Stochastic Processes & Asset Pricing

1. The Limits of Static Models

Before we define a stochastic process, let's first understand why we need it. Imagine you take all the daily returns of Bitcoin over the past five years and plot a histogram. You'll get a single distribution. However, this single picture hides a crucial fact: the nature of those returns was not the same in 2020 as it was in 2023. The underlying "data-generating process" changes.

The following video segment provides an excellent visualization of this very problem.

As the video shows, by collapsing a time-varying process into a single distribution, we lose all information about its evolution. We can't distinguish between different market regimes (e.g., bull market vs. bear market, high volatility vs. low volatility). This is why a more dynamic approach is necessary.

2. Defining a Stochastic Process

A stochastic process is the mathematical tool for describing randomly evolving systems. There are a couple of ways to define it.

The Formal and Intuitive Definitions

Let's start with a formal definition from an MIT OpenCourseWare lecture, which will appeal to your preference for precise, foundational concepts.

To build on that, let's get a more intuitive feel for what this means. What is a "path"? How does a collection of random variables generate one?

Let's synthesize these two perspectives.

• Definition: A stochastic process, $\{X_t\}$, is a family of random variables indexed by time. For any specific time $t$, $X_t$ is a random variable with its own probability distribution.
• Sample Path: A single realization of a stochastic process over a period of time is called a sample path. If you plot the price of BTC from yesterday to today, you are looking at one sample path.
• The Process: The stochastic process itself is the complete description of all possible paths that could have occurred, and the probability of each one.

From your perspective as a developer, you can think of a stochastic process as a generative algorithm. Imagine a for loop that iterates through time steps. In each iteration t, the loop calls a function that returns a random number drawn from a distribution P(t). The sequence of numbers generated by this loop is one sample path. The full specification of the loop and the P(t) function for all t defines the stochastic process.

3. Fundamental Examples

To make this concept tangible, let's look at the two most important stochastic processes in finance.

a) The Simple Random Walk (Discrete Time)

The simplest and most intuitive stochastic process is the simple random walk. It's the foundation for many more complex models.

• Imagine a game where you toss a fair coin at discrete time steps $t=1,2,3,\dots$.
• If it's heads, you take a step up (+1). If it's tails, you take a step down (-1).
• Your position at time $T$, let's call it $X_T$, is the sum of all your steps up to that point: $X_T={\sum }_{i=1}^T{Y}_i$, where $Y_i$ is the outcome of the $i$-th coin toss.

The sequence of positions $\{X_0,X_1,X_2,\dots \}$ is a simple random walk.

The following video from the MIT course formally introduces this process and discusses its properties.

b) Brownian Motion (Continuous Time)

The random walk operates in discrete time steps. If we take the conceptual limit of this process as the time steps become infinitesimally small, we arrive at its continuous-time counterpart: Brownian motion (also known as a Wiener process). This is arguably the most important stochastic process in all of finance.

Brownian motion, denoted $W_t$, has the following key properties:

1. $W_0=0$.
2. It has independent increments.
3. The increment $W_t-W_s$ (for $t>s$) is normally distributed with a mean of 0 and a variance of $t-s$. That is, $W_t-W_s\sim \mathcal{N}(0,t-s)$.
4. Its sample paths are continuous everywhere but differentiable nowhere.

The third property is the most critical. It means that the uncertainty, or variance, of the process grows linearly with time. The further you try to predict into the future, the wider the range of possible outcomes becomes. This aligns perfectly with our real-world experience of financial markets.

4. Application: Modeling Asset Prices

So, how do we use this to model an asset price? A common (though simplified) first approach is to model the price process using Arithmetic Brownian Motion. This model proposes that the change in a stock price $S_t$ is composed of two parts: a deterministic "drift" and a random "shock".

The model is written in differential form as:

Where:

• $d{S}_t$ is the infinitesimal change in the asset price.
• $\mu$ is the drift parameter, representing the average rate of return.
• $dt$ is the infinitesimal change in time.
• $\sigma$ is the volatility parameter, representing the magnitude of the random fluctuations.
• $d{W}_t$ is the increment of a standard Brownian motion process, representing the source of randomness.

This simple model beautifully captures the essence of why stochastic processes are used: it decomposes price movement into a predictable trend ($\mu dt$) and an unpredictable, random component ($\sigma d{W}_t$). While real-world models are more sophisticated (e.g., Geometric Brownian Motion, which we will encounter later), they are all built upon this fundamental idea.

The next clip shows this exact application and explains how such a model is used to price financial instruments, connecting the abstract theory to a concrete financial problem.

Conclusion

This lesson concludes our foundational module on probability and statistics. We've moved from basic concepts like expected value and variance to understanding how assets relate via correlation, and finally, to a dynamic framework for modeling their evolution over time.

Key Takeaways:

• Asset price behavior is non-stationary; its statistical properties like mean and variance change over time, making static models insufficient.
• A stochastic process is a collection of random variables indexed by time, which provides a mathematical framework for modeling such time-varying systems.
• A sample path is a single realization of a stochastic process, like a historical price chart. The process itself describes all possible paths.
• The Simple Random Walk (discrete-time) and Brownian Motion (continuous-time) are the foundational stochastic processes for finance.
• Brownian motion is crucial because its variance grows linearly with time, capturing the idea that uncertainty increases the further we look into the future.
• Asset prices are modeled as stochastic processes (like Arithmetic Brownian Motion) by representing their movements as a combination of a deterministic drift and a random shock.

Preview of the Next Module:

We now have the essential statistical vocabulary to describe and analyze financial data. In our next module, The Quant's Toolkit: Python for Finance, we will shift gears and begin mastering the practical tools needed to implement these ideas. We'll start by diving into NumPy, the fundamental package for scientific computing in Python, and explore how its powerful, vectorized array operations allow for the high-performance computations required in quantitative analysis.