Why the Exposure Base Matters
A pure premium is a measure of loss cost per unit of exposure. The exposure base is therefore not a bookkeeping detail: it defines what one unit of the loss cost represents. A dollar amount per vehicle-year answers a different question from a dollar amount per policy-year, per $100 of payroll, or per participant-session. Choosing the right base makes the result interpretable and useful for the intended rate unit.
The appropriate base depends on the coverage being priced and on how the insured risk is measured. For example, vehicle-years can suit an automobile coverage whose exposure accumulates with the time vehicles are insured. Payroll can suit a workers compensation coverage whose expected losses vary with covered payroll. A count of policies may be reasonable when policies have similar coverage and duration, but can mislead when policy sizes or terms differ substantially.
Selecting a base requires more than finding a number that can be counted. The base should relate to the covered risk, use the same coverage and loss measure as the numerator, and correspond to the unit on which a rate is intended to be charged. For this tutorial, losses and exposures are assumed to describe the same hypothetical portfolio and coverage. The detailed treatment of earned and written exposure timing is a separate topic.
Candidate Exposure Bases
Common bases include counts of insured objects, time-adjusted counts, amounts of insured activity, and monetary measures of risk. Each can be appropriate under the right conditions, but none is universally best.
| Exposure base | Typical unit | When it may be useful | Potential limitation |
|---|---|---|---|
| Vehicle-years | One vehicle insured for one year | Automobile coverage when vehicle count and insured duration represent the risk | Does not by itself reflect differences in vehicle use or coverage amount |
| Policy-years | One policy in force for one year | Coverage priced per policy when policy size and covered objects are reasonably comparable | Can treat a small policy and a much larger policy as equal exposure |
| Payroll | $100 or $1,000 of covered payroll | Workers compensation when payroll is the prescribed or relevant measure of covered work | Payroll amount alone does not distinguish different work classifications |
| Insured value | $1,000 of covered property value | Property coverage when value insured is an appropriate scale for potential loss | Value may not capture all differences in construction, location, or hazards |
| Activity count | One participant-session, trip, or operating hour | Liability coverage whose risk accumulates with the covered activity | Activity counts need consistent definitions and coverage boundaries |
A base can also be a practical proxy rather than a direct physical measure of risk. A proxy is defensible when it is consistently defined and is sufficiently related to expected covered losses for the purpose at hand. A convenient count is not automatically a good proxy: if the amount of risk represented by each counted item varies greatly, the resulting average can obscure that variation.
How the Denominator Sets the Rate Unit
For a selected base, the calculation is straightforward: divide the relevant aggregate loss amount by the number of units of that base. The arithmetic does not decide whether the base is appropriate. That is a judgment about the relationship between the coverage, the losses, the exposures, and the intended rate unit.
The relationship between aggregate loss and the denominator also explains why two correctly calculated loss costs can have different numerical values. Suppose total losses are \(L\), with \(E\) units of a first exposure base. If another base has \(E'=cE\) units for a positive constant \(c\), then the loss cost per unit of the second base is \(L/E'=L/(cE)\). The aggregate is preserved when the per-unit amount is multiplied by its corresponding exposure count:
This identity justifies changing units when the conversion is known and consistent. It does not mean that any two bases can be substituted. If exposure per policy varies, for example, there is no single conversion factor from policies to vehicle-years that applies to every portfolio. The resulting average per policy may change as the portfolio mix changes, even when the per-vehicle loss cost does not.
Worked Example: Vehicle-Years Versus Policy-Years
A hypothetical automobile portfolio has 180 single-vehicle policies insured for a full year and 60 two-vehicle policies insured for half a year. Aggregate covered losses are $294,000. Calculate the loss cost per vehicle-year and per policy-year, and identify which base is appropriate if the intended rate is charged per vehicle-year.
Method. Calculate each exposure count from the number of covered objects and the insured duration, then divide the same aggregate losses by each count. This avoids treating a half-year policy as a full year or a two-vehicle policy as one vehicle of exposure.
The 180 single-vehicle policies contribute \(180 \times 1 \times 1=180\) vehicle-years and 180 policy-years. The 60 two-vehicle policies contribute \(60 \times 2 \times 0.5=60\) vehicle-years and \(60 \times 0.5=30\) policy-years.
Total vehicle-years are \(180+60=240\). Total policy-years are \(180+30=210\). Thus the loss cost per vehicle-year is \(\$294{,}000/240=\$1{,}225\). The loss cost per policy-year is \(\$294{,}000/210=\$1{,}400\).
If the intended rate unit is one vehicle-year, use the $1,225 per vehicle-year result. The policy-year result is also arithmetically correct, but it represents a different unit and reflects the average number of vehicles per policy in this portfolio.
Check. \(\$1{,}225 \times 240=\$294{,}000\), and \(\$1{,}400 \times 210=\$294{,}000\). The totals reconcile. The common mistake would be dividing by all 240 policies as if each were a full-year policy; that would give $1,225 per policy, but the portfolio has only 210 policy-years.
Match the Base to the Risk Being Measured
A well-chosen exposure base has an understandable connection to the covered loss process. This does not require the base to explain every difference in risk. It does require the analyst to know what the denominator counts and what the resulting per-unit figure means.
State the unit in plain language, such as dollars per $100 of covered payroll or dollars per participant-session.
Determine whether the relevant unit is an object, a period of insurance, a monetary amount, or an instance of covered activity.
Ask whether policies, objects, or activities differ enough in size or duration that a simple count would be misleading.
