Why Period Matching Matters
A loss ratio compares losses with the premium associated with the exposures that produced those losses. The arithmetic is simple; the harder step is confirming that both amounts describe the same experience. A numerator drawn from one period and a denominator drawn from another can produce a precise-looking ratio that does not measure the intended experience.
Loss records and premium records may be organized differently. Losses can be grouped by when an accident occurred or by the policy period to which a claim is assigned. Premium may be reported by when coverage was earned, by when policies were written, or by another stated period. A year in a report title is not enough to establish that two amounts are comparable: the actuary must identify what the year means for each amount.
This tutorial treats the supplied amounts as deterministic totals; no interest rate, mortality assumption, probability distribution, or claim model is required. The examples assume that each loss total is already on the stated loss basis and is sufficiently complete for the comparison requested. They do not estimate claim development. If losses are paid, incurred, or otherwise selected on different bases, those bases must also be reconciled before interpreting the ratio.
Identify the Period Behind Each Amount
An accident-period loss grouping assigns claims according to when the covered accident or event occurred. For example, accident-year 2025 losses are losses from covered events occurring during 2025, regardless of when the policies were written. Accident-period earned premium, when used, is the premium earned on coverage exposed during that same calendar period.
A policy-period grouping assigns experience to the policies or coverage periods that generated it. Policy-year losses are attributed to the relevant policy-year cohort, even if an accident occurs in a different calendar year. The premium denominator must be earned on that same policy-year cohort. A claim can therefore belong to accident year 2025 and to a policy year that began in 2024.
A calendar-period label describes a date interval, but the underlying loss measure still matters. Calendar-year paid losses, for instance, are payments made during the calendar year, not necessarily losses from accidents occurring in that year. The calendar label alone does not make paid losses suitable for comparison with accident-year earned premium.
| Label or measure | What determines inclusion? | Matching denominator to check |
|---|---|---|
| Accident-year losses | Date of the covered accident or event | Earned premium for exposures during the same accident period |
| Policy-year losses | Policy or policy-period assignment | Earned premium for that same policy-year cohort |
| Calendar-year paid losses | Date the claim payment was made | A premium and loss comparison explicitly defined for a compatible calendar-period purpose; not automatically accident-year premium |
| Written premium | Date or period in which policies are written | Not a substitute for earned premium when the intended measure is losses per earned premium |
The basic diagnostic is to ask two questions before dividing: Which exposures are represented by the losses? and Which exposures generated the premium? If the answers differ, either obtain amounts on a common basis or map the underlying records to a common experience period. Do not repair a mismatch merely by changing the year label.
Why the Ratio Requires Matched Experience
The loss ratio is meaningful as a measure of losses per premium dollar only when the numerator and denominator refer to a common experience base. To see why, suppose an intended period contains losses \(L_A\) and earned premium \(P_A\), while an adjacent period contains losses \(L_B\) and earned premium \(P_B\). The intended ratio for period A is \(L_A/P_A\). Replacing its denominator with \(P_B\) gives \(L_A/P_B\), which changes the comparison even though the losses are unchanged.
For multiple periods that are each matched internally, the combined ratio follows directly from adding the amounts before dividing:
The equality holds because multiplying each period ratio \(L_j/P_j\) by its premium share \(P_j/\sum_k P_k\) gives \(L_j/\sum_k P_k\); summing those terms gives total losses divided by total premium. Thus the combined ratio is a premium-weighted average of the period ratios, not generally their simple arithmetic average. This identity assumes each period’s losses and premium are matched and that the amounts are on a consistent basis. It cannot make mismatched inputs valid.
Worked Examples
Worked Example: Accident-Year Losses and the Wrong Premium Measure
A hypothetical insurer is reviewing accident-year 2025 experience. Its report shows $2,100,000 of accident-year losses, $3,500,000 of earned premium during calendar 2025, and $4,000,000 of written premium during calendar 2025. Calculate the actual loss ratio for accident-year experience using the appropriate premium amount. Also show the result of the common incorrect calculation.
Method. The losses are explicitly grouped by accident date. The intended comparison is with premium earned on exposure during the corresponding period, not with premium merely written during that period. Assume the stated earned premium covers the same 2025 exposure population as the accident-year loss measure.
Correct calculation. Use the $3,500,000 of earned premium:
Incorrect calculation and correction. Dividing by written premium gives
That 52.50% is not the requested accident-year loss ratio on earned premium: written premium does not necessarily represent the coverage earned during the period. The correct result is 60.00%. As a check, 60% of $3,500,000 is $2,100,000.
Worked Example: Separating Accident-Year and Policy-Year Views
In a hypothetical, fully reconciled data extract, losses attributed to policy year 2024 total $1,800,000. Of that amount, $1,000,000 arose from accidents in 2024 and $800,000 from accidents in 2025. Earned premium for the policy-year 2024 cohort is $3,000,000. Separately, accident-year 2024 losses total $1,600,000, including $600,000 attributed to policies incepted in 2023 and the $1,000,000 attributed to policy year 2024. Calculate the policy-year 2024 loss ratio and diagnose the result from dividing accident-year 2024 losses by policy-year 2024 premium.
Method. For a policy-year ratio, add the losses assigned to that policy-year cohort across the accident years shown, then divide by earned premium for that cohort. The example assumes that all listed losses are on the same selected basis and that the policy-year totals are sufficiently complete for the requested comparison.
Step 1: Assemble policy-year 2024 losses. The two accident-year components assigned to policy year 2024 are $1,000,000 and $800,000:
Step 2: Divide by matching policy-year earned premium.
