Methods for Sparse and Immature Periods

Why Sparse and Immature Periods Need Care

An accident period with little reported experience may have a small observed loss amount simply because much of its claim activity has not yet emerged. Treating that amount as a reliable indication of ultimate losses can produce an unstable estimate. Bornhuetter-Ferguson (BF) and Cape Cod (CC) address this problem by using the proportion of losses expected to be reported by the valuation date, together with an expected loss level.

The distinction between the methods matters most when periods are immature or their reported losses are sparse. BF uses an expected loss ratio selected independently of the period’s reported experience. Cape Cod estimates a loss ratio from the portfolio’s reported experience, adjusted for how much of each period is expected to have emerged. In both methods, a selected development pattern determines the reported proportion and therefore affects the estimate.

Definition: Let $E_j$ be the exposure measure for accident period $j$, $R_j$ its reported losses at the valuation date, and $p_j$ the selected proportion of ultimate losses expected to be reported by that date. The expected unpaid amount for the period under an expected loss ratio $r$ is $E_j r(1-p_j)$. The corresponding ultimate estimate is $R_j+E_jr(1-p_j)$.

The quantities must be on a consistent basis. For example, reported losses and the selected development proportions should both refer to paid losses, or both refer to incurred losses. The expected loss ratio must be compatible with the exposure measure and loss basis. If a cumulative development factor is used, it is an ultimate-to-current factor, so the reported proportion is its reciprocal: $p_j=1/\text{development factor}$.

Bornhuetter-Ferguson for a Sparse Period

For BF, the expected loss ratio $r$ is selected in advance of the period’s reported loss experience. The method does not ask the small observed amount to determine the expected loss level. Instead, it uses reported losses as observed so far and adds the expected portion that remains unreported.

\text{BF unpaid}_j=E_jr(1-p_j), \qquad \text{BF ultimate}_j=R_j+E_jr(1-p_j).

This approach can be useful when a period is very immature, or when its observed losses are too sparse to be a stable guide to ultimate losses. It does not mean that the reported amount is ignored: reported losses enter the ultimate estimate directly. The expected loss ratio determines only the unpaid component.

Worked Example: BF When Only a Small Proportion Is Reported

A hypothetical accident period has earned premium of $1,200,000. Its selected expected loss ratio is 65%, and its reported losses are $90,000. The selected cumulative development factor to ultimate is 4.00. Calculate the BF unpaid estimate and ultimate estimate.

Method. Find the reported proportion from the development factor, then apply the BF expected-loss calculation to the unreported proportion.

Step 1: Reported proportion. The factor is ultimate-to-current, so $p=1/4.00=0.25$. Thus 25% is expected to be reported and $1-0.25=0.75$, or 75%, remains unreported.

Step 2: Expected ultimate losses at the selected loss ratio. $E r=\$1,200,000(0.65)=\$780,000$.

Step 3: BF unpaid estimate. $\$780,000(0.75)=\$585,000$.

Step 4: BF ultimate estimate. Add the observed reported losses to the expected unpaid amount: $\$90,000+\$585,000=\$675,000$.

Check. The expected losses not yet reported are 75% of the $780,000 expected ultimate, or $585,000. The ultimate estimate is $90,000 reported plus $585,000 unpaid. The method avoids treating the sparse reported amount as a complete estimate of the period’s ultimate loss.

Cape Cod: Estimating the Loss Ratio From Portfolio Experience

Cape Cod replaces the independently selected BF loss ratio with a portfolio estimate. Each period contributes reported losses to the numerator and exposure multiplied by its reported proportion to the denominator. More mature periods generally contribute more denominator weight per unit of exposure because a larger share of their expected losses has emerged.

\widehat{r}_{CC} = \frac{\sum_j R_j}{\sum_j E_jp_j}, \qquad \text{CC unpaid}_j = E_j\widehat{r}_{CC}(1-p_j).

The Cape Cod ratio is an estimate of ultimate losses per unit of exposure. Its denominator is not simply total exposure: it is exposure weighted by the proportion reported. The unpaid estimate for each period then applies that portfolio ratio to the period’s exposure and unreported proportion.

Worked Example: Cape Cod With Two Periods at Different Maturities

A hypothetical portfolio has two accident periods, each with earned premium of $1,000 thousand. Period 1 has reported losses of $80 thousand and a selected reported proportion of 0.20. Period 2 has reported losses of $420 thousand and a selected reported proportion of 0.60. Calculate the Cape Cod loss ratio, each period’s expected unpaid amount, and the combined unpaid estimate.

