Aligning Premium and Exposure with Bornhuetter-Ferguson Projections

Why Alignment Matters

Bornhuetter-Ferguson (BF) combines losses already observed with an estimate of the losses not yet emerged. Its arithmetic is straightforward; the harder judgment is often whether the expected ultimate loss estimate is on the same basis as the accident year being projected. Earned premium can reflect historical rate levels, while the selected expected loss ratio may reflect a different rate level or a different loss-cost date. Exposure counts can have the same issue when the expected loss cost is stated at another date.

The key is to define the target basis before calculating. Identify the accident years and earned periods represented, the rate level and loss-cost date underlying the expected loss assumption, and the loss definition used for both the expected and observed amounts. Only then should the expected ultimate losses be combined with the unreported share.

Aligning Earned Premium

Suppose earned premium for an accident year was written at a historical rate level, but the selected a priori loss ratio is intended for a reference rate level. A rate-level index expresses the average rate level relative to a chosen base. If the historical index is \(R_h\) and the reference index is \(R_r\), the simple on-level premium adjustment is:

This adjustment estimates the premium that the same earned exposure would have produced at the reference rates. It assumes the rate-level indices appropriately represent the business mix and that the adjustment is proportional. If the accident year contains materially different rate changes across its earned periods or classes, use an earned-premium-weighted on-level calculation rather than treating the whole year as if it had one rate level.

After on-leveling, multiply by an expected loss ratio only if that ratio is defined for the same premium basis and loss definition. A loss ratio selected at the reference rate level can be applied to premium on that reference level. A separate loss trend adjustment is needed only if the expected loss ratio’s underlying loss costs are not already at the intended loss-cost date.

Aligning Exposure and Loss Costs

When expected losses are built from exposure, the basic relationship is expected loss equals exposure multiplied by expected loss cost per exposure unit. If the loss cost is stated at a different date from the accident-year average loss date, trend the loss cost over the appropriate interval. For an annual trend rate \(g\) over \(n\) years, the factor is \((1+g)^n\). The years must represent the time between the loss-cost basis date and the date for which losses are being estimated, not simply the accident-year label difference.

Here \(X_{\text{AY}}\) is exposure for the accident year, \(LC_{\text{base}}\) is the expected loss cost per exposure unit at its stated base date, and \(n\) is the elapsed time to the target loss date. This form assumes the annual trend rate is constant over the interval and that the exposure definition is stable. If the selected loss cost is already stated at the target date, do not trend it again.

How Alignment Enters the BF Formula

Let \(U\) denote the selected unreported share at the valuation age, and let \(E_{\text{ult}}\) be expected ultimate losses on the aligned basis. If \(L_{\text{obs}}\) is the observed loss amount on a consistent basis, the BF ultimate and unpaid estimate are:

The expected amount in the unpaid component must match the accident year and loss definition of the observed amount. For example, if observed losses are net of reinsurance, the expected losses should generally be net on a consistent basis as well. Likewise, do not apply an expected loss ratio that includes allocated loss adjustment expenses to observed losses that exclude them without making the definitions consistent.

The reason the BF unpaid formula uses the expected amount, rather than the observed amount divided by emergence, follows from the model’s decomposition. Write ultimate loss as emerged loss plus unreported loss. Under the BF assumption, expected unreported loss is the a priori expected ultimate multiplied by the unreported share \(U\). The observed loss supplies the emerged component; it does not change the separately estimated expected unreported component. Therefore:

This decomposition is meaningful only if \(E_{\text{ult}}\), \(L_{\text{obs}}\), and \(U\) refer to the same loss basis and accident-year population. Rate-level and trend adjustments are part of establishing that expected amount, not extra BF terms to apply automatically afterward.

Combining Rate-Level and Loss-Trend Adjustments

Rate-level adjustment and loss trend address different quantities. On-leveling changes premium to reflect a chosen rate level. Loss trend changes expected loss costs to reflect a change in the level of losses over time. They may both be needed, but only when the selected expected loss ratio or loss cost does not already embody the corresponding adjustment.

For example, an expected loss ratio may be selected at a particular historical loss-cost level but intended to apply to premium at a reference rate level. If the accident-year exposure is expected to experience two years of loss trend relative to that loss-cost basis, then on-level premium and trended losses can both enter the expected loss calculation. In contrast, if the selected expected loss ratio is already prospective and already reflects that trend, applying another trend factor double counts the change.

Practical Alignment Checks and Exam Traps

Before calculating expected losses, write down the basis of each input. A compact check prevents many otherwise plausible errors:

A frequent exam trap is to treat every stated trend rate as an instruction to trend all losses. The trend interval and the item being trended matter. Another is to adjust the premium for a rate-level change and then also alter the expected loss ratio as though it were still on the old premium basis. These operations can both be appropriate only when each adjustment has a distinct, stated purpose.