Choosing Between Bornhuetter-Ferguson and Cape Cod

Why Method Selection Depends on the Portfolio

Bornhuetter-Ferguson (BF) and Cape Cod both combine observed losses with an expectation of losses not yet observed. Their difference is the source of that expectation. BF uses an expected ultimate loss assumption selected independently of the current experience being reserved. Cape Cod estimates a common expected loss ratio from the portfolio’s observed losses, adjusted for how much of each period is expected to have emerged.

Neither method is automatically preferable. A well-supported external expectation can be valuable when a portfolio has little credible experience or unstable recent results. Conversely, when the portfolio has substantial, relevant, stable experience but the external expected loss ratio is questionable, Cape Cod can let the experience inform the estimate. The selection is only as sound as the fit between the method’s assumptions and the portfolio.

Definition: For accident period $i$, let $E_i$ be the exposure measure (such as premium, on a consistent basis), $p_i$ the expected proportion of ultimate losses reported or paid by the valuation date on the selected loss basis, and $L_i$ the observed cumulative losses on that same basis. Let $q_i=1-p_i$. Let $L_i^0$ denote an independently selected expected ultimate loss for BF. The BF reserve is $L_i^0q_i$. For Cape Cod, the common experience-based loss ratio is $\widehat{\theta}_{CC}=\sum_i L_i/\sum_i E_ip_i$, and the reserve for period $i$ is $E_i\widehat{\theta}_{CC}q_i$.

These formulas require consistent inputs. The exposure and losses must refer to compatible populations, periods, and financial bases; development proportions must be appropriate for those periods; and $p_i$ must be between 0 and 1. The Cape Cod formula additionally assumes that one common expected loss ratio is suitable across the periods being pooled. If risk levels or exposure definitions differ materially, the pooled ratio may not represent any period well.

What Each Method Asks You to Trust

Question	Bornhuetter-Ferguson	Cape Cod
Where does the expected loss level come from?	An independently selected expected ultimate loss or expected loss ratio	The portfolio’s observed losses, adjusted for expected emergence
What is especially important?	Credibility and relevance of the external prior	Credibility, stability, and comparability of the experience being pooled
When can it be weak?	When the expected loss assumption is poorly supported or no longer relevant	When experience is sparse, volatile, or drawn from changing risk mixes

“Large volume” alone does not guarantee that Cape Cod is suitable. Experience must also be relevant to the risks being estimated and reasonably stable across the periods used to estimate the common loss ratio. Likewise, a prior does not become reliable merely because it is labeled an expected loss ratio: it needs a defensible basis, compatible exposure, and continued relevance.

Why the Cape Cod Adjustment Uses Developed Exposure

The Cape Cod estimate adjusts the exposure in each period by its expected developed proportion. The reason is that the observed losses represent only the part of ultimate losses expected to have emerged by the valuation date. If the expected ultimate loss per unit of exposure is a common value $\theta$, then the expected observed loss in period $i$ is $E_i\theta p_i$. Summing across periods gives the expected portfolio losses observed to date.

\sum_i L_i \approx \sum_i E_i\theta p_i =\theta\sum_i E_ip_i \quad\Longrightarrow\quad \widehat{\theta}_{CC}=\frac{\sum_i L_i}{\sum_i E_ip_i}.

This is an experience-based estimate of the common loss ratio. It is not the ratio of observed losses to total exposure: periods with less expected emergence contribute less developed exposure to the denominator. After estimating the loss ratio, Cape Cod assigns expected ultimate losses $E_i\widehat{\theta}_{CC}$ and reserves the unreported or unpaid share $q_i$.

Justification: Under the model that each period has expected ultimate losses $E_i\theta$ and that proportion $p_i$ is expected to have emerged, expected observed losses equal $E_i\theta p_i$. The estimator above equates total observed losses to their model expectation and solves for $\theta$. The adjustment is appropriate only if the loss and exposure data are aligned and a common $\theta$ is a reasonable portfolio assumption.

Worked Comparisons

Worked Example: A Defensible Prior and Limited Experience

A hypothetical specialty portfolio has two accident periods. Exposure is measured in thousands of dollars of premium, losses are in thousands of dollars, and all values are on the same selected cumulative loss basis. For the two periods, $E_1=1{,}000$, $E_2=800$, $p_1=0.25$, $p_2=0.50$, and observed losses are $L_1=120$, $L_2=300$. A recent, independently reviewed expected loss ratio of 0.60 is considered applicable. Calculate both reserves and recommend a method, given that the portfolio is small and recent results are volatile.

Method. Calculate BF using the selected prior, then calculate Cape Cod to show how the observed experience would set its own loss ratio. The method choice should reflect the stated credibility and prior evidence, not simply which reserve is larger.

BF calculation. Expected ultimate losses from the 0.60 expected loss ratio are $1{,}000(0.60)=600$ and $800(0.60)=480$. The unreported proportions are $q_1=1-0.25=0.75$ and $q_2=1-0.50=0.50$. Therefore the reserves are $600(0.75)=450$ and $480(0.50)=240$, for a total BF reserve of $450+240=690$ thousand dollars.

