Why Basis Alignment Matters
Before using an expected loss ratio or an exposure-based loss estimate in Bornhuetter-Ferguson or Cape Cod, the inputs must describe the same business on the same basis. A premium amount may reflect several historical rate levels; an expected loss cost may reflect a different accident period; and observed losses may be net of deductibles while the selected loss ratio is gross of them. Combining such quantities can produce a precise calculation that estimates the wrong thing.
The objective is not to make every input identical. It is to identify what each input represents and adjust it, where appropriate, to a common basis. In particular, distinguish accident period, premium rate level, loss trend level, and loss definition. A rate-level adjustment to premium does not replace a trend adjustment to losses: they account for different changes.
Accident Period, Earned Premium, and Exposure
An accident period groups losses by when the insured events occurred, not necessarily by when policies were written, premiums were collected, or claims were reported. The premium or exposure used to estimate expected losses should represent the business whose losses are being estimated. For an accident-year estimate, earned premium and earned exposures associated with that accident year are generally more relevant than the year's written premium or policies in force at year-end.
Premium is earned over the period during which coverage is provided. Consequently, written premium in a calendar year can include coverage that earns in a later year, while earned premium can include portions of policies written earlier. The appropriate premium measure depends on the reserving analysis and the available data, but the chosen period must match the loss period.
An exposure measure counts units of risk, such as earned vehicle-years or insured property-years. An exposure-based estimate combines those units with an expected loss cost per unit. It is not interchangeable with premium: premium reflects both exposure and the price charged for that exposure. If rates have changed, premium may need to be adjusted to a common rate level before it is used as the basis for an expected loss ratio.
On-Leveling Premium
On-leveling expresses premium earned at different historical rate levels as if it had been earned at one selected target rate level. Let \(P_j\) be earned premium attributable to rate level \(r_j\), and let \(r_T\) be the target rate level. When premium is proportional to the rate level, the adjusted premium is:
Rate levels can be represented relative to a common index, such as 0.90, 0.95, and 1.00. The ratio \(r_T/r_j\) converts premium at the historical level to its equivalent at the target level. The calculation should use the rate level applicable to the premium being adjusted, accounting for earning patterns if the premium contains business written at multiple levels.
Worked Example: On-Leveling Premium Earned at Three Rate Levels
A hypothetical accident-year portfolio has $300,000 of earned premium at rate level 0.90, $500,000 at rate level 0.95, and $400,000 at rate level 1.00. The selected target rate level is 1.00. A selected expected loss ratio is 60% on that target-level premium basis. Calculate on-level earned premium and the corresponding expected losses. State the method and a common mistake it avoids.
Method. Adjust each earned-premium component separately to rate level 1.00, then apply the expected loss ratio to the adjusted total. This avoids treating a portfolio with mixed rate levels as though all its premium were earned at one level.
At rate level 0.90, the adjustment factor is \(1.00/0.90=1.111111\ldots\), so the adjusted component is:
At rate level 0.95, the factor is \(1.00/0.95=1.052631\ldots\), giving:
The premium already at the target level remains unchanged: \(\$400{,}000(1.00/1.00)=\$400{,}000.00\). Thus:
Expected losses on the stated basis are \(0.60(\$1{,}259{,}649.12)=\$755{,}789.47\), to the nearest cent.
Check. Each adjustment factor is at least 1 because the target level is higher than or equal to every historical level. The adjusted total, $1,259,649.12, therefore exceeds the unadjusted $1,200,000, as expected. Rounding was performed only for the final monetary amounts, which are reported to the cent.
Exposure-Based Loss Costs and Trend
When an expected loss cost is stated per exposure unit, its time basis must correspond to the accident period being estimated. If the loss cost is stated at an earlier level and a selected annual loss trend applies, compound it for the number of years between the loss-cost level and the target accident-period level. With annual trend rate \(t\) over \(n\) years, the adjustment is \((1+t)^n\). The time interval should be defined carefully; it is not automatically the number of calendar years between the valuation date and the estimate.
The resulting expected losses equal the aligned loss cost per unit multiplied by the matching exposure units. Do not also multiply by on-level premium factors: that would apply a rate-level adjustment to an exposure-based calculation that does not use premium as its exposure measure.
Worked Example: Trending a Loss Cost to the Accident Period
A hypothetical book has 1,600 earned vehicle-years in accident year 2025. A selected ultimate loss cost is $520 per vehicle-year at the 2023 accident-year level. Assume annual loss trend of 4% compounded for each year from 2023 to 2025. Estimate expected ultimate losses. Name the method and the common mistake it avoids.
Method. Trend the per-unit loss cost to the 2025 accident-year level, then multiply by 2025 earned vehicle-years. This avoids using a loss cost from an earlier period without adjustment or trending premium instead of losses.
The interval from the 2023 to the 2025 accident-year level is two years, so the trend factor is:
The 2025 loss cost is \( \$520(1.0816)=\$562.432\) per vehicle-year. Multiplying by 1,600 vehicle-years gives:
The estimated expected ultimate losses are $899,891.20. As a check, the trended cost is greater than $520 because the assumed trend is positive, and multiplying by 1,600 gives a total greater than 1,600 times the untrended cost, or $832,000.
