CS2 – Risk Modelling and Survival Analysis · IFoA (Institute and Faculty of Actuaries) · 6 min read
Introduction: Making Sense of Randomness
Welcome to the start of your journey into Stochastic Processes! If the name sounds a bit intimidating, don't worry—you already understand the concept intuitively. In the world of Actuarial Science, we rarely know exactly what will happen next. We don't know the exact day a claim will arrive or what a stock price will be tomorrow.
A stochastic process is simply a mathematical way to model how a random variable changes over time. Think of it as a "random movie" rather than a "random snapshot." By the end of this chapter, you’ll be able to classify different types of processes and understand the fundamental properties that make them tick.
1. What is a Stochastic Process?
At its simplest, a stochastic process is a collection of random variables indexed by time. We usually write this as \( \{X_t : t \in T\} \).
The Ingredients:
1. The State Space (\(S\)): This is the set of all possible values the random variable can take. If we are counting the number of people in a queue, the state space is \( \{0, 1, 2, ...\} \).
2. The Index Set (\(T\)): This represents time. It could be discrete (e.g., every day at noon) or continuous (e.g., any moment in time).
Real-World Analogy:
Imagine you are tracking the height of a tide. The "State Space" is any possible height (e.g., 0 to 10 meters). The "Index Set" is the time of day. Because the tide can be measured at any micro-second, this is a continuous time process.
2. Classifying Stochastic Processes
One of the most common exam tasks is identifying the type of process. We classify them based on whether Time and State Space are Discrete or Continuous.
A. Discrete Time, Discrete State Space
Example: The number of claims received by an insurer each day.
Time moves in steps (Day 1, Day 2), and the count of claims is a whole number.
B. Continuous Time, Discrete State Space
Example: The total number of claims received from the start of the year up to time \(t\).
Claims can happen at any moment (continuous time), but the total count is still a whole number (discrete state).
C. Discrete Time, Continuous State Space
Example: The closing price of a share at the end of each day.
Time is discrete (daily), but the price can be any decimal value (continuous).
D. Continuous Time, Continuous State Space
Example: The yield on a government bond measured constantly throughout the day.
Both the time and the value can take any value within a range.
Quick Review: The 2x2 Grid
- Discrete Time: Time is a sequence of points: \( t = 0, 1, 2... \)
- Continuous Time: Time is an interval: \( [0, \infty) \)
- Discrete State: Values are countable (0, 1, 2...).
- Continuous State: Values can be any real number in a range.
3. Key Properties of Stochastic Processes
To make these processes easier to work with, we look for specific "behaviors" or properties.
The Markov Property
This is arguably the most important concept in CS2. A process has the Markov Property if the future depends only on the present state, not on the path it took to get there.
The Goldfish Analogy:
Imagine a goldfish in a bowl. To predict where the goldfish will be in 10 seconds, you only need to know where it is right now. Knowing where it was 5 minutes ago doesn't help. The "memory" of the process is only one step deep.
Mathematically:
\( P(X_{t+n} = y | X_t = x, X_{t-1} = x_{t-1}, ..., X_0 = x_0) = P(X_{t+n} = y | X_t = x) \)
Stationarity
A process is "stationary" if its statistical properties do not change as time passes. There are two main types you need to know:
1. Strict Stationarity: The entire joint distribution remains the same regardless of a shift in time. This is very restrictive and rare in the real world.
2. Weak (or Second-Order) Stationarity: We only care about the first two "moments":
- The mean \( E[X_t] \) is constant (it doesn't depend on \(t\)).
- The covariance between two points \( Cov(X_t, X_{t+k}) \) depends only on the lag (\(k\)) between them, not the actual time (\(t\)).
Don't worry if this seems tricky...
Think of a stationary process like a fountain. While the individual water droplets are moving (randomness), the overall "shape" and "height" of the fountain look the same over time.
4. Increments: Changes Over Time
We often study the increments of a process, which is the change over a period: \( X_{t+s} - X_t \).
Independent Increments:
The changes in the process over non-overlapping time intervals are independent. For example, the number of claims that arrive between 9 AM and 10 AM has no impact on how many arrive between 2 PM and 3 PM.
Stationary Increments:
The distribution of the change depends only on the length of the time interval, not the start time. The number of claims in any 1-hour window follows the same distribution, whether it’s Monday morning or Friday night.
5. Common Examples to Remember
The Random Walk
Imagine a person standing on a line. Every second, they flip a coin. If it's Heads, they step right (+1). If it's Tails, they step left (-1).
- This is a discrete time, discrete state process.
- It is a classic example of a Markov Process.
The Poisson Process
This is used to model the arrival of events (like insurance claims or phone calls).
- It has independent increments.
- It is a continuous time, discrete state process.
6. Common Mistakes to Avoid
- Mixing up State and Time: Always ask yourself: "What am I measuring?" (State) and "When am I measuring it?" (Time).
- Assuming all processes are Markov: Some processes have "long memory" where the past 10 steps matter. These are NOT Markovian.
- Confusing Increments: Remember, Independent increments refer to the relationship between intervals, while Stationary increments refer to the distribution of the change itself.
Summary & Key Takeaways
- A stochastic process is a sequence of random variables over time.
- Processes are classified by their Time (Discrete/Continuous) and State Space (Discrete/Continuous).
- The Markov Property means the future is independent of the past, given the present.
- Weak Stationarity requires a constant mean and a covariance that only depends on the time lag.
- Independent increments mean what happens in one time slot doesn't affect another.
Pro-Tip for Revision: When looking at a new process, always try to sketch it on a graph. If the x-axis (time) is dots, it's discrete time. If the y-axis (values) is dots, it's a discrete state space!
Quick check
Can you answer these now?
Open each question to check the key ideas from this chapter.
How is a stochastic process formally defined in actuarial modelling?+
A stochastic process is a collection of random variables indexed by time, typically denoted as \( \{X_t : t \in T\} \).
What are the two fundamental components used to classify a stochastic process?+
The State Space (\(S\)), which is the set of possible values the variables can take, and the Index Set (\(T\)), which represents time.
On what basis is the 2x2 grid classification of stochastic processes formed?+
Processes are classified by whether Time (Index Set) is Discrete or Continuous and whether the State Space is Discrete or Continuous.
What does the Markov property imply about the memory of a stochastic process?+
The Markov property means the future of the process depends only on its current state, not on the path it took to get there.
What are the two conditions for a stochastic process to be weakly (second-order) stationary?+
The process must have a constant mean \( E[X_t] \) and a covariance \( Cov(X_t, X_{t+k}) \) that depends only on the lag \( k \), not on time \( t \).
What is the definition of independent increments in a stochastic process?+
A process has independent increments if the changes over non-overlapping time intervals are statistically independent of each other.
What defines a stochastic process with stationary increments?+
A process has stationary increments if the distribution of the change \( X_{t+s} - X_t \) depends only on the length of the interval \( s \), not the start time \( t \).
How is a simple Random Walk classified in terms of its time and state space?+
A simple Random Walk is a discrete-time, discrete-state space Markov process.
What are the key classification features of a Poisson process?+
A Poisson process is a continuous-time, discrete-state space process that features independent increments.
Ready to test yourself?
Turn these notes into exam-style practice. Get unlimited AI questions on this topic with instant marking and explanations.
* The content provided by thinka is generated by AI and may not always be accurate or up-to-date. Please use it as a supplementary resource and verify with official materials.
Put These Notes into Practice
Reading the notes is a great start. Now practise with unlimited AI-generated questions and get instant feedback. 100,000+ students are already improving their grades.