Building A Priori Estimates From Historical Ultimate Experience

Introduction

An expected claims estimate for a new period can be informed by the insurer’s own historical experience. The starting point is not simply the losses paid or reported so far: historical losses must first be estimated at ultimate, so the periods being compared include an estimate of their full claim costs. Those developed amounts can then be related to historical premium or exposure volume.

This tutorial develops two historical benchmarks: an ultimate loss ratio and an ultimate loss cost. Each can support an a priori estimate for a new period when the historical data and the new-period basis are suitably comparable. The calculations assume that the ultimate losses, premiums, and exposures have been placed on a consistent basis where necessary. Adjusting a loss ratio to a different premium basis is a separate issue.

Definition: For historical period $j$, let $U_j$ be the estimated ultimate losses, $P_j$ the corresponding earned premium, and $E_j$ the corresponding exposure units. An ultimate loss ratio is ultimate losses divided by earned premium. An ultimate loss cost is ultimate losses divided by exposure units. All losses in a calculation must have the same scope, such as consistently including or excluding allocated loss adjustment expenses.

From Developed Losses to a Historical Benchmark

An ultimate loss estimate is an estimate of the total losses attributable to a period after claims have fully developed. It may come from a selected development method, such as a chain ladder estimate, or from another appropriate actuarial selection. The development method is not the subject here; the important input is the selected ultimate amount, rather than the amount currently paid or reported.

For periods with earned premium, the aggregate historical ultimate loss ratio is calculated by dividing total ultimate losses by total earned premium. For periods with measured exposures, the aggregate ultimate loss cost is total ultimate losses divided by total exposures.

\widehat{LR}_{\text{hist}} = \frac{\sum_j U_j}{\sum_j P_j}, \qquad \widehat{LC}_{\text{hist}} = \frac{\sum_j U_j}{\sum_j E_j}.

The subscripts indicate that these are historical indications. They are not automatically the right selections for a future period. Their usefulness depends on whether the underlying experience is relevant and comparable, and on whether the selected benchmark is appropriate for the new period.

Why Aggregate Ratios and Costs Are Weighted

An aggregate loss ratio is equivalent to a premium-weighted average of the individual period loss ratios. To see why, write $LR_j=U_j/P_j$. Then:

\frac{\sum_j U_j}{\sum_j P_j} = \frac{\sum_j P_j(U_j/P_j)}{\sum_j P_j} = \frac{\sum_j P_j LR_j}{\sum_j P_j}.

Thus, a period with more premium receives more weight in the aggregate. Likewise, the aggregate loss cost is an exposure-weighted average of the individual period loss costs. This follows by writing $LC_j=U_j/E_j$:

\frac{\sum_j U_j}{\sum_j E_j} = \frac{\sum_j E_j(U_j/E_j)}{\sum_j E_j} = \frac{\sum_j E_j LC_j}{\sum_j E_j}.

These identities explain why a simple average of annual ratios or costs will generally differ from the aggregate result. A simple average gives every period equal weight, whether it represents a small or large amount of premium or exposure.

Assumptions and limits: Applying a historical benchmark to a new period assumes that it is a reasonable guide to expected claims on the new period’s basis. Material changes in claim costs, coverage, business mix, or the premium or exposure basis can make unadjusted historical results unsuitable. The formulas calculate historical indications; they do not establish that the periods are comparable.

Worked Example: A Premium-Weighted Historical Loss Ratio

A hypothetical commercial insurer has selected ultimate losses of $840,000, $1,050,000, and $1,320,000 for three historical accident years. The corresponding earned premiums are $1,200,000, $1,500,000, and $1,800,000. Assume all amounts are on a consistent basis and that the ultimate losses have the same loss scope. Calculate the aggregate historical ultimate loss ratio and estimate ultimate losses for a new period with $2,100,000 of earned premium.

Method. Calculate each year’s ultimate loss ratio, then calculate the aggregate ratio from total ultimate losses divided by total premium. Apply the aggregate ratio to the new-period premium.

Historical period ratios. The three individual ratios are:

\frac{\$840{,}000}{\$1{,}200{,}000}=0.70, \qquad \frac{\$1{,}050{,}000}{\$1{,}500{,}000}=0.70, \qquad \frac{\$1{,}320{,}000}{\$1{,}800{,}000}=0.733333\ldots.

