Why Period Alignment Comes Before Calculation
Exposure, premium, and loss totals are meaningful together only when their period definitions are understood. Three records can all be labelled “2023” while describing different things: risks in force during calendar 2023, policies that began in 2023, and loss transactions recorded in calendar 2023. Adding those figures to an indication without checking what “2023” means can produce a precise calculation from mismatched inputs.
The practical task is not simply to match labels. It is to identify how each measure was assigned to a period, check the dates and valuation basis, and obtain comparable records. A period mismatch may be repairable from detailed records. If only incompatible totals remain, an exact correction may be impossible without additional data.
The exact convention used by a source system must be verified rather than inferred from a column heading. For example, a premium “year” might mean policy effective year, the year in which premium was earned, or a financial accounting year. Likewise, a loss amount might be paid loss, incurred loss, or another measure. Period alignment does not make different loss measures interchangeable.
A Period-Control Check
Before calculating or interpreting an indication, record the basis of each input. At a minimum, check the period start and end dates, the assignment rule, the unit of exposure, the premium earning or accounting rule, the loss measure, and the loss valuation date. Confirm that the data also cover the intended product, territory, coverage, and population. Matching time periods does not fix a mismatch in scope.
| Input | Questions to ask | Example of a period mismatch |
|---|---|---|
| Exposure | What unit is counted, and when is it assigned to the period? | Policy-years by effective year are compared with losses by accident date |
| Premium | Is it written, earned, or another premium measure? What dates determine the year? | Calendar-year earned premium is compared with policy-year exposure |
| Loss | What loss measure is used, how is it assigned to a period, and when was it valued? | Calendar-year accounting activity is compared with accident-year premium |
For example, specify accident year 2023, policy effective years 2022–2023, or calendar-year accounting activity in 2024. Include the exact start and end dates where necessary.
Check field descriptions and query rules, not just labels. Confirm how each exposure, premium, and loss record is assigned to a period.
Check the measure, valuation date, unit, and business scope. Losses valued at different dates may differ even if their accident-year labels match.
Filter or reaggregate detailed records using the intended assignment rule. Reconcile the corrected subtotal to the source records.
If only aggregate totals are available and they do not contain enough information to convert bases, report the limitation and obtain a suitable extract instead of inventing an allocation.
Why Matching Bases Matters
A quotient of two totals can always be computed, but its interpretation depends on what those totals describe. Let \(L^{A}\) be losses assigned to period \(A\), and let \(P^{B}\) be premium assigned to period \(B\). The ratio \(L^{A}/P^{B}\) is not automatically a loss-to-premium measure for either period. If the periods differ, its numerator and denominator need not represent the same experience.
The reason is visible by separating the periods into their components. Suppose a combined loss total includes losses from periods \(A\) and \(C\), while the premium total covers only period \(A\). The quotient then includes a loss contribution from \(C\) without including the premium associated with that contribution:
The second term is not a period-\(A\) loss-to-premium relationship. This is why selecting a total because its year label looks right is not sufficient. The period assignment rule has to match. This period-control discipline is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
Worked Examples
Worked Example: Calendar-Year Loss Activity in an Accident-Year Dataset
A hypothetical insurer is preparing an accident-year 2023 data set. It has 8,400 earned car-years assigned to accident year 2023 and $1,344,000 of earned premium assigned to the same accident year. The loss file supplied by an analyst is labelled “2023 losses” and totals $780,000. The field definition says this is incurred-loss accounting activity recorded during calendar year 2023. A claim-level extract, valued at December 31, 2024, separately reports $672,000 of incurred losses for accidents occurring in 2023. Identify the valid accident-year loss amount and quantify the effect of using the wrong amount on a simple loss-to-premium diagnostic.
Method: verify assignment basis, then reaggregate to the target accident year. The exposure and premium are assigned by accident year. The $780,000 loss total is assigned by the calendar year in which accounting activity was recorded. Its year label does not make it an accident-year total.
Step 1: select the matching loss measure. The claim-level extract identifies accidents occurring in 2023 and gives their incurred losses at a stated valuation date. The matching accident-year loss amount is therefore $672,000. The $780,000 calendar-year total is not substituted for it.
Step 2: check the diagnostic calculation. With the correct loss amount, the loss-to-premium diagnostic is:
Using the calendar-year amount instead would give:
The difference is about 8.04 percentage points. This difference reflects the incompatible loss basis, not a change in the accident-year premium. The exposure and premium are both already assigned to accident year 2023; the available accident-year loss subtotal is the one to use, with its valuation date retained in the documentation.
Check. The selected accident-year loss is $672,000, and $672,000 divided by $1,344,000 is exactly 0.50. The common mistake is treating “2023 losses” as an adequate period definition without checking whether 2023 means accident year or calendar-year accounting activity.
Worked Example: A Policy-Year Loss Extract with an Accident-Year Label
A hypothetical portfolio is being summarized by policy effective year. For policy year 2022, the source reports 6,250 exposure units and $1,125,000 of earned premium. The loss summary supplied with the file reports $590,000 and is defined as losses for accidents occurring in 2021. A separate claim-policy extract, valued at December 31, 2024, reports $645,000 for claims arising under policies effective in 2022. Determine which loss amount aligns with the stated portfolio and show the diagnostic ratios produced by the two choices.
