Welcome to the World of Distribution Parameters!

In your journey through Exam ASTAM, you’ve likely noticed that actuarial models aren't just static equations; they are living tools that must adapt to a changing world. One of the most common things an actuary has to deal with is change—inflation, currency shifts, or policy adjustments.

In this chapter, we explore how changing the parameters of a distribution changes the "shape" and "size" of the claims we are modeling. This is crucial because it allows us to answer questions like: "If inflation increases all claims by 5%, what happens to our total expected loss?" Let’s dive in and make these concepts feel like second nature.


1. The "Scale" Parameter: Our Zoom Lens

When we talk about the effects of parameters, the most important concept to master is the Scale Parameter. Think of a scale parameter like the zoom lens on a camera. When you zoom in or out, the shape of the object doesn't change—your dog still looks like a dog—but the size of everything in the frame changes proportionally.

What is a Scale Parameter?

A parameter \( \theta \) is considered a scale parameter for a random variable \( X \) if the distribution of the new variable \( Y = X / \theta \) does not depend on \( \theta \). In simpler terms, if you can "factor out" the parameter by dividing, it's a scale parameter.

Why this matters for Severity Models:

In insurance, many distributions we use (like the Exponential, Gamma, and Pareto) have a scale parameter. Usually, this parameter is denoted as \( \theta \). If you change \( \theta \), you are effectively stretching or shrinking the horizontal axis of your probability distribution graph.

Key Properties of Scale Parameters:
  • If \( X \) has a scale parameter \( \theta \), then the random variable \( Y = cX \) (where \( c > 0 \)) will have the same distribution, but with the scale parameter changed to \( c\theta \).
  • Did you know? This is exactly how we model inflation. If \( c = 1.05 \), it represents a 5% inflation rate across all claim sizes.

Quick Summary: Changing a scale parameter moves the "weight" of the distribution but keeps its fundamental shape intact. It's like changing the units of measurement from dollars to euros.


2. Mathematical Effects on the Distribution Functions

Don't worry if the math looks intimidating at first! There is a very logical pattern to how parameters affect the Cumulative Distribution Function (CDF) and the Probability Density Function (PDF).

The Transformation Rule

Suppose we have a random variable \( X \) and we create a new variable \( Y = cX \). To find the new functions, follow these steps:

1. The New CDF:
\( F_Y(y) = F_X(y/c) \)
Think of it this way: To get the same probability level for the "inflated" variable \( Y \), you have to look at the original value divided by the inflation factor.

2. The New PDF:
\( f_Y(y) = \frac{1}{c} f_X(y/c) \)
Notice the \( 1/c \) in front! Because the distribution is "stretched" wider, the height must decrease so that the total area under the curve still equals 1.

Real-World Analogy:

Imagine a piece of pizza dough. If you stretch the dough to be twice as wide (\( c = 2 \)), the dough becomes thinner (the \( 1/c \) factor). The total amount of dough (the total probability of 1.0) stays exactly the same!


3. Impact on Moments (Mean, Variance, etc.)

One of the most common tasks in ASTAM is calculating how the mean and variance change when a parameter is adjusted. This is actually the easiest part!

Linearity of Moments

If we multiply our random variable by a constant \( c \) (representing a scale change):

  • The Mean: \( E[cX] = c \cdot E[X] \). (The mean moves directly with the scale).
  • The Variance: \( Var(cX) = c^2 \cdot Var(X) \). (Variance moves with the square of the scale).
  • The Raw Moments: \( E[(cX)^k] = c^k \cdot E[X^k] \).

The Coefficient of Variation (CV)

The CV is defined as \( \frac{\sigma}{\mu} \). Crucial Point: If you only change the scale parameter, the CV does not change.
\( CV(cX) = \frac{\sqrt{c^2 Var(X)}}{c E[X]} = \frac{c \sigma}{c \mu} = \frac{\sigma}{\mu} \).
This makes sense: if everything increases by 10%, the "relative" risk remains the same.

Key Takeaway: Scale parameters affect the size of the numbers but not the relative volatility of the distribution.


4. Recognizing Parameters in the Loss Models Table

For Exam ASTAM, you will have access to the "Table of Distributions." You don't need to memorize every PDF, but you must be able to identify which parameter is the scale parameter.

Common Scale Parameters (\( \theta \)):
  • Exponential: \( \theta \) is the mean and the scale parameter.
  • Gamma: \( \theta \) is the scale parameter; \( \alpha \) is the shape parameter.
  • Pareto: \( \theta \) is the scale parameter; \( \alpha \) is the shape parameter.
  • Weibull: \( \theta \) is the scale parameter; \( \tau \) is the shape parameter.

Quick Trick: In almost all ASTAM distributions, if you see an \( x \) in the formula, it is usually divided by \( \theta \) (like \( x/\theta \)). This is a dead giveaway that \( \theta \) is a scale parameter!


5. Shape Parameters: The "Personality" of the Distribution

While the scale parameter handles the size, the shape parameter (often \( \alpha \) or \( \tau \)) determines the "personality" of the distribution. It dictates how heavy the tails are.

Unlike scale parameters:

  • Changing a shape parameter does change the Coefficient of Variation.
  • Changing a shape parameter can make a distribution "heavy-tailed" (more likely to have huge claims) or "light-tailed" (more predictable).
  • Shape parameters are usually exponents in the PDF or CDF formulas.

Common Mistake to Avoid: When a problem says "inflation affects claims," you only adjust the scale parameter. You almost never change the shape parameter for simple inflation. If you change the shape, you are changing the underlying nature of the risk, not just the cost of the claims.


6. Summary and Quick Review

Let’s wrap up what we’ve learned to keep it fresh in your mind:

Quick Review Box:

  • Scale Parameter (\( \theta \)): Adjusts the size/magnitude. Found as \( x/\theta \) in formulas.
  • Inflation (\( c \)): To model a \( 10\% \) increase, replace \( \theta \) with \( 1.10\theta \).
  • CDF Effect: \( F_{new}(x) = F_{old}(x/c) \).
  • Mean Effect: Increases by factor \( c \).
  • CV Effect: No change!
  • Shape Parameter: Changes the "heaviness" of the tail and the CV.

Final Encouragement: Understanding how parameters interact is like learning the rules of a game. Once you know how the pieces (parameters) move, you can handle any inflation or transformation problem the SOA throws at you. You've got this!