Why Cape Cod Weights Expected Loss by Emergence
An accident year’s observed losses are only the portion of its ultimate losses that has emerged by the valuation date. A recent year may therefore have few observed losses even when its ultimate claims are expected to be substantial. The Cape Cod method uses both the losses observed so far and the amount of expected loss exposure those observations represent to estimate a common expected loss ratio.
The key is to weight each year’s premium by its expected emergence percentage. A year with high premium and substantial emergence provides more information about the common loss ratio than a small, immature year. Cape Cod then applies the estimated common ratio to each accident year’s premium. This is an exposure-weighted expected loss principle, not a simple average of the accident years’ individual loss ratios.
The product \(P_iw_i\) is the year’s emergence-weighted premium exposure. It is not the full premium and it is not the premium multiplied by the unreported proportion. The emergence percentage describes the share expected to have emerged; the unreported share is \(1-w_i\).
For example, if an accident year has premium of $1,000,000 and 60% of its ultimate losses are expected to have emerged, its emergence-weighted premium is $600,000. It contributes its observed losses to the numerator and $600,000 to the denominator. It does not contribute $400,000, which is the premium multiplied by the unreported share.
How the Common Ratio Uses the Data
Cape Cod estimates one ratio across the included accident years, then applies that ratio to each year’s premium. The same ratio is used because the method assumes, for this calculation, that the years share a common underlying expected loss ratio after the selected premium and loss adjustments. Different emergence percentages account for different maturity levels.
An individual year’s observed loss ratio, \(L_i^{obs}/P_i\), is not itself an ultimate loss ratio: it reflects both the underlying loss ratio and the proportion of loss expected to have emerged. Dividing its observed losses by \(P_iw_i\) gives an emergence-adjusted indication for that year, but Cape Cod combines information across years by using total losses divided by total weighted premium. This is equivalent to weighting those year-specific indications by \(P_iw_i\), rather than giving every accident year equal weight.
This identity is useful for interpretation when every included year has \(P_iw_i>0\): the overall ratio is a weighted average of the emergence-adjusted year ratios. The weights are the emergence-weighted premiums, not equal weights and not the premiums alone. If \(w_i=0\), that year has no emergence-weighted exposure and cannot form an individual adjusted ratio; it contributes no denominator exposure.
Why the Formula Is Exposure-Weighted
The central justification is that the observed loss amount from a year is expected to reflect both the premium exposure and the fraction of ultimate losses that has emerged. Suppose the expected ultimate loss for year \(i\) is \(P_i r\), where \(r\) is the common expected loss ratio. If the selected emergence proportion is \(w_i\), then the expected observed loss is \(P_i r w_i\).
Assume the expected ultimate loss for accident year \(i\) is \(P_i r\), and the expected observed loss at the valuation date is \(P_i r w_i\). Treat the selected premiums and emergence proportions as fixed for this calculation.
Summing across years gives \(\sum_i E[L_i^{obs}]=r\sum_i P_iw_i\).
Estimating the common ratio by equating the observed aggregate loss with its expected value gives \(\hat r_{\mathrm{CC}}=\sum_i L_i^{obs}/\sum_i P_iw_i\).
This justification explains the role of each term: observed losses estimate the emerged amount, while \(\sum_i P_iw_i\) measures how much premium exposure is represented in those emerged losses. It relies on the common-ratio assumption and suitable emergence selections. If expected loss ratios differ materially by year, or if the premiums and losses are not made comparable, the resulting common ratio may not be a suitable estimate.
The method does not require each accident year’s observed losses to equal its expected losses. Actual claims fluctuate. The aggregate ratio is an estimate, not a guarantee that each year’s projected ultimate will match its eventual result.
Worked Example: Combining Two Accident Years
A hypothetical insurer has the following data, all amounts in thousands of dollars. The premiums are on the selected basis, the observed losses use one consistent loss basis, and the selected emergence proportions are 50% and 80%. Estimate the Cape Cod expected loss ratio and the expected ultimate losses assigned to each year.
| Accident year | Premium \(P_i\) | Emergence \(w_i\) | Observed losses \(L_i^{obs}\) |
|---|---|---|---|
| 2023 | 2,000 | 0.50 | 720 |
| 2024 | 3,000 | 0.80 | 1,200 |
Step 1: Calculate weighted premium exposure. For 2023, \(2{,}000(0.50)=1{,}000\). For 2024, \(3{,}000(0.80)=2{,}400\). Total emergence-weighted premium is \(1{,}000+2{,}400=3{,}400\).
Step 2: Add observed losses. Total observed losses are \(720+1{,}200=1{,}920\).
Step 3: Estimate the common ratio. $$ \hat r_{\mathrm{CC}}=\frac{1{,}920}{3{,}400}=0.5647059\ldots=56.4706\%. $$
Step 4: Apply the ratio to each year’s premium. For 2023, \(2{,}000(0.5647059)=1{,}129.412\). For 2024, \(3{,}000(0.5647059)=1{,}694.118\). These are expected ultimate losses in thousands of dollars, rounded to three decimal places.
Check. Applying the selected emergence proportions to those expected ultimate losses gives \(1{,}129.412(0.50)=564.706\) and \(1{,}694.118(0.80)=1{,}355.294\). Their sum is \(1{,}920.000\), matching the observed total, apart from rounding. This is the exposure-weighted principle in action.
Method and trap avoided: This uses the aggregate Cape Cod ratio. Averaging the two years’ observed loss ratios, \(720/2{,}000=36\%\) and \(1{,}200/3{,}000=40\%\), would ignore their different emergence levels and would not estimate the common ultimate ratio.