Cape Cod calculation. Developed exposure is $1{,}000(0.25)+800(0.50)=250+400=650$. Total observed losses are $120+300=420$. Thus $\widehat{\theta}_{CC}=420/650=0.6461538\ldots$. The period reserves are $1{,}000(0.6461538\ldots)(0.75)=484.6154\ldots$ and $800(0.6461538\ldots)(0.50)=258.4615\ldots$. The total is $743.0769\ldots$ thousand dollars.

Selection and check. The stated small volume and volatile experience make the portfolio loss ratio a weak basis for replacing the independently reviewed prior. BF is therefore better supported by the facts given, although the difference between estimates is a reason to examine the assumptions. The BF reserve plus observed losses is $690+420=1{,}110$ thousand dollars, its implied total estimated ultimate losses. Rounding only at the final step, to the nearest thousand dollars, the reserves are $690,000 for BF and $743,000 for Cape Cod.

Common mistake avoided: selecting Cape Cod just because it uses current experience. Current experience is not necessarily credible; here the stated limited volume and volatility weaken it.

Worked Example: Credible Experience and a Questionable Prior

A hypothetical mature, stable portfolio has two periods, each with exposure of 2,000 thousand dollars of premium. The expected emergence proportions are $p_1=0.50$ and $p_2=0.75$, with observed losses $L_1=720$ and $L_2=1{,}140$, in thousands of dollars. An older external expected loss ratio of 0.60 has not been validated for current conditions. Assume the periods are comparable and the experience is sufficiently large to be credible. Calculate both reserves and identify the more supportable basis.

BF calculation. Applying the stated external ratio gives expected ultimate losses of $2{,}000(0.60)=1{,}200$ in each period. The unreported proportions are $q_1=0.50$ and $q_2=0.25$. The BF reserves are $1{,}200(0.50)=600$ and $1{,}200(0.25)=300$, totaling 900 thousand dollars.

Cape Cod calculation. Developed exposure is $2{,}000(0.50)+2{,}000(0.75)=1{,}000+1{,}500=2{,}500$. Total observed losses are $720+1{,}140=1{,}860$. Therefore $\widehat{\theta}_{CC}=1{,}860/2{,}500=0.744$. The reserves are $2{,}000(0.744)(0.50)=744$ and $2{,}000(0.744)(0.25)=372$, totaling 1,116 thousand dollars.

Selection and check. Since the experience is assumed credible, stable, and comparable, while the external ratio is old and unvalidated, Cape Cod has the stronger stated basis. Its estimated ultimate losses are $1{,}116+1{,}860=2{,}976$ thousand dollars. BF’s are $900+1{,}860=2{,}760$ thousand dollars. Rounding only at the final step, to the nearest thousand dollars, the reserves are $900,000 for BF and $1,116,000 for Cape Cod.

Common mistake avoided: treating the external expected loss ratio as authoritative simply because it was selected previously. An assumption can lose relevance as the portfolio or conditions change.

Worked Example: Changing Risk Mix Weakens a Pooled Ratio

A hypothetical insurer has two distinct classes with stable, class-specific expected loss ratios. Exposure is in thousands of dollars of premium, and observed losses are in thousands of dollars on a consistent basis. Class A has $E_A=1{,}000$, $p_A=0.50$, $L_A=400$, and a well-supported expected loss ratio of 0.50. Class B has $E_B=500$, $p_B=0.25$, $L_B=180$, and a well-supported expected loss ratio of 0.80. Calculate the sum of class-specific BF reserves and the reserve from applying a single Cape Cod ratio across both classes. Explain which method better respects the available risk information.

Class-specific BF calculation. For A, expected ultimate losses are $1{,}000(0.50)=500$, and $q_A=1-0.50=0.50$, giving a reserve of $500(0.50)=250$. For B, expected ultimate losses are $500(0.80)=400$, and $q_B=1-0.25=0.75$, giving a reserve of $400(0.75)=300$. The total class-specific BF reserve is $250+300=550$ thousand dollars.

Pooled Cape Cod calculation. Total observed losses are $400+180=580$. Developed exposure is $1{,}000(0.50)+500(0.25)=500+125=625$. The pooled ratio is $\widehat{\theta}_{CC}=580/625=0.928$. Using this common ratio, the A reserve is $1{,}000(0.928)(0.50)=464$; the B reserve is $500(0.928)(0.75)=348$. The pooled total is $464+348=812$ thousand dollars.

Selection and check. The common Cape Cod ratio treats the two classes as if they shared one expected loss ratio, despite the stated class-specific levels of 0.50 and 0.80. Because those distinct levels are well supported, class-specific BF better preserves the available risk information. The pooled result’s estimated ultimate losses are $812+580=1{,}392$ thousand dollars. The BF reserve plus observed losses is $550+580=1{,}130$ thousand dollars. Rounding only at the final step, to the nearest thousand dollars, the reserves are $550,000 and $812,000, respectively.

Common mistake avoided: pooling all periods or classes without checking whether they share a meaningful common loss ratio. A larger pooled data set does not correct a mismatch in risk level.

