Welcome to the World of Reserving!
Hello future actuaries! Today, we are diving into one of the most critical tasks in the life of a short-term actuary: Estimating Outstanding Claims (also known as Reserving).
Think of an insurance company like a big party. People are constantly arriving (claims being reported) and leaving (claims being paid). At any given moment, the company needs to know how much money to set aside to pay for the guests who are still at the party or those who haven't even walked through the door yet!
Don't worry if this seems tricky at first. We’re going to break down the five major methods you need to know for the ASTAM exam into bite-sized, digestible pieces. Let’s get started!
1. The Expected Loss Ratio (ELR) Method
The Expected Loss Ratio (ELR) method is the simplest approach. It doesn't look at how individual claims are developing; instead, it looks at the big picture based on what we expected to happen when we first wrote the policies.
How it Works
We assume that a certain percentage of the Earned Premium will eventually be paid out as losses.
The formula is:
\( \text{Estimated Ultimate Losses} = \text{Earned Premium} \times \text{Expected Loss Ratio} \)
To find the Reserve (the money we need to set aside), we just subtract what we have already paid:
\( \text{Reserve} = \text{Estimated Ultimate Losses} - \text{Paid Losses to Date} \)
Real-World Analogy
Imagine you are planning a wedding and you expect it to cost $100 per guest. If you have 100 guests, your "Ultimate Loss" is $10,000. If you have already paid the caterer $4,000, your "Reserve" is $6,000. You don’t care how much the appetizers cost specifically; you just stick to your original budget plan.
When to Use It
It’s great for new lines of business where you don't have much historical data, or for very unstable lines where current data might be misleading.
Key Takeaway: The ELR method ignores actual claim development patterns and relies entirely on prior expectations.
2. The Chain-Ladder (CL) Method
The Chain-Ladder method (also called the Loss Development Method) is the most famous tool in the actuarial toolbox. It assumes that the future will look like the past.
How it Works (Step-by-Step)
1. Organize the Data: We arrange losses into a "triangle" by accident year and development year.
2. Calculate Age-to-Age Factors (Link Ratios): For each year, we see how much the losses grew from one period to the next.
\( f_j = \frac{\sum \text{Cumulative Losses at Year } j+1}{\sum \text{Cumulative Losses at Year } j} \)
3. Select Development Factors: Usually, we take an average of these ratios.
4. Project to Ultimate: We multiply the current losses by the Cumulative Development Factor (CDF) to get the final estimated cost.
Quick Review: Cumulative vs. Incremental
Cumulative: The total amount paid/reported from the start until now.
Incremental: Just the amount paid/reported in one specific time window.
Common Mistake: Always make sure your triangle is cumulative before calculating link ratios!
The "Expanding Cake" Analogy
Think of a claim like a cake in the oven. In the first 10 minutes, it rises by 50%. In the next 10 minutes, it rises by 10%. If we know how much a cake usually rises at each stage, we can look at a half-baked cake and predict its final size!
Key Takeaway: The Chain-Ladder method relies 100% on actual data development and ignores original expectations.
3. The Bornhuetter-Ferguson (BF) Method
What if the ELR is too rigid and the Chain-Ladder is too volatile? Enter the Bornhuetter-Ferguson (BF) method. It’s the "Goldilocks" of reserving—it blends both approaches.
How it Works
The BF method says: "I will keep the losses I’ve already seen (Actual), but for the future, I’ll go back to my original plan (ELR)."
The formula for the BF Reserve is:
\( \text{Reserve} = \text{Earned Premium} \times \text{Expected Loss Ratio} \times (1 - \frac{1}{\text{CDF}}) \)
Where \( (1 - \frac{1}{\text{CDF}}) \) represents the percentage of losses that are still unreported or unpaid.
Memory Aid: The Two Voices
Imagine two voices in your head.
Voice A (Chain-Ladder): "Look at the data! The data is everything!"
Voice B (ELR): "Stick to the budget! The budget is everything!"
The BF Method listens to Voice A for the past, but follows Voice B for the future.
Key Takeaway: BF is more stable than Chain-Ladder because it doesn't "overreact" to a single large claim early in a year's development.
4. The Bayesian Method
The Bayesian method is like the BF method’s more mathematical older sibling. It uses Credibility Theory to decide exactly how much we should trust our data versus our prior beliefs.
How it Works
In the Bayesian framework, we treat the ultimate losses as a random variable. We start with a Prior Distribution (what we think before seeing data) and update it with Actual Data to get a Posterior Distribution.
The estimate is often a weighted average:
\( \text{Estimate} = Z \times (\text{Data-Based Estimate}) + (1-Z) \times (\text{Prior Expectation}) \)
Where \( Z \) is the Credibility Factor (a number between 0 and 1).
Did You Know?
The BF method is actually a special case of a Bayesian/Credibility approach! The math can get complex with Gamma and Poisson distributions, but the core idea is always the same: Update your opinion as you get more information.
Key Takeaway: Bayesian methods allow us to incorporate statistical uncertainty and formal "prior" knowledge into our reserves.
5. The Frequency-Severity Method
Sometimes, looking at just "Total Dollars" isn't enough. The Frequency-Severity method breaks the problem into two parts.
The Formula
\( \text{Ultimate Losses} = (\text{Ultimate Number of Claims}) \times (\text{Ultimate Average Cost per Claim}) \)
The Step-by-Step
1. Project Frequency: Use a Chain-Ladder on the number of claims to find out how many total claims will eventually be reported.
2. Project Severity: Look at the average cost per claim. You might need to adjust this for inflation!
3. Multiply: Multiply the projected total claims by the projected average cost.
Why use this?
It helps you understand why reserves are changing. Are there more accidents (Frequency), or are doctors and repair shops getting more expensive (Severity)? This is very useful for explaining results to management!
Key Takeaway: Frequency-Severity provides a deeper "under the hood" look at claim trends compared to aggregate methods.
Summary Quick-Check
Which method should I use?
• No data / New business? Use ELR.
• Stable, mature data? Use Chain-Ladder.
• Volatile data / Middle ground? Use Bornhuetter-Ferguson.
• Want to understand trends? Use Frequency-Severity.
• Want to use formal statistical priors? Use Bayesian.
Don't worry if the formulas for the Bayesian method look intimidating—focus on the logic of "Weighting" and "Updating," and the math will follow! You are one step closer to mastering ASTAM!