Why Severity Trends Need Careful Period Matching
A severity trend projects the average cost of a claim from one cost period to another. The arithmetic can be brief, but two details matter: whether a stated trend is an annual rate or a cumulative change, and how many years separate the base and future cost periods. Confusing either can materially misstate projected claim costs.
This topic concerns the severity component of loss estimation. It does not project claim counts. To make a useful comparison, the base and projected severities must refer to a consistent definition of claim loss and a comparable claim population. The examples below use hypothetical data and treat each selected trend as an assumption, not as an estimate from actual industry experience. The methods are aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
The calculation begins with the base severity. If the data provide aggregate losses \(L_0\) and claim count \(N_0\), then \(S_0=L_0/N_0\). The trend factor is then applied to that average. Keep the loss measure and claim definition consistent: for example, do not divide losses that include one set of claim components by a count defined on a materially different basis.
Annual Rates and Cumulative Factors
A constant annual trend rate \(r\) means that severity increases by the same percentage in each year. Under the assumption of annual geometric compounding, the factor for \(n\) years is \((1+r)^n\). A rate of 6% per year over two years therefore produces a cumulative factor of \(1.06^2\), not \(1+2(0.06)\).
If the annual rate varies by year, multiply the separate annual factors. If the selected trend is instead supplied as a cumulative factor for the whole interval, apply that factor once. These are two ways to describe the same total change only when the rates and periods correspond.
| Trend information supplied | Factor to apply | Interpretation |
|---|---|---|
| Constant annual rate \(r\), over \(n\) years | \((1+r)^n\) | Apply one annual factor for each year. |
| Different annual rates \(r_1,\ldots,r_n\) | \(\prod_{j=1}^{n}(1+r_j)\) | Apply each year’s rate to that year’s starting severity. |
| Cumulative factor \(C\) for the full interval | \(C\) | Apply the total factor once, not once per year. |
Why the Compounding Formula Holds
Suppose the selected annual trend is a constant effective rate \(r\), and each year’s severity is measured on a consistent basis. By definition, severity one year later is the preceding severity multiplied by \(1+r\). Applying that relationship successively gives the projection formula:
The exponent counts the annual compounding intervals between the base and future cost periods. It is not automatically the number of calendar years named in a problem: compare the actual period endpoints. For a fractional number of years \(t\), the factor \((1+r)^t\) is appropriate when the selected annual trend is assumed to compound geometrically over fractional years. If the stated trend convention specifies a different treatment of partial periods, use that convention instead.
Worked Example: Two Years at a Constant Annual Trend
A hypothetical portfolio has aggregate claim loss of $756,000 for 63 claims in a base cost period beginning January 1, 2024. Project its average severity to a cost period beginning January 1, 2026, using a constant 6% annual severity trend. Assume the claim definition and loss basis remain unchanged, and that the trend compounds annually.
Step 1: Calculate base severity. Divide aggregate loss by claim count:
Step 2: Count the annual periods. January 1, 2024, to January 1, 2026, is two years, so \(n=2\). The cumulative factor is:
Step 3: Apply the factor. The first year’s projected severity is \(\$12{,}000(1.06)=\$12{,}720\). The second year’s projection is \(\$12{,}720(1.06)=\$13{,}483.20\). Equivalently, \(\$12{,}000(1.1236)=\$13{,}483.20\).
The projected average severity is $13,483.20, reported to the nearest cent. As a check, the cumulative increase is \(\$13{,}483.20-\$12{,}000=\$1{,}483.20\), or 12.36% of the base severity, consistent with a factor of 1.1236. The method is annual compounding; it avoids treating a two-year projection as though the base were increased by only one year’s trend.
Varying Rates and Cumulative Trend Information
The same period-counting discipline applies when the annual rates are different. Each year’s rate applies to the severity at the start of that year. Alternatively, a cumulative factor already summarizes the full interval. Do not convert between these descriptions by adding annual percentages unless a simple, noncompounding convention is explicitly specified.
Worked Example: A Different Trend Rate in Each Year
A hypothetical book has base-period average severity of $8,400. A selected trend assumption is 4% in the first year, 5% in the second year, and 3% in the third year. Project the severity at the end of those three years. Assume each rate is an effective annual rate and the claim and loss bases stay consistent.
Step 1: Apply the first year’s rate.
Step 2: Apply the second year’s rate to the updated severity.
Step 3: Apply the third year’s rate.
The equivalent cumulative factor is \(1.04(1.05)(1.03)=1.12476\). Thus the projected average severity is $9,447.98, rounded to the nearest cent. The check is \(\$8{,}400(1.12476)=\$9{,}447.984\), which agrees before rounding. The method is successive compounding; adding 4%, 5%, and 3% and applying 12% once would not reproduce the selected annual-rate assumptions.
