Girsanov's Theorem

Reweight the scenarios and a drift appears — but the volatility cannot be touched.

The idea

Theorem (Girsanov's theorem).

Let $W$ be a standard Brownian motion under $\mathbb{P}$ and fix a constant $\theta$. Reweight each path by $Z_{t} = \exp\!\left(-\theta W_{t} - \tfrac{1}{2}\theta^{2} t\right).$ Under the resulting measure $Q$, the process $W_{t} + \theta t$ is a standard Brownian motion.

Girsanov's theorem describes what changing the probability measure does to a Brownian motion. The question comes from risk-neutral pricing in finance, which values a claim as a discounted average under an invented probability — one under which every asset drifts at the risk-free rate — and leaves open both what justifies inventing a probability and what the invention is allowed to change.

Changing measure supplies the first answer. Keep every scenario that was possible and the value it produces; change only how much each scenario counts. Nothing is added and nothing is deleted, so the whole operation is described by one number per scenario: the factor by which its weight is multiplied.

Girsanov's theorem supplies the second answer, in continuous time. Reweighting the paths of a Brownian motion shifts its drift, and choosing $\theta$ places the drift wherever we need it. Equally important is what the theorem does not permit: the reweighted process is still a standard Brownian motion, with the same volatility as before. This asymmetry is what makes risk-neutral pricing work — the drift depends on which measure we price under, while the volatility is a property of the paths themselves, and no reweighting of paths can alter it.

Ways to work on it

Not sure where to start? Take the ten-question placement test.