Risk-neutral pricing¶
Derivative pricing in modern quantitative finance is primarily an arbitrage story, not a forecasting story. The key idea is that (under standard assumptions) we can price without ever estimating the real-world drift \(\mu\).
The martingale pricing principle¶
Fix a strictly positive traded asset as a numéraire. In the simplest single-currency setting we use the money-market account
and with constant \(r\),
In an arbitrage-free market, there exists an equivalent martingale measure \(\mathbb{Q}\) such that every traded asset price, when discounted by the numéraire, is a \(\mathbb{Q}\)-martingale. For a stock \(S_t\),
Equivalently, for \(0\le t\le T\),
Pricing formula¶
Let a claim pay \(H_T\) at time \(T\). If the market is arbitrage-free (and complete in the classic Black–Scholes setup), then its price process \(V_t\) satisfies
where \(D(t,T)=\frac{B_t}{B_T}=\exp\big(-\int_t^T r_u\,du\big)\) is the stochastic discount factor.
That is the “one-line” recipe behind most models: simulate / compute under \(\mathbb{Q}\) and discount.
Change of measure and the Radon–Nikodym derivative¶
A change of measure from \(\mathbb{P}\) to \(\mathbb{Q}\) is described by a Radon–Nikodym derivative (density) process \((Z_t)_{t\ge 0}\) such that
Intuitively: \(Z_t\) reweights paths so that the discounted traded assets lose their drift and become martingales.
From \(\mu\) to \(r\) in Black–Scholes (Girsanov in one page)¶
Under the real-world measure \(\mathbb{P}\), the Black–Scholes model assumes
with constant \(\sigma>0\).
Define the (constant) market price of risk
Then the exponential martingale
defines an equivalent measure \(\mathbb{Q}\) via \(d\mathbb{Q}=Z_T\,d\mathbb{P}\) (for each fixed horizon \(T\)). Under \(\mathbb{Q}\), Girsanov’s theorem implies
is a Brownian motion, and substituting into the SDE yields
So the drift becomes \(r\) under the pricing measure.
Practical notes¶
- Dividends: with a continuous dividend yield \(q\), the risk-neutral drift is \(r-q\).
- Different numeraires: changing numéraire (e.g., to a forward measure) changes \(\mathbb{Q}\) but keeps prices invariant.
- Curves: in real markets you typically use the appropriate discount curve for \(D(t,T)\); the martingale idea remains the same.
Where to go next¶
- The change-of-measure step uses Brownian motion; see Brownian motion and Itô.
- The resulting stock dynamics under \(\mathbb{Q}\) are solved explicitly in GBM.