Choosing Loss Measures for Trend Analysis

Why the Choice of Loss Measure Matters

A change in aggregate loss does not, by itself, show whether claim costs are changing. Aggregate dollars can move because there are more exposures, more claims per exposure, larger payments per claim, or changes in claim settlement and reserving. The measure used for analysis should separate the effects relevant to the question.

For example, a pricing analysis may focus on expected loss cost per exposure, while an investigation of claim handling may focus on amounts paid or claim counts. These measures are related, but they are not interchangeable. A useful comparison requires a consistent definition of claims, losses, exposures, and valuation date.

Definition: Paid loss is the amount of claim payments made by a specified valuation date. Incurred loss is paid loss plus the case reserves for outstanding claims at that date. Claim count is the number of claims meeting a specified counting rule. Average severity is loss divided by claim count. Pure premium is loss divided by exposure.

These definitions need a clear basis. For instance, a count might mean reported claims, claims with positive payments, or claims closed during a period; each choice can produce a different result. Loss and count should refer to the same claims and the same coverage basis. Exposures should also be measured consistently. For a comparison intended to describe changes in claim cost, different exposure volumes should not be mistaken for changes in cost per exposure.

What Each Measure Shows

Measure	Useful for	Main limitation to check
Paid loss	Cash payments made; analyses focused on payment activity	Depends on payment timing and claim maturity
Incurred loss	Paid amounts plus current case estimates of unpaid claim amounts	Can change when case reserves are revised
Claim count	How often claims occur relative to a suitable exposure base	Depends on the claim definition and reporting practices
Average severity	Loss per claim for the defined claim population	Can change with claim mix, limits, and loss valuation
Pure premium	Loss cost per exposure, combining claim frequency and severity	Can change with exposure mix as well as claim cost

Paid loss is directly tied to cash outflow, but payment timing makes it sensitive to how long claims have had to develop. A payment total for a calendar period may include payments on claims from several earlier periods. If the question concerns the cost of claims from a particular group of exposures, a calendar-period payment total may not represent that group without further alignment.

Incurred loss includes case estimates for amounts not yet paid. It is therefore often more complete than paid loss at an early valuation, but it is not final: case reserves can be increased, reduced, or removed as information changes. Neither paid nor incurred loss automatically equals ultimate claim cost.

Claim count and average severity help distinguish the number of claims from their dollar size. Pure premium combines these effects into a single loss-per-exposure measure. The frequency-severity relationship gives a useful check on the calculation.

Result: If $N$ is the number of claims, $E$ is the number of exposures, and $L$ is the total loss for those same claims on the same valuation and coverage basis, then frequency is $N/E$, average severity is $L/N$ when $N>0$, and pure premium is $L/E$.

Why Pure Premium Equals Frequency Times Severity

This identity follows from the definitions rather than from a special distributional model. Assume the claim count, loss total, and exposure count refer to the same population and period, and use the same loss basis throughout. If there is at least one claim, multiply frequency by average severity:

\frac{N}{E}\times\frac{L}{N}=\frac{L}{E}.

The $N$ in the numerator of frequency cancels the $N$ in the denominator of severity, leaving total loss per exposure. Thus, a pure-premium change can reflect a change in frequency, severity, or both. If $N=0$, average severity $L/N$ is undefined, but pure premium remains defined as $L/E$; in that case the decomposition cannot be calculated from the observed average severity.

The identity is valid only when the ingredients match. For example, dividing incurred losses for a set of reported claims by a count of closed claims does not produce a consistent average severity. Similarly, a pure premium calculated from one loss basis cannot be reconciled to a frequency-severity product using a different basis.

Worked Examples

Worked Example: Choosing Between Paid and Incurred Loss

A hypothetical insurer compares two groups, each with 1,000 earned exposure units. Group A is valued 12 months after the end of its claim period; Group B is valued 6 months after its claim period. The reported data are:

Group A: paid loss $=\$600{,}000$, case reserves $=\$300{,}000$, and 90 reported claims.
Group B: paid loss $=\$540{,}000$, case reserves $=\$360{,}000$, and 90 reported claims.

An analyst wants to compare estimated claim cost per exposure. Calculate the observed paid and incurred amounts per exposure, then determine which measure is more suitable for the stated comparison.

Method and assumptions. Use incurred loss as paid loss plus case reserves, and divide each loss total by the stated 1,000 exposures. This is a comparison of current recorded amounts, not a claim that incurred loss is ultimate loss. The unequal valuation ages are important because Group B has had less time for payments and reserve estimates to develop.

Step 1: Calculate incurred loss. Group A has $\$600{,}000+\$300{,}000=\$900{,}000$. Group B has $\$540{,}000+\$360{,}000=\$900{,}000$.

