Problem 2 - A Course in Financial Calculus

Question

Let \(\left\\{\mathcal{F}_{t}\right\\}_{t \geq 0}\) be the natural filtration associated with a \(\mathbb{P}\)-Brownian motion \(\left\\{W_{t}\right\\}_{t \geq 0}\). Show that if \(\mathbb{Q}\) is a probability measure equivalent to \(\mathbb{P}\) and \(H_{T}\) is an \(\mathcal{F}_{T}\)-measurable random variable with \(\mathbb{E}^{\mathbb{Q}}\left[H_{T}^{2}\right]<\infty\) then $$ M_{t} \triangleq \mathbb{E}^{\mathbb{Q}}\left[H_{T} \mid \mathcal{F}_{t}\right] $$ defines a square-integrable \(\mathbb{Q}\)-martingale.

Step-by-Step Solution

Answer

\(M_t\) is a square-integrable \(\mathbb{Q}\)-martingale by the conditional expectation properties and given conditions.

Step 1: Understand the Problem

We need to show that the process \(M_t\), defined by \(M_{t} = \mathbb{E}^{\mathbb{Q}}[H_{T} \mid \mathcal{F}_{t}]\), is a square-integrable \(\mathbb{Q}\)-martingale. This involves verifying the martingale properties under the measure \(\mathbb{Q}\).

Step 2: Recall Martingale Properties

A process \(M_t\) is a \(\mathbb{Q}\)-martingale if for all \(0 \leq s \leq t \leq T\), \(\mathbb{E}^{\mathbb{Q}}[M_t \mid \mathcal{F}_s] = M_s\) and \(M_t\) is \(\mathcal{F}_t\)-measurable. Additionally, \(M_t\) is square-integrable if \(\mathbb{E}^{\mathbb{Q}}[M_t^2] < \infty\) for all \(t \leq T\).

Step 3: Apply the Martingale Conditional Expectation Property

To confirm \(M_t\) is a martingale, use the tower property of conditional expectation: \(\mathbb{E}^{\mathbb{Q}}[M_t \mid \mathcal{F}_s] = \mathbb{E}^{\mathbb{Q}}[\mathbb{E}^{\mathbb{Q}}[H_T \mid \mathcal{F}_t] \mid \mathcal{F}_s] = \mathbb{E}^{\mathbb{Q}}[H_T \mid \mathcal{F}_s] = M_s\). This shows \(M_t\) satisfies the martingale property.

Step 4: Verify \(\mathbb{Q}\)-Measurability and Square Integrability

Since \(H_T\) is \(\mathcal{F}_T\)-measurable, \(M_t = \mathbb{E}^{\mathbb{Q}}[H_T | \mathcal{F}_t]\) is \(\mathcal{F}_t\)-measurable. The condition \(\mathbb{E}^{\mathbb{Q}}[H_T^2]<\infty\) implies \(\mathbb{E}^{\mathbb{Q}}[M_t^2] \leq \mathbb{E}^{\mathbb{Q}}[H_T^2]<\infty\) ensuring square-integrability.

Key Concepts

Understanding Martingale

A martingale is a special kind of stochastic process that models a fair game.This means that the expectation of your future winnings, given all past information, is equal to your current winnings.For a process \( M_t \) to be a \( \mathbb{Q} \)-martingale:

• \( M_t \) must be \( \mathcal{F}_t \)-measurable, ensuring that the present value depends only on the current information.
• The conditional expectation \( \mathbb{E}^{\mathbb{Q}}[M_t \mid \mathcal{F}_s] \) must equal \( M_s \) for any times \( s \leq t \).
• The process should be square-integrable, meaning \( \mathbb{E}^{\mathbb{Q}}[M_t^2] < \infty \) for all \( t \).

This property is useful because it helps model situations where knowledge of past events doesn't affect the expected future gain.Martingales play a crucial role in finance, often used in pricing derivative securities and risk management.

Deep Dive into Conditional Expectation

Conditional expectation is like finding an average value of a random variable, given some known information.In the context of our problem, the process \( M_t = \mathbb{E}^{\mathbb{Q}}[H_T \mid \mathcal{F}_t] \) uses conditional expectation to "predict" \( H_T \) based on the current information set \( \mathcal{F}_t \).

• It captures the expected value of \( H_T \) conditioned on the filtration up to time \( t \).
• The process uses the 'tower property', meaning the expectation can be further conditioned on a prior time, keeping it consistent with past information.

This expectation principle is foundational in both probability theory and its applications, allowing us to make informed predictions in uncertain environments.

Demystifying Probability Measure

A probability measure is a function that assigns probabilities to events within a given probability space.It must satisfy specific properties such as non-negativity, normalization (total probability is 1), and countable additivity.In our context, \( \mathbb{P} \) and \( \mathbb{Q} \) represent different probability measures.

• \( \mathbb{P} \) is associated with the original Brownian motion, defining its inherent randomness.
• \( \mathbb{Q} \) is an equivalent measure to \( \mathbb{P} \), but might reflect different underlying "pricing" or "risk-adjusted" scenarios.
• Equivalent measures have the same null sets but differ in assigning probabilities to other events.

These measures allow us to translate theoretical probability concepts into practical scenarios, such as financial modeling and risk assessment.

Exploring Filtration in Probability

Filtration embodies the idea of information accumulation over time in stochastic processes.\( \left\{\mathcal{F}_{t}\right\}_{t \geq 0} \) is a family of \( \sigma \)-algebras, each \( \mathcal{F}_t \) representing all information available up to time \( t \).

• A filtration must satisfy that if \( s \leq t \), then \( \mathcal{F}_s \subseteq \mathcal{F}_t \), meaning information is non-decreasing over time.
• In our exercise, filtration allows us to understand how our predictions (the martingale) work throughout time, conditioned on increasing information.
• This concept ensures that stochastic processes like Brownian motion adhere to a strict mathematical structure that reflects the evolution of information.

Filtrations are indispensable for understanding complex systems evolving over time, providing a framework for modeling and analysis in statistics and finance.