Method: match the policy cohort, not the accident-date label. The intended portfolio is grouped by policy effective year. The appropriate loss cohort is claims arising under the policies effective in 2022, provided the extract uses the same coverage and population scope as the exposure and premium.
Step 1: reject the unmatched subtotal. The $590,000 amount is grouped by accident year 2021. It does not identify losses arising under policies effective in 2022. A claim may occur in a policy term that began in one year and therefore fall into a different accident year, so these assignments are not interchangeable.
Step 2: use the matching policy-year subtotal. The claim-policy extract gives $645,000 for the policy-year 2022 cohort at the stated valuation date. Use this amount with the 6,250 exposure units and $1,125,000 premium.
The aligned diagnostic is:
The mismatched calculation would be:
Check. The selected numerator refers to policies effective in 2022, as do the stated exposure and premium. The loss-to-premium result is approximately 57.33% when rounded to two decimal places. The common mistake is assuming the accident year of a claim must equal the effective year of the policy that covers it.
Worked Example: Calendar-Year Accounting Data and an Accident-Year Loss Total
A hypothetical insurer is reviewing calendar-year 2024 results. It has 9,600 exposure units and $1,440,000 of earned premium, both assigned to calendar 2024. The loss file includes $840,000 for accidents occurring in 2024, valued at December 31, 2025. The accounting system separately reports $765,000 of incurred-loss activity recorded during calendar 2024, including activity on claims from multiple accident years. Which loss total matches the stated calendar-year basis, and what is the aligned diagnostic?
Method: follow the accounting-period definition for the target data set. Here the requested loss measure is calendar-year accounting activity, not the cumulative incurred amount for accidents occurring in 2024. The $765,000 total is the calendar-year amount. The $840,000 amount has a different assignment rule and is not used as the calendar-year loss total.
Step 1: retain the calendar-year inputs. The calendar-year exposure is 9,600 units, the calendar-year earned premium is $1,440,000, and the calendar-year incurred-loss activity is $765,000.
Step 2: calculate the diagnostic.
If the accident-year 2024 loss total were used instead, the quotient would be:
Check. The aligned calendar-year diagnostic is 53.125%, or 53.13% rounded to two decimal places. The accident-year loss amount is not “more correct” in general; it is simply not the amount specified by this calendar-year accounting convention. The common mistake is to assume all loss summaries labelled with the same year include the same claims or transactions.
When Reaggregation Is Not Possible
A reliable correction usually needs records detailed enough to apply the intended assignment rule. For example, a claim-level file may include accident date, policy effective date, accounting transaction date, and valuation information. Those fields can support different summaries, but only if the loss definition and claim scope are also consistent. A policy-level file may similarly allow exposure and premium to be assigned to a common period.
An aggregate calendar-year loss total generally cannot be converted exactly into an accident-year loss total just by subtracting a guess for prior-year claims. Calendar-year incurred activity can include payments, changes in case reserves, and other accounting movements for claims from multiple accident years. Conversely, an accident-year total does not reveal how much of the amount was recorded in any particular calendar year. Without the underlying detail or a valid, documented allocation method, the source data do not determine an exact conversion.
Common Mistakes
- Trusting the column heading. “2023 losses” is incomplete unless the loss assignment rule is known.
- Confusing policy year and accident year. A claim’s accident date need not be in the calendar year when the covering policy began.
- Ignoring valuation date. Two accident-year loss totals can have the same assignment basis but different valuation dates, so they are not necessarily comparable.
- Mixing time bases while keeping other inputs fixed. A numerically plausible quotient can still combine unlike periods.
- Prorating an aggregate without support. A time-based allocation is not an exact conversion unless its assumptions and the required data justify it.
Check Your Understanding
For each problem, identify the period basis that controls the correction before calculating.
- A target accident-year 2024 file has 4,800 exposure units and $720,000 of earned premium. A calendar-year 2024 loss total is $390,000, while the claim extract reports $336,000 for accidents in 2024 at the selected valuation date. Which loss amount belongs in the accident-year summary? (Answer: $336,000)
- A policy-year 2023 summary has $960,000 of earned premium. The loss extract gives $510,000 by accident year 2023 and $552,000 for claims arising under policies effective in 2023. Which amount aligns with policy year 2023, and what is its loss-to-premium diagnostic? (Answer: $552,000; 57.50%)
- A calendar-year 2025 data set has $2,000,000 earned premium. Its supplied $1,080,000 loss amount is by policy effective year 2025; a correctly defined calendar-year accounting extract gives $980,000. What is the calendar-year diagnostic? (Answer: 49.00%)
- A report lists exposure and premium by policy effective year, but its loss field is labelled only “year incurred.” What should be verified before the loss field is used? (Answer: Verify the field’s assignment rule, loss measure, valuation date, and scope; the label alone does not establish that it is policy-year data.)
- A calendar-year 2024 file contains exposure of 7,200 units and $900,000 premium for January through September, plus 2,400 exposure units and $330,000 premium for October through December. The calendar-year loss total is $645,000. What are the full-year exposure, premium, and diagnostic? (Answer: 9,600 exposure units; $1,230,000 premium; approximately 52.44%)
- Only an aggregate calendar-year incurred-loss total is available, but the intended summary is by accident year. No claim-level dates or allocation data are available. Can the exact accident-year total be recovered from that aggregate alone? (Answer: No; obtain suitable detailed data or document that an exact reaggregation is unavailable.)