Aggregate ratio. Total ultimate losses are $840,000 + $1,050,000 + $1,320,000 = $3,210,000. Total earned premium is $1,200,000 + $1,500,000 + $1,800,000 = $4,500,000. Therefore:

\widehat{LR}_{\text{hist}} = \frac{\$3{,}210{,}000}{\$4{,}500{,}000} = 0.713333\ldots = 71.333\ldots\%.

New-period estimate. Apply the historical ratio to the new premium:

\$2{,}100{,}000 \times 0.713333\ldots = \$1{,}498{,}000.

The estimated ultimate losses are $1,498,000. As a check, the selected ratio is below 1, so the estimate is below the $2,100,000 premium. That relationship is consistent with the calculated 71.333% ratio. The check does not prove the selection is suitable; it confirms the arithmetic and interpretation.

Common mistake avoided. The simple average of the three annual ratios is $(70\%+70\%+73.333\ldots\%)/3=71.111\ldots\%$. Applying it would give $1,493,333.33, not $1,498,000. The aggregate ratio is the appropriate premium-weighted historical result because the years have different premium volumes.

Calculating and Applying Historical Loss Costs

When exposure units are available and suitable for the coverage, a loss cost expresses ultimate losses per unit of exposure. Examples of exposure units include policy-years, vehicle-years, and earned units of property coverage. The units must be defined consistently across historical and prospective periods.

Once a historical loss cost has been selected, the corresponding new-period expected losses are calculated by multiplying by the new period’s exposure volume. This is a direct application of the per-exposure estimate: expected losses per unit multiplied by the number of units.

Worked Example: An Exposure-Weighted Historical Loss Cost

A hypothetical insurer estimates historical ultimate losses of $360,000, $475,000, and $600,000 for three periods with 400, 500, and 600 earned policy-years, respectively. Calculate the aggregate ultimate loss cost and use it to estimate losses for a new period with 720 policy-years. Assume all periods have comparable coverage and consistent exposure definitions.

Method. Divide total ultimate losses by total policy-years, then multiply the resulting cost per policy-year by the new exposure volume.

Historical totals. The losses total $360,000 + $475,000 + $600,000 = $1,435,000. The exposures total 400 + 500 + 600 = 1,500 policy-years. Therefore:

\widehat{LC}_{\text{hist}} = \frac{\$1{,}435{,}000}{1{,}500} = \$956.666\ldots \text{ per policy-year}.

As a check on the weighting, the separate period costs are $360,000/400 = $900, $475,000/500 = $950, and $600,000/600 = $1,000 per policy-year. The aggregate cost, $956.666..., lies between the smallest and largest annual costs.

New-period estimate. Multiply the historical cost by 720 policy-years:

720 \times \$956.666\ldots = \$688{,}800.

The estimated ultimate losses are $688,800. The result is consistent with the scale of the inputs: 720 exposures are 48% of the 1,500 historical exposures, and 48% of $1,435,000 is $688,800.

Common mistake avoided. Averaging the three annual costs equally gives $(\$900+\$950+\$1,000)/3=\$950$, which would produce $684,000 for the new period. That calculation gives equal weight to each year rather than to each policy-year. The aggregate cost uses the total exposure denominator and is the exposure-weighted historical result.

Developed Ultimate Experience as an A Priori Estimate

The word “a priori” distinguishes the selected expected-loss benchmark from losses that have emerged so far in the period being estimated. Historical ultimate experience can provide that benchmark, while the new period’s premium or exposure volume determines its scale. In practice, the historical ultimate amounts may themselves be estimates, so the output inherits uncertainty from the loss development and from the selection of historical periods.

A loss ratio and a loss cost describe related but different bases. A loss ratio relates losses to premium; a loss cost relates losses to exposure units. Use the measure that fits the available data and the intended prospective basis. Do not switch between them without accounting for the relationship between premium and exposure.

Assemble comparable historical inputs.
For each period, identify the selected ultimate losses, the earned premium or exposure units, and the loss scope. Confirm the denominators use consistent definitions.

