Constructing Aggregate Pure Premium

What Aggregate Pure Premium Measures

A ratemaking target must pair a loss amount with the exposure that generated the opportunity for that loss. Aggregate pure premium is the eligible aggregate loss divided by the matching earned exposure. It is a loss-cost measure per unit of exposure, not the average amount paid per claim.

The word eligible matters. The loss numerator must use the intended coverage, loss basis, and experience period. The exposure denominator must describe the same coverage and period, expressed in the intended exposure unit. A precise division cannot fix mismatched inputs.

Definition: For records $j=1,\ldots,n$, let $L_j$ be the eligible loss assigned to record $j$, and let $e_j$ be its earned exposure in the selected exposure unit. If the total exposure is positive, aggregate pure premium is $$ \widehat{P}=\frac{\sum_{j=1}^{n}L_j}{\sum_{j=1}^{n}e_j}. $$ Its units are dollars per exposure unit. For example, if exposure is measured in policy-years, the result is dollars per policy-year.

A record might represent a policy, a vehicle within a policy, or another unit appropriate to the rating exposure base. The numerator and denominator need not be stored at the same data grain, but they must be aggregated consistently to the same target population. If losses are first summarized by policy, for example, the denominator must cover those same policies and the same period.

Constructing a Matching Numerator and Denominator

Before dividing, specify what belongs in each total. The loss definition may depend on coverage, valuation basis, and the experience period. Exposure may be measured in policy-years, vehicle-years, or another unit. The selection must be consistent: a loss for one coverage cannot be paired indiscriminately with exposure for a broader or different coverage.

Define the target scope.
Specify the coverage, population, experience period, loss basis, and exposure unit the target is intended to represent.

Assemble eligible losses.
Include only losses meeting the stated scope and assign them to the records or aggregate for which the matching exposure is available.

Measure earned exposure on the same scope.
Use earned, rather than merely written, exposure for the selected period, and convert all records to a common unit before summing.

Divide and verify.
Divide total eligible loss by total matching earned exposure. Check that the answer has the intended units and that no eligible loss or exposure was omitted or duplicated.

If the denominator is zero, the ratio is undefined; it is not a zero pure premium. A positive numerator with zero matching exposure is a data or allocation warning that should be investigated. More generally, losses assigned to records with no exposure require particular care, as the weighted-average identity discussed below does not automatically accommodate them.

Worked Example: Annual Policy-Year Target

A hypothetical insurer is constructing a target for one coverage over a single experience year. Three policies have matching earned exposures of 1.0, 0.5, and 1.5 policy-years. Their eligible losses are $1,200, $900, and $0. Calculate aggregate pure premium per policy-year.

Method. Sum the eligible losses and matching policy-year exposures separately: $$ \sum_j L_j=1{,}200+900+0=2{,}100, \qquad \sum_j e_j=1.0+0.5+1.5=3.0. $$

The aggregate pure premium is $$ \widehat{P}=\frac{2{,}100}{3.0}=\$700.00\text{ per policy-year}. $$

Check. Multiplying the result by total exposure gives $700.00\times3.0=\$2{,}100.00$, the eligible aggregate loss. The zero-loss policy still contributes its 1.5 policy-years to the denominator. The common mistake this avoids is averaging policy loss amounts without weighting for their different earned exposures.

Why the Exposure-Weighted Form Sometimes Works

For a record with positive exposure, its record-level loss cost is $L_j/e_j$. The aggregate ratio can be written as an exposure-weighted average of those record-level loss costs only when the positive-exposure records account for the full aggregate loss numerator. In particular, every zero-exposure record must have zero eligible loss, unless such loss has been appropriately allocated to positive exposure.

When each $e_j>0$, the justification is an algebraic rearrangement:

\frac{\sum_j L_j}{\sum_j e_j} = \frac{\sum_j e_j\left(L_j/e_j\right)}{\sum_j e_j} = \sum_j \left(\frac{e_j}{\sum_k e_k}\right)\frac{L_j}{e_j}.

The weights are nonnegative and sum to one, so this is an exposure-weighted average. If some records have zero exposure and zero loss, they can be omitted from both sums without changing the result. But if a zero-exposure record has positive eligible loss, the rearrangement does not represent the full numerator: $L_j/e_j$ is undefined, and the positive-exposure records' weighted average leaves out that loss. The aggregate ratio can still be computed mechanically when total exposure is positive, but the data assignment should be checked before treating the result as a coherent loss-cost target.

Worked Example: Auditing a Zero-Exposure Loss

A hypothetical data extract reports two records for the same coverage period: record A has exposure of 1.0 vehicle-year and eligible loss of $1,000; record B has exposure of 0.5 vehicle-year and eligible loss of $500. A third entry has zero exposure and $200 of eligible loss. An audit confirms that this $200 belongs to record A but was separated by a coding error. Find the corrected aggregate pure premium and explain the weighted-average check.

Correct the assignment first. Allocate the $200 to record A. Its corrected loss is $1{,}000+200=\$1{,}200$. Record B remains at $500. Total eligible loss is $1{,}200+500=\$1{,}700$, and total exposure is $1.0+0.5=1.5$ vehicle-years.

