Why Development Weights Matter in Cape Cod
A Cape Cod estimate uses the amount of loss already observed and the expected amount still to emerge. The development weight measures the first part of that relationship: it is the selected proportion of ultimate loss expected to have emerged by the valuation date. Because the weight determines how much premium contributes to the expected loss ratio denominator, even a modest change in the selected emergence pattern can change both that ratio and the resulting ultimate estimates.
A tail factor is part of the development selection when selected development factors end before ultimate. It extends the development pattern from the latest modeled age to ultimate. The tail is not itself the percentage of ultimate loss already observed: the weight is the reciprocal of the cumulative age-to-ultimate factor.
Paid weights must be applied to cumulative paid losses, and incurred weights to cumulative incurred losses. The premium and loss amounts must refer to compatible accident years and valuation dates. The selections are hypothetical actuarial assumptions; they are not universal constants.
From Development Factors to Emergence Weights
A loss development factor projects a cumulative amount at one age to a later age. If the selected age-to-age factors from the current age through the modeled horizon are \(f_1,\ldots,f_k\), and the selected tail factor from that horizon to ultimate is \(T\), then the cumulative age-to-ultimate factor is their product. The corresponding emergence weight is its reciprocal.
Equivalently, if \(q_i\) is the selected percentage emerged through the modeled horizon before adding the tail, then the tail factor projects that amount forward: ultimate is the horizon amount multiplied by \(T\). Therefore the percentage of ultimate already emerged is \(q_i/T\), not \(q_iT\).
Worked Example: Converting a Horizon Percentage and Tail
A hypothetical incurred-loss selection indicates that 84% of ultimate has emerged through the latest modeled age. A selected tail factor of 1.20 applies from that age to ultimate. Find the ultimate emergence weight. State the common mistake this calculation avoids.
Method: convert the horizon emergence percentage to an ultimate emergence percentage. Let the cumulative incurred amount at the horizon be \(C\). The selected horizon percentage means \(C=0.84U_H\), where \(U_H\) is the amount projected to the modeled horizon. The tail selection gives ultimate \(U=1.20U_H\). Thus:
The ultimate emergence weight is 70%. As a check, if the observed cumulative amount is 84 units, the implied horizon amount is \(84/0.84=100\), and ultimate is \(100(1.20)=120\). The observed share of ultimate is \(84/120=0.70\).
Common mistake avoided: Multiplying \(0.84\) by \(1.20\) gives 1.008, or 100.8% emerged, which is inconsistent with a positive tail still to develop. Because the tail increases projected ultimate, the emergence percentage must decrease.
When selected age-to-age factors and a tail are available instead of a horizon percentage, multiply all factors from the current age through ultimate and take the reciprocal. For example, factors of 1.25 and 1.10 followed by a tail of 1.05 give a cumulative age-to-ultimate factor of \(1.25(1.10)(1.05)=1.44375\), so the emergence weight is \(1/1.44375=0.6926\) to four decimals.
Cape Cod Expected Loss Ratio and Estimates
The Cape Cod expected loss ratio is calibrated so that the premium-weighted expected loss emerging by the valuation date matches the observed cumulative loss total. In this calculation, \(P_iw_i\) is the portion of premium corresponding to the selected proportion of ultimate expected to have emerged.
The expected ultimate loss for year \(i\) is \(P_i\widehat{R}_{CC}\). The estimated amount not yet emerged is that expected ultimate multiplied by \(1-w_i\). Thus the Cape Cod ultimate estimate can be written either as expected ultimate less nothing (the full expected ultimate) plus observed experience only through the calibrated formula, or directly as observed loss plus estimated unreported loss:
The first expression for ultimate is the operative Cape Cod estimate: it retains the observed cumulative amount and adds expected unreported loss. It is not generally equal to \(P_i\widehat{R}_{CC}\), because the observed amount \(C_i\) need not equal \(P_i\widehat{R}_{CC}w_i\) for each individual year. The calibration matches those quantities only in aggregate.
