Selecting the Exposure Basis for Experience Periods

Why the Exposure Denominator Matters

A historical loss total has limited meaning on its own. A hypothetical fleet with $90,000 of losses might have insured 30 vehicles for a year or 300 vehicles for a year. Those totals describe very different experience. An exposure measure supplies the denominator needed to interpret losses relative to the amount of risk insured.

Selecting that denominator is a design decision, not a clerical one. The measure should reflect the insured risk and the way losses can arise under the coverage. For a coverage that responds to employee injury, payroll or employee time may be informative. For a fleet, vehicle-years or miles may be useful. For property damage, insured value-years may be more informative than a count of buildings when the buildings differ substantially in value.

Definition: An exposure basis is the measure used to quantify the amount of insured risk over a specified experience period. An exposure amount is the total quantity measured on that basis. A loss rate per exposure unit is historical losses divided by exposure amount, with the result expressed per a stated, standardized unit. The exposure basis does not change the losses; it changes the scale on which those losses are interpreted.

An exposure unit must be specified clearly. “Payroll” is incomplete unless its currency and scale are stated; “$100 of payroll” is more precise. Similarly, “miles” might be stated as miles or as units of 10,000 miles. Standardizing the unit makes the resulting rate easier to interpret and use.

Choosing a Basis for the Covered Risk

The best basis is not necessarily the measure that is easiest to retrieve or the one that produces a more convenient rate. Start with the risk unit defined for the coverage, then ask what quantity best represents the amount of risk insured. The relationship need not be perfect, but it should be defensible for the intended use.

Choose a measure connected to the coverage. A basis should have a plausible relationship to the opportunities for covered loss. For example, miles driven may be informative for vehicle liability exposure when more driving means more opportunities for an accident.
Choose a measure that represents the insured risk. If properties differ widely in value, counting each property as one identical exposure may hide important differences in the amount insured.
Use a well-defined unit. State whether the measure is per employee-year, vehicle-year, $100 of payroll, or another explicit unit. A count without a time component can be ambiguous when policies begin or end during the period.
Consider consistency and availability. A conceptually strong measure may be impractical if it is recorded inconsistently. A weaker but consistently measured basis may be more useful, provided its limitations are understood.
Keep the interpretation tied to the chosen basis. A rate per vehicle-year is not the same quantity as a rate per mile. The values cannot be compared as if they had the same denominator.

The calculation of a loss rate is straightforward once the basis is selected. If total historical losses are $L$, the exposure amount is $E$, and each standardized unit represents $c$ exposure units, then the rate per standardized unit is $cL/E$. For a count-based measure, $c$ may be 1,000 exposures; for a dollar-based measure, it may be 100 dollars of payroll or 1,000 dollars of insured value.

\text{Loss rate per }c\text{ exposure units} = \frac{cL}{E}

The units should cancel as intended. If losses are in dollars and exposure is in payroll dollars, the resulting rate has units of loss dollars per payroll dollar. Multiplying by 100 expresses it per $100 of payroll. Explicit unit labels are a practical check against denominator errors.

Why Loss Divided by Exposure Has Meaning

The loss-per-exposure calculation has a simple justification when the expected loss grows proportionally with exposure. Suppose $E$ is a fixed amount of exposure and $\lambda$ is the expected loss per exposure unit for a homogeneous risk group. Under the assumption that expected losses are proportional to exposure, expected aggregate loss is $\lambda E$. Dividing that expected amount by $E$ recovers the rate $\lambda$.

E[L\mid E]=\lambda E \qquad\Longrightarrow\qquad E\left[\frac{L}{E}\,\middle|\,E\right] = \frac{E[L\mid E]}{E} = \lambda, \quad E>0

This derivation explains the role of the denominator: it scales the loss amount by the quantity of insured risk. It also identifies an important limitation. If risk per exposure unit varies materially across the portfolio, a single aggregate rate can change when the mix of risks changes, even if the underlying rates within groups do not. A rate per exposure unit is therefore not automatically a universal rate for every portfolio.

