Problem 13 - A Course in Financial Calculus

Question

Suppose that \(\left\\{N_{t}\right\\}_{t \geq 0}\) is a Poisson process whose intensity under \(\mathbb{P}\) is \(\left\\{\lambda_{t}\right\\}_{t \geq 0} .\) Show that \(\left\\{M_{t}\right\\}_{t \geq 0}\) defined by $$ M_{t}=N_{t}-\int_{0}^{t} \lambda_{s} d s $$ is a \(\mathbb{P}\)-martingale with respect to the \(\sigma\)-field generated by \(\left\\{N_{t}\right\\}_{t \geq 0}\)

Step-by-Step Solution

Answer

\( \{M_t\} \) is a \( \mathbb{P} \)-martingale with respect to the filtration generated by the Poisson process.

Step 1: Define a Martingale

A process \( \{M_t\}_{t \geq 0} \) is a martingale with respect to a filtration \( \{ \mathcal{F}_t\}_{t \geq 0} \) if for all \( s < t \), \( \mathbb{E}[M_t | \mathcal{F}_s] = M_s \). Here, the filtration is generated by \( \{N_t\}_{t \geq 0} \), the Poisson process.

Step 2: Express the Expectation of the Increment

Consider the increment \( M_t - M_s = N_t - N_s - \int_{s}^{t} \lambda_u du \). By the properties of the Poisson process, \( N_t - N_s \) is a Poisson random variable with mean equal to \( \int_{s}^{t} \lambda_u du \).

Step 3: Compute the Conditional Expectation

Given \( \mathcal{F}_s \), the expected increment is \( \mathbb{E}[N_t - N_s | \mathcal{F}_s] = \int_{s}^{t} \lambda_u du \). Therefore, \( \mathbb{E}[M_t - M_s | \mathcal{F}_s] = \mathbb{E}[N_t - N_s - \int_{s}^{t} \lambda_u du | \mathcal{F}_s] = 0 \).

Step 4: Verify Martingale Property

Since \( \mathbb{E}[M_t | \mathcal{F}_s] = M_s + \mathbb{E}[M_t - M_s | \mathcal{F}_s] = M_s \), this satisfies the martingale requirement \( \mathbb{E}[M_t | \mathcal{F}_s] = M_s \). Hence, \( \{M_t\}_{t \geq 0} \) is a \( \mathbb{P} \)-martingale with respect to \( \{ \mathcal{F}_t \}_{t \geq 0} \).

Key Concepts

Poisson process

A Poisson process is an essential concept in probability theory and stochastic processes. It models a sequence of events occurring randomly over time. These events are ideal for phenomena like the number of phone calls received at a call center or the arrival of buses at a stop. One important feature of a Poisson process is that the number of events in different intervals are independent. This means observing events in one time interval doesn't affect the likelihood of events in another.
A standard Poisson process has a rate parameter, denoted usually by \( \lambda \), representing the average number of occurrences in a unit time interval. The number of events that occur in time interval \([0, t]\) is a Poisson random variable with mean \( \lambda t \).

• **Independent Increments**: The numbers of events in disjoint time intervals are independently distributed.
• **Stationary Increments**: The distribution of the number of events depends only on the length of the time interval.

Understanding these properties is crucial as they form the basis for further analysis and usage in various probabilistic models.

conditional expectation

Conditional expectation is a central tool in probability and statistics, representing the expected value of a random variable given some condition or event. In simple terms, it's like saying, "What do we expect to happen, knowing this new information?"
For instance, if you're expecting rain tomorrow based on a weather forecast, the conditional expectation would be your predicted chance of rain considering this updated forecast.
Mathematically, conditional expectation \( \mathbb{E}[X | Y] \) tells us the expected or mean value of a random variable \( X \) when the variable \( Y \) provides additional information.
In the context of Poisson processes and martingales, conditional expectations are employed to ensure the increment of the process between times \( s \) and \( t \) is zero, basing this expectation on past events described by the filtration. This concept is vital to ensure a process like \( M_t \) remains a martingale.
Key aspects include:

• The expectation is "conditional" on the information available up to a certain point, usually represented by a \( \sigma \)-field.
• Useful in refining predictions by incorporating known events into the calculation of future expectations.

filtration

Filtration in probability theory refers to the increasing sequence of \( \sigma \)-fields that represent the accumulation of information over time. This concept is fundamental to the study of stochastic processes, as it provides a structured way to describe the progression of what's known at any given point.
In the case of a Poisson process, the filtration generated by this process—denoted as \( \{ \mathcal{F}_t \}_{t \geq 0} \)—represents all the information available up to time \( t \). With each increment of time, \( \mathcal{F}_t \) includes all earlier data points collected from the process.
The role of filtration is crucial in defining and understanding martingales. With correct filtration, martingales can detect changes that are considered irrelevant or noise relative to the history of the process.
Important points include:

• Each \( \mathcal{F}_t \) contains all information known up to time \( t \), and \( \mathcal{F}_t \subset \mathcal{F}_s \) for \( t < s \).
• Used extensively in financial mathematics for modeling stock prices and risk assessment.

stochastic processes

Stochastic processes are mathematical objects used to predict the evolution of systems over time that are inherently random. These processes are widely applicable in fields such as finance, physics, and biology, essentially anywhere you must model uncertain, dynamic systems.
They're like functions that add randomness to a sequence of events. For example, in the real world, determining the future price of a stock, predicting weather changes, or estimating the spread of a disease employ stochastic processes.
The Poisson process, as discussed, is a good example of a simple stochastic process, as it describes counting events randomly scattered over time. More complex examples involve Brownian motion or Markov processes, each with distinct properties and applications.

• **Random Sequences**: A sequence or collection of random variables indexed by time or space.
• **Applications**: Includes predicting stock prices in finance, modeling populations in ecology, or determining the behavior of particles in physics.

Each stochastic process has certain defined rules or equations governing its distribution and evolution, helping researchers and professionals make informed decisions based on the probabilistic models. Understanding the different processes and their properties can greatly aid in analyzing systems where uncertainty is a dominating feature.