Matching Losses and Exposures to the Same Basis

Why Matching Matters

A pure premium is a loss amount measured per exposure unit. The arithmetic is simple, but it is meaningful only when the losses and exposures describe the same underlying risk. If a numerator includes losses from one coverage or population while the denominator counts exposures from another, their ratio may be numerically correct and actuarially misleading.

The earlier material on selecting an exposure base and on earned and written exposures distinguishes exposure units and experience periods. The practical task here is to check that the particular losses being analyzed correspond to those units. Before dividing, ask: What coverage produced these losses? Which risks produced them? What period do they belong to?

Definition: Losses and exposures are on the same basis when they refer to the same coverage, the same eligible population of insured risks, and the same experience period, using compatible exposure units. A pure premium calculated on matched data is the loss amount for that basis divided by its corresponding exposure amount.

“Same basis” does not mean that every loss must be paired with one specific exposure record. It means that both totals describe the same defined portfolio slice. A portfolio slice might be, for example, collision coverage for a specified class of vehicles during a specified accident year. If the losses include additional coverages, risks, or periods, those components must be removed or the exposure denominator must be expanded to match them.

A Consistency Check Before Division

For any proposed calculation, write down the basis in words before selecting the numerator or denominator. Record the coverage, population definition, experience period, loss measure, and exposure unit. The loss measure might be paid losses, incurred losses, or a selected ultimate amount; the choice must be explicit and appropriate to the analysis. The key point here is that the selected measure must correspond to the exposures included.

Define the target.
State the coverage, eligible risks, and experience period for which a loss cost is wanted.

Trace the losses.
Confirm that the numerator includes only losses for that coverage, those risks, and that period. Check what the loss amounts represent.

Trace the exposures.
Confirm that the denominator counts the same risks over the corresponding period and uses the intended exposure unit.

Reconcile before calculating.
Remove unmatched components or obtain the corresponding exposures or losses. Then divide the matched loss total by the matched exposure total.

The reason for this check can be seen algebraically. Suppose the target portfolio consists of risks $j=1,\ldots,n$, with loss $L_j$ and exposure $e_j$ for each risk. If every loss and exposure in the totals belongs to this same set and basis, then:

\frac{\sum_{j=1}^{n} L_j}{\sum_{j=1}^{n} e_j} = \sum_{j=1}^{n} \left(\frac{e_j}{\sum_{k=1}^{n} e_k}\right) \left(\frac{L_j}{e_j}\right), \quad e_j>0.

Thus, the total loss divided by total exposure is an exposure-weighted average of the individual loss costs for the matched risks. If the loss total also contains a different coverage or an additional period, it has extra terms that are not represented in the denominator. The ratio then no longer describes the target portfolio. This identity is an accounting relationship; by itself, it does not guarantee that the observed experience will predict future losses.

Worked Example: Separating Two Coverages

A hypothetical insurer is reviewing water-damage coverage for a defined group of commercial buildings during one experience period. Its records show 1,200 earned building-years for water damage and 800 earned building-years for equipment breakdown. Selected water-damage losses are $540,000, and selected equipment-breakdown losses are $210,000. Calculate the water-damage loss cost per earned building-year. Then quantify the result of incorrectly combining both coverages. All amounts are already selected on a consistent loss basis; no development or trend adjustment is requested.

Method and basis. The target is water-damage coverage for the specified buildings and period. Therefore, use water-damage losses with water-damage exposures. The two coverages are separate, even though the same buildings may have purchased both.

Step 1: Calculate the matched loss cost.

\frac{\$540{,}000}{1{,}200\text{ earned building-years}} =\$450\text{ per earned building-year}.

Step 2: Calculate the mismatched result. Combining the coverages gives total losses of $\$540{,}000+\$210{,}000=\$750{,}000$. Combining the exposures gives $1{,}200+800=2{,}000$ building-years. The resulting ratio is:

\frac{\$750{,}000}{2{,}000\text{ building-years}} =\$375\text{ per building-year}.