Divide by the selected exposure count and retain the exact unit in the answer so that the result cannot be mistaken for a different rate basis.
Worked Example: Payroll as the Workers Compensation Base
A hypothetical insurer has two groups of covered businesses. Group A has $4,800,000 of covered payroll and $72,000 of covered losses. Group B has $3,200,000 of covered payroll and $96,000 of covered losses. Calculate the pooled loss cost per $100 of payroll and compare it with the group-specific loss costs.
Method. Use covered payroll because the intended exposure unit is $100 of payroll. Convert each payroll amount into $100 units before dividing. This avoids reporting a rate per dollar when the requested unit is per $100.
Group A has \(\$4{,}800{,}000/\$100=48{,}000\) exposure units. Its loss cost is \(\$72{,}000/48{,}000=\$1.50\) per $100 of payroll. Group B has \(\$3{,}200{,}000/\$100=32{,}000\) units. Its loss cost is \(\$96{,}000/32{,}000=\$3.00\) per $100.
For the pooled result, total payroll is \(\$4{,}800{,}000+\$3{,}200{,}000=\$8{,}000{,}000\), or 80,000 units of $100. Total losses are \(\$72{,}000+\$96{,}000=\$168{,}000\). The pooled loss cost is \(\$168{,}000/80{,}000=\$2.10\) per $100 of payroll.
Check. The group calculations reconstruct the total: \(\$1.50 \times 48{,}000=\$72{,}000\) and \(\$3.00 \times 32{,}000=\$96{,}000\). The pooled result also gives \(\$2.10 \times 80{,}000=\$168{,}000\).
A common unit error is to divide $168,000 by $8,000,000 and report $0.021 per $100. That quotient is $0.021 per dollar of payroll. Converting to the requested unit gives \(0.021 \times 100=\$2.10\) per $100. The answer must always state the denominator unit.
Worked Example: Participant-Sessions for Activity-Based Liability Coverage
A hypothetical guided-kayaking program records 1,200 customers who each take one covered session and 800 customers who each take five covered sessions. Aggregate covered liability losses are $104,000. Find the loss cost per participant-session and per unique participant. If a proposed rate is charged for each session, select the appropriate base.
Method. Count both unique participants and covered sessions, because they represent different possible rate units. The mistake to avoid is treating a customer who takes five sessions as the same exposure as a customer who takes one when the rate is intended to reflect each session.
The first group generates \(1{,}200 \times 1=1{,}200\) participant-sessions. The second generates \(800 \times 5=4{,}000\). Total exposure is \(1{,}200+4{,}000=5{,}200\) participant-sessions. There are \(1{,}200+800=2{,}000\) unique participants.
Loss cost per participant-session is \(\$104{,}000/5{,}200=\$20\). Loss cost per unique participant is \(\$104{,}000/2{,}000=\$52\). For a rate charged per session, the appropriate result is $20 per participant-session.
Check. \(\$20 \times 5{,}200=\$104{,}000\), and \(\$52 \times 2{,}000=\$104{,}000\). Both figures reproduce the same aggregate losses, but their units differ. The per-participant figure depends on the observed average of \(5{,}200/2{,}000=2.6\) sessions per person; a different session mix could change it.
Common Exposure-Base Traps
Exposure selection errors often survive the arithmetic because the division itself is correct. A useful final review is to check whether the numerator and denominator describe the same covered risk and whether the answer’s unit matches the question.
- Using a convenient count without checking size. Policy count may be unsuitable when policies insure different numbers of objects or different amounts of risk.
- Ignoring duration. A half-year exposure is not automatically one full year of exposure. Use the time adjustment appropriate to the defined unit.
- Dropping the unit scale. Dividing by payroll dollars and dividing by $100 of payroll differ by a factor of 100.
- Comparing unlike units as if they were contradictory. A per-policy loss cost and a per-vehicle loss cost can both be correct. They are not directly interchangeable without a valid conversion.
- Choosing a base that does not fit the intended charge. If the rate will be charged per activity session, a loss cost per unique customer does not directly answer the pricing question.
Check Your Understanding
For each hypothetical case, identify or calculate the loss cost using the exposure unit that fits the stated rate basis.
- A property portfolio has $45,000 of covered losses and $3,000,000 of insured value. The intended unit is $1,000 of insured value. How many exposure units are there, and what is the loss cost per unit? (Answer: 3,000 units; $15 per $1,000 of insured value.)
- A hypothetical workers compensation portfolio has $250,000 of covered losses and $12,500,000 of covered payroll. Calculate the loss cost per $100 of payroll. (Answer: 125,000 units; $2 per $100 of payroll.)
- A program has 900 participants who each attend one covered session and 300 who each attend four. Covered losses are $63,000. Calculate the loss cost per participant-session and per unique participant, then select the base for a rate charged per session. (Answer: 2,100 participant-sessions; $30 per participant-session; 1,200 unique participants; $52.50 per participant. Use participant-sessions for the per-session rate.)
- A hypothetical auto portfolio has 300 single-vehicle policies insured for a full year and 100 single-vehicle policies insured for half a year. Losses are $70,000. Calculate vehicle-years and the loss cost per vehicle-year. (Answer: 350 vehicle-years; $200 per vehicle-year.)
- A seasonal liability program has 600 covered workers and $9,000,000 of covered payroll, with $180,000 of covered losses. Find the loss cost per $100 of payroll and per worker. Which unit is appropriate if the rate is to vary with payroll? (Answer: 90,000 payroll units; $2 per $100 of payroll; $300 per worker. Use payroll units for a payroll-based rate.)