Worked Applications

Worked Example: Selecting a Workers’ Compensation Exposure Basis

A hypothetical employer reports $105,000 of workers’ compensation losses for an experience period. It had $5,250,000 of payroll and an average of 150 employees, each insured for a full year. For interpreting this experience and projecting a similar payroll-based account, calculate the loss rate per $100 of payroll and per employee-year. Explain which measure is more appropriate if the covered operations and wage levels vary. Assume the losses and exposures refer to the same stated experience period.

Method. Calculate each rate using its own denominator, then consider what each measure represents. The example does not assume that either rate is a universal rate for other employers.

Payroll basis. The number of $100 payroll units is $5{,}250{,}000/100=52{,}500$. Thus the loss rate is $105{,}000/52{,}500=\$2.00$ per $100 of payroll.

Employee-year basis. Since 150 employees were each insured for a full year, exposure is 150 employee-years. The rate is $105{,}000/150=\$700.00$ per employee-year.

Interpretation and check. Both calculations use the same $105,000 loss total, but their denominators describe different quantities. If payroll is the chosen basis, the rate is $2.00 for every $100 of payroll. For example, applying it to $5,250,000 of payroll gives $52{,}500\times \$2.00=\$105{,}000$, recovering the original total. Payroll may be preferable where wage levels differ because it reflects payroll volume; employee-years give each employee-year equal weight regardless of wages. The common mistake is to treat $700 per employee-year as interchangeable with $2.00 per $100 of payroll. They are different rates with different units and assumptions.

Worked Example: Choosing Between Vehicle-Years and Mileage

A hypothetical delivery fleet has $96,000 of liability losses, 80 insured vehicle-years, and 2,400,000 miles driven during an experience period. Management expects the next period to involve 2,700,000 miles and 105 vehicle-years. Calculate historical loss rates per vehicle-year and per 100,000 miles, then project losses under each basis. If driving distance is judged to be the better measure of the liability exposure for this fleet, which projection should be used as the basis for interpretation?

Method. Normalize the historical losses separately by each exposure measure. Then apply each rate to its corresponding projected exposure; do not apply a per-mile rate to vehicle-years or vice versa.

Vehicle-year rate. The rate is $96{,}000/80=\$1{,}200.00$ per vehicle-year. Applied to 105 projected vehicle-years, the projection is $105\times \$1{,}200=\$126{,}000$.

Mileage rate. The historical mileage exposure is $2{,}400{,}000/100{,}000=24$ units of 100,000 miles. The rate is $96{,}000/24=\$4{,}000.00$ per 100,000 miles. The projected mileage is $2{,}700{,}000/100{,}000=27$ such units, giving $27\times \$4{,}000=\$108{,}000$.

Interpretation and check. The two projections differ because the projected relationship between miles and vehicle-years differs from the historical relationship. The calculations are internally consistent: each historical rate multiplied by its historical exposure recovers $96,000. If mileage is the selected basis because it better represents the fleet’s liability exposure, the mileage-based projection is $108,000. The common mistake is to assume that both measures must produce the same prospective amount. They need not, especially when the portfolio’s use of vehicles changes.

Worked Example: Property Value Versus Property Count

A hypothetical property portfolio has $72,000 of losses during an experience period. The summed insured values, each counted for its insured time in the period, total $18,000,000 of insured value-years. The portfolio also has 240 insured location-years. Calculate the loss rate per $1,000 of insured value-year and per location-year. For a prospective period, assume 20,000 units of $1,000 insured value-years and 250 location-years. Calculate the loss projection under each historical rate, and explain why one basis may be preferable.

Method. Express the insured-value exposure in $1,000 units before dividing. The values and location counts are alternative exposure measures, not interchangeable units.

Insured-value rate. The historical exposure is $18{,}000{,}000/1{,}000=18{,}000$ units of $1,000 insured value-years. The rate is $72{,}000/18{,}000=\$4.00$ per $1,000 insured value-year. Applying it to 20,000 prospective units gives $20{,}000\times \$4.00=\$80{,}000$.

