Aggregating Expected Losses Across Segments

Why Aggregation Starts With Expected Loss Dollars

A portfolio may contain segments with different premiums, exposure volumes, and expected loss rates. The portfolio estimate must reflect both each segment’s expected loss level and its size. A segment with a higher expected loss ratio does not necessarily contribute more expected loss than a much larger segment with a lower ratio.

The reliable starting point is to calculate the expected loss for each segment, then add those dollar estimates. Once the dollar total is established, divide by the portfolio premium or exposure total that matches the measure being reported. This order avoids treating a small segment’s rate as though it had the same weight as a large segment’s rate.

Definition: For segments $j=1,\ldots,n$, let $L_j$ be the future loss amount for segment $j$, and let $\widehat{E[L_j]}$ be its selected expected loss estimate. The aggregate expected loss estimate is

\widehat{E[L_{\mathrm{portfolio}}]}=\sum_{j=1}^{n}\widehat{E[L_j]}.

If the estimate is based on expected loss ratios $r_j$ and premiums $P_j$, then $\widehat{E[L_j]}=P_jr_j$. If it is based on expected loss costs $c_j$ and exposure units $e_j$, then $\widehat{E[L_j]}=e_jc_j$.

Here, an expected loss ratio is expected losses divided by premium on a consistent basis. An expected loss cost is expected losses per exposure unit, such as dollars per vehicle-year. The segment definitions, loss scope, valuation basis, and premium or exposure basis must be compatible before estimates are added. For example, estimates that include loss adjustment expense should not be added to estimates that exclude it without first reconciling the scope.

Aggregate Ratios and Loss Costs Are Weighted Averages

If all segment estimates use expected loss ratios, the aggregate estimate is $\sum_j P_jr_j$. The corresponding aggregate expected loss ratio divides that sum by total premium. Rearranging shows that the portfolio ratio is a premium-weighted average of the segment ratios.

r_{\mathrm{portfolio}} = \frac{\sum_{j=1}^{n}P_jr_j}{\sum_{j=1}^{n}P_j} = \sum_{j=1}^{n} \left(\frac{P_j}{\sum_{k=1}^{n}P_k}\right)r_j.

The weight for segment $j$ is its share of total premium. The weights are nonnegative and sum to one, so the aggregate ratio lies between the smallest and largest segment ratios when all premiums are positive. The same reasoning applies to loss costs, except that exposure units—not premiums—are the weights.

c_{\mathrm{portfolio}} = \frac{\sum_{j=1}^{n}e_jc_j}{\sum_{j=1}^{n}e_j} = \sum_{j=1}^{n} \left(\frac{e_j}{\sum_{k=1}^{n}e_k}\right)c_j.

A simple average gives every segment equal weight. It is appropriate only when the relevant bases are equal across segments, or when the intended statistic is specifically the unweighted mean of segment rates. It is generally not the portfolio rate implied by the segment expected losses.

Why the Summation Formula Holds

The aggregation identity follows from linearity of expectation. The total portfolio loss is the sum of the segment losses when the segments partition the portfolio without overlap. For a finite sum of random variables with finite expectations, expectation distributes over addition.

Express total loss.
Because each covered loss belongs to one segment, $L_{\mathrm{portfolio}}=L_1+\cdots+L_n$.

Take expectations.
By linearity, $E[L_{\mathrm{portfolio}}]=E[L_1+\cdots+L_n]=\sum_{j=1}^{n}E[L_j]$.

Apply the selected segment estimates.
Replacing each unknown segment expectation with its selected estimate gives $\widehat{E[L_{\mathrm{portfolio}}]}=\sum_{j=1}^{n}\widehat{E[L_j]}$.

This derivation does not require the segment losses to be independent. Independence can matter when estimating the variability of total losses, but it is not needed to add their expected values. The identity does require a sound partition: overlapping segments could count the same loss more than once, while omitted segments would leave losses out.

Worked Applications

Worked Example: Premium-Weighted Aggregate Expected Loss Ratio

A hypothetical insurer estimates losses for three regions using expected loss ratios on the same premium and loss basis. Region North has premium of $1,200,000 and an expected loss ratio of 62%. Region Central has premium of $800,000 and an expected loss ratio of 70%. Region South has premium of $500,000 and an expected loss ratio of 84%. Calculate the aggregate expected loss and aggregate expected loss ratio. Compare the result with the simple average of the three ratios.

Method. Calculate expected loss by region as premium multiplied by expected loss ratio, sum the dollar amounts, and divide by total premium.

North. Expected loss is $1{,}200{,}000(0.62)=\$744{,}000$.

Central. Expected loss is $800{,}000(0.70)=\$560{,}000$.

South. Expected loss is $500{,}000(0.84)=\$420{,}000$.

The aggregate expected loss is $744{,}000+560{,}000+420{,}000=\$1{,}724{,}000$. Total premium is $1{,}200{,}000+800{,}000+500{,}000=\$2{,}500{,}000$. Therefore, the aggregate expected loss ratio is $1{,}724{,}000/2{,}500{,}000=0.6896$, or 68.96%.