Step 2: Divide by exposure. Group A paid loss per exposure is $\$600{,}000/1{,}000=\$600$, and incurred loss per exposure is $\$900{,}000/1{,}000=\$900$. Group B paid loss per exposure is $\$540{,}000/1{,}000=\$540$, and incurred loss per exposure is $\$900{,}000/1{,}000=\$900$.

Step 3: Select a measure for the goal. The paid amounts differ by $\$60$ per exposure, but the groups have different valuation ages, so that difference may reflect payment timing rather than a difference in claim cost. Incurred loss is the more informative recorded measure here because it includes case reserves, although it remains subject to reserve development. The equal claim counts do not establish that ultimate claim costs are equal.

Check and common mistake avoided. For each group, paid loss plus case reserves equals $\$900{,}000$, and dividing by 1,000 gives $\$900$ per exposure. Treating the paid difference as a direct cost difference would ignore the unequal maturity of the data.

Worked Example: Separating Frequency, Severity, and Pure Premium

A hypothetical equipment insurer reviews two policy groups on the same incurred-loss basis. Group A has 2,000 exposures, 100 claims, and $\$1{,}200{,}000$ of incurred loss. Group B has 2,500 exposures, 125 claims, and $\$1{,}625{,}000$ of incurred loss. The claim definition, coverage, and valuation age are the same for both groups. Calculate frequency, average severity, and pure premium for each group, and identify what differs.

Method and assumptions. Frequency is claims divided by exposures, severity is incurred loss divided by claims, and pure premium is incurred loss divided by exposures. The common claim and loss basis allows the measures to be compared and reconciled.

Step 1: Calculate frequency. Group A frequency is $100/2{,}000=0.05$ claims per exposure. Group B frequency is $125/2{,}500=0.05$ claims per exposure.

Step 2: Calculate average severity. Group A severity is $\$1{,}200{,}000/100=\$12{,}000$ per claim. Group B severity is $\$1{,}625{,}000/125=\$13{,}000$ per claim.

Step 3: Calculate pure premium. Group A pure premium is $\$1{,}200{,}000/2{,}000=\$600$ per exposure. Group B pure premium is $\$1{,}625{,}000/2{,}500=\$650$ per exposure.

Step 4: Reconcile the results. For Group A, $0.05\times\$12{,}000=\$600$. For Group B, $0.05\times\$13{,}000=\$650$. Frequency is the same, while severity and pure premium are higher in Group B. This supports focusing the comparison on claim size, while still checking whether differences in claim mix explain the severity result.

Check and common mistake avoided. Group B has more total claims and more total loss, but it also has more exposures. Comparing only the totals would confuse group size with claim cost per exposure. The per-exposure results reconcile exactly to frequency multiplied by severity.

Worked Example: Recognizing a Case-Reserve Change

A hypothetical insurer has two equally sized books, each with 1,000 exposures and 100 reported claims. On the same claim population, Book A has $\$800{,}000$ paid and $\$200{,}000$ in case reserves. Book B has $\$800{,}000$ paid and $\$400{,}000$ in case reserves. The insurer reports that a new reserving review increased case estimates for Book B. Calculate paid and incurred loss per exposure and explain which measure is appropriate for an analysis of current cash payments versus an analysis of recorded claim cost.

Method and assumptions. Incurred loss is paid loss plus case reserves. The exposures and claim counts are the same, and the paid amounts are reported on a consistent basis. The stated change in reserving practice means the incurred comparison needs an explicit reserve-method caveat.

Step 1: Calculate incurred loss. Book A incurred loss is $\$800{,}000+\$200{,}000=\$1{,}000{,}000$. Book B incurred loss is $\$800{,}000+\$400{,}000=\$1{,}200{,}000$.

Step 2: Calculate loss per exposure. Paid loss per exposure is $\$800{,}000/1{,}000=\$800$ for both books. Incurred loss per exposure is $\$1{,}000{,}000/1{,}000=\$1{,}000$ for Book A and $\$1{,}200{,}000/1{,}000=\$1{,}200$ for Book B.

Step 3: Interpret according to the goal. For current cash payments, paid loss per exposure shows no difference. For current recorded claim cost, incurred loss per exposure shows a $\$200$ difference, but the known case-reserve change may account for it. Before interpreting that difference as a change in underlying claim cost, an analyst should make the reserve basis comparable or use an analysis suited to the question.

Check and common mistake avoided. Both books have the same paid amount and exposure count, so paid loss per exposure is $\$800$ in each. The $\$200$ incurred difference comes entirely from case reserves; treating it as automatically representing higher ultimate loss would mistake a reserve-estimate change for a demonstrated change in claim cost.

