Premium Selection in Cape Cod Estimation

Why Premium Selection Matters in Cape Cod

Cape Cod estimation uses earned premium to represent the amount of business exposed to loss. But a premium figure is useful only if it represents the same business and period as the losses being estimated, and if its rate level is consistent across periods. Selecting the wrong premium basis changes the estimate, even when the loss data and development factors are unchanged.

The key feature of Cape Cod is that premium is paired with a development weight. A less mature accident period contributes less of its premium to the denominator because a smaller share of its ultimate losses is expected to have emerged by the valuation date. Selecting premiums therefore involves more than finding a convenient total: the premium for each period must match its losses and its weight.

Definition: For accident period $i$, let $P_i$ be earned premium on the selected, consistent rate level; let $L_i$ be the cumulative loss measure at the valuation date; and let $w_i$ be the proportion of ultimate losses expected to have emerged by that date. The Cape Cod expected loss ratio is $$ \widehat{R}_{CC}=\frac{\sum_i L_i}{\sum_i P_i w_i}. $$

The loss measure and development weights must use a consistent basis. For example, paid losses require weights representing the proportion of ultimate paid losses expected to have been paid; incurred losses require weights for the incurred measure being used. Premiums must correspond to the same accident periods, covered business, and coverage scope as those losses. The selected ratio is then applied to premium to estimate ultimate losses.

What Makes a Premium Selection Suitable?

For Cape Cod, the selected $P_i$ is generally earned premium for the accident period, adjusted to a common rate level when needed. Written premium is not automatically a substitute: it relates to policies written during a period, whereas earned premium represents coverage provided during a period. A policy-year or calendar-year premium total can include business that does not match an accident-year loss total. Use a different premium basis only when it has been converted or demonstrated to correspond to the losses being analyzed.

Rate-level adjustment puts premiums from different historical rate levels on a common basis. When an on-level factor $F_i$ is supplied for period $i$, the calculation is

P_i^{\text{on-level}}=P_i^{\text{earned}}F_i.

The factor should reflect the rate changes applicable to the earned exposure, not simply the most recent filed rate change. If a period's earned business spans multiple rate levels, its factor should reflect the relevant mix of exposure or premium at those levels. The exact calculation depends on the information supplied and the on-level method used. Do not treat an arbitrary average of rate changes as an on-level factor.

Rate level is not the same as loss trend. On-leveling adjusts premium to a common price level; loss trending adjusts losses to a common cost level when that is part of the analysis. Do not change the premium base to compensate for a loss trend adjustment. The premiums and losses should each be on the basis required by the estimation method.

Selection check	Question to ask	Why it matters
Period	Does this earned premium correspond to the accident period of the losses?	A period mismatch pairs different business in the numerator and denominator.
Rate level	Have historical premiums been adjusted to the selected common rate level?	Unadjusted changes in price can look like changes in expected loss cost.
Business scope	Do premiums and losses cover the same product, territory, limits, and other relevant segments?	Different mixes can have different expected loss ratios.
Development basis	Does each $w_i$ correspond to the loss measure and valuation age of $L_i$?	The denominator gives premium credit in proportion to expected loss emergence.

Why the Cape Cod Denominator Uses Weighted Premium

The weighted denominator follows from the model rather than from a simple total-premium average. Suppose each period has a common expected ultimate loss ratio $R$, its selected earned premium is $P_i$, and its expected ultimate loss is $RP_i$. If the development weight $w_i$ is the expected proportion of ultimate losses emerged by the valuation date, expected observed losses for period $i$ are $RP_iw_i$.

The estimate chooses $R$ so that the sum of expected observed losses agrees with the observed losses across periods. Equating those quantities and solving gives the Cape Cod ratio:

\sum_i L_i=R\sum_i P_iw_i \qquad\Longrightarrow\qquad \widehat{R}_{CC}=\frac{\sum_i L_i}{\sum_i P_iw_i}.

This argument assumes that the selected ratio is applicable across the periods being pooled, that premiums represent the relevant exposures at a consistent level, and that the weights correctly represent development of the loss measure. If the periods contain materially different business or the weights are inappropriate, a precise calculation does not make the result appropriate. Premium selection cannot repair a mismatch in the underlying data.

Worked Example: On-Level Premiums Across Three Accident Years

A hypothetical insurer is estimating ultimate losses using a Cape Cod incurred-loss method. All amounts below are in thousands of dollars. The earned premiums have already been adjusted to a common rate level and correspond to the same coverage as the losses. At the valuation date, the selected development weights and incurred losses are: Year 1, $P_1=1{,}000$, $w_1=0.80$, $L_1=560$; Year 2, $P_2=1{,}200$, $w_2=0.60$, $L_2=390$; and Year 3, $P_3=900$, $w_3=0.40$, $L_3=140$. Calculate the Cape Cod expected loss ratio and each year's estimated ultimate loss.

