Stochastic Process
A stochastic process is a mathematical object describing the evolution of a random variable over time — formally, a collection of random variables {X_t, t ∈ T} indexed by time on a probability space, where each X_t represents the uncertain state of the system at time t. In finance, stochastic processes model asset prices, interest rates, volatility, and other quantities whose future values are uncertain, forming the mathematical foundation of derivatives pricing, risk management, and quantitative investment theory.
Key takeaways
- The most fundamental stochastic process in finance is Brownian motion (Wiener process) — a continuous-time process with independent Gaussian increments, forming the core of the Black-Scholes options pricing model.
- Geometric Brownian motion (GBM), where asset prices follow dS = μS dt + σS dW, combines a drift (deterministic trend) with diffusion (random Brownian motion component) to model equity prices with positive prices and lognormal returns.
- Mean-reverting processes (Ornstein-Uhlenbeck, Cox-Ingersoll-Ross) model interest rates, volatility, and commodity prices that tend to revert toward a long-run equilibrium — more realistic than GBM for interest rate term structure models.
- Jump processes (Poisson jumps, Levy processes) extend continuous diffusion models to allow discrete, sudden price movements — better capturing crash risk and fat tails observed in real asset prices.
- Itô's lemma is the stochastic calculus tool for finding the differential of a function of a stochastic process, enabling derivation of the Black-Scholes PDE and other fundamental results in derivatives pricing.
Explanation
Stochastic processes are the mathematical language of uncertainty in financial modeling. Every quantitative finance application — from options pricing to risk management VaR to algorithmic trading signal generation — ultimately rests on assumptions about the stochastic process governing asset prices, rates, or other market variables. The choice of process determines which market behaviors can and cannot be captured, and understanding the properties of different process types is essential for model selection and risk management.
The classification of stochastic processes follows several dimensions. Discrete vs. continuous time: discrete-time processes evolve at fixed intervals (ARIMA models, GARCH) while continuous-time processes (Brownian motion, diffusion processes) are defined for all instants. Discrete vs. continuous state space: binomial trees use discrete price states while geometric Brownian motion allows continuous price evolution. Markov vs. non-Markov: a Markov process has the property that future evolution depends only on the current state, not on the history of how the current state was reached — most financial models are Markov, simplifying computation enormously.
Geometric Brownian motion is the workhorse of financial modeling. Its key properties include: non-negativity of prices (a critical property — stock prices cannot go negative); lognormal distribution of returns over any time horizon; scaling of variance with time (variance = σ²T for horizon T); and the martingale property under the risk-neutral measure used for derivatives pricing. The Black-Scholes option pricing model derives its famous formula from the assumption that the underlying follows GBM — specifically, from solving the Black-Scholes PDE derived using Itô's lemma and no-arbitrage arguments.
Mean-reverting processes are essential for modeling interest rates, volatility, and commodity prices. The Ornstein-Uhlenbeck (OU) process — the continuous-time limit of an AR(1) process — has the form: dX = θ(μ - X)dt + σdW, where θ is the speed of mean reversion, μ is the long-run equilibrium level, and σ is volatility. This process is stationary (its distribution stabilizes over time) and captures the empirical tendency of interest rates and implied volatility to cycle around a central tendency rather than drift without bound. The Vasicek model applies the OU process to short-term interest rates; the CIR model adds square-root diffusion to ensure non-negative rates.
Stochastic volatility models address the well-documented empirical failure of constant-volatility GBM: real markets exhibit volatility clustering (large moves follow large moves), volatility mean reversion, and a negative correlation between price and volatility for equities (the leverage effect). The Heston model introduces a second stochastic process for variance: dV = κ(θ - V)dt + ξ√V dW_V, correlated with the asset price process — allowing for the volatility smile observed in equity options markets while maintaining analytical tractability through a characteristic function solution.
Formula
Geometric Brownian Motion: dS = μS dt + σS dW, where W is Brownian motion; S(T) = S(0)·exp((μ - σ²/2)T + σ√T·Z), Z ~ N(0,1)
Example
An options trader at a hedge fund needs to price a 6-month at-the-money call option on an equity with current price $100, risk-free rate 4%, and historical volatility 25%. Under geometric Brownian motion: dS = 0.04 × S × dt + 0.25 × S × dW. Using Black-Scholes (the analytical solution of the GBM-based pricing PDE), the call option price is $10.20. The trader also wants to assess the probability of the stock reaching $130 within 6 months. Under GBM, log(S/S₀) ~ N(0.0325 × 0.5, 0.25² × 0.5) = N(0.01625, 0.03125), so log(130/100) = 0.2624. P(S > 130) = P(Z > (0.2624 - 0.01625)/0.1768) = P(Z > 1.39) ≈ 8.2%. This probability estimate is used for stress testing and scenario analysis. The trader then extends the model to a Heston stochastic volatility framework to capture the observed implied volatility skew, finding that the same option prices at $10.85 under Heston — the 65-cent difference reflecting the premium for stochastic volatility risk.
Related terms
Algorithmic Trading Arbitrage At The Money Autoregressive Model Brownian Motion Call Option Correlation Equity Fundamental Law Of Active Management Geometric Brownian Motion Hedge Fund Historical Volatility