Cape Cod with Exposure-Based Expected Loss Costs

From Expected Loss Ratios to Expected Loss Costs

Cape Cod estimation uses observed losses and development information to estimate an expected loss level for each accident year. The expected level can be expressed as an expected loss ratio applied to premium, or as an expected loss cost applied to exposure units. The exposure form is useful when the number of insured units is the natural measure of volume, and it makes the assumption about expected losses especially clear: comparable exposure units are assigned a common expected loss cost.

The calculation still depends on a consistent loss basis and development pattern. For each accident year, use the cumulative loss amount observed at the valuation date and the selected proportion of ultimate losses expected to be observed by that date. The selected proportion must match the triangle basis: for example, paid losses require paid development proportions, while incurred losses require incurred development proportions.

Definition: Let $E_i$ be the exposure units for accident year $i$, $L_i$ its observed cumulative losses at the valuation date, and $p_i$ the selected proportion of ultimate losses expected to be observed by that date. The exposure-based Cape Cod expected loss cost is $$ \widehat{c}=\frac{\sum_i L_i}{\sum_i E_i p_i}. $$ The estimated ultimate loss for year $i$ is $\widehat{U}_i=E_i\widehat{c}$. The estimated amount beyond observed losses is $\widehat{U}_i-L_i$; when $L_i$ is paid loss, this is an estimate of unpaid losses, and when $L_i$ is incurred loss, it is the estimated development beyond incurred losses.

An exposure unit must be defined consistently across the years being combined: it might be a vehicle-year, policy-year, payroll unit, or another measure of insured volume. If exposures differ in coverage, limit, deductible, rate level, or expected risk, a common loss cost may not be appropriate without adjustments or segmentation. Exposures should also be measured on the basis relevant to the losses being estimated, such as earned rather than written exposure when the loss experience is organized by earned coverage.

Why the Exposure-Weighted Formula Works

The estimator follows from the expected-loss model. Suppose each year has the same expected ultimate loss cost $c$ per exposure unit. Then the expected ultimate loss for year $i$ is $E_i c$. If $p_i$ is the expected fraction of ultimate loss observed by the valuation date, the expected observed loss is $p_iE_ic$. Summing over years gives the model's expected observed losses across the portfolio.

\operatorname{E}\!\left[\sum_i L_i\right] =\sum_i p_iE_ic =c\sum_i E_ip_i.

Solving this aggregate relationship for $c$ gives $\sum_i L_i/\sum_i E_ip_i$, the Cape Cod estimate. The denominator is the exposure volume adjusted for maturity: an exposure with only 40% of ultimate losses expected to be observed contributes $0.40$ exposure units to the estimate, while a fully mature exposure contributes one. This is an aggregate estimating equation, not a claim that each year's observed loss must equal its expectation.

The implied ultimate losses are then $E_i\widehat{c}$. The amount beyond observed losses can equivalently be written as $$ \widehat{U}_i-L_i =E_i\widehat{c}-L_i =E_i\widehat{c}(1-p_i) $$ when the observed losses are set equal to the model's expected observed losses for that year, $L_i=E_i\widehat{c}p_i$. In an actual data set, $L_i$ will generally not equal that expected value, so the direct estimate beyond observed losses is $E_i\widehat{c}-L_i$. The formula $E_i\widehat{c}(1-p_i)$ is the expected unpaid or unobserved portion under the model, not necessarily the difference between the selected ultimate and the actual observed amount.

Worked Example: Estimating a Common Loss Cost from Exposures

A hypothetical book has 10,000 vehicle-years in accident year 1 and 8,000 in accident year 2. At the valuation date, the selected paid development proportions are $p_1=0.80$ and $p_2=0.50$. Observed cumulative paid losses are $1,200,000 and $900,000, respectively. Estimate the common Cape Cod expected loss cost, ultimate losses, and unpaid losses by year.

Method. Use the exposure-weighted Cape Cod estimator. The paid development proportions are used because the observations are paid losses.

First calculate the maturity-adjusted exposure denominator:

E_1p_1=10{,}000(0.80)=8{,}000,\qquad E_2p_2=8{,}000(0.50)=4{,}000,

so the total is $8{,}000+4{,}000=12{,}000$ adjusted exposure units. Total observed paid losses are $1{,}200{,}000+900{,}000=$2{,}100{,}000$. Therefore,

\widehat{c}=\frac{\$2{,}100{,}000}{12{,}000} =\$175\text{ per vehicle-year}.

The estimated ultimate loss for year 1 is $10{,}000(\$175)=$1{,}750{,}000$; its unpaid estimate is $$1{,}750{,}000-$1{,}200{,}000=$550{,}000$. For year 2, estimated ultimate losses are $8{,}000(\$175)=$1{,}400{,}000$, and unpaid losses are $$1{,}400{,}000-$900{,}000=$500{,}000$.