The combined ratio is not the water-damage loss cost: it includes equipment-breakdown losses and equipment-breakdown exposures. It could only describe a deliberately defined combined-coverage measure if that measure and its exposure basis were appropriate. For the requested water-damage measure, the checked answer is $450 per earned building-year.

Common mistake avoided: Adding all available losses and all available exposures because they are reported in the same file. Shared data storage does not make separate coverages the same basis.

Experience Period and Loss Timing

Loss records can be organized by when an accident occurred, when a claim was reported, when a payment was made, or when a valuation was prepared. Exposure records also have a period definition, such as earned exposure during a calendar year or exposure assigned to an accident year. A numerator defined by one timing basis cannot be casually divided by a denominator defined by another.

For instance, calendar-year payments can include payments on accidents from several prior years. Those payments do not automatically correspond to the exposures earned in the payment year. Conversely, a loss amount assigned to an accident year should be compared with exposures for that accident year, subject to the selected loss measure and the intended analysis. The period label alone is not enough: inspect how the records were assigned.

No interest or mortality assumptions are involved in the examples in this tutorial. The calculations use deterministic loss and exposure totals, not a probability distribution. Their main assumption is a clearly defined, consistently measured basis.

Worked Example: Avoiding a Calendar-Payment Mismatch

A hypothetical insurer wants the selected loss cost for accident year 2023. It has 1,200 earned exposures assigned to accident year 2023 and a selected ultimate loss total of $660,000 for claims arising from 2023 accidents. During calendar year 2024, it paid $90,000 on those 2023 claims and $120,000 on claims arising from 2024 accidents. Calculate the 2023 loss cost and explain why the calendar-year 2024 payments are not its numerator.

Method and basis. The target is claims arising in accident year 2023, paired with exposures assigned to accident year 2023. The stated $660,000 is already the selected ultimate loss total for that accident year. The calendar payment figures describe when cash was paid, not the accident year of every loss.

Step 1: Use the 2023 accident-year amounts.

\frac{\$660{,}000}{1{,}200\text{ earned exposures}} =\$550\text{ per earned exposure}.

Step 2: Identify the tempting but mismatched numerator. Calendar-year 2024 payments total $\$90{,}000+\$120{,}000=\$210{,}000$. The $90,000 relates to 2023 accidents, but the $120,000 relates to 2024 accidents. Neither is the stated selected ultimate loss total for all 2023 accidents. Dividing $210,000 by 1,200 would give $175 per exposure, but that quotient mixes calendar-year payments with accident-year exposures and does not answer the question.

The correct 2023 result is $550 per earned exposure. The calendar-year payment information should not replace the selected accident-year loss amount in this calculation.

Common mistake avoided: Treating the year in which a payment is made as though it were the year in which the insured event occurred. Always establish which time basis the loss and exposure records use.

Population and Exposure-Unit Alignment

The population definition determines which risks belong in both totals. A rate review might include only new business, a specified territory, a particular vehicle class, or risks meeting a defined underwriting rule. If losses from outside that group remain in the numerator, the calculated loss cost is not for the intended population. If the exposure total includes risks whose losses were excluded, the same problem occurs in the opposite direction.

The exposure unit must also be compatible. A vehicle-year is not automatically interchangeable with a policy-year: one policy may cover several vehicles. If the two totals use different units, do not combine them as if the units were identical. First obtain a consistent basis or a valid conversion. Likewise, be alert to partial-year exposures: the numerator and denominator must both represent the intended period, rather than one counting a full year while the other reflects only the time at risk.

Worked Example: Matching a Fleet Class and Coverage

A hypothetical insurer is measuring collision loss cost for a specified fleet class during one experience period. The eligible class has 500 earned vehicle-years and $275,000 of selected collision losses. The same vehicles also have $90,000 of selected liability losses. A separate, excluded vehicle class has 100 earned vehicle-years and $60,000 of collision losses. Calculate collision loss cost for the eligible class, and show the result of including all available records by mistake.

Method and basis. The target is collision coverage for the specified eligible class during the stated period. Select only its collision losses and its earned vehicle-years. Liability losses belong to another coverage; the excluded class’s collision losses and exposures belong to another population.

Step 1: Calculate the matched result.

