Exposure and Loss Bases in Rate Indications

Why Exposure and Loss Bases Must Match

A rate indication is meaningful only when its numerator and denominator describe the same coverage, population, and experience period. For example, losses for one year of homeowners coverage should not be divided by a count of policies that includes policies earning only part of that year. Nor should capped losses be divided by exposures while uncapped losses are used in a comparison intended to measure the capped coverage. These mismatches can produce a precise-looking result that does not measure the risk being priced.

This tutorial focuses on selecting and reconciling the exposure and loss bases used to prepare filing calculations. Earlier material on selecting experience periods, on-leveling premiums, loss development, and trend selections supplies important adjustments. Here, assume those selections have already been made when an example says that losses are developed or projected. The remaining task is to ensure that the adjusted data still use compatible units and definitions.

Definition: An exposure measure is a quantified unit of insured risk for a specified coverage and period, such as earned vehicle-years, earned building-years, or payroll per stated dollar amount. Earned exposure is the portion of the exposure attributable to coverage actually provided during the period. A loss basis specifies which claims and loss amounts are included, how they are valued, and whether contractual limits or other stated adjustments have been applied.

Choosing the Exposure Measure

The exposure unit should correspond to the coverage being priced and the way risk varies for that coverage. An auto liability indication might use earned vehicle-years; a property indication might use earned insured locations or building-years; and a workers compensation indication might use payroll in units such as $1,000 of payroll. The unit must be stated explicitly. “One exposure” is not sufficiently precise unless the filing defines what one exposure represents.

Written policies or written units are not automatically the correct denominator for losses from an experience period. A policy written for a full year but in force for only six months contributes approximately one-half of a policy-year of earned exposure, assuming exposure accrues uniformly over the insured period. A policy that renews during the period may contribute earned exposure before and after renewal, but the exposure must be counted consistently with the coverage and rating basis being analyzed.

The following distinctions help keep the denominator clear:

Written policies or units count business issued during a period. They do not, by themselves, measure how long coverage was provided.
Earned exposures measure the coverage provided during the experience period, using the chosen exposure unit.
Claim counts are not exposures. They are event counts used with exposures to calculate claim frequency.
Premium is a monetary amount, not an exposure unit. A premium-based loss ratio calculation and an exposure-based loss-cost calculation answer related but different questions.

If risks have different coverage terms or exposure amounts, aggregate earned exposure rather than simply counting contracts. For example, two half-year policies may contribute one earned policy-year in total if a policy-year is the chosen unit. Conversely, if the chosen unit is a location-year, a contract covering several locations may contribute more than one location-year per year.

Claim Counts, Incurred Losses, and Limited Losses

Claim counts and loss amounts should refer to the same coverage, experience period, and claim population. Depending on the data and the filing definition, claims may be assigned to the period by accident date, report date, or another specified basis. The basis must be stated and used consistently. A count of reported claims paired with losses from a differently defined population can distort both frequency and average severity.

Definition: At a stated evaluation date, incurred losses are paid losses plus outstanding loss reserves. Outstanding reserves include case reserves for reported claims and any applicable incurred-but-not-reported (IBNR) reserves. If a selected data basis deliberately excludes IBNR or another reserve component, that exclusion must be stated; it is not the general definition of incurred losses.

Paid losses reflect amounts paid by the evaluation date; they omit amounts still expected to be paid on known or unreported claims. Incurred losses include the applicable outstanding reserves and therefore provide an estimate of the ultimate obligation as of the selected evaluation date. The loss development selections used to estimate ultimate losses must be applied on a basis consistent with the claims and loss amounts being summarized.

A limited-loss basis applies the policy’s relevant per-claim payment limit to individual claim amounts before aggregating them. For a covered loss $X$, deductible $d$, and per-claim payment limit $u$, where $u$ is the maximum insurer payment after the deductible, the payment per loss is

Y^L=\min\bigl((X-d)_+,u\bigr),\qquad (z)_+=\max(z,0).

For a zero-deductible policy with a limit $u$, this reduces to $Y^L=\min(X,u)$. Limited losses must be calculated claim by claim before they are summed. Capping the aggregate total once is not equivalent: a per-claim limit applies separately to each claim. Also, use the same loss basis when comparing periods or calculating a loss cost. Do not divide limited losses by exposures and then inadvertently compare the result with an uncapped loss cost.