Worked Example: Converting a Cumulative Change to an Annual Rate
A hypothetical product has base-period average severity of $20,000. The selected cumulative trend factor from the base period to the future period, two years later, is 1.21. Project the future severity and state the equivalent constant annual trend rate. Assume the cumulative factor applies to the full two-year interval.
Step 1: Apply the cumulative factor once.
Step 2: Find the equivalent annual rate. Let \(r\) be the constant annual rate that produces the same two-year factor. It must satisfy \((1+r)^2=1.21\). Taking the positive square root gives \(1+r=1.10\), so \(r=0.10\), or 10% per year.
The projected severity is $24,200.00, and the equivalent annual rate is 10%. The check is \(\$20{,}000(1.10)^2=\$24{,}200\). A common wrong calculation is to treat the 21% cumulative increase as an annual rate and calculate \(\$20{,}000(1.21)^2=\$29{,}282\). That applies the full two-year change twice. Use the cumulative factor once, or first convert it to the appropriate annual rate.
Worked Example: Projecting Across Half a Year
A hypothetical group has base average severity of $9,500. Project it six months forward using a constant 12% effective annual trend. Assume the trend compounds geometrically during fractional years and the claim and loss bases are unchanged.
Step 1: Express the interval in years. Six months is \(t=0.5\) years.
Step 2: Calculate the fractional-year factor.
Step 3: Project the severity.
The projected severity is $10,053.85, rounded to the nearest cent. The factor represents about a 5.83% increase over six months; it is not a full 12% increase. This result follows the stated geometric fractional-year assumption.
Applying the Projection Consistently
A severity projection is meaningful only when the base statistic and the target statistic describe comparable costs. Before calculating, identify the source and target cost periods, the loss measure, the claim-count basis, and whether the trend is annual or cumulative. The selected trend should correspond to the severity being projected. For example, a severity measure affected by policy deductibles or limits may not change in the same way as an unlimited underlying loss measure.
Keep the claim population in view as well. Average severity can change because claim costs change, but it can also change when the mix of claims shifts. A trend factor is an assumption about the change to apply to the chosen base average; it does not, by itself, explain the source of observed changes or correct a mismatch in claim mix. The earlier material on diagnosing severity changes is useful context when assessing whether a trend assumption is appropriate.
Severity is loss per claim, not loss per exposure. A severity trend alone does not account for a change in claim frequency or exposure. When the objective is to project total loss cost, count and severity effects must be handled consistently; projecting them together is a separate calculation.
Use aggregate loss divided by claim count, or use a supplied average, on the defined loss basis.
Count the time from the base cost period to the future cost period, including fractional years when appropriate.
Distinguish an annual rate from a cumulative factor; multiply separate annual factors when rates vary.
Multiply base severity by the applicable factor and verify that the result moves in the expected direction and reflects the stated period.
Common Exam Traps
- Using one year of trend for a two-year interval. With a constant annual rate \(r\), two years require \((1+r)^2\), not \(1+r\).
- Applying an annual rate as if it were cumulative. A 6% annual rate over two years gives \(1.06^2\); the two-year factor is not simply \(1.06\).
- Applying a cumulative factor repeatedly. A factor of 1.21 for the entire interval is applied once. It is not squared unless 1.21 is explicitly the annual factor.
- Adding annual rates when compounding is intended. Different annual factors multiply. Their rates do not generally add to the same cumulative change.
- Ignoring period endpoints or units. Confirm whether the stated interval is one year, two years, or a fraction of a year before selecting the exponent.
- Mixing severity bases. Check that the loss measure and claim count used to form the base average correspond to the measure being projected.
Check Your Understanding
For each hypothetical problem, identify the time interval and trend interpretation before calculating. Answers are rounded to the nearest cent where money is requested.
- A group has aggregate loss of $960,000 for 80 claims in the base period. Project its average severity three years forward at a constant 4% effective annual trend. (Answer: Base severity $12,000; projected severity $13,498.37.)
- A product’s base average severity is $5,000. The selected rates are 2% for year one and 6% for year two. Find the average severity after two years, assuming annual compounding. (Answer: $5,406.00.)
- A selected cumulative trend factor is 1.15 over a three-year interval, and base average severity is $24,000. What is the projected severity? (Answer: $27,600.00.)
- Project a base average severity of $10,000 forward six months using an 8% effective annual trend and geometric fractional-year compounding. (Answer: $10,392.30.)
- A hypothetical portfolio has aggregate loss of $1,050,000 for 75 claims. Project its average severity two years forward at a constant 5% annual trend. (Answer: Base severity $14,000; projected severity $15,435.00.)