Why the Exposure Denominator Deserves an Audit
A pure premium divides losses by exposures. That ratio is only meaningful if the losses and exposures refer to compatible coverage, experience periods, and exposure units. A denominator that is numerically plausible can still be wrong: it may count vehicle-months as vehicle-years, omit policies with no claims, include a policy twice, or come from a different year than the losses.
The practical consequence is direct. With the loss amount held fixed, an exposure total that is too large makes the calculated pure premium too small; an exposure total that is too small makes it too large. Before interpreting an indication, verify not only the arithmetic but also what each exposure record represents and how it was selected.
The selected loss amount might be an incurred or ultimate loss measure, depending on the analysis. That choice does not change the denominator audit: the exposure total must still correspond to the same experience being measured. This tutorial focuses on errors in that total, not on selecting or projecting losses.
An exposure unit is not automatically the same as a policy. A policy lasting six months may contribute half a policy-year; a fleet policy may contribute several vehicle-years. The governing exposure definition determines the conversion. A label such as “policies” is insufficient unless the calculation also establishes how policy duration and any other relevant exposure basis are handled.
How a Denominator Error Changes the Ratio
Suppose the correct loss amount \(L\) is divided by the correct exposure \(E\), giving pure premium \(P=L/E\). If the denominator used in the calculation is \(E^*\), then the calculated result is \(P^*=L/E^*\). Comparing the two makes the direction of the error clear.
Thus, if \(E^*\) is 10% greater than \(E\), the calculated pure premium is \(1/1.10\) times the correct value—not 10% lower. Conversely, if \(E^*\) is 10% less than \(E\), the calculated pure premium is \(1/0.90\) times the correct value. The relationship is reciprocal because exposure is in the denominator.
Worked Example: Converting Vehicle-Months to Vehicle-Years
A hypothetical delivery-van program has selected losses of $500,000 for a year of experience. Its exposure file reports 30,000 vehicle-months. Calculate the pure premium per vehicle-year. Then identify the result of incorrectly treating vehicle-months as vehicle-years.
Method and assumptions. The stated losses and exposures refer to the same coverage and experience period. One vehicle-year is defined as 12 vehicle-months. No interest, mortality, or probability distribution assumption is needed: this is a deterministic ratio of the stated totals.
Step 1: Convert the exposure to vehicle-years.
Step 2: Divide losses by exposure in the requested unit.
Step 3: Check the result. Multiplying the pure premium by the exposure returns the selected losses: \(2{,}500\times \$200.00=\$500{,}000\).
Common mistake avoided. Dividing by 30,000 and reporting \(\$16.67\) as the pure premium per vehicle-year mislabels the unit. The division gives \(\$16.67\) per vehicle-month when rounded to cents. Converting that monthly amount to an annual amount gives \(\$16.666\ldots\times12=\$200.00\) per vehicle-year. The final amount is rounded to two decimals.
Reconcile Records Before Summing Exposure
A total can be wrong even when every individual exposure value is in the correct unit. Common data-processing causes include an omitted record, a record included twice, or a join that keeps only policies with claims. A sound reconciliation compares the exposure records used in the calculation with the expected policy or risk population for the experience period.
A claims file is not, by itself, a complete exposure file: policies with no claims may have valid earned exposure. If a calculation takes its denominator only from records that appear in a claims file, it can discard precisely those no-claim exposures that belong in the total. Conversely, a many-to-many data join can repeat an exposure record when one policy has multiple claim records.
Worked Example: Finding an Omitted or Duplicated Policy Record
A hypothetical small-business portfolio has four verified earned-exposure records, measured in policy-years. The selected losses for the same coverage and period are $70,000.
| Record | Earned exposure (policy-years) |
|---|---|
| A | 120 |
| B | 80 |
| C | 100 |
| D | 50 |
Method and assumptions. Each listed record represents a distinct policy exposure, each is earned in the stated experience period, and the selected losses have the same scope. No interest, mortality, or probability distribution assumption is used.
Step 1: Reconcile the verified exposure total.
Step 2: Calculate the correct pure premium.
Step 3: Quantify the effect of omitting record D. The denominator would be \(120+80+100=300\) policy-years, so the erroneous result would be \(\$70{,}000/300=\$233.33\) per policy-year, rounded to two decimals. The missing 50 policy-years make the indicated amount too high.
Step 4: Quantify the effect of duplicating record B instead. The denominator would be \(350+80=430\) policy-years, so the erroneous result would be \(\$70{,}000/430=\$162.79\) per policy-year, rounded to two decimals. The duplicate exposure makes the indicated amount too low.
Check and common mistake avoided. The verified total gives \(350\times \$200.00=\$70{,}000\). Do not assume that a plausible-looking total is complete: compare record identifiers and exposure amounts with the expected population. Omission and duplication have opposite effects here, but neither error should be “corrected” by adjusting losses to force a preferred ratio.
Match the Exposure Period to the Loss Period
The exposure period is the time interval in which the policies earned the exposure used in the denominator. It must align with the experience period represented by the losses. Using a prior-year denominator with current-year losses changes the ratio even if the exposure units are identical.
For example, a portfolio that wrote fewer policies this year may have lower current-year exposure than in the prior year. Dividing current-year losses by prior-year exposure then makes the calculated ratio reflect a mixture of two periods rather than the stated current-year experience. Confirm the dates and the earning basis, not just the year printed in a file name.
Worked Example: Detecting a Prior-Year Exposure Total
A hypothetical equipment-insurance portfolio has $360,000 of selected losses for calendar year 2025. Its earned exposures are 1,800 equipment-years in 2025 and 2,000 equipment-years in 2024. An analyst divides the 2025 losses by the 2024 exposure total. Calculate the correct 2025 pure premium, the analyst’s result, and the relative understatement.
