, rates, dividends, and moneyness.
Vega measures sensitivity to volatility:
for European calls and puts. This expression is per unit change in volatility expressed as a decimal. Sensitivity to a one-percentage-point change is . Positive vega means that, within this model, greater volatility raises the value of standard European calls and puts. That conclusion does not automatically apply to every exotic payoff.
A useful local hedging calculation separates assumed volatility from realized variation. For a non-dividend-paying stock, suppose an option value is computed with volatility , but the stock has instantaneous variance rate . Ignoring jumps and costs, the financing-adjusted gain of a long option hedged with shares is locally
This illustrates why gamma-weighted realized variation matters to a delta-hedged option position. It is not a general guarantee of profit from comparing two volatility numbers: actual outcomes also depend on rebalancing times, volatility changes, execution costs, and discontinuous price moves.
Numerical methods, early exercise, and contract variation
Closed-form solutions exist for important contracts, but the differential equation is also a computational framework. Finite-difference methods approximate its time and price derivatives on a grid. Values are initialized using the contractual payoff at maturity, then propagated backward to the valuation date. Explicit schemes calculate new values directly from previously known grid values but require restrictions on grid spacing and time steps for stability. Implicit schemes solve systems of linear equations and have more favorable stability properties for the standard linear problem.
The Crank–Nicolson method averages explicit and implicit time stepping. For sufficiently smooth problems it provides second-order accuracy in time, but nonsmooth payoffs can cause oscillations, especially near expiry or a payoff discontinuity. Initial implicit steps, often called Rannacher smoothing, can reduce this difficulty. Practical accuracy also depends on the placement of the strike on the grid, the truncation of the unbounded stock-price domain, and the treatment of boundary conditions.
An American option permits exercise before maturity. Its value must be at least its immediate exercise payoff , because its holder can choose to exercise. With the operator
the problem can be represented by the variational inequality
In the continuation region the pricing equation holds, while in the exercise region . The boundary between them is unknown in advance and must be found as part of the valuation.
For a non-dividend-paying stock with a nonnegative interest rate, early exercise of an American call is not optimal under the standard assumptions. Exercising sacrifices remaining optionality and requires paying the strike earlier without gaining dividend income. An American put can be optimally exercised early because receiving the strike sooner can outweigh the remaining optionality. Dividends can also make early exercise relevant for calls. These results concern the specified model and contractual rights rather than a universal exercise rule for every market setting.
Other numerical approaches express the same economics differently. The binomial option-pricing model, associated especially with Cox, Ross, and Rubinstein’s 1979 formulation, replaces continuous price paths with a recombining tree of up and down moves. Appropriate refinements converge to diffusion-based prices. Monte Carlo simulation estimates discounted payoffs under the pricing measure and is particularly useful for multiple risk factors and path-dependent claims, although early exercise requires additional methods.
Contract design determines the state variables and boundaries. A barrier option may terminate or become active when a specified price level is reached. An Asian option depends on an average price, typically requiring the accumulated average or sum as an additional state variable. A lookback option depends on a running maximum or minimum. Such contracts demonstrate the distinction between a pricing principle and a single formula: replication and risk-neutral valuation remain central, while the governing mathematical problem changes with the information that determines payment.
Empirical departures, extensions, and practical use
The most visible empirical departure from the constant-volatility model is that observed options on the same underlying asset do not generally imply one common volatility. Implied volatility is the value of that makes the Black–Scholes formula match a specified market price. Across strikes, this often produces a smile or skew; across maturities, a term structure. Together these form an implied-volatility surface. Implied volatility is a quotation derived from prices, not a directly observed physical parameter or an unqualified forecast of future realized volatility.
Equity-index options commonly display higher implied volatilities for lower strikes than for comparable higher strikes, a pattern especially prominent after the October 1987 stock-market crash. More generally, observed returns exhibit heavy tails, changing volatility, and occasional jumps that a constant-parameter lognormal diffusion cannot reproduce. A surface of different implied volatilities can be used to quote market prices, but treating each option as if it belonged to a separate constant-volatility world does not by itself define a consistent joint model for future price movements.
Several extensions address these differences. Merton’s 1976 jump-diffusion model adds sudden price changes to continuous fluctuations. Stochastic-volatility models, including Steven Heston’s 1993 model, make volatility or variance an additional random state variable. Local-volatility approaches, associated with work by Bruno Dupire and Emanuel Derman and Iraj Kani in 1994, use a diffusion coefficient depending on time and asset price. Under appropriate conditions, such models can reproduce a sufficiently regular surface of European option prices, although matching current vanilla prices does not determine the future evolution of that surface uniquely.
These changes affect hedging as well as valuation. A single stock holding cannot generally cancel both continuous price risk and every possible jump. An additional untraded volatility shock likewise introduces risk that cannot ordinarily be eliminated with only the stock and risk-free account. In such incomplete markets, absence of arbitrage may allow more than one pricing measure. Valuation then requires additional traded instruments, restrictions, or assumptions about the pricing of unhedgeable risks. Different models can agree closely on standard European options while assigning different values to path-dependent or early-exercise claims.
Market frictions create a separate departure. Continuous rebalancing would be prohibitively costly with nonzero spreads and fees. Actual hedges are adjusted at discrete times, leaving residual exposure to price moves between trades. Limited liquidity, market impact, stock-borrowing constraints, and differing funding rates can further separate a theoretical replication value from an executable trading price. For contracts traded in the over-the-counter market, counterparty credit risk, collateral arrangements, and funding conventions may also matter.
Despite these limitations, the equation remains widely used as a benchmark, a source of standardized sensitivities, and a convenient mapping between option prices and volatility quotations. Its replication logic also extends beyond listed stock options. In the Merton structural model of corporate debt, shareholders’ limited-liability claim resembles a call option on a firm’s assets, while risky debt can be decomposed using an option-like default exposure. The mathematical analogy is powerful, but practical implementation depends on how firm asset values, volatility, debt obligations, and default provisions are modeled.