Frequency, Severity, and Loss Cost

Let $E$ denote earned exposure units, $N$ the number of claims on the selected claim-count basis, and $L$ the total loss amount on the selected loss basis. When $E>0$ and $N>0$, the observed claim frequency, average severity, and loss cost per exposure are

\text{Frequency}=\frac{N}{E},\qquad \text{Average severity}=\frac{L}{N},\qquad \text{Loss cost per exposure}=\frac{L}{E}.

The units are important. If $E$ is earned vehicle-years, frequency is claims per vehicle-year, severity is dollars per claim, and loss cost is dollars per vehicle-year. If losses are limited, $L$ is the sum of limited claim amounts, so average severity and loss cost are limited-basis measures. If the count is zero, severity $L/N$ is undefined; the total loss cost $L/E$ can still be calculated.

Derivation: Provided the same $N$, $L$, and $E$ are used in all three ratios, multiplying frequency by average severity cancels the claim count:

\frac{N}{E}\times\frac{L}{N}=\frac{L}{E},\qquad N>0.

Thus, loss cost per exposure is exactly the product of claim frequency and average severity on a consistent basis. This is an arithmetic identity, not a claim that frequency and severity are statistically independent. It fails as a reconciliation check if the count and losses do not describe the same claims or if different exposure denominators are used.

Worked Applications

Worked Example: Reconciling Earned Exposure and Incurred Losses

A hypothetical auto coverage segment has 1,200 earned vehicle-years and 96 claims on the selected accident-period claim-count basis. At the evaluation date, the claims have $600,000 paid, $120,000 in case reserves, and $48,000 in IBNR reserves. Calculate incurred losses, frequency, average severity, and loss cost per earned vehicle-year. Assume the stated amounts are already on the selected development and trend basis. No discounting or mortality model is used; the amounts are empirical aggregate data, with no parametric loss distribution assumed.

Method. First construct incurred losses by adding paid losses and all applicable outstanding reserves. Then divide the resulting total by claims and earned exposure using the stated, consistent bases.

The outstanding reserves total $120{,}000+48{,}000=\$168{,}000$. Therefore:

L=600{,}000+120{,}000+48{,}000=\$768{,}000.

Frequency is $96/1{,}200=0.08$ claims per vehicle-year. Average severity is $\$768{,}000/96=\$8{,}000$ per claim. Loss cost is $\$768{,}000/1{,}200=\$640$ per vehicle-year. As a check, $0.08\times \$8{,}000=\$640$ per vehicle-year.

Common mistake avoided: Using only paid losses would produce $\$600{,}000/1{,}200=\$500$, understating the selected incurred loss cost by $\$140$ per vehicle-year. Omitting IBNR would also be wrong here: incurred losses include the applicable $48,000 IBNR reserve as well as case reserves.

Worked Example: Converting Written Policies to Earned Policy-Years

A hypothetical commercial coverage has 500 policies written during the experience period. Of these, 300 are in force for a full year; the other 200 are in force for one-half year. The selected basis is earned policy-years. There are 54 claims and $270,000 of developed incurred losses, including applicable outstanding reserves. Calculate frequency, average severity, and loss cost per earned policy-year. Assume the claims and losses match the same coverage and experience period, no additional trend adjustment is required, and no discounting, mortality model, or parametric loss distribution is used.

Method. Weight each policy by its earned term rather than treating every written policy as a full year of exposure. The 300 full-year policies contribute $300\times1=300$ earned policy-years. The 200 half-year policies contribute $200\times0.5=100$. Thus total earned exposure is $300+100=400$ policy-years.

Frequency is $54/400=0.135$ claims per earned policy-year. Average severity is $\$270{,}000/54=\$5{,}000$ per claim. Loss cost is $\$270{,}000/400=\$675$ per earned policy-year. The check is $0.135\times \$5{,}000=\$675$.

Common mistake avoided: Dividing by 500 written policies as if each earned a full year gives $\$270{,}000/500=\$540$. That is not a loss cost per earned policy-year: the actual denominator is 400 earned policy-years. The written-policy count may be reported as a separate statistic, but it is not interchangeable with earned exposure.