Method. Use the ratio of total incurred losses to total earned premium weighted by the incurred development weights. The stated premiums are already on-level, so do not apply another rate adjustment.

Step 1: Calculate the weighted premium for each year.

P_1w_1=1{,}000(0.80)=800,\quad P_2w_2=1{,}200(0.60)=720,\quad P_3w_3=900(0.40)=360.

The denominator is $800+720+360=1{,}880$ thousand. Total incurred losses are $560+390+140=1{,}090$ thousand.

Step 2: Estimate the ratio.

\widehat{R}_{CC}=\frac{1{,}090}{1{,}880}=0.579787\ldots

The selected expected loss ratio is $57.9787\ldots\%$, or 58.0% rounded to one decimal place.

Step 3: Estimate ultimate losses. For each year, add the observed incurred loss to the expected loss emergence still remaining, $\widehat{R}_{CC}P_i(1-w_i)$.

\begin{aligned} \widehat{U}_1&=560+0.579787(1{,}000)(0.20)=675.957,\\ \widehat{U}_2&=390+0.579787(1{,}200)(0.40)=668.298,\\ \widehat{U}_3&=140+0.579787(900)(0.60)=453.085. \end{aligned}

The estimated ultimate losses are approximately $676.0$, $668.3$, and $453.1$ thousand. The check is that their sum, $1{,}797.34$ thousand, equals $0.579787(1{,}000+1{,}200+900)=1{,}797.34$ thousand, subject to displayed rounding.

Common mistake avoided: using total unweighted premium of $3{,}100$ in the ratio denominator. That would give $1{,}090/3{,}100=35.16\%$, which does not allow for the different proportions of ultimate losses expected to have emerged.

Rate-Level Adjustment and a Mismatched Premium

A common premium-selection error is to combine historical earned premium at its actual rate level with losses from the same period, but then treat all periods as though premiums were at one common level. The numerical effect can be substantial. The following example isolates that effect: periods and coverage match, but one period's premium needs an on-level adjustment.

Worked Example: Correcting a Historical Rate Level

A hypothetical book has two accident years. Year 1 earned premium is $800$ thousand at a rate index of $0.80$, and Year 2 earned premium is $1{,}200$ thousand at the selected current index of $1.00$. Assume the Year 1 premium is fully earned at the stated index. The same incurred-loss segment has losses of $300$ and $500$ thousand, with incurred development weights $0.50$ and $0.80$. Calculate the ratio using premiums on the current level, and compare it with the result from using unadjusted premium.

Step 1: On-level Year 1 premium. Convert the Year 1 premium from index $0.80$ to index $1.00$:

F_1=\frac{1.00}{0.80}=1.25,\qquad P_1^{\text{on-level}}=800(1.25)=1{,}000.

Year 2 is already at the selected level, so $P_2^{\text{on-level}}=1{,}200$.

Step 2: Calculate the matched denominator and ratio.

\sum_i P_iw_i=1{,}000(0.50)+1{,}200(0.80)=500+960=1{,}460,

Total losses are $300+500=800$ thousand. Therefore,

\widehat{R}_{CC}=\frac{800}{1{,}460}=0.547945\ldots,

or 54.8% rounded to one decimal place.

Step 3: Quantify the error from using raw premium. The unadjusted denominator would be $800(0.50)+1{,}200(0.80)=400+960=1{,}360$ thousand. The resulting ratio would be $800/1{,}360=0.588235\ldots$, or $58.8\%$. It is higher because Year 1's historical premium is lower than the selected level; its weighted premium is understated.

The incorrect result does not indicate a different loss experience. It comes from mixing rate levels in the premium denominator. The common mistake is to assume that matching the accident-year labels is sufficient; the premium must also be on the intended rate level.

Period Alignment and Experience Consistency

A premium selection also needs to be checked against the numerator and the development weights, not just against a rate-level index. First confirm that the losses and premiums describe the same accident periods. Then confirm that both represent the same covered business. For example, including premium from a separate product while retaining losses from only one product will inflate the denominator and lower the indicated ratio. Conversely, omitting premium for business whose losses remain in the numerator can raise it.

Next, check that each weight is appropriate for the loss measure and period age. A weight for paid losses should not be paired with incurred losses merely because both are available. Nor should an accident year be assigned a weight for a different valuation age. The weight determines how much of that period's premium contributes to the estimate of losses emerged so far.

Finally, examine whether the combined experience is reasonably consistent with the assumption of a common expected loss ratio. This does not mean every period's observed loss ratio must be identical. Loss experience varies, and less mature periods have less emerged experience. But a striking difference in business mix, premium basis, or period results may warrant checking the data and selections before pooling the periods. Do not force agreement by changing a valid premium merely to obtain a preferred ratio.