Calculate an aggregate historical benchmark.
Divide total ultimate losses by total premium for a loss ratio, or by total exposures for a loss cost. This automatically weights periods by premium or exposure volume.

Apply the selected benchmark.
Multiply the loss ratio by the relevant new-period premium, or multiply the loss cost by the new-period exposure volume.

Review relevance and arithmetic.
Check that the historical experience is suitable for the intended estimate, that units match, and that the result is consistent with the calculated ratio or cost.

Worked Example: Developing Historical Losses Before Estimating a New Period

A hypothetical insurer has two historical accident years. For the first year, the selected current incurred amount is $640,000 and the selected development factor to ultimate is 1.25. For the second year, those amounts are $900,000 and 1.10. Earned premiums are $1,000,000 and $1,200,000. Assume the factors are appropriate for the respective periods, and that no separate trend or premium-basis adjustment is needed for this illustration. Estimate ultimate losses for a new period with $1,500,000 of premium.

Method. First develop each historical amount to ultimate by multiplying by its selected factor. Then calculate the aggregate historical ultimate loss ratio and apply it to the new premium.

Develop the first year. $640,000 × 1.25 = $800,000.

Develop the second year. $900,000 × 1.10 = $990,000.

Calculate the aggregate ratio. The ultimate losses total $800,000 + $990,000 = $1,790,000, and the earned premiums total $1,000,000 + $1,200,000 = $2,200,000:

\widehat{LR}_{\text{hist}} = \frac{\$1{,}790{,}000}{\$2{,}200{,}000} = 0.81363636\ldots = 81.363636\ldots\%.

Estimate new-period ultimate losses.

\$1{,}500{,}000 \times 0.81363636\ldots = \$1{,}220{,}454.55.

Rounded to the nearest dollar, the estimate is $1,220,455. As an arithmetic check, the result is 81.3636...% of the new premium, matching the historical ratio. The method avoids the mistake of applying a development factor to the new period’s premium or treating the historical current incurred amounts as if they were ultimate losses.

Common Traps and Practical Checks

Using emerged losses instead of ultimate losses. A partially developed period may not yet include its full claim cost. Use the selected ultimate estimate for each historical period.
Taking an unweighted average without a reason. A simple average can be useful only when equal period weighting is deliberately intended. It is not the aggregate loss ratio or aggregate loss cost when period volumes differ.
Mixing units. A loss cost per vehicle-year must be applied to vehicle-years, not policies or vehicles at a point in time unless those are demonstrably the same exposure measure.
Ignoring comparability. A precisely calculated historical ratio can still be a poor prospective estimate if the data do not represent the new period’s conditions.
Rounding too early. Keep intermediate ratios and costs at adequate precision, then round the final result to the required level.

Key takeaway: Develop historical losses to ultimate, relate their total to the corresponding total premium or exposure, and apply the selected aggregate benchmark to the new period’s matching volume. Weighting, unit consistency, and comparability are essential to interpreting the resulting estimate.

Check Your Understanding

Work each problem using the aggregate historical benchmark unless the question states otherwise. Round final percentages to two decimal places and money to the nearest cent.

A hypothetical insurer has ultimate losses of $480,000, $600,000, and $750,000, with earned premiums of $600,000, $750,000, and $1,000,000. What is the aggregate historical ultimate loss ratio? (Answer: 77.87%)
Historical ultimate losses total $210,000, $300,000, and $440,000 for 300, 400, and 500 exposure units. Estimate losses for 650 comparable new-period exposure units using the aggregate loss cost. (Answer: $514,583.33)
Two historical periods have current incurred losses of $720,000 and $950,000, selected factors to ultimate of 1.15 and 1.08, and earned premiums of $1,200,000 and $1,400,000. Estimate ultimate losses for a new period with $900,000 of premium. (Answer: $641,769.23)
Three periods have ultimate losses of $400,000, $600,000, and $900,000 and premiums of $500,000, $1,000,000, and $1,500,000. Calculate the aggregate ratio and the estimate for a new period with $2,000,000 premium. (Answer: 63.33%; $1,266,666.67)
Historical ultimate losses are $960,000 for 800 exposure units and $1,100,000 for 1,000 exposure units. Estimate losses for a new period with 450 comparable exposure units. (Answer: $515,000.00)

What You'll Learn