Problem 5 - A Course in Financial Calculus

Question

Suppose that the martingale \(\left\\{M_{t}\right\\}_{0 \leq t \leq T}\) has bounded quadratic variation and \(\left\\{A_{t}\right\\}_{0 \leq t \leq T}\) is Lipschitz-continuous. Let \(S_{t}=M_{t}+A_{t} .\) By analogy with Definition 4.2.2, we define the quadratic variation of \(\left\\{S_{t}\right\\}_{0 \leq t \leq T}\) over \([0, T]\) to be the random variable \([S]_{T}\) such that for any sequence of partitions \(\left\\{\pi_{n}\right\\}_{n \geq 1}\) of \([0, T]\) with \(\delta\left(\pi_{n}\right) \rightarrow 0\) as \(n \rightarrow \infty\) $$ \mathbb{E}\left[\left|\sum_{j=1}^{N(\pi)}\right| S_{t_{j}}-\left.S_{t_{j-1}}\right|^{2}-\left.[S]_{T}\right|^{2}\right] \rightarrow 0 \quad \text { as } n \rightarrow \infty $$

Step-by-Step Solution

Answer

The quadratic variation of \( S_t = M_t + A_t \) is \([S]_T = [M]_T\).

Step 1: Understanding the Problem

We need to define the quadratic variation of the process \( \{S_t\} \) where \( S_t = M_t + A_t \). The process \( M_t \) is a martingale with bounded quadratic variation, and \( A_t \) is Lipschitz-continuous.

Step 2: Quadratic Variation of a Martingale

Since \( M_t \) is a martingale with bounded quadratic variation, we denote its quadratic variation by \([M]_t\). This is the random variable that satisfies \( \sum (M_{t_{j}} - M_{t_{j-1}})^2 \rightarrow [M]_T \) as the grid gets finer.

Step 3: Property of Lipschitz Function

Given \( A_t \) is Lipschitz-continuous, the change \( A_{t_j} - A_{t_{j-1}} \) is bounded by a constant times \( |t_j - t_{j-1}| \). Therefore, its quadratic variation tends to 0.

Step 4: Quadratic Variation of \( S_t \)

To find the quadratic variation \([S]_T\), note \( S_t = M_t + A_t \). Since the quadratic variation of a continuous function like \( A_t \) contributes nothing and \( M_t \) already has bounded quadratic variation, \([S]_T = [M]_T\).

Step 5: Conclusion

Thus, for partitions \( \pi_n \) satisfying \( \delta(\pi_n) \to 0 \), we have \( \mathbb{E}[|\sum (S_{t_j} - S_{t_{j-1}})^2 - [M]_T|] \rightarrow 0 \) as \( n \to \infty \).

Key Concepts

Martingale

In probability theory, a martingale is a model describing a fair game. Imagine flipping a fair coin multiple times, where you win $1 for heads and lose $1 for tails. A martingale is a sequence of random variables that captures this idea of a fair game. Formally,

• the expectation of the next outcome, given all prior outcomes, equals the present outcome.
• This means that, if you gamble over time, your expected wealth at the next stage is equal to your current wealth.
• The mathematical significance of a martingale is in its use in various fields such as finance, in modeling fair price processes.

Martingales have an interesting property known as quadratic variation. This represents how much the process "wiggles" over time. Simply put, it measures the total accumulated change squared, highlighting volatility without worrying about the direction. Understanding martingales is crucial for grasping complex stochastic processes and their applications.

Lipschitz Continuity

Lipschitz continuity is a refinement of the concept of continuity for functions. It ensures that a function does not oscillate too wildly. More technically, a function \( f(x) \) is Lipschitz continuous if there exists a constant \( L \) such that for every pair of points \( x \) and \( y \), \(|f(x) - f(y)| \leq L |x-y|\).

• The smaller the constant \( L \), the more control we have over the changes in the output relative to changes in the input.
• This property is crucial in ensuring stability in solutions to differential equations and other mathematical systems.
• For example, a graph of a Lipschitz continuous function exhibits a bounded slope, which does not exceed the predefined constant \( L \).

Understanding Lipschitz continuity helps establish limits on how a process like \( A_t \) changes over time, assuring that its contribution to quadratic variation becomes negligible in contrast with martingales.

Partitioning Sequences

Partitioning sequences is a method used in mathematical analysis to break down intervals into smaller segments. Imagine you want to measure the distance traveled by an object over time. Instead of looking at the entire journey, you split this time into smaller parts and analyze each segment. In the context of quadratic variation:

• A partition \( \pi_n \) of an interval \( [0, T] \) is a sequence \( 0=t_0
• These partitions help approximate continuous-time stochastic processes by breaking them into discrete, manageable pieces.
• As the sequence becomes finer (i.e., the largest gap \( \delta(\pi_n) \) decreases), it provides a more precise representation of the continuous process.

When combined with martingales, partitioning sequences allow us to approximate the quadratic variation of complex processes. Understanding this concept aids in computations and ensuring approximations are accurate as limits approach infinity.