Step 3: Diagnose the tempting mismatch. Accident-year 2024 losses of $1,600,000 include $600,000 assigned to policy year 2023. They exclude the $800,000 of policy-year 2024 losses from accidents in 2025. Dividing those accident-year losses by policy-year 2024 premium yields
This 53.33% is not the policy-year 2024 loss ratio: the numerator contains losses from the wrong cohort and omits losses from the requested cohort. It is also not, by itself, a correctly matched accident-year ratio, because the denominator is policy-year 2024 premium. The correct policy-year result is 60.00%. The assignment check is that the two components attributed to policy year 2024 sum to $1,800,000, matching the stated policy-year loss total.
Worked Example: Correcting a Twelve-Month Boundary Mismatch
A hypothetical analyst wants the calendar-year 2025 loss ratio. Accident-year 2025 losses are $4,400,000. A premium report contains earned premium by calendar quarter: $1,000,000 for the first quarter of 2025, $1,200,000 for the second, $1,400,000 for the third, $1,600,000 for the fourth, and $1,800,000 for the first quarter of 2026. The analyst mistakenly uses the April 2025–March 2026 premium total. Find the correctly matched ratio and quantify the incorrect one.
Method. The numerator is for January through December 2025, so select earned premium for those same four calendar quarters. Do not use an adjacent twelve-month window simply because it also contains four quarters.
Step 1: Sum calendar-year 2025 earned premium.
Step 2: Calculate the matched calendar-year ratio.
Step 3: Show the boundary error. The April 2025–March 2026 premium is the second, third, and fourth quarters of 2025 plus the first quarter of 2026:
Using that amount gives \( \$4{,}400{,}000/\$6{,}000{,}000=0.7333\ldots=73.33\% \), but this combines calendar-year 2025 losses with premium from a different twelve-month interval. The matched answer is 84.62%. As a check, the correct four quarters sum to $5,200,000, while the shifted window replaces the first-quarter 2025 premium with first-quarter 2026 premium, increasing the denominator by $800,000.
A Practical Reconciliation Before Division
When a report’s period labels are ambiguous, reconcile the inputs before calculating or interpreting a ratio. The following sequence makes the key decisions explicit.
Name the exact period and whether the requested view is accident-period or policy-period. A bare label such as “2025” is not sufficiently specific.
Determine whether the claim amounts are classified by accident date, policy cohort, payment date, or another stated basis. Do not infer the basis from the report’s year heading alone.
Confirm whether it is earned or written and which exposure dates or policy cohorts it represents. For a loss ratio on earned premium, do not replace the earned-premium denominator with written premium without an explicit reason and matching purpose.
Use a consistent period and exposure set; also confirm that losses and premium refer to compatible coverage and are measured consistently. If the records do not align, obtain or construct a reconciled comparison rather than dividing the unmatched totals.
Divide matched losses by matched earned premium. Recheck the numerator and denominator against the source totals and confirm that the interpretation names the period and basis used.
This reconciliation matters because a period mismatch can look like a change in experience when it is only a change in which policies or exposures appear in the numerator or denominator. The resulting ratio may be arithmetically correct but not answer the actuarial question. Matching periods is therefore a prerequisite to interpreting an actual loss ratio as a measure of rate adequacy, not a cosmetic reporting preference.
Common Exam Trap
A frequent error is to assume that two totals with the same calendar-year label necessarily cover the same experience. They may not: accident-year losses are classified by event date, while calendar-year paid losses are classified by payment date. Another trap is matching policy-year losses with accident-year premium because both are described as “2025.” The correction is to state the assignment rule for each amount, then use amounts tied to the same experience population and period.
Check Your Understanding
For each problem, identify the relevant period basis before calculating. Amounts are hypothetical and stated on compatible loss bases unless the problem says otherwise.
- A review of accident-year 2026 experience reports $1,560,000 of accident-year losses, $2,600,000 of earned premium during 2026, and $2,900,000 of written premium during 2026. What is the loss ratio on the earned-premium basis? (Answer: 60.00%.)
- A policy-year 2025 cohort has $720,000 of losses from 2025 accidents and $480,000 from 2026 accidents. Its earned premium is $2,400,000. What is the policy-year 2025 loss ratio? (Answer: 50.00%.)
- A report shows calendar-year 2027 losses of $2,250,000. Earned premium is $700,000 in each of the first three quarters of 2027 and $900,000 in the fourth quarter. An adjacent first-quarter 2028 amount is $1,000,000. Calculate the calendar-year 2027 loss ratio, and explain whether the first-quarter 2028 amount belongs in its denominator. (Answer: $2,250,000 / $3,000,000 = 75.00%; the first-quarter 2028 premium is outside calendar 2027.)
- Policy-year 2026 losses total $1,050,000, consisting of $650,000 from accidents in 2026 and $400,000 from accidents in 2027. Policy-year 2026 earned premium is $1,750,000. Accident-year 2026 losses are $900,000 because they also include $250,000 assigned to policy year 2025. What is the policy-year 2026 ratio, and why is dividing accident-year 2026 losses by policy-year 2026 premium inappropriate? (Answer: $1,050,000 / $1,750,000 = 60.00%; the accident-year numerator includes losses assigned to another policy cohort and omits policy-year 2026 losses from 2027 accidents.)
- Two matched periods have losses and earned premium of $900,000 and $1,500,000 in period A, and $1,600,000 and $2,500,000 in period B. What is the combined loss ratio, and why is a simple average of the individual ratios not generally the combined ratio? (Answer: $2,500,000 / $4,000,000 = 62.50%; the combined ratio weights each period by its earned premium, whereas a simple average gives the periods equal weight.)