Method. Estimate the portfolio loss ratio using total reported losses divided by exposure-weighted reported proportions, then apply that ratio to each period’s unreported proportion.

Step 1: Exposure-weighted reported amount in the denominator. Period 1 contributes $1{,}000(0.20)=200$ thousand. Period 2 contributes $1{,}000(0.60)=600$ thousand. The denominator is $200+600=800$ thousand.

Step 2: Cape Cod loss ratio. Total reported losses are $80+420=500$ thousand. Therefore, $\widehat{r}_{CC}=500/800=0.625$, or 62.50%.

Step 3: Period 1 unpaid estimate. The unreported proportion is $1-0.20=0.80$. Thus the estimate is $1{,}000(0.625)(0.80)=500$ thousand.

Step 4: Period 2 unpaid estimate. The unreported proportion is $1-0.60=0.40$. Thus the estimate is $1{,}000(0.625)(0.40)=250$ thousand.

Step 5: Combined unpaid estimate. The total is $500+250=750$ thousand.

Check. The total exposure-weighted unreported amount is $1{,}000(0.80)+1{,}000(0.40)=1{,}200$ thousand. Multiplying by 62.50% gives $1{,}200(0.625)=750$ thousand, consistent with the period-by-period calculation. The reported losses of $500 thousand are not included in this unpaid total.

Why the Formulas Work

The key relationship follows by separating expected ultimate losses into the portions reported and unreported. Suppose a period has exposure $E_j$, a common expected loss ratio $r$, and reported proportion $p_j$. Under that expected-loss model, expected ultimate losses are $E_jr$. If the development pattern allocates proportion $p_j$ of ultimate losses to the reported portion, expected reported losses are $E_jrp_j$, leaving expected unpaid losses $E_jr-E_jrp_j$.

$$ E_jr-E_jrp_j=E_jr(1-p_j). $$

For Cape Cod, the same relationship gives expected reported losses across the portfolio of $r\sum_j E_jp_j$. Equating this to the observed portfolio reported losses and solving for $r$ gives the Cape Cod ratio. This is a moment-based rationale, not a guarantee that observed reported losses exactly follow their expectations. In particular, if the portfolio experience is sparse, the estimated ratio can be sensitive to individual reported amounts and to the selected proportions.

Check the loss basis.
Confirm that reported amounts and development proportions refer to the same paid or incurred basis, valuation date, and loss definition.

Convert development factors to reported proportions.
For an ultimate-to-current factor $f_j$, use $p_j=1/f_j$, not $f_j$ itself.

Select the loss-ratio basis.
For BF, use the selected expected loss ratio. For Cape Cod, calculate the ratio from portfolio reported losses and $\sum_jE_jp_j$.

Calculate unpaid losses, then ultimate losses if needed.
Apply $E_jr(1-p_j)$ to obtain unpaid losses. Add $R_j$ only when the requested quantity is ultimate losses.

How Sparse Periods Affect the Estimates

A period with a small $p_j$ has a large unreported proportion. Under BF, its unpaid estimate is therefore driven mainly by exposure, the selected expected loss ratio, and the development pattern; its reported amount affects the ultimate estimate only through the direct addition of $R_j$. Under Cape Cod, the period’s own reported amount is part of the portfolio numerator, but its contribution to the denominator is only $E_jp_j$. A very immature period consequently has relatively little influence on the estimated loss ratio per unit of exposure.

That does not make Cape Cod immune to sparse experience. If the portfolio itself has little reported experience, its ratio can still be unstable. Also, a period with $p_j=0$ contributes no denominator weight. If every period has $p_j=0$, the Cape Cod ratio is undefined; there is no reported experience from which to estimate it using this formula. A selected external ratio may then be needed for an estimate such as BF.

Worked Example: A Development-Pattern Change Alters Cape Cod

A hypothetical three-period portfolio has earned premium of $1,000 thousand in each period. Reported losses are $30 thousand, $260 thousand, and $500 thousand, respectively. The selected reported proportions are initially 0.10, 0.50, and 0.90. Recalculate the Cape Cod ratio and total unpaid estimate if review of the first period’s development pattern changes its reported proportion from 0.10 to 0.20, leaving all other inputs unchanged.

Method. Recalculate both the denominator and the period unpaid estimates. Changing one reported proportion affects not only that period’s unreported percentage but also the portfolio loss ratio.