A Practical Selection Framework

A method recommendation should explain which information is more credible for the unpaid losses and why. A useful sequence is to evaluate the expected loss assumption, the experience, and then the comparability of the risks being combined.

Check the prior.
For BF, ask whether the expected ultimate losses or expected loss ratio are supported, current, and based on exposure consistent with the losses being estimated.

Check experience credibility.
For Cape Cod, consider whether the observed volume and period maturity provide meaningful information. Sparse or highly volatile experience may not be a dependable replacement for a sound prior.

Check stability and comparability.
Determine whether periods have sufficiently similar underlying risk levels for one experience-based ratio to be appropriate. A changing mix can undermine a pooled Cape Cod estimate even when total volume is high.

State the recommendation conditionally.
Connect the method to the evidence: for example, “BF is better supported because the external prior is current while the portfolio experience is sparse,” rather than asserting that one method is always superior.

The figures from the two methods are not interchangeable truths. They are estimates based on different assumptions. A substantial difference may signal that the current experience and external expectation imply different loss levels, but the difference alone does not identify which estimate is correct. The conclusion should follow from the credibility, stability, and relevance of the evidence described in the problem.

Common Exam Traps

Using total exposure in Cape Cod’s denominator. The denominator is developed exposure, $\sum_i E_ip_i$, not $\sum_i E_i$. Omitting $p_i$ ignores differences in maturity.
Confusing the methods’ expected loss sources. BF takes its expected ultimate losses from an independent selection; Cape Cod estimates a common loss ratio from observed losses and developed exposure.
Choosing solely from the numerical result. A higher reserve does not make a method more appropriate. Explain the method choice from the quality of its assumptions and data.
Assuming portfolio volume guarantees stability. Large experience can still be unrepresentative when the risk mix changes or unlike classes are pooled.
Mixing units or bases. Exposure, losses, and development proportions must correspond to the same population and chosen loss basis. A ratio calculated from inconsistent inputs has no sound interpretation.

Key takeaway: Prefer BF when a relevant independent prior is more dependable than the available experience. Prefer Cape Cod when experience is credible and stable, a common loss ratio is defensible, and the external prior is weak or outdated. In either case, justify the choice from the evidence, not from the direction of the reserve difference.

Check Your Understanding

For each problem, calculate the requested quantities and connect the method choice to the stated evidence. All amounts are in thousands of dollars unless stated otherwise.

A hypothetical portfolio has $E_1=500$, $E_2=700$, $p_1=0.40$, $p_2=0.80$, and observed losses $L_1=140$, $L_2=490$. Calculate the Cape Cod ratio and total reserve. Then calculate BF reserve using an independently selected expected loss ratio of 0.75. Which reserve is larger? (Answer: Cape Cod ratio $=630/(200+560)=0.8289$; Cape Cod reserve $=500(0.8289)(0.60)+700(0.8289)(0.20)=364.74$; BF reserve $=500(0.75)(0.60)+700(0.75)(0.20)=330.00$; Cape Cod is larger.)
A small, volatile portfolio has exposure $E=800$, $p=0.25$, and observed losses $L=170$. A current, well-supported expected loss ratio is 0.50. Calculate BF and Cape Cod reserves and recommend a basis if the experience is explicitly described as not credible. (Answer: BF reserve $=800(0.50)(0.75)=300$; Cape Cod ratio $=170/(800(0.25))=0.85$, giving reserve $=800(0.85)(0.75)=510$; BF is better supported by the stated evidence.)
A large stable portfolio has total exposure $E=3{,}000$, expected emergence $p=0.60$, and observed losses $L=1{,}260$. An external expected loss ratio of 0.40 is known to be outdated. Calculate both reserves and state which estimate is more supported if the experience is credible and the risk mix is stable. (Answer: Cape Cod ratio $=1{,}260/(3{,}000(0.60))=0.70$, so its reserve is $3{,}000(0.70)(0.40)=840$; BF reserve $=3{,}000(0.40)(0.40)=480$; Cape Cod is better supported under the stated assumptions.)
Two stable classes have well-supported expected loss ratios. Class A has $E_A=600$, $p_A=0.50$, $L_A=180$, and expected loss ratio 0.40. Class B has $E_B=400$, $p_B=0.25$, $L_B=100$, and expected loss ratio 0.80. Calculate class-specific BF reserve and pooled Cape Cod reserve. Which method respects the stated class differences? (Answer: BF $=600(0.40)(0.50)+400(0.80)(0.75)=120+240=360$; pooled Cape Cod ratio $=280/(300+100)=0.70$, reserve $=600(0.70)(0.50)+400(0.70)(0.75)=210+210=420$; class-specific BF respects the stated class differences.)
A common Cape Cod ratio is calculated by dividing total observed losses by total exposure, without multiplying each exposure by its emergence proportion. Explain the error and give the correct denominator for periods with exposures $E_1=900$, $E_2=600$ and emergence proportions $p_1=0.20$, $p_2=0.75$. (Answer: the denominator must be developed exposure, not total exposure; it is $900(0.20)+600(0.75)=180+450=630$, rather than $900+600=1{,}500$.)

What You'll Learn