Match the Loss Definition
An expected loss ratio is meaningful only when its numerator and the losses to which it is applied use compatible definitions. Check whether amounts are gross or net of deductibles, policy limits, salvage, and subrogation. Also check whether loss adjustment expenses are included, and whether the data are paid, incurred, or otherwise defined. A mismatch can arise even when the accident period and premium rate level are correct.
For example, a loss ratio selected net of a deductible should not be applied to a premium basis to estimate ground-up losses without an explicit adjustment. Conversely, a ground-up expected loss ratio does not directly estimate losses payable after deductibles. The treatment of expenses must be aligned as well: if the selected ratio includes allocated or unallocated loss adjustment expenses, the target loss measure must include them on the same basis.
Worked Example: Matching a Net-of-Deductible Loss Ratio
A hypothetical portfolio has 40 claims of $2,000 each and 10 claims of $8,000 each, all stated as ground-up covered losses. The coverage has a $1,000 deductible per claim and no per-claim limit. Earned premium is $200,000. A selected expected loss ratio of 55% is defined net of the deductible and excludes loss adjustment expenses. Calculate the net losses in the claim set and the expected losses implied by the selected ratio. Explain the basis check.
Method. Apply the deductible separately to each claim, then compare the resulting net losses with expected losses calculated from the net-of-deductible ratio. This avoids applying the net ratio to ground-up losses or adding expenses that the selected ratio excludes.
For each of the 40 claims of $2,000, the amount after the deductible is \(\$2{,}000-\$1{,}000=\$1{,}000\). Their total is \(40(\$1{,}000)=\$40{,}000\).
For each of the 10 claims of $8,000, the amount after the deductible is \(\$8{,}000-\$1{,}000=\$7{,}000\). Their total is \(10(\$7{,}000)=\$70{,}000\). Therefore net losses are:
The expected losses implied by the selected net ratio are \(0.55(\$200{,}000)=\$110{,}000\). The two amounts agree in this constructed example, and both exclude loss adjustment expenses. For comparison, ground-up losses are \(40(\$2{,}000)+10(\$8{,}000)=\$160{,}000\), which is not the appropriate amount to compare with the selected net ratio.
Check. The deductible reduces each claim by $1,000, so total ground-up losses of $160,000 fall by \(50(\$1,000)=\$50,000\), leaving $110,000. The calculation treats the deductible per claim, not once per portfolio.
A Practical Alignment Sequence
Before inserting an expected loss amount or ratio into a reserving calculation, document the basis of each input. The following sequence helps identify mismatches before they affect an estimate.
Confirm which accident period the losses represent and use earned premium or exposure corresponding to that period.
Decide whether the calculation uses earned premium at a selected rate level or exposure units paired with a loss cost per unit.
On-level premium for rate changes when needed. Trend loss costs or losses for claim-cost changes when needed, using the appropriate accident-period interval.
Check coverage, deductibles, limits, recoveries, expenses, and paid or incurred status against the expected loss ratio or loss cost.
Check units, adjustment direction, and whether the adjusted quantity should be larger or smaller than its original value.
Why the Premium Adjustment Works
The on-leveling formula follows directly from the meaning of a rate level. Suppose premium \(P_j\) was earned at relative rate level \(r_j\). Under the proportional-rate assumption, the exposure represented by that premium is proportional to \(P_j/r_j\): dividing out the rate level leaves a measure of the underlying exposure. Pricing that same exposure at target level \(r_T\) gives \((P_j/r_j)r_T\), which is \(P_j(r_T/r_j)\). Summing over components gives the total on-level premium:
This justification depends on the assumption that the relative rate-level index applies to the premium being adjusted and that premium changes proportionally with that index. If coverage, exposure mix, rating variables, or earning patterns have changed materially, a simple level factor may not capture the full change. The calculation adjusts the specified rate level; it does not correct unrelated changes in the business.
Common Exam Traps
- Using written premium automatically. Written premium and accident-period losses may cover different time intervals. Check the earning and accident-period basis first.
- Applying one factor to a mixed premium total. If the earned premium contains several rate levels, adjust each component using its own factor before summing.
- Trending to the wrong date. Trend the loss cost to the relevant accident-period level, not mechanically to the evaluation date.
- Double-adjusting. Do not trend premium as though it were a loss cost, or apply both an exposure-based loss-cost calculation and a premium adjustment without a defined reason.
- Mixing gross and net losses. A loss ratio net of deductibles, limits, or recoveries must be matched to losses on that same basis.
Check Your Understanding
For each problem, identify the relevant adjustment and check that the answer uses the stated basis.
- A hypothetical portfolio has $720,000 of earned premium at rate level 0.96. On-level it to target rate level 1.08. (Answer: $810,000)
- A hypothetical book has 900 earned exposure units in the target accident period. Its expected loss cost is $300 per unit at the level two years earlier, and annual loss trend is 3%. Calculate expected losses. (Answer: $286,443.00)
- Earned premium consists of $200,000 at rate level 0.80 and $300,000 at rate level 1.00. On-level the total to target rate level 1.00. (Answer: $550,000)
- A hypothetical set has 25 ground-up claims of $4,000 each, each subject to a $500 deductible. Calculate the total losses net of deductibles. (Answer: $87,500)
- A selected expected loss ratio is 40% net of salvage and subrogation, while the loss amount to be estimated is defined gross of those recoveries. Is direct application consistent? State the needed correction in basis. (Answer: No; adjust the expected loss ratio or loss measure so both are either gross or net of salvage and subrogation.)