Worked Example: A Mixed Rate-Level Premium and a Consistency Check

A hypothetical insurer estimates losses for two accident years in the same coverage segment. Year 1 earned premium is $1{,}000$ thousand: 60% of that premium was earned at rate index $0.80$, and 40% at index $1.00$. The selected common index is $1.10$. Year 2 earned premium is $900$ thousand and is already at index $1.00$; its on-level factor to the selected level is supplied as $1.10$. Incurred losses are $300$ and $400$ thousand, and the matching incurred development weights are $0.40$ and $0.75$. Calculate the on-level Cape Cod ratio and explain one useful check on the selection.

Step 1: Calculate Year 1's factor from its earned-premium mix. A dollar of premium earned at index $0.80$ is multiplied by $1.10/0.80=1.375$; a dollar earned at index $1.00$ is multiplied by $1.10/1.00=1.10$. Weight these factors by the stated shares of earned premium:

$$ F_1=0.60(1.375)+0.40(1.10)=0.825+0.440=1.265. $$

Thus $P_1^{\text{on-level}}=1{,}000(1.265)=1{,}265$ thousand. Year 2 on-level premium is $P_2^{\text{on-level}}=900(1.10)=990$ thousand.

Step 2: Apply the appropriate weights.

P_1w_1=1{,}265(0.40)=506,\qquad P_2w_2=990(0.75)=742.5.

The weighted denominator is $506+742.5=1{,}248.5$ thousand. Total incurred losses are $300+400=700$ thousand, so

\widehat{R}_{CC}=\frac{700}{1{,}248.5}=0.560673\ldots,

or 56.1% rounded to one decimal place.

Step 3: Check the period-level information without expecting identical results. The ratios of incurred loss to weighted premium are $300/506=0.5929$ for Year 1 and $400/742.5=0.5387$ for Year 2. They differ, but the combined Cape Cod ratio is the total loss divided by total weighted premium, not the simple average of these two figures. The difference is a prompt to confirm period, rate-level, and weight selections; it is not, by itself, proof that either premium is wrong.

The method avoids two common errors: applying a single rate factor to all Year 1 business despite its mixed rate levels, and averaging the two period ratios without weighting them by their denominators. The final combined ratio is based on the supplied data and assumptions; any further selection would require a reason grounded in the business or data.

Common Traps in Premium Selection

Using premium from a different period basis. Written, policy-year, calendar-year, and accident-year amounts are not interchangeable by name alone. Confirm what exposures each amount represents.
Adjusting only some periods to a common rate level. Apply a consistent selected basis to all relevant periods, including periods already at the target level where the factor may be 1.00.
Using the wrong development weight. Match the weight to the loss measure, accident-period age, and valuation date used for the losses.
Including a different segment in premium or losses. The numerator, premium, and weights must describe a compatible portfolio.
Treating period ratios as equally weighted. Cape Cod combines losses over the sum of weighted premiums; it does not generally take a simple average of period-level ratios.

Key takeaway: A Cape Cod premium selection is defensible when each earned premium matches its accident-period losses and development weight, reflects a consistent rate level, and represents compatible business. The ratio is only as meaningful as that alignment.

Check Your Understanding

For each problem, use the stated loss measure and weights. Unless indicated otherwise, amounts are in thousands of dollars.

A single accident year has earned premium of $750$, an incurred development weight of $0.40$, and incurred losses of $180$. What is the Cape Cod expected loss ratio? (Answer: $60.00\%$)
Two years have on-level earned premiums $500$ and $800$, weights $0.60$ and $0.25$, and matching losses $210$ and $160$. Calculate the combined Cape Cod ratio. (Answer: $74.00\%$)
Year 1 has earned premium $900$ at rate index $0.90$, and Year 2 has on-level premium $600$. The common index is $1.08$. Their matching weights are $0.50$ and $0.75$, and losses are $250$ and $300$. Calculate the ratio using on-level premium. (Answer: $55.56\%$)
A target coverage segment has on-level premiums $1{,}000$ and $1{,}200$, weights $0.50$ and $0.75$, and losses $280$ and $420$. A separate coverage segment has $400$ of premium but no losses included in this analysis. What ratio should be calculated for the target segment? (Answer: $50.00\%$)
Year 1 earned premium is $1{,}000$, with 40% earned at index $0.80$ and 60% at index $1.00$; the common index is $1.10$. Year 2 has on-level premium $900$. Weights are $0.50$ and $0.80$, and losses are $350$ and $400$. Calculate the Cape Cod ratio. (Answer: $56.60\%$)

What You'll Learn