Method and assumptions. The losses and 2025 exposures represent the same coverage and 2025 experience period; the 2024 exposure total is included only as a deliberately mismatched denominator. No interest, mortality, or probability distribution assumption is used.
Step 1: Use the exposure from the loss period.
Step 2: Calculate the mismatched result.
Step 3: Measure the understatement relative to the correct result.
Check and common mistake avoided. The correct result satisfies \(1{,}800\times \$200.00=\$360{,}000\); the mismatched calculation does not describe the 2025 exposure base. The mistaken use of the larger 2024 denominator produces a result 10% below the correct 2025 pure premium. Amounts are shown to two decimals and the relative difference to one percentage point.
A Record-Selection Trap: Policies With No Claims
An omitted record is not always a record that disappears from the exposure source. It may be lost when two files are combined. For example, an inner join between a policy file and a claim file can retain only policies that have a matching claim. If the pure premium numerator includes all losses but the denominator is restricted to claim-bearing policies, the ratio is not the portfolio pure premium.
Worked Example: Restoring No-Claim Exposure After a File Join
A hypothetical rental-equipment pool has three policy records, each with 100 equipment-years of earned exposure. The loss file contains $20,000 of selected losses associated with only one of those records. A joined analysis retains just the record with a claim. Calculate the correct pool pure premium and the joined-file result.
Method and assumptions. All three policy records belong to the same pool, coverage, and experience period. The selected $20,000 is the total loss amount for that pool. No interest, mortality, or probability distribution assumption is needed.
Step 1: Include all eligible earned exposure, including the two records with no claims.
Step 2: Calculate the correct pure premium.
where the amount is rounded to two decimals. The check is \(300\times(\$20{,}000/300)=\$20{,}000\), before rounding.
Step 3: Calculate the result if the joined file supplies the denominator. The joined file retains only 100 equipment-years, producing \(\$20{,}000/100=\$200.00\) per equipment-year. This is three times the correct unrounded pure premium because it retains one-third of the eligible exposure.
Common mistake avoided. A policy with no claims contributes zero to the loss numerator, not zero to earned exposure. Excluding its exposure changes the denominator and overstates the pure premium.
A Practical Denominator Audit
The following checks separate arithmetic errors from definition and data-selection errors. They can be applied before the pure premium is used in an indication.
State the unit explicitly, such as vehicle-year or payroll unit, and describe how the unit is earned. Do not rely on an unlabeled total.
Check that the exposure dates match the experience period represented by the selected losses. Investigate mismatched file periods rather than assuming they are interchangeable.
Compare distinct policy or risk identifiers, counts, and exposure amounts with the expected population. Check for omissions, repeated records, and filters that retain only claim-bearing risks.
Show the conversion factor in the calculation. A rate per vehicle-month and a rate per vehicle-year differ by a factor of 12.
Multiply the calculated pure premium by the reconciled exposure, using unrounded values, and verify that it returns the selected losses. This checks arithmetic; it does not replace validation of the data and definitions.
Common Mistakes and Exam Traps
- Using the right number with the wrong unit. Dividing by vehicle-months can produce a valid monthly cost, but it is not a cost per vehicle-year unless converted.
- Using a claims-only population as the exposure population. No-claim policies still contribute their eligible earned exposure.
- Assuming duplicate records cancel out. Repeating a denominator record while keeping the loss numerator unchanged lowers the ratio; the record must be corrected, not balanced by an unrelated adjustment.
- Combining different experience periods. A prior-year exposure total does not become a current-year denominator merely because the unit label matches.
- Checking only the final division. A correct calculation using an incomplete or mismatched total is still an incorrect pure premium.
The essential exam discipline is to name the unit and period before dividing, then establish that every eligible exposure record appears once and only once. This topic is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
Check Your Understanding
For each problem, identify the denominator issue, calculate the appropriate pure premium, and compare it with any stated erroneous result.
- A hypothetical pet-insurance pool has $84,000 of selected losses and 7,200 earned pet-months. Convert exposure to pet-years and calculate the pure premium per pet-year. (Answer: 600 pet-years; $140.00 per pet-year.)
- A hypothetical contractor portfolio has $96,000 of selected losses and four verified exposure records of 90, 60, 40, and 50 policy-years. A data extract omits the 40-policy-year record. Calculate the correct and omitted-record pure premiums. (Answer: Correct exposure 240 policy-years and $400.00 per policy-year; omitted-record result $480.00 per policy-year.)
- A hypothetical marine portfolio has 2026 losses of $225,000, 2026 exposure of 1,500 vessel-years, and 2025 exposure of 1,800 vessel-years. Calculate the 2026 pure premium and the result from using the 2025 denominator. State the percentage understatement relative to the correct result. (Answer: $150.00 per vessel-year; $125.00 per vessel-year; 16.67% understatement.)
- A hypothetical appliance-protection pool has $30,000 of selected losses and three eligible records of 40, 30, and 30 appliance-years. Only the 40-year record appears in a claim-file join. Calculate the correct pure premium and the join-based result, and state the direction of error. (Answer: Correct $300.00 per appliance-year; join-based $750.00 per appliance-year; the join-based result is too high.)
- A hypothetical transit-fleet analysis reports $180,000 of losses for one year. The exposure file lists 1,000 bus-years for that year and 1,200 bus-years for the prior year. A separate audit finds that one current-year record of 100 bus-years was counted twice in the 1,000 total. Calculate the corrected pure premium and the result if the duplicated total is used. (Answer: Correct exposure 900 bus-years and $200.00 per bus-year; duplicated-total result $180.00 per bus-year.)