Why Matching Periods Matters
A loss measure is meaningful only when its numerator and denominator describe compatible experience. Claim losses assigned to one period cannot automatically be divided by exposures assigned to another. The same rule applies to claim counts: if counts and losses refer to different groups of claims, their frequency and average severity do not describe one consistent experience set.
The earlier material on accident-year, policy-year, and calendar-year loss measures distinguishes ways to assign experience to periods. Here the practical task is to make sure the claim counts and loss amounts use the selected basis and that the exposure denominator represents that same period. Dates such as occurrence, report, payment, and policy effective dates can fall in different periods; choosing one basis means applying its assignment rule consistently.
The exposure unit must also be clear. It might be one policy-year, one vehicle-year, or another stated measure of insured risk. A count of policies is not necessarily the same as earned exposure: a policy in force for half of a year contributes half a policy-year if exposures are measured in policy-years. These measures answer different questions, so state the unit before interpreting a result.
What Must Be Matched
Three ingredients belong together:
- Claim count: count each claim once in the selected period, according to the chosen claim assignment date or rule.
- Loss amount: include the loss amounts for the claims in that same cohort, with the amount and valuation basis clearly defined.
- Exposure: include the insured exposure corresponding to that same period and exposure definition.
A period label alone is not enough. For example, “2025 losses” could mean losses assigned by occurrence date, by policy period, or by payment date. Likewise, “2025 exposure” could mean earned exposure during calendar year 2025 or exposure attached to policies designated as policy year 2025. The labels are compatible only if the underlying assignment rules are compatible.
| Selected basis | Claim assignment question | Exposure question |
|---|---|---|
| Accident period | In which period did the covered accident or occurrence happen? | What exposure is assigned to that accident period? |
| Policy period | To which policy period is the claim assigned under the selected rule? | What exposure belongs to those policy periods? |
| Calendar period | Which calendar-period rule assigns the claim or loss amount? | What exposure was earned during that calendar period? |
For example, a claim that occurs late in December but is reported in January can belong to one accident period and a different report period. If its loss amount is grouped by accident date but its count is grouped by report date, the numerator and count do not describe the same claim cohort. Similarly, payment in a later year does not, by itself, change the occurrence date. Always identify the date or rule used for each measure rather than assuming that all dates coincide.
Calculations from Matched Data
Once the period basis and exposure unit are consistent, the basic measures can be interpreted directly. Frequency is claims per exposure unit; average loss per claim is loss dollars per claim; and loss cost is loss dollars per exposure unit. These are descriptive ratios, not guarantees that a future period will have the same results.
The relationship between the three measures explains why the counts and losses must refer to the same claims. If \(N_j>0\) and \(E_j>0\), multiply frequency by average loss per claim:
The \(N_j\) terms cancel, leaving loss cost. This decomposition is valid only when the \(N_j\) claims used in the count are exactly the claims whose losses make up \(L_j\). If the count instead includes claims from a different period, the cancellation is merely algebraic; the resulting frequency and severity do not combine to explain the stated loss cost.
Worked Example: Accident-Year Claims and Exposure
In a hypothetical coverage portfolio, a company selects accident year 2024 as its basis. The year has 1,200 earned exposure units. Three claims are assigned to 2024 by occurrence date, with loss amounts of $14,000, $9,000, and $7,000. One claim occurred on December 30, 2024, but was reported in January 2025. Calculate claim frequency, average loss per claim, and loss cost. Use the accident-year basis for both claims and losses.
Method. Assign claims by occurrence date, add the loss amounts for those claims, and use the 2024 accident-year exposure denominator. The January report date does not move the December occurrence to a different accident year.
Step 1: Count the 2024 claims. There are \(N_{2024}=3\) claims, including the claim reported in January 2025.
Step 2: Add their losses.
Step 3: Calculate frequency.
This is 0.0025 claims per exposure unit, or \(0.0025\times1{,}000=2.5\) claims per 1,000 exposure units.
Step 4: Calculate average loss per claim.
Step 5: Calculate loss cost.
Check. Frequency times average loss per claim is \(0.0025\times\$10{,}000=\$25.00\) per exposure unit, agreeing with the loss cost. The common mistake avoided is moving the late-December claim into 2025 just because it was reported in January.
Diagnosing a Period-Basis Mismatch
A mismatch often produces a plausible-looking quotient. The arithmetic may be correct while the interpretation is not. For example, dividing an accident-year loss total by policy-year exposures gives a number, but it does not give an accident-year loss cost or a policy-year loss cost unless the underlying data have been consistently reassigned to that basis.
When a source extract contains different bases, first identify what each total represents. If the desired measure is accident-year, obtain or construct the accident-year claim count, accident-year loss total, and accident-year exposure. Do not “fix” a mismatch by changing only the denominator or only the claim count. If the underlying records do not support the required reassignment, the matched measure cannot be calculated from those totals alone.
Worked Example: Accident-Year Losses Divided by Policy-Year Exposure
A hypothetical insurer has the following two sets of 2025 experience. On an accident-year basis, there are 12 claims, $96,000 of losses, and 800 exposure units. On a policy-year basis, there are 10 claims, $75,000 of losses, and 500 exposure units. Calculate frequency, average loss per claim, and loss cost for each basis. Then show the result of incorrectly combining the accident-year losses and counts with the policy-year exposures.
Method. Calculate each basis using only its own claim count, loss amount, and exposure. Then use the mismatched quotient to show why a correct division does not make an incompatible comparison valid.
