Welcome to the World of Trend Analysis!
Hello future actuaries! Today, we are diving into one of the most practical parts of the ASTAM syllabus: Projected Losses using Trend Analysis. If you’ve ever wondered how insurance companies decide how much to charge you next year when everything seems to be getting more expensive, you’re in the right place!
In this chapter, we learn how to take data from the past and "fast-forward" it to the future. Because of things like inflation, safer cars, or changes in medical costs, a claim that cost \$1,000 three years ago won't cost the same tomorrow. Our job is to bridge that gap.
Don’t worry if this seems a bit math-heavy at first. We’re going to break it down step-by-step so it feels like second nature!
1. The "Big Two": Frequency and Severity
When we talk about losses trending upward or downward, we usually look at two different "levers" that move the total cost:
- Frequency: How often do claims happen? (Claims per exposure unit). For example, if people start driving less because they work from home, frequency might go down.
- Severity: When a claim happens, how much does it cost? (Average cost per claim). If car parts become more expensive to replace, severity goes up.
The Golden Rule: To find the total change in Pure Premium (the total cost of loss per exposure), we multiply the frequency and severity trends together.
Formula: \( (1 + \text{Pure Premium Trend}) = (1 + \text{Frequency Trend}) \times (1 + \text{Severity Trend}) \)
Quick Review: If frequency increases by 2% and severity increases by 5%, the total loss trend is not 7%! It is \( 1.02 \times 1.05 = 1.071 \), or 7.1%.
2. Exponential vs. Linear Trend
In the actuarial world, we have two main ways to model how costs change over time:
A. Exponential Trend (The Most Common)
This assumes costs change by a constant percentage every year (like compound interest). This is the "industry standard" for ASTAM because inflation behaves exponentially.
The Math: \( \text{Future Value} = \text{Past Value} \times (1 + r)^n \)
Where \( r \) is the annual trend rate and \( n \) is the number of years.
B. Linear Trend
This assumes costs change by a fixed dollar amount every year. While simpler, it's less common for long-term projections because it doesn't account for the "snowball effect" of inflation.
Analogy: Imagine your favorite coffee. An exponential trend means it goes up by 5% every year. A linear trend means it goes up by exactly 10 cents every year. Over time, the 5% increase will eventually become much larger than the 10-cent increase!
3. Mastering the "Trend Period"
The trickiest part of ASTAM trend problems is figuring out exactly how much time (\( n \)) has passed. Actuaries use the Average Accident Date (the midpoint of a period) to measure time.
Key Rule: Midpoint to Midpoint
To trend losses from an old "Experience Period" to a new "Forecast Period," you measure the time from the midpoint of the historical period to the midpoint of the future period.
Step-by-Step Calculation:
- Find the Average Accident Date of the Historical Data: If the data is from the calendar year 2022, the midpoint is July 1, 2022.
- Find the Average Accident Date of the Forecast Period: If you are pricing policies that will be issued during 2024 (assuming 1-year policy terms), the accidents will happen throughout 2024 and 2025. The midpoint of these accidents is usually one year after the start of the policy issuance period.
- Calculate the difference in years between these two midpoints. That is your \( n \).
Common Mistake: Forgetting that policies have a "term" (usually 6 months or 1 year). If a company issues policies all through 2024, the last policy issued on Dec 31, 2024, won't expire until Dec 31, 2025. You must account for that entire window of time!
4. Internal vs. External Data
Where do we get our trend numbers?
- Internal Data: The insurance company’s own history. It’s specific to their customers but can be "noisy" (unstable) if the company is small.
- External Data: Using indices like the Consumer Price Index (CPI) or industry-wide data. This is more stable but might not perfectly match the company’s specific niche.
Did you know? Actuaries often use weighted averages of internal and external data. This is a concept related to Credibility, which you’ll see elsewhere in your studies!
5. Adjusting for One-Time Changes
Sometimes, a change isn't a "trend" (a gradual slope) but a "step" (a sudden jump). This happens when laws change or when there's a specific shift in policy benefits.
Example: If a state passes a law that increases the minimum liability limit on January 1, 2023, you apply an On-Level Factor to adjust all data before that date to the current level before you apply your trend factor.
The Workflow:
1. Bring historical data to "current levels" (On-leveling).
2. Apply Trend to move from "current levels" to "future levels."
Summary and Key Takeaways
Quick Review Box:
- Trend is the process of adjusting historical loss costs to reflect the cost levels of the period when the insurance will be in effect.
- Frequency Trend \(\times\) Severity Trend = Pure Premium Trend. (Remember to use the \( 1 + r \) format!)
- Always use the midpoints of the accident periods to calculate the length of the trend period (\( n \)).
- Exponential trend is the standard model: \( \text{Factor} = (1+r)^n \).
Keep practicing those timeline diagrams! Drawing a physical line and marking the midpoints is the best way to ensure you never get the time period \( n \) wrong. You’ve got this!