Why the ratio formula holds. The calibration principle assigns expected emerged loss \(P_iRw_i\) to year \(i\), where \(R\) is the common expected loss ratio for this calculation. Across years, the fitted emerged amount is \(\sum_i P_iRw_i\). Equating this to the observed total and solving for \(R\) gives:
This calculation requires a positive denominator and consistent loss, premium, and emergence bases. If a weight is wrong, its premium contribution to the denominator is wrong as well. Consequently, a change in weights can alter the expected loss ratio for every year, not merely the estimate for the year whose weight changed.
Worked Cape Cod Calculation and Tail Sensitivity
Worked Example: Measuring the Effect of a Longer Tail
A hypothetical liability portfolio has the following cumulative paid losses and aligned premiums, all in thousands of dollars. The selected percentages emerged through the latest modeled horizon are 90%, 60%, and 30% for the three accident years, respectively. A tail factor of 1.10 applies to each. Calculate the Cape Cod expected loss ratio, ultimate estimates, and unpaid estimates. Then repeat the calculation using a tail factor of 1.25.
| Accident year | Premium \(P_i\) | Paid loss \(C_i\) | Horizon percentage |
|---|---|---|---|
| Year 1 | 1,000 | 300 | 90% |
| Year 2 | 1,200 | 420 | 60% |
| Year 3 | 800 | 400 | 30% |
Method: convert horizon percentages to ultimate weights, calibrate the ratio, then estimate unreported loss. With tail 1.10, the weights are \(0.90/1.10=9/11\), \(0.60/1.10=6/11\), and \(0.30/1.10=3/11\). The observed paid total is \(300+420+400=1{,}120\). The weighted premium denominator is:
Therefore \(\widehat{R}_{CC}=1{,}120/1{,}690.9091=0.6624\), or 66.24% to two decimal places. The unreported estimates, using the unrounded ratio, are:
Adding each amount to its observed paid loss gives ultimate estimates of \(420.43\), \(781.29\), and \(785.38\), respectively. Total unpaid is \(120.43+361.29+385.38=867.10\), and total ultimate is \(1{,}120+867.10=1{,}987.10\). All monetary results are in thousands and are rounded to two decimals.
For a tail of 1.25, the weights become \(0.90/1.25=0.72\), \(0.60/1.25=0.48\), and \(0.30/1.25=0.24\). The weighted premium denominator is \(1{,}000(0.72)+1{,}200(0.48)+800(0.24)=1{,}488\). The new expected loss ratio is \(1{,}120/1{,}488=0.7526882\), or 75.27%. Using that unrounded ratio, unpaid estimates are \(1{,}000(0.7526882)(0.28)=210.75\), \(1{,}200(0.7526882)(0.52)=469.68\), and \(800(0.7526882)(0.76)=457.63\). The corresponding ultimates are \(510.75\), \(889.68\), and \(857.63\); total unpaid is \(1{,}138.06\), and total ultimate is \(2{,}258.06\).
The longer tail lowers each weight. That reduces the weighted-premium denominator and raises the fitted expected loss ratio in this example; it also raises the estimated unreported amounts. The direction and size of the change depend on the observed losses, premiums, and which weights change, so the calculation should be repeated rather than adjusted by a simple multiplier.
Paid and Incurred Selections Must Stay on Their Own Bases
Paid and incurred development patterns can differ because payments and case estimates emerge differently. A paid Cape Cod calculation pairs cumulative paid losses with paid emergence weights. An incurred calculation pairs cumulative incurred losses with incurred emergence weights. Using paid losses with incurred weights, or combining the two loss totals in one numerator, breaks the calibration basis.
Worked Example: Comparing Paid and Incurred Weights
A hypothetical portfolio has premiums of 500, 700, and 600 thousand dollars. Its cumulative paid losses are 220, 330, and 260; cumulative incurred losses are 250, 420, and 360, all in thousands. For incurred losses, horizon emergence percentages are 95%, 75%, and 45%; for paid losses, they are 90%, 60%, and 35%. A tail factor of 1.05 applies to both selections. Calculate each basis separately and compare the resulting expected loss ratios and total ultimate estimates.