Step 1: Total reported losses. The numerator is $30+260+500=790$ thousand.

Step 2: Revised denominator. The contributions are $1{,}000(0.20)=200$, $1{,}000(0.50)=500$, and $1{,}000(0.90)=900$ thousand. The revised denominator is $200+500+900=1{,}600$ thousand.

Step 3: Revised Cape Cod ratio. $\widehat{r}_{CC}=790/1{,}600=0.49375$, or 49.375%.

Step 4: Revised unpaid estimates. Period 1 has unreported proportion 0.80, giving $1{,}000(0.49375)(0.80)=395$ thousand. Period 2 has unreported proportion 0.50, giving $1{,}000(0.49375)(0.50)=246.875$ thousand. Period 3 has unreported proportion 0.10, giving $1{,}000(0.49375)(0.10)=49.375$ thousand.

Step 5: Revised total and comparison. The revised combined unpaid estimate is $395+246.875+49.375=691.25$ thousand. Under the original proportions, the denominator was $100+500+900=1{,}500$ thousand, so the ratio was $790/1{,}500=0.526666\ldots$. The original total unpaid estimate was $0.526666\ldots[1{,}000(0.90)+1{,}000(0.50)+1{,}000(0.10)]=790$ thousand. The revised estimate is 98.75 thousand lower.

Check. Under the revised pattern, total exposure-weighted unreported amount is $800+500+100=1{,}400$ thousand. Multiplying by the revised ratio gives $1{,}400(0.49375)=691.25$ thousand, matching the sum of the three period estimates. All money amounts are in thousands; no intermediate rounding was used.

Common Exam Trap

A frequent error is to confuse a development factor with a reported proportion. If an ultimate-to-current factor is 4, the reported proportion is $1/4=0.25$, not 4. Another error is to calculate an unpaid estimate and then subtract reported losses from it. The BF unpaid formula already measures the amount expected to remain unpaid; reported losses are added only when calculating an ultimate estimate.

For example, in the BF calculation above, the expected-loss amount was $780,000 and the unreported proportion was 0.75. The correct unpaid estimate is $\$780,000(0.75)=\$585,000$. The tempting calculation $\$780,000-\$90,000=\$690,000$ is not the BF unpaid estimate: it subtracts actual reported losses from expected ultimate losses, rather than applying the selected development proportion. The correct BF ultimate estimate is $\$90,000+\$585,000=\$675,000$.

A separate Cape Cod trap is using total exposure rather than $\sum_jE_jp_j$ in the loss-ratio denominator. That would fail to adjust for maturity and would not be consistent with the model’s expected reported-loss relationship.

Key takeaway: For sparse or immature periods, identify the reported proportion first and keep the unpaid and ultimate quantities distinct. BF applies an independently selected loss ratio; Cape Cod estimates its ratio from reported portfolio losses using exposure weighted by reported proportions. Changes in the development pattern can affect both the unreported share and the Cape Cod ratio.

Check Your Understanding

For each problem, calculate with the stated units and distinguish unpaid losses from ultimate losses.

A hypothetical period has earned premium of $800,000, an expected loss ratio of 55%, reported losses of $70,000, and a cumulative development factor of 4.00. Find the BF unpaid and ultimate estimates. (Answer: BF unpaid $330,000; BF ultimate $400,000.)
Two periods have earned premiums of $600 thousand and $900 thousand, reported proportions of 0.50 and 0.80, and reported losses of $150 thousand and $500 thousand. Find the Cape Cod loss ratio and combined unpaid estimate. (Answer: 63.73%; $305.88 thousand.)
A BF period has premium of $2,000,000, expected loss ratio 60%, and reported proportion 0.20. If a revised development selection raises the reported proportion to 0.25, by how much does the unpaid estimate decrease? (Answer: $60,000.)
A two-period Cape Cod portfolio has exposures of $500 thousand and $1,000 thousand, reported proportions of 0.20 and 0.80, and reported losses of $20 thousand and $400 thousand. Find the Cape Cod loss ratio and the first period’s unpaid estimate. (Answer: 46.67%; $186.67 thousand.)
A portfolio has total reported losses of $360 thousand, total exposure of $1,200 thousand, and exposure-weighted reported proportions summing to $600 thousand. What Cape Cod loss ratio results, and what is the combined unpaid amount if total exposure-weighted unreported exposure is $600 thousand? (Answer: 60.00%; $360 thousand.)

What You'll Learn