Accident-year measures. The count and loss total both refer to accident-year claims:
The frequency is 0.015 claims per exposure unit. Checking, \(0.015\times\$8{,}000=\$120.00\) per exposure unit.
Policy-year measures. Use the policy-year data as a separate, internally matched set:
The check is \(0.02\times\$7{,}500=\$150.00\) per exposure unit.
Incorrect combination. Dividing accident-year losses by policy-year exposure gives
Using the accident-year claim count with that same denominator gives \(12/500=0.024\) claims per exposure unit. Neither result is a valid measure for either stated basis: the numerator and denominator, and the count and denominator, refer to different sets of experience. The correct results are $120.00 for accident year and $150.00 for policy year. The common mistake avoided is treating a shared calendar label as proof that the assignment bases match.
Exposure Earned Across Calendar Periods
A calendar period may contain only part of the exposure from policies that span more than one calendar year. If the chosen exposure unit is a full policy-year, allocate the earned exposure to the calendar periods during which the coverage is in force. Do not use the number of policies in force as though every policy supplied a full year of exposure in each period.
The same care applies to claims. If the selected basis assigns claims by occurrence date, use that rule for the count and the associated losses, even when a claim is reported or paid later. In particular, a calendar-year payment total should not be paired with an occurrence-year count as though both described the same set of claims.
Worked Example: Allocating Policy-Year Exposure to Calendar Years
A hypothetical insurer writes 300 one-year policies effective July 1, 2025, and another 300 one-year policies effective July 1, 2026. Measure exposure in policy-years and assign claims and losses by occurrence date. In calendar year 2025, the assigned claims total 6 and their losses total $18,000. In calendar year 2026, the assigned claims total 12 and their losses total $42,000. Calculate the earned exposure and loss cost for each calendar year. Also calculate the 2025 result if someone incorrectly uses 300 policies as the exposure denominator.
Method. Allocate each six-month portion of a one-year policy term as one-half policy-year, then divide each calendar year's occurrence-assigned loss total by that year's earned exposure.
Step 1: Calculate 2025 exposure. The July 1, 2025 policies earn six months of exposure in 2025. Thus,
Step 2: Calculate 2026 exposure. The first group earns the other six months in 2026, and the second group earns its first six months:
Step 3: Calculate the calendar-year measures. For 2025, frequency is \(6/150=0.04\) claims per policy-year, and loss cost is
For 2026, frequency is \(12/300=0.04\) claims per policy-year, and loss cost is
Step 4: Check by average loss per claim. The average losses are \(\$18{,}000/6=\$3{,}000\) in 2025 and \(\$42{,}000/12=\$3{,}500\) in 2026. Then \(0.04\times\$3{,}000=\$120.00\) and \(0.04\times\$3{,}500=\$140.00\), as above.
Step 5: Show the denominator error. Using 300 policies instead of the 150 earned policy-years in 2025 would produce a frequency of \(6/300=0.02\) and a loss cost of \(\$18{,}000/300=\$60.00\). Both are half the correct 2025 values because the denominator incorrectly treats each policy as providing a full policy-year in 2025. The common mistake avoided is substituting a policy count for earned exposure.
A Practical Matching Workflow
Before calculating a frequency or loss cost, make the assignment choices explicit. This short sequence helps reveal mismatches before they affect interpretation.
State whether the intended measure is accident-period, policy-period, or calendar-period experience.
Identify the date or rule used to place a claim and its loss amount into a period. Do not assume occurrence, report, and payment dates are interchangeable.
Specify what one exposure represents and how partial periods contribute to the total.
Check that the count includes the claims represented by the loss total, and that the exposure corresponds to the same experience period.
Compute frequency, average loss per claim, or loss cost only after the preceding definitions agree.
This matching discipline matters because a ratio can look precise while measuring no coherent experience. The essential exam skill is not merely performing the division; it is recognizing whether the data justify the measure being requested. That discipline is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
Check Your Understanding
For each problem, use the stated period basis and exposure unit. Check that the count and loss amount refer to the same claims.
- A hypothetical accident-year cohort has 8 claims, $48,000 of losses, and 400 exposure units. Calculate frequency, average loss per claim, and loss cost. (Answer: 0.02 claims per exposure unit; $6,000 per claim; $120.00 per exposure unit.)
- Accident-year losses are $105,000 on 700 accident-year exposure units. Policy-year exposure is 500 units. What is the accident-year loss cost, and what incorrect result comes from dividing the accident-year losses by the policy-year exposure? (Answer: $150.00 per accident-year exposure unit; the mismatched quotient is $210.00 and is not a policy-year or accident-year loss cost.)
- Two claims occur in accident year 2024, with losses of $4,000 and $6,000. The second is reported in 2025. There are 250 accident-year exposure units. Calculate the 2024 accident-year frequency and loss cost. (Answer: 0.008 claims per exposure unit; $40.00 per exposure unit.)
- A group of 240 one-year policies is effective October 1. Assume exposure is earned uniformly over each policy term. The three months of coverage in the calendar year produce 3 occurrence-assigned claims and $9,000 of losses. Calculate calendar-year exposure, frequency, and loss cost. (Answer: 60 policy-years; 0.05 claims per policy-year; $150.00 per policy-year.)
- For accident year 2024, there are 6 claims, $24,000 of losses, and 300 exposure units. For accident year 2025, there are 8 claims, $40,000 of losses, and 400 exposure units. Calculate the combined loss cost across both years using the combined totals. (Answer: \((\$24{,}000+\$40{,}000)/(300+400)=\$91.43\) per exposure unit.)