\frac{\$275{,}000}{500\text{ earned vehicle-years}} =\$550\text{ per earned vehicle-year}.

Step 2: Calculate the incorrectly combined result. Including the other coverage and the excluded class gives losses of $\$275{,}000+\$90{,}000+\$60{,}000=\$425{,}000$. Including both classes’ exposures gives $500+100=600$ vehicle-years. Thus:

\frac{\$425{,}000}{600\text{ vehicle-years}} =\$708.333\ldots \approx \$708.33\text{ per vehicle-year}.

The incorrect calculation is higher, but the direction of distortion is not guaranteed in other cases. It depends on the added losses and exposures. The amount is rounded to the nearest cent only at the final step. The correct result for the requested class and coverage is $550 per earned vehicle-year.

Common mistake avoided: Assuming that a numerator and denominator match merely because both refer to a fleet. Coverage and class eligibility still have to agree.

Common Traps and a Practical Audit

Many mismatches are not obvious arithmetic errors. They arise from labels that sound similar or from totals that have already been combined. Before accepting a pure premium calculation, check the definitions behind each number. In particular:

Coverage scope: Confirm that losses and exposures refer to the same coverage or explicitly defined combination of coverages.
Population scope: Confirm that both totals use the same eligibility, territory, class, and other relevant group definitions.
Time scope: Confirm that the loss attribution period matches the exposure period; do not substitute payment timing without a reasoned basis.
Loss measure: Label whether losses are paid, incurred, or selected ultimate amounts. Do not treat these as interchangeable without specifying the adjustment or purpose.
Exposure unit: Verify the unit and duration. A policy-year, vehicle-year, and building-year are not interchangeable just because each is called an exposure.

A useful exam habit is to write the units beside the totals before dividing. For example, $\$275{,}000$ of eligible collision losses divided by 500 eligible earned vehicle-years produces dollars per earned vehicle-year. If the numerator or denominator cannot be described with the same coverage, population, and period, pause and reconcile the data.

Key takeaway: A pure premium is interpretable only for the basis represented by both its numerator and denominator. Match coverage, population, period, and exposure unit first; perform the division second.

Check Your Understanding

For each problem, identify the matched basis before calculating or evaluating the proposed ratio.

A hypothetical insurer wants a loss cost for glass coverage in one territory during a specified year. It reports $84,000 of glass losses and 280 earned vehicle-years in that territory. It also reports $36,000 of collision losses for the same vehicles. What is the glass loss cost, and should the collision losses be included? (Answer: $300 per earned vehicle-year; no, the collision losses are a different coverage.)
A hypothetical book has 900 earned policy-years for accident year 2025 and selected ultimate accident-year losses of $495,000. Calendar-year 2026 payments include $70,000 on 2025 accidents and $50,000 on 2026 accidents. What is the 2025 loss cost, and why is $120,000 not the appropriate numerator? (Answer: $550 per earned policy-year; the $120,000 is calendar-year payment data, not the selected ultimate losses for 2025 accidents.)
A target class has $132,000 of selected theft losses and 240 earned vehicle-years. A different class has $48,000 of theft losses and 160 earned vehicle-years. What is the target-class loss cost? What quotient results from improperly combining the classes? (Answer: target class, $550 per earned vehicle-year; combined, $180,000 divided by 400, or $450 per vehicle-year.)
A hypothetical insurer reports $96,000 of losses for one coverage and 300 earned policy-years. Each policy covers an average of two insured units, but no exposure conversion has been applied. Can this information alone support a loss cost per insured-unit-year? State the matched calculation that is supported by the figures. (Answer: no; the figures support $320 per earned policy-year, but a per-insured-unit-year measure requires a corresponding unit-year exposure basis.)
For a defined coverage and eligible population, a report shows $210,000 of selected losses for accident year 2024 and 350 earned exposures for calendar year 2024. What must be checked before calculating a loss cost, and what is the ratio if the exposure records are confirmed to represent the same risks and period? (Answer: verify that the exposure year and population correspond to the 2024 accident-year losses; if confirmed, $600 per exposure.)

What You'll Learn