Worked Example: Applying a Per-Claim Limit Before Aggregation

A hypothetical liability coverage has a $15,000 insurer payment limit per claim and no deductible. Five developed claim amounts are $8,000, $22,000, $40,000, $12,000, and $15,000. The segment has 500 earned exposure units. Calculate the total limited losses and the limited loss cost per exposure. Assume the amounts are incurred claim amounts on the same selected basis, including applicable outstanding reserves. No discounting, mortality model, or parametric loss distribution is used.

Method. Apply the $15,000 per-claim cap separately to each claim using $Y^L=\min(X,15{,}000)$. The limited amounts are, in the original order, $8,000, $15,000, $15,000, $12,000, and $15,000. Their sum is

L^L=8{,}000+15{,}000+15{,}000+12{,}000+15{,}000=\$65{,}000.

The limited loss cost is $\$65{,}000/500=\$130$ per exposure. As a cross-check, the limited average severity is $\$65{,}000/5=\$13{,}000$ per claim, and frequency is $5/500=0.01$ claims per exposure; their product is $0.01\times \$13{,}000=\$130$.

Common mistake avoided: The uncapped sum is $8{,}000+22{,}000+40{,}000+12{,}000+15{,}000=\$97{,}000$, which would give an uncapped loss cost of $194 per exposure. That is not the requested limited-basis result. The cap applies to each claim before summation; it is not applied once to the aggregate.

A Practical Basis-Alignment Check

Before calculating an indication, document what each data field represents. The exposure definition, claim-count definition, loss valuation, coverage, and experience-period assignment should be explicit. A useful review is to trace the units from the source data through each calculation and ask whether every ratio has an interpretable unit.

Define the coverage and exposure unit.
State the policy coverage and whether exposure is, for example, an earned vehicle-year, location-year, or specified payroll unit.

Determine earned exposure for the period.
Account for the portion of coverage actually provided. Do not substitute a written-policy count unless that is deliberately the defined measure.

Align claim counts and loss amounts.
Confirm that they refer to the same coverage, claim population, and period assignment, and state the evaluation date and reserve basis.

Apply the selected loss basis consistently.
Include applicable outstanding reserves in incurred losses. If the analysis uses limited losses, apply the limit claim by claim before aggregation.

Check the resulting units and arithmetic.
Verify that frequency times severity equals loss cost when the same counts, losses, and exposures are used.

The loss cost calculated here is a loss measure per exposure, not automatically a final rate. A filed rate may also reflect other selected components and adjustments. Those components must be handled on compatible bases, but they do not change the basic requirement that each loss-cost ratio use matching exposure and loss definitions.

Key takeaway: An exposure-based loss indication is interpretable only when the exposure denominator, claim count, and loss numerator share a clearly defined coverage and experience basis. Incurred losses include paid losses plus applicable outstanding case and IBNR reserves; limited losses are capped claim by claim.

Check Your Understanding

For each problem, identify the stated basis and keep the units consistent.

A hypothetical segment has 800 earned exposure units, 64 claims, and $352,000 of incurred losses, including all applicable outstanding reserves. Calculate frequency, average severity, and loss cost per exposure. (Answer: Frequency $=0.08$ claims per exposure; severity $=\$5,500$ per claim; loss cost $=\$440$ per exposure.)
A book has 400 written policies, each in force for six months, and $90,000 of matching incurred losses from 30 claims. If the unit is an earned policy-year, calculate earned exposure and loss cost per earned policy-year. Assume exposure accrues uniformly during each policy term. (Answer: 200 earned policy-years; $\$450$ per earned policy-year.)
Four incurred claim amounts are $6,000, $14,000, $25,000, and $9,000. Apply a $10,000 per-claim limit with no deductible, then calculate limited losses per 250 earned exposure units. (Answer: Limited losses total $35,000; limited loss cost $=\$140$ per exposure.)
At the evaluation date, a segment has $410,000 paid, $55,000 in case reserves, and $35,000 in IBNR reserves. If it has 1,000 earned exposure units on the same basis, calculate incurred losses and incurred loss cost per exposure. (Answer: Incurred losses $=\$500,000$; loss cost $=\$500$ per exposure.)
A report gives 40 claims and $180,000 of losses for a coverage segment, but the claims are assigned by report date while the losses are assigned by accident date. Can the resulting quotient be treated as average severity for the accident-period claims? State the necessary correction. (Answer: Not without reconciling the bases; align the count and losses to the same claim population and accident-period definition before calculating severity.)

What You'll Learn