Incurred method. The incurred weights are \(0.95/1.05=19/21\), \(0.75/1.05=5/7\), and \(0.45/1.05=3/7\). The incurred total is \(250+420+360=1{,}030\). The weighted premium denominator is:
Thus \(\widehat{R}_{CC}=1{,}030/1{,}209.5238=0.8515748\), or 85.16%. The unreported estimates are \(500(0.8515748)(2/21)=40.55\), \(700(0.8515748)(2/7)=170.31\), and \(600(0.8515748)(4/7)=291.97\). Adding the incurred amounts gives ultimates of 290.55, 590.31, and 651.97. Total unpaid is 502.83 and total ultimate is \(1{,}030+502.83=1{,}532.83\).
Paid method. The paid weights are \(0.90/1.05=6/7\), \(0.60/1.05=4/7\), and \(0.35/1.05=1/3\). The paid total is \(220+330+260=810\), and the weighted premium denominator is \(500(6/7)+700(4/7)+600(1/3)=1{,}028.5714\). Thus the paid expected loss ratio is \(810/1{,}028.5714=0.7875\), or 78.75%. Unreported estimates are \(500(0.7875)(1/7)=56.25\), \(700(0.7875)(3/7)=236.25\), and \(600(0.7875)(2/3)=315.00\). Adding paid losses gives ultimates of 276.25, 566.25, and 575.00; total unpaid is 607.50 and total ultimate is 1,417.50.
The results are internally checked: each ultimate equals the corresponding observed loss plus its unreported estimate, and the separate totals equal observed totals plus unpaid totals. The two ratios and ultimate totals are not competing calculations on identical inputs; each is calibrated to a different loss basis and its selected weights.
Selection Checks and Common Traps
A selected weight is a judgment about the proportion of ultimate loss expected to have emerged. Review its source and confirm that its development factors, tail, and loss basis fit the data being estimated. The calculation is sensitive to selections, but it does not establish that a chosen development pattern is appropriate merely because the arithmetic is correct.
- Using a development factor as a weight. An age-to-ultimate factor greater than one projects observed loss upward; its reciprocal is the emergence weight, which is at most one.
- Applying the tail in the wrong direction. A tail factor greater than one increases projected ultimate and therefore reduces the percentage emerged at the latest modeled horizon.
- Omitting the tail. A percentage through the modeled horizon is not an ultimate emergence percentage when a tail remains. Convert it before using it in \(P_iw_i\).
- Mixing paid and incurred bases. Keep each cumulative loss amount, emergence pattern, and calculation together on its own basis.
- Using a different weight in the denominator than in the reserve calculation. The same selected \(w_i\) must determine both the weighted-premium denominator and the unreported share \(1-w_i\).
Check Your Understanding
For each problem, keep the development basis and the calculation steps explicit. Monetary amounts are in thousands unless stated otherwise.
- A cumulative paid percentage through the latest modeled age is 84%, and the tail factor is 1.20. Find the ultimate emergence weight. (Answer: 70.00%)
- From the current age, selected age-to-age factors are 1.25 and 1.10, followed by a 1.10 tail factor. Find the cumulative age-to-ultimate factor and the emergence weight, rounding the weight to four decimals. (Answer: Factor 1.5125; weight 0.6612)
- Two accident years have premiums of 400 and 600, cumulative losses of 160 and 300, and emergence weights of 0.80 and 0.50. Calculate the Cape Cod expected loss ratio and each year's unreported estimate, rounding monetary amounts to two decimals. (Answer: Expected loss ratio 74.19%; unreported estimates 59.35 and 222.58)
- A single accident year has premium 300, observed cumulative loss 180, selected expected loss ratio 70%, and emergence weight 0.60. Calculate the unreported estimate and ultimate estimate. (Answer: Unreported estimate 84; ultimate estimate 264)
- A tail selection changes from 1.10 to 1.25 while the horizon emergence percentage is 75%. Find the weight under each tail and the change in percentage points. (Answer: Weights 68.18% and 60.00%; decrease 8.18 percentage points)