Check. Total estimated ultimate losses are $$1{,}750{,}000+$1{,}400{,}000=$3{,}150{,}000$. Total unpaid losses are $$550{,}000+$500{,}000=$1{,}050{,}000$, and observed paid losses plus unpaid losses equal $$2{,}100{,}000+$1{,}050{,}000=$3{,}150{,}000$. A common mistake is to divide observed losses by the unadjusted 18,000 exposure units; that would ignore that the years are at different levels of maturity.

Relationship to the Premium-Based Cape Cod Loss Ratio

In a premium formulation, let $P_i$ denote the premium corresponding to accident year $i$, measured consistently with the exposures and adjusted to an appropriate rate level. The premium-based Cape Cod expected loss ratio is

\widehat{\lambda} =\frac{\sum_i L_i}{\sum_i P_i p_i}.

Here $\widehat{\lambda}$ is an expected loss ratio, while $\widehat{c}$ is an expected loss cost per exposure unit. If each year has the same premium per exposure unit $r$, so that $P_i=rE_i$, then the two estimates are directly convertible:

\sum_i P_ip_i=r\sum_i E_ip_i,\qquad \widehat{\lambda}=\frac{\widehat{c}}{r},\qquad \widehat{c}=r\widehat{\lambda}.

This equivalence also carries through to each accident year's expected ultimate loss: $P_i\widehat{\lambda}=rE_i(\widehat{c}/r)=E_i\widehat{c}$. The underlying loss estimate is unchanged; only its unit of expression differs.

The condition that premium per exposure is common matters. When $P_i/E_i$ varies by year, a common loss cost and a common loss ratio are different assumptions. The exposure estimator weights maturity-adjusted exposure units, whereas the premium estimator weights them by premium per exposure as well. They need not produce the same expected loss for each year. Differences may reflect changing rates, coverage mix, or other premium variation; they should not be removed by silently substituting one denominator for the other.

Worked Example: Converting Loss Cost to an Expected Loss Ratio

A hypothetical liability portfolio has 4,000 exposure units in year 1 and 5,000 in year 2. Both years have premium of $650 per exposure unit. Selected incurred development proportions are $p_1=0.75$ and $p_2=0.40$, and observed incurred losses are $540,000 and $460,000. Calculate the exposure-based loss cost, the premium-based expected loss ratio, and the resulting ultimate losses and amounts beyond incurred losses.

Method. Calculate both maturity-adjusted denominators, then verify the conversion using the common premium of $650 per exposure unit.

Adjusted exposure is $4{,}000(0.75)=3{,}000$ units for year 1 and $5{,}000(0.40)=2{,}000$ units for year 2, or 5,000 units total. Observed losses total $$540{,}000+$460{,}000=$1{,}000{,}000$, so

\widehat{c} =\frac{\$1{,}000{,}000}{5{,}000} =\$200\text{ per exposure unit}.

Total premium is $4{,}000(\$650)=$2{,}600{,}000$ in year 1 and $5{,}000(\$650)=$3{,}250{,}000$ in year 2. The maturity-adjusted premiums are $$2{,}600{,}000(0.75)=$1{,}950{,}000$ and $$3{,}250{,}000(0.40)=$1{,}300{,}000$. Thus the denominator is $3,250,000 and

\widehat{\lambda} =\frac{\$1{,}000{,}000}{\$3{,}250{,}000} =\frac{4}{13} \approx 0.3077.

The conversion check is $$650(4/13)=$200$ per exposure unit. Estimated ultimate losses are $4{,}000(\$200)=$800{,}000$ for year 1 and $5{,}000(\$200)=$1{,}000{,}000$ for year 2. Amounts beyond incurred losses are $$800{,}000-$540{,}000=$260{,}000$ and $$1{,}000{,}000-$460{,}000=$540{,}000$.

Check. Premium-based estimates give $$2{,}600{,}000(4/13)=$800{,}000$ and $$3{,}250{,}000(4/13)=$1{,}000{,}000$, matching the exposure estimates. Using the exact fraction $4/13$ verifies the equality; multiplying by the displayed rounded ratio gives an approximate result.

Worked Example: When Premium per Exposure Differs

Consider two hypothetical accident years. Year 1 has 6,000 exposure units, $400 premium per unit, $p_1=0.75$, and $900,000 of observed losses. Year 2 has 4,000 exposure units, $600 premium per unit, $p_2=0.40$, and $625,000 of observed losses. Compare the exposure-based expected loss cost with the premium-based expected loss ratio, and show how the implied ultimate estimates differ by year.

Method. Compute each denominator independently. Since premium per exposure differs, do not assume that the two methods will give identical year-by-year ultimates.

