geometric Brownian motion
A mathematical model assuming stock prices fluctuate randomly and continuously, like smoke diffusing, with returns following a normal distribution; used as the risk assumption underlying CAPM and Black-Scholes.
As defined in Emanuel Derman on Theories, Models, and Science
Where it comes up · 1
Most important claims · 5 of 5Search all claims for “geometric Brownian motion” →
- The brilliant engineering idea of Black-Scholes is that you can create the risk of an option by borrowing money and buying the stock—building a recipe that replicates an option's payoff—which is correct given geometric Brownian motion; the science component (defining the stock's risk via Brownian motion) is the flawed part, but the replication insight is sound.Emanuel Derman on Theories, Models, and Science
- CAPM holds that you should only get paid for taking risk you have to take—any risk that can be diversified or hedged away should not be compensated because you didn't have to take it; since individual stock risk can be canceled out in a large portfolio, the only irreducible risk is the whole market's risk, captured by a stock's beta or correlation with the market.Emanuel Derman on Theories, Models, and Science
- The faulty part of CAPM and related models is the assumption that stock returns satisfy geometric Brownian motion and that risk is a well-defined quantity captured by a single standard deviation or volatility; while it has the qualities of something risky, real-world risk—with jumps, panics, and contagion—has attributes that standard deviation cannot capture.Emanuel Derman on Theories, Models, and Science
- Using high-level mathematical concepts like Brownian motion to fancy-up financial models is scientism rather than science—it gives the patina or verisimilitude of science to what may be a bald and unconvincing narrative, dressing it up to look fancy without making it true.Emanuel Derman on Theories, Models, and Science
- Theories are attempts to describe the world the way it really is, framed mathematically but not as analogies (like Newton's Laws), whereas models are essentially metaphors—limited analogies mapping something you want to understand onto something you already understand, like comparing stock price diffusion to smoke diffusing.Emanuel Derman on Theories, Models, and Science