Chapter 5 Risk-Neutral Pricing and Martingales

5.1 Martingales: Fair Games

A stochastic process \(M(t)\) is a martingale if:

\[E[M(t) \mid \mathcal{F}_s] = M(s) \quad \text{for all } t \geq s\]

where \(\mathcal{F}_s\) represents all information up to time \(s\).

Intuitive meaning: The best prediction of tomorrow’s value is today’s value. No drift, just random fluctuations. It’s a “fair game” - you can’t make money on average.

Key property: If \(M(t)\) is a martingale, then \(E[M(t)] = E[M(0)]\) for all \(t\).

5.1.1 Examples

  • Brownian motion: \(B(t)\) is a martingale
  • Itô integral: \(\int_0^t f(s) \, dB(s)\) is a martingale (if \(f\) satisfies certain conditions)
  • Exponential martingale: \(\exp(\sigma B(t) - \frac{1}{2}\sigma^2 t)\) is a martingale
  • Casino winnings: Your wealth in a fair game

5.2 The Fundamental Theorem of Asset Pricing

This is the nuclear bomb of mathematical finance:

Theorem: A market is arbitrage-free if and only if there exists a probability measure \(\mathbb{Q}\) (the risk-neutral measure) under which all discounted asset prices are martingales.

Let me unpack this carefully.

5.2.1 What’s Arbitrage?

An arbitrage is a “free lunch” - a trading strategy that: 1. Costs nothing to set up 2. Never loses money 3. Sometimes makes money

Markets without arbitrage are called arbitrage-free. This is a minimal rationality assumption.

5.2.2 The Risk-Neutral Measure

In the real world (measure \(\mathbb{P}\)), a stock might follow:

\[dS = \mu S \, dt + \sigma S \, dB\]

where \(\mu\) is the real drift (could be 8%, 12%, whatever).

The discounted price is \(\tilde{S}(t) = e^{-rt}S(t)\) where \(r\) is the risk-free rate.

By Itô’s lemma: \[d\tilde{S} = \tilde{S}[(\mu - r)dt + \sigma \, dB]\]

This has drift \(\mu - r\). If \(\mu \neq r\), this is NOT a martingale under \(\mathbb{P}\).

The trick: Change to a different probability measure \(\mathbb{Q}\) where the drift vanishes!

5.2.3 Girsanov’s Theorem

Under \(\mathbb{Q}\), we define a new Brownian motion:

\[\tilde{B}(t) = B(t) + \frac{\mu - r}{\sigma}t\]

By Girsanov’s theorem (deep measure theory), \(\tilde{B}(t)\) is a Brownian motion under \(\mathbb{Q}\).

Then under \(\mathbb{Q}\): \[d\tilde{S} = \tilde{S}\sigma \, d\tilde{B}\]

No drift! The discounted stock price is a martingale under \(\mathbb{Q}\).

5.3 Risk-Neutral Pricing Formula

Since \(\tilde{S}(t) = e^{-rt}S(t)\) is a \(\mathbb{Q}\)-martingale:

\[e^{-rt}S(t) = E^{\mathbb{Q}}[e^{-rT}S(T) \mid \mathcal{F}_t]\]

For a derivative with payoff \(g(S(T))\) at time \(T\), no-arbitrage requires:

\[\boxed{V(t) = e^{-r(T-t)}E^{\mathbb{Q}}[g(S(T)) \mid S(t)]}\]

This is the universal pricing formula!

5.3.1 The Miracle

  • The real drift \(\mu\) doesn’t appear!
  • Risk preferences are already encoded in \(S(t)\)
  • Only need: \(S(t)\), \(r\), \(\sigma\), and payoff \(g\)

Under \(\mathbb{Q}\), assets grow at rate \(r\) (not \(\mu\)):

\[dS = rS \, dt + \sigma S \, d\tilde{B}\]

This is why it’s called “risk-neutral” - it’s as if investors are indifferent to risk.

5.4 Why Does This Work?

The key insight: replication.

If you can replicate the derivative’s payoff using the underlying asset, then to avoid arbitrage, the derivative’s price must equal the replication cost.

The risk-neutral measure emerges from imposing no-arbitrage across all possible replication strategies.

5.5 Example: Forward Contracts

A forward contract obliges you to buy stock at price \(K\) at time \(T\). Payoff: \(S(T) - K\).

Price: \[V(t) = e^{-r(T-t)}E^{\mathbb{Q}}[S(T) - K]\] \[= e^{-r(T-t)}[E^{\mathbb{Q}}[S(T)] - K]\] \[= e^{-r(T-t)}[S(t)e^{r(T-t)} - K]\] \[= S(t) - Ke^{-r(T-t)}\]

The forward price (fair delivery price) is \(F = S(t)e^{r(T-t)}\).

5.6 Change of Numeraire

You can use any traded asset as the “numeraire” (unit of account).

If you use asset \(N(t)\) as numeraire, there exists a measure \(\mathbb{Q}^N\) under which all prices relative to \(N(t)\) are martingales:

\[\frac{S(t)}{N(t)} \text{ is a } \mathbb{Q}^N\text{-martingale}\]

Common choices: - Money market account: \(N(t) = e^{rt}\) → standard risk-neutral measure - Zero-coupon bond: Useful for interest rate derivatives - Stock itself: Useful for options on options

This technique is powerful for simplifying derivative pricing.

5.7 Market Price of Risk

The drift shift from \(\mathbb{P}\) to \(\mathbb{Q}\) is governed by the market price of risk \(\lambda\):

\[\lambda = \frac{\mu - r}{\sigma}\]

This measures the excess return per unit of volatility. Under \(\mathbb{Q}\):

\[d\tilde{B} = dB + \lambda \, dt\]

Different assets may have different market prices of risk. In a complete market, all risks are spanned by traded assets, so there’s a unique \(\mathbb{Q}\).

5.8 Incomplete Markets

If not all risks can be hedged (e.g., jumps, stochastic volatility without tradable volatility derivatives), the market is incomplete.

Then: - Multiple risk-neutral measures exist - No unique price for derivatives - Need to choose \(\mathbb{Q}\) based on additional criteria (e.g., minimize risk, match market prices)

This is a major research area in quantitative finance.

5.9 Summary

The martingale approach to pricing:

  1. Identify the source of randomness (Brownian motions)
  2. Find the risk-neutral measure (Girsanov theorem)
  3. Price as discounted expectation under \(\mathbb{Q}\)

This elegant framework unifies all of derivative pricing and connects probability theory to finance in a profound way.