Geometric Brownian motion and the price of an option
Geometric Brownian motion is the standard stochastic model of a share price. It explains why the typical outcome of an investment lies below the average one, and it leads to the Black-Scholes price of an option, which a Monte Carlo simulation reproduces.
A saver who places €10,000 in a stock-index fund for ten years usually asks two questions: what can be expected, and how bad could it be? A third question arises for anyone who holds employee share options, or who considers insuring a portfolio against a fall: what is the right to buy or sell at a fixed price in the future worth today?
All three questions need a model of how prices move. The standard one is geometric Brownian motion, a continuous-time random walk in which percentage changes, not absolute changes, are random. This article shows what the model implies for the saver, and how it leads to the most famous formula in finance, the Black-Scholes price of an option.
The relevant literature is as follows.
- Bachelier (1900), in his thesis Théorie de la spéculation, modelled prices on the Paris Bourse as a random walk in continuous time, five years before Einstein's paper on Brownian motion.
- Samuelson (1965) proposed modelling the logarithm of a price as a random walk, so that prices stay positive and percentage changes have a stable distribution: geometric Brownian motion.
- Black and Scholes (1973) and Merton (1973) derived the price of a European option by constructing a portfolio of the share and a bond that replicates the option's payoff. Merton and Scholes received the Nobel memorial prize in economics in 1997; Black had died in 1995.
- Boyle (1977) introduced Monte Carlo simulation for option pricing, the method used when no formula exists.
- Rubinstein (1994) documented that, after the crash of October 1987, the volatilities implied by option prices differ systematically by strike, the "volatility smile" that the basic model cannot produce.
- Hull's textbook Options, Futures, and Other Derivatives is the standard practitioner reference.
The model
Geometric Brownian motion describes a price by
where is the expected rate of return, the volatility and a Brownian motion. Its solution is explicit:
The logarithm of the price is normal, so the price itself is lognormal. Over successive days the log-returns are independent and normal, which is the continuous-time version of the random walk discussed for autoregressive models.
The mean and the median
The lognormal distribution is skewed to the right, and its mean and median differ:
With an expected return of 7% a year, a volatility of 20% and a horizon of ten years, the average outcome is a doubling of the investment (a factor of 2.01), but the typical outcome, the median, is a factor of 1.65, and in about one case in five the investment is worth less after ten years than at the start. The average is pulled up by a minority of very good paths.
The gap between the two is the "volatility drag" . A simple example shows its origin: an investment that gains 50% one year and loses 50% the next has an average annual return of zero, yet it ends at 75% of its starting value. Higher volatility with the same expected return lowers the typical outcome, which is why the volatility of a pension portfolio matters to a saver even when its expected return is unchanged.
Pricing an option
A European call option gives the right, but not the obligation, to buy the share at the strike price on date . Its payoff is . A put option, the right to sell at , pays and works like an insurance policy on the share, with the strike playing the role of the deductible.
The central result of Black, Scholes and Merton is that the option can be replicated by continuously adjusting a holding of the share and of a risk-free bond. Its price therefore cannot depend on the investor's attitude to risk, and it equals the discounted expected payoff computed as if the share earned the risk-free rate instead of :
which gives the Black-Scholes formula
For a share at 100, a strike of 100, one year to expiry, a risk-free rate of 2% and a volatility of 20%, the call is worth about 8.92. The expected return does not appear at all, which is one of the least intuitive results in finance and one of the most consequential.
When the payoff is more complicated, for example when it depends on the average price over a year, there is often no formula, and the expectation is computed by simulation. The Monte Carlo estimate is an average of simulated payoffs, and its standard error falls like in the number of paths: a hundred times more paths give one more decimal place.
Limitations
The model assumes a constant volatility and normal log-returns. Neither holds exactly. Volatility clusters, as shown in volatility clustering and the GARCH model, and daily returns have heavier tails than the normal, as discussed in the γ-order generalized normal distribution. Since 1987, options that protect against large falls have been priced with a higher volatility than the model would assign them, which is the market's own correction for heavy tails. The formula nevertheless remains the common language of the market: traders quote options by the volatility that, entered into Black-Scholes, reproduces the price.
The figure below simulates the model. The median outcome lies below the mean by the factor ; the Monte Carlo price of a call converges to the Black-Scholes value at the rate .
Suggested settings:
- Mean and median. With the defaults (σ = 20%, μ = 7%, ten years), the readouts show a mean of 2.01 and a median of 1.65 times the initial price, and a 21% probability of ending below it. In the histogram the mean lies to the right of the peak.
- Volatility drag. Raise σ to 40% with μ unchanged. The mean is still 2.01, but the median falls to about 0.90, and the probability of ending below the initial price rises to about 53%. A riskier asset with the same expected return leaves the typical investor worse off.
- Option price. Press "Option preset: 1 year, at the money". The Black-Scholes value is 8.92. After 1,000 paths the Monte Carlo estimate is typically within about ±0.86 of it; after 100,000 paths within about ±0.09.
- Volatility and option value. Raise σ from 20% to 40% in the option preset. The call price rises to about 16.70. An option gains from volatility because its loss is limited to the price paid while its gain is not.
Why the median lies below the mean. Write . The median of is , because the exponential is increasing. The mean uses the moment-generating function of the normal: . Their ratio is , which grows with both volatility and horizon. The Black-Scholes formula follows from the same integral, , split at and evaluated with the normal distribution function.
Geometric Brownian motion is a simplification, but it captures two facts that matter in practice:
- For a long-term saver, the typical outcome is below the average one, and volatility widens the gap.
- The value of an option depends on volatility and not on the expected return, which is why volatility is the quantity that option markets effectively trade.
References
- L. Bachelier, "Théorie de la spéculation", Annales scientifiques de l'École Normale Supérieure, 17, 21-86, 1900.
- P. A. Samuelson, "Rational theory of warrant pricing", Industrial Management Review, 6(2), 13-39, 1965.
- F. Black and M. Scholes, "The pricing of options and corporate liabilities", Journal of Political Economy, 81(3), 637-654, 1973.
- R. C. Merton, "Theory of rational option pricing", Bell Journal of Economics and Management Science, 4(1), 141-183, 1973.
- P. P. Boyle, "Options: a Monte Carlo approach", Journal of Financial Economics, 4(3), 323-338, 1977.
- M. Rubinstein, "Implied binomial trees", Journal of Finance, 49(3), 771-818, 1994.
- J. C. Hull, Options, Futures, and Other Derivatives, Pearson.