The adjusted exposures are $6{,}000(0.75)=4{,}500$ and $4{,}000(0.40)=1{,}600$, totaling 6,100. Observed losses total $1,525,000, so the exposure-based loss cost is

\widehat{c} =\frac{\$1{,}525{,}000}{6{,}100} =\$250\text{ per exposure unit}.

Premiums are $6{,}000(\$400)=$2{,}400{,}000$ and $4{,}000(\$600)=$2{,}400{,}000$. Their maturity-adjusted values are $1,800,000 and $960,000, totaling $2,760,000. The premium-based ratio is

\widehat{\lambda} =\frac{\$1{,}525{,}000}{\$2{,}760{,}000} \approx 0.5525.

The exposure method estimates ultimate losses of $6{,}000(\$250)=$1{,}500{,}000$ and $4{,}000(\$250)=$1{,}000{,}000$. The premium method estimates approximately $0.552536( $2{,}400{,}000)=$1{,}326{,}087$ in each year. The rounded ratio is used here for display; using the exact ratio $1{,}525/2{,}760$ gives the same values to the nearest dollar.

Check. The premium-based aggregate ultimate is approximately $2,652,174, while the exposure-based aggregate ultimate is $2,500,000. The methods differ because their total maturity-adjusted denominators differ: 6,100 exposure units versus $2,760,000 premium. A common expected loss cost assigns the same expected loss per exposure to both years; a common expected loss ratio assigns the same expected loss per premium dollar. Neither assumption implies the other when premium per exposure varies. The trap is to convert the exposure estimate using one arbitrary premium rate and then claim year-by-year equivalence.

Practical Checks and Interpretation

Exposure-based presentation can make the selected expected level easier to interpret: it states the expected loss in dollars per insured unit. It also makes the role of the exposure base visible. Before relying on a result, check that the numerator and denominator correspond to the same populations, periods, and coverage basis, and that any adjustment for rate level, trend, or exposure mix has been applied consistently.

Confirm the loss basis.
Use observed paid or incurred losses consistently, with development proportions selected on that same basis.

Build the maturity-adjusted denominator.
For exposure estimation, sum $E_ip_i$. For premium estimation, sum $P_ip_i$. Do not use unadjusted exposure or premium.

Apply the selected common level.
Multiply each year's exposures by the estimated loss cost, or its premium by the estimated loss ratio, according to the model being used.

Check the implied assumption.
A common loss cost requires comparable exposure units; a common loss ratio requires comparable loss-to-premium relationships. If premiums per exposure vary, compare the assumptions before treating the results as interchangeable.

For a paid-loss estimate, ultimate less paid is an estimate of unpaid losses. For an incurred-loss estimate, ultimate less incurred is the estimated development beyond current incurred losses; it should not be described as total unpaid claim payments without further information. In either case, the Cape Cod calculation relies on selected development proportions and on the stability of the expected loss level across the years being pooled.

Key takeaway: Exposure-based Cape Cod estimates a common expected loss cost as observed losses divided by maturity-adjusted exposures. It is equivalent to a premium-based expected loss ratio calculation when premium is proportional to exposure at the same rate per unit; varying premium per exposure generally changes the weighting and the year-by-year estimates.

Check Your Understanding

For each problem, use the stated development proportions and keep the loss and exposure bases consistent.

A hypothetical property portfolio has 5,000 exposure units with $p_1=0.60$ and $450,000 of observed losses, and 3,000 units with $p_2=0.30$ and $240,000 of observed losses. Find the common expected loss cost and total estimated ultimate losses. (Answer: $176.92 per exposure unit; total ultimate losses approximately $1,415,384.62.)
A hypothetical book has 4,000 exposure units at $p_1=0.80$ with $720,000 observed losses, and 6,000 units at $p_2=0.50$ with $396,000 observed losses. Premium is $400 per exposure unit in both years. Find the exposure-based loss cost and premium-based expected loss ratio. (Answer: $180 per exposure unit and 0.4500.)
For the same two years in question 2, calculate the estimated ultimate loss and the amount beyond observed losses for each year using the exposure method. (Answer: Year 1 ultimate $720,000 and amount beyond observed losses $0; year 2 ultimate $1,080,000 and amount beyond observed losses $684,000.)
A hypothetical book has 2,000 units at $p_1=0.70$ with $315,000 observed losses and 1,500 units at $p_2=0.40$ with $135,000 observed losses. Premium per unit is $500 in the first year and $800 in the second. Calculate the exposure-based loss cost and the premium-based expected loss ratio. (Answer: $225 per unit and approximately 0.3814.)
Using question 4, what is the exposure-weighted average premium per unit, calculated as total maturity-adjusted premium divided by total maturity-adjusted exposure? Verify that multiplying this rate by the premium-based ratio recovers the exposure-based loss cost. (Answer: $590 per unit; $0.3813559(\$590)\approx \$225$.)

What You'll Learn