A Priori Loss Estimates for Bornhuetter-Ferguson

Purpose of an A Priori Loss Estimate

The Bornhuetter-Ferguson method needs an expected-loss assumption in addition to observed claims and an estimate of how much loss has emerged. This tutorial focuses on constructing that expected-loss input. A priori means the estimate is selected from assumptions established independently of the current period’s observed loss emergence; it does not mean the assumptions cannot be informed by historical data, pricing analyses, or other relevant experience.

There are two standard ways to express the expected amount: apply a selected expected loss ratio to premium, or apply a selected expected loss cost to exposure. Both estimate the same kind of quantity: the expected ultimate losses for the period. Neither method, by itself, measures the observed amount of claims reported or paid to date.

The estimate is an expected amount, not a claim count and not an estimate of unpaid losses on its own. In Bornhuetter-Ferguson, the expected ultimate is combined with the unreported share implied by the selected emergence pattern. The emergence pattern is a separate assumption: it does not create or change the a priori expected ultimate. Here, the methods for selecting that expected amount are the subject.

Premium-Based Expected Ultimate Loss

Under the premium-based method, select an a priori expected loss ratio and apply it to the premium for the same policies and coverage period. If \(P^{on}\) is earned premium adjusted to the chosen rate level, and \(\widehat{LR}^{a\ priori}\) is the selected expected loss ratio on a compatible basis, then the expected ultimate loss is:

The superscript “on” indicates that premium has been adjusted to the intended rate level when such an adjustment is needed. If the selected loss ratio is already defined for the premium basis being used, no further adjustment should be made. The key requirement is compatibility: the premium and expected loss ratio must refer to the same business, period, and loss definition.

For example, if expected loss ratio includes loss adjustment expense, the resulting expected ultimate includes that expense. If the estimate is intended to be net of reinsurance, both the expected loss ratio and premium basis must support a net estimate. Do not combine a gross loss ratio with net premium, or a loss ratio for one coverage with premium for a broader package, without an explicit adjustment that makes the bases consistent.

Exposure-Based Expected Ultimate Loss

An exposure-based method uses an expected loss cost per exposure unit. An exposure unit might be a vehicle-year, payroll amount, insured value, or another measure appropriate to the coverage. If \(E_j\) is the number of exposure units in class \(j\), and \(\widehat{LC}^{a\ priori}_j\) is expected ultimate loss per unit for that class, then:

For one homogeneous class, the sum reduces to exposure multiplied by expected loss cost. Units must be carried through the calculation. If loss cost is stated per 100 exposure units, first convert the exposure to groups of 100, or divide the exposure by 100. If loss cost is per vehicle-year, use vehicle-years rather than the number of vehicles at a single point in time.

The expected loss cost must also match the period and coverage being estimated. If it is a selected cost at an earlier cost level, any required trend or coverage adjustment must be made before multiplying by current exposure. Conversely, do not trend a loss cost that is already selected at the target level. The selected expected cost may reflect assumptions about claim frequency and severity, but calculating or selecting those components is not required to apply this method.

Why These Formulas Estimate Expected Ultimate Loss

The premium-based formula follows from the definition of an expected loss ratio. For a defined portfolio on a defined basis, the expected loss ratio is expected ultimate loss divided by the corresponding premium. If the selected ratio is \(\widehat{LR}^{a\ priori}\), rearranging that relationship gives the premium-based estimate:

The exposure-based formula follows by partitioning the portfolio into exposure units. If the expected ultimate loss per unit in class \(j\) is \(\widehat{LC}^{a\ priori}_j\), then the expected loss for \(E_j\) units is \(E_j\widehat{LC}^{a\ priori}_j\). Adding across classes gives the total. These relationships require the selected ratio or cost to be appropriate for the corresponding premium or exposure. They are not identities that make a mismatched assumption valid.

In practical estimation, the selected ratio or cost is an actuarial assumption. It may come from a pricing indication, historical experience adjusted for relevant changes, or another justified source. The formula calculates the expected amount conditional on that selection; it does not prove that the selected assumption is appropriate. An expected value also does not imply that actual losses will equal the estimate.

Basis Checks and Common Exam Traps

Before accepting an a priori expected ultimate, check the connection between the selected assumption and the quantity being estimated. A clean multiplication can still produce an unsuitable estimate if the inputs refer to different periods or scopes.

• Premium measure: determine whether the calculation calls for earned or written premium, and whether that premium is at the required rate level. Do not silently substitute one measure for another.
• Loss definition: establish whether the target includes loss adjustment expense, particular coverages, or reinsurance effects. The selected ratio or cost must reflect the same scope.
• Exposure units: state what one unit represents and confirm that the loss cost is quoted per that unit. Convert “per 100 units” or “per $1,000 of payroll” explicitly.
• Time basis: align the exposure period with the loss cost and any trend or other adjustment. An annual cost per vehicle-year is not automatically appropriate for a partial-year exposure.
• Expected versus observed: do not replace the selected a priori expected loss with reported or paid losses. Those observed amounts belong to the loss-emergence information used elsewhere in the Bornhuetter-Ferguson calculation.

If premium-based and exposure-based estimates differ, first compare their bases and assumptions. Differences can result from inconsistent premium periods, class mix, loss components, trend levels, or exposure definitions. Do not average two outputs merely because they disagree. Identify and resolve the mismatch, or explain why the underlying selections genuinely represent different views.