Selection Checks and Common Traps

A measure is useful only if its data basis supports the intended interpretation. Before comparing results, ask what each number includes and what event or population it describes. The following checks are especially important:

Match the claim population. The count used for severity must correspond to the claims underlying the loss total. A count of claims closed in a year is generally not a matching denominator for losses paid in that year, because those payments can relate to claims from other periods.
Match the loss basis. Paid and incurred amounts answer different questions. If case reserves or valuation ages differ, an incurred comparison may reflect estimation and maturity effects; a paid comparison may reflect timing.
Use an appropriate exposure denominator. Pure premium requires a meaningful, consistently measured exposure base. A total-loss comparison alone does not control for different amounts of business.
Check the mix. Average severity and pure premium can move when the distribution of risks, coverages, limits, or claim types changes. An aggregate measure may not isolate a change within otherwise comparable risks.
State what “claim” means. Changes in reporting, claim splitting, or counting conventions can alter claim count and average severity even when the underlying insured events have not changed.

A frequent exam trap is to see higher aggregate losses and conclude that severity increased. For example, suppose a book's losses rise from $\$1{,}000{,}000$ to $\$1{,}200{,}000$, while exposures rise from 1,000 to 1,200. The wrong comparison treats the 20% increase in total loss as a 20% increase in cost. The correction is to calculate loss per exposure: $\$1{,}000{,}000/1{,}000=\$1{,}000$ and $\$1{,}200{,}000/1{,}200=\$1{,}000$. On this measure, cost per exposure is unchanged. This does not prove that frequency and severity are each unchanged; the claim counts and average severities would be needed to distinguish them.

The practical choice follows the question. Use paid loss when cash payments are the target and payment timing is understood; incurred loss when current recorded claim cost is useful and case-reserve practices are comparable; claim count when claim occurrence is the focus; average severity when loss per claim is the focus; and pure premium when the target is loss per exposure. These measures can be examined together, but their definitions and limitations should remain explicit. This approach is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Check Your Understanding

For each hypothetical situation, choose or calculate the measure that best fits the stated goal. Keep the claim, loss, and exposure bases aligned.

A portfolio has 800 exposures and 40 claims in one period, with $\$360{,}000$ of incurred loss. In the next period it has 800 exposures, 48 claims, and $\$504{,}000$ of incurred loss. Calculate frequency, average severity, and pure premium in both periods. (Answer: Period 1: frequency $0.05$, severity $\$9{,}000$, pure premium $\$450$; Period 2: frequency $0.06$, severity $\$10{,}500$, pure premium $\$630$.)
A hypothetical book has $\$420{,}000$ paid loss, $\$180{,}000$ case reserves, and 600 exposures. A second book has $\$480{,}000$ paid loss, $\$120{,}000$ case reserves, and 600 exposures. Calculate each incurred loss per exposure and state one caution before interpreting the comparison. (Answer: Both are $\$1{,}000$ per exposure; check that valuation age and case-reserving practices are comparable.)
A hypothetical insurer reports 70 claims and $\$840{,}000$ incurred loss for one group, and 90 claims and $\$990{,}000$ for another. Each group has 1,000 exposures, and the data are on the same basis. Calculate frequency, severity, and pure premium for each group; which component differs? (Answer: Group 1 frequency $0.07$, severity $\$12{,}000$, pure premium $\$840$; Group 2 frequency $0.09$, severity $\$11{,}000$, pure premium $\$990$. Frequency and pure premium differ; severity is lower in Group 2.)
A calendar-year payment report includes $\$150{,}000$ of payments on older claims and $\$450{,}000$ of payments on the current claim cohort. The current cohort has 500 exposures. An analyst wants a measure of current-cohort paid loss per exposure. Calculate it and identify the incorrect alternative. (Answer: $\$450{,}000/500=\$900$ per exposure; using the full $\$600{,}000$ calendar-year payment total would give $\$1{,}200$ and include payments on older claims.)
A hypothetical portfolio has 1,500 exposures, 75 claims, and $\$900{,}000$ incurred loss. After a coverage change, a comparable-sized portfolio has 1,500 exposures, 75 claims, and $\$1{,}050{,}000$ incurred loss. Calculate the two severities and pure premiums, then state why the comparison may not isolate underlying claim-cost change. (Answer: First severity $\$12{,}000$ and pure premium $\$600$; second severity $\$14{,}000$ and pure premium $\$700$. The coverage change can affect measured loss and claim mix, so the result may not represent a change on a constant coverage basis.)

What You'll Learn