Thus, $$ \widehat{P}=\frac{1{,}700}{1.5} =\$1{,}133.333\ldots =\$1{,}133.33\text{ per vehicle-year}, $$ rounded to the nearest cent.

The corrected record-level costs are $1{,}200/1.0=\$1{,}200$ and $500/0.5=\$1{,}000$. Their exposure-weighted average is $$ \frac{1.0(1{,}200)+0.5(1{,}000)}{1.5} =\frac{1{,}700}{1.5} =\$1{,}133.33\text{ per vehicle-year}. $$ Check. The weighted average now includes the full $1,700 numerator. Before correction, averaging only the two positive-exposure record costs would give $(1.0(1{,}000)+0.5(1{,}000))/1.5=\$1{,}000$, omitting the $200. The trap is to assume that a positive-exposure weighted average always captures aggregate pure premium.

Pure Premium Is Not Claim Severity

The denominator determines what a loss ratio per unit means. Aggregate pure premium uses earned exposure, so it measures eligible loss per exposure unit. Claim-level severity instead uses a count of claims as its denominator and measures loss per claim. Those quantities answer different questions and generally have different units.

For example, a portfolio can have a loss total of $4,500 and 6 claims over 3.5 vehicle-years. Its aggregate pure premium is $4,500 divided by 3.5 vehicle-years, while its average claim severity is $4,500 divided by 6 claims. Neither denominator can be substituted for the other. Calling the claim average a pure premium would change the target from a policy- or exposure-level loss cost to a claim-level amount.

Worked Example: Multiple Exposure Units and Claim Severity

A hypothetical fleet program has three records for the same coverage and period. Record A covers two vehicles for one full year and has $2,400 of eligible loss from two claims. Record B covers one vehicle for six months and has $600 of eligible loss from one claim. Record C covers three vehicles for four months and has $1,500 of eligible loss from three claims. Calculate aggregate pure premium per vehicle-year and claim-level average severity.

Convert exposure to vehicle-years. Record A has $2\times1=2.0$ vehicle-years. Record B has $1\times0.5=0.5$ vehicle-years. Record C has $3\times(4/12)=1.0$ vehicle-years. Therefore total exposure is $2.0+0.5+1.0=3.5$ vehicle-years.

Aggregate eligible losses. Total loss is $2{,}400+600+1{,}500=\$4{,}500$. Hence aggregate pure premium is $$ \frac{4{,}500}{3.5} =\$1{,}285.714\ldots =\$1{,}285.71\text{ per vehicle-year}, $$ rounded to the nearest cent.

Use the claim denominator for severity. There are $2+1+3=6$ claims, so claim-level average severity is $$ \frac{4{,}500}{6}=\$750.00\text{ per claim}. $$ Check. The two results use the same $4,500 loss total but different denominators: 3.5 vehicle-years versus 6 claims. The common mistake is to report $750 per claim as the exposure-based pure premium.

Practical Checks and Common Traps

Mixing coverage or periods: losses and earned exposure must refer to the same intended coverage and experience period.
Mixing exposure units: convert months to years, or vehicles to vehicle-years, as needed before adding exposures.
Using written exposure by mistake: the target here uses earned exposure for the period being measured.
Dropping zero-loss exposure: a record with no eligible loss can still contribute earned exposure and must not be removed solely because its loss is zero.
Forcing a weighted-average interpretation: first verify that zero-exposure records have no eligible loss, or that their losses have been allocated appropriately. Otherwise the positive-exposure weighted average need not equal the aggregate ratio.
Confusing loss cost and severity: state the denominator and units explicitly; dollars per exposure unit and dollars per claim are not interchangeable.

Key takeaway: Construct aggregate pure premium from the complete eligible loss total and its matching earned exposure total. Verify scope, units, and allocation before interpreting the ratio as a loss cost per exposure unit.

Check Your Understanding

For each problem, identify the correct denominator and keep the stated exposure units consistent.

A hypothetical book has eligible losses of $900 and $1,100 on 0.75 and 1.25 policy-years, respectively. What is aggregate pure premium per policy-year? (Answer: $1,000.00 per policy-year)
A hypothetical coverage has $3,600 eligible loss over 4 vehicle-years and 8 claims. Calculate pure premium per vehicle-year and average severity per claim. (Answer: $900.00 per vehicle-year; $450.00 per claim)
Two records have losses of $800 and $300 with exposures of 1.0 and 0.5 policy-years. A third record has $100 loss and zero exposure, and the loss is confirmed eligible but has not been allocated. What is the mechanical aggregate ratio using all losses, and can the positive-exposure weighted average represent it? (Answer: $800.00 per policy-year mechanically; no, the weighted average of the positive-exposure records omits the $100 zero-exposure loss)
A hypothetical policy covers four vehicles for three months and has $1,200 of eligible loss. Convert exposure to vehicle-years and calculate its aggregate pure premium if it is the only record. (Answer: 1.0 vehicle-year; $1,200.00 per vehicle-year)
A hypothetical portfolio reports $2,700 of eligible loss, 3.0 earned policy-years, and 9 claims. A draft labels $300 per claim as pure premium. What is the correct pure premium, and what quantity is $300? (Answer: $900.00 per policy-year; $300.00 is average claim severity)

What You'll Learn