Comparing the Method With an Unweighted Average
An unweighted average gives a small accident year the same influence as a large one. It also fails to account for the fact that a less mature year has had less opportunity for its losses to emerge. Cape Cod accounts for both effects through \(P_iw_i\). The table below summarizes the different quantities.
| Quantity | Calculation | What it represents |
|---|---|---|
| Observed loss ratio | \(L_i^{obs}/P_i\) | Observed losses relative to full premium; affected by emergence. |
| Year-specific adjusted indication | \(L_i^{obs}/(P_iw_i)\), if \(w_i>0\) | An emergence-adjusted indication for one year. |
| Cape Cod common ratio | \(\sum_iL_i^{obs}/\sum_iP_iw_i\) | Aggregate indication weighted by each year’s emergence-weighted premium. |
In practice, premiums must represent comparable exposure and be on the basis appropriate for the loss estimate. For example, if the selected loss estimate is adjusted for a change in claim cost level, a premium basis that does not reflect the corresponding intended comparison can distort the ratio. Likewise, \(w_i\) must be chosen for the loss measure being analyzed: a paid emergence pattern and an incurred emergence pattern need not be interchangeable. These are inputs to the calculation, not quantities inferred from the Cape Cod ratio itself.
Worked Example: Different Maturities and Unequal Premiums
A hypothetical portfolio has three accident years. Premiums and losses are in thousands of dollars. The selected emergence proportions are 20%, 50%, and 100%. Calculate the Cape Cod ratio, the expected ultimate loss for each year, and the unweighted average of the three year-specific adjusted indications to see why that average is different.
| Accident year | \(P_i\) | \(w_i\) | \(L_i^{obs}\) |
|---|---|---|---|
| Year A | 1,000 | 0.20 | 100 |
| Year B | 500 | 0.50 | 160 |
| Year C | 2,000 | 1.00 | 1,300 |
Step 1: Calculate each weighted premium. Year A contributes \(1{,}000(0.20)=200\); Year B contributes \(500(0.50)=250\); Year C contributes \(2{,}000(1.00)=2{,}000\). The total is \(200+250+2{,}000=2{,}450\).
Step 2: Calculate the aggregate ratio. Total observed losses are \(100+160+1{,}300=1{,}560\), so $$ \hat r_{\mathrm{CC}}=\frac{1{,}560}{2{,}450}=0.6367347\ldots=63.6735\%. $$
Step 3: Apply the common ratio to each premium. The expected ultimate losses are \(1{,}000(0.6367347)=636.735\), \(500(0.6367347)=318.367\), and \(2{,}000(0.6367347)=1{,}273.469\), all in thousands of dollars and rounded to three decimal places.
Step 4: Compare with an unweighted average. The individual adjusted indications are \(100/200=50\%\), \(160/250=64\%\), and \(1{,}300/2{,}000=65\%\). Their unweighted average is \((50\%+64\%+65\%)/3=59.6667\%\). It differs from the Cape Cod ratio because it gives each year equal weight. Cape Cod weights those indications by 200, 250, and 2,000.
Check. The three expected ultimate losses sum to \(2{,}228.571\) thousand, which equals \(2{,}450(0.6367347)\) to rounding. Multiplying each by its emergence proportion gives \(127.347+159.184+1{,}273.469=1{,}560.000\) thousand, matching total observed losses.
Method and trap avoided: The aggregate ratio gives the large, fully emerged Year C more influence because it represents much more emerged exposure. The unweighted average would treat Year A, with only 200 thousand of weighted premium, as equally informative.
Common Calculation Traps
- Using the unreported share. The denominator uses \(P_iw_i\), not \(P_i(1-w_i)\). The former is the premium exposure associated with expected emerged losses.
- Averaging year-specific indications without weights. If the inputs are comparable and all \(w_i>0\), the equivalent weighted average uses weights \(P_iw_i\).
- Using observed loss ratio as if it were ultimate. \(L_i^{obs}/P_i\) does not account for losses that have not yet emerged.
- Mixing bases. Losses, premiums, and emergence proportions must refer to compatible definitions and adjustments. A selected emergence percentage for one loss measure should not be applied automatically to another.
- Assuming a zero-emergence year contributes information. If \(w_i=0\), the year adds no weighted premium to the denominator. With no observed losses attributable to that period, it provides no direct information about the common ratio in this formula.
Check Your Understanding
For each problem, use the stated emergence proportions and keep all dollar amounts in the units given.
- An accident year has premium of $800,000, observed losses of $120,000, and selected emergence of 25%. What is its Cape Cod indication if it is the only year included? (Answer: 60.0000%)
- Two accident years have premiums of 1,000 and 2,000 thousand, emergence proportions of 0.40 and 0.75, and observed losses of 300 and 900 thousand. Calculate the common ratio and the expected ultimate loss for each year. (Answer: 63.1579%; 631.579 thousand and 1,263.158 thousand)
- Three years have premiums of 500, 1,500, and 1,000 thousand; emergence proportions of 0.20, 0.60, and 0.90; and observed losses of 80, 600, and 720 thousand. Find the common ratio and the three expected ultimate losses. (Answer: 73.6842%; 368.421, 1,105.263, and 736.842 thousand)
- Year A has premium of 1,000 thousand, emergence of 0.50, and observed losses of 300 thousand. Year B has premium of 2,000 thousand, emergence of zero, and observed losses of zero. Calculate the common ratio and the expected ultimate losses by year. (Answer: 60.0000%; 600 thousand and 1,200 thousand)
- Two years have premiums of 2,000 and 1,000 thousand, emergence proportions of 0.80 and 0.25, and observed losses of 1,040 and 200 thousand. Calculate the Cape Cod ratio and expected ultimate losses. State the incorrect ratio obtained if the denominator mistakenly uses the unreported shares. (Answer: 67.0270%; 1,340.541 and 670.270 thousand; incorrect ratio 107.8261%)