Location rate. The rate is $72{,}000/240=\$300.00$ per location-year. Applying it to 250 prospective location-years gives $250\times \$300=\$75{,}000$.

Interpretation and check. Insured value may be a more informative basis when the amount of property at risk varies substantially across locations and loss potential is expected to relate to that amount. A location-year treats each location-year equally and may be more useful if locations are reasonably similar. Each historical rate checks back to the same loss total: $18{,}000\times \$4=\$72{,}000$ and $240\times \$300=\$72{,}000$. The prospective amounts differ because the new period’s value-to-location mix differs. The common mistake is to report $4.00 as a rate per $1,000 of property without stating the time component; here, the unit is specifically a $1,000 insured value-year.

Portfolio Mix and Common Traps

An aggregate loss rate is an exposure-weighted result. It is not generally the unweighted average of subgroup rates. This matters whenever groups have different amounts of exposure or different loss rates. If the portfolio’s composition changes, the overall rate can change even when each subgroup’s own rate stays the same.

For example, suppose one group has 10,000 exposure units and a $3 loss rate per unit, while another has 2,000 units and an $8 rate. Their aggregate losses are $10{,}000\times \$3=\$30{,}000$ and $2{,}000\times \$8=\$16{,}000$, for $46,000 total. The aggregate rate is $46{,}000/12{,}000=\$3.83$ per exposure unit, rounded to the nearest cent. The simple average of the subgroup rates, $(3+8)/2=\$5.50$, is not the portfolio rate because it gives equal weight to groups with unequal exposure.

Before interpreting any rate, check the basis, scale, and time unit. In particular, do not divide by a current exposure figure simply because it is available if the intended measure is exposure over the experience period. The denominator must be defined for the period and risk being interpreted. This tutorial focuses on selecting and interpreting that denominator; it does not establish that the losses and exposure records have been adjusted for every possible timing or coverage difference.

Key takeaway: Choose an exposure measure that represents the insured risk, state its unit precisely, and divide losses by the corresponding exposure amount. A rate is meaningful only together with its denominator, and alternative bases can yield different projections when exposure mix changes.

Check Your Understanding

For each problem, state the relevant basis and keep its units attached to the rate.

A hypothetical equipment policy has $42,000 of losses, 1,400 equipment-months of exposure, and 70 machine-years. If utilization is the main risk driver, calculate the loss rate per 100 equipment-months and project losses for 1,600 equipment-months. (Answer: $3,000 per 100 equipment-months; projected losses are $48,000.)
A hypothetical professional liability portfolio has $18,000 of losses and $2,400,000 of revenue exposure. Calculate the rate per $1,000 of revenue and project losses for $3,000,000 of revenue at that rate. (Answer: $7.50 per $1,000 revenue; projected losses are $22,500.)
A hypothetical property group has $96,000 of losses, $24,000,000 of insured value-years, and 300 location-years. Calculate the historical rates per $1,000 insured value-year and per location-year. If the prospective exposure is 26,000 units of $1,000 insured value-years, calculate the projection using the value basis. (Answer: $4.00 per $1,000 insured value-year; $320 per location-year; value-based projection is $104,000.)
A hypothetical delivery fleet has $126,000 of losses, 90 vehicle-years, and 1,800,000 miles. Calculate the rates per vehicle-year and per 10,000 miles. If next year has 95 vehicle-years and 2,100,000 miles, calculate both projections and select the mileage-based projection if miles are judged the better measure. (Answer: $1,400 per vehicle-year and $700 per 10,000 miles; projections are $133,000 and $147,000, respectively; use $147,000 under the stated mileage-basis assumption.)
Two hypothetical groups have loss rates of $3 and $8 per exposure unit, with 10,000 and 2,000 units, respectively. Calculate the aggregate rate and the projection for 15,000 units at that aggregate rate. Explain why the unweighted average rate is not the aggregate rate. (Answer: $46,000 historical losses over 12,000 units gives $3.83 per unit, rounded to the nearest cent; the projection is $57,500. The unweighted $5.50 average ignores unequal exposure amounts.)

What You'll Learn