As a check using premium weights, the weights are $1{,}200{,}000/2{,}500{,}000=0.48$, $0.32$, and $0.20$. Thus $0.48(0.62)+0.32(0.70)+0.20(0.84)=0.2976+0.2240+0.1680=0.6896$, agreeing with the dollar calculation.

The simple average is $(62\%+70\%+84\%)/3=72\%$, which is not the aggregate portfolio ratio. It gives the smallest-premium region the same weight as the largest. The common mistake is to average segment ratios without considering premium volume; the premium-weighted calculation avoids it.

Worked Example: Exposure-Weighted Aggregate Expected Loss Cost

A hypothetical equipment insurer estimates expected losses for three equipment classes. Class A has 1,800 insured-equipment years and an expected loss cost of $410 per equipment-year. Class B has 900 equipment-years at $560 per equipment-year. Class C has 300 equipment-years at $730 per equipment-year. Find the aggregate expected loss and portfolio expected loss cost.

Method. Multiply each class’s exposure units by its expected loss cost, add the expected losses, and divide by the total number of equipment-years.

Class A expected loss is $1{,}800(410)=\$738{,}000$. Class B expected loss is $900(560)=\$504{,}000$. Class C expected loss is $300(730)=\$219{,}000$.

Total expected loss is $738{,}000+504{,}000+219{,}000=\$1{,}461{,}000$. Total exposure is $1{,}800+900+300=3{,}000$ equipment-years. The aggregate loss cost is $1{,}461{,}000/3{,}000=\mathbf{\$487}$ per equipment-year.

The exposure weights are $1{,}800/3{,}000=0.60$, $0.30$, and $0.10$. The weighted calculation gives $0.60(410)+0.30(560)+0.10(730)=246+168+73=\$487$. By contrast, the simple mean is $(410+560+730)/3=\$566.67$ per equipment-year. It overstates the portfolio loss cost because it gives the smallest class as much influence as the largest. The expected loss total and the weighted result provide an arithmetic check.

Worked Example: Combining Segment Estimates Built on Different Bases

A hypothetical insurer estimates expected loss for two segments from expected loss ratios and a third from an expected loss cost. All three estimates have the same loss scope and valuation basis. Segment A has $900,000 of premium and a 58% expected loss ratio. Segment B has $600,000 of premium and a 73% expected loss ratio. Segment C has 2,000 exposure units and an expected loss cost of $315 per unit; its premium on the common premium basis is $1,000,000. Calculate total expected loss and the aggregate expected loss ratio.

Method. Convert each segment estimate to expected loss dollars. Since the loss scope is consistent, add those dollar estimates. Then divide by the total premium for all three segments.

Segment A expected loss is $900{,}000(0.58)=\$522{,}000$. Segment B expected loss is $600{,}000(0.73)=\$438{,}000$. Segment C expected loss is $2{,}000(315)=\$630{,}000$.

Total expected loss is $522{,}000+438{,}000+630{,}000=\$1{,}590{,}000$. Total premium is $900{,}000+600{,}000+1{,}000{,}000=\$2{,}500{,}000$. The aggregate expected loss ratio is $1{,}590{,}000/2{,}500{,}000=0.636$, or 63.6%.

As a cross-check, Segment C’s implied expected loss ratio is $630{,}000/1{,}000{,}000=63\%$. The premium weights are $0.36$, $0.24$, and $0.40$; the weighted ratio is $0.36(0.58)+0.24(0.73)+0.40(0.63)=0.2088+0.1752+0.2520=0.636$. The simple average of 58%, 73%, and 63% is 64.67%, not the aggregate ratio. The key condition is that the segment dollar estimates and total premium refer to compatible loss and premium bases; a simple average cannot repair a basis mismatch.

Checks and Common Traps

Aggregation is often a short calculation, but several conceptual checks determine whether the result is meaningful. First, confirm that segments are mutually exclusive and collectively cover the portfolio being measured. Second, confirm that each segment estimate represents the same type of loss at a comparable valuation point. Third, use the denominator that matches the requested rate: total premium for an expected loss ratio, or total exposure units for an expected loss cost.

A weighted aggregate must also be consistent with its components. With positive premium weights, an aggregate expected loss ratio should lie between the smallest and largest segment ratios. With positive exposure weights, the corresponding statement holds for expected loss costs. If a result falls outside that range, check for arithmetic errors, inconsistent units, or incompatible data bases.

Common exam trap: Adding segment ratios or costs and dividing by the number of segments treats all segments as equally large. Instead, calculate the expected loss dollars by segment, add them, and divide by total premium or total exposure. Do not average rates first and multiply by a total unless the correct weights have already been applied.

The aggregate expected loss is a sum of selected segment expectations; it is not automatically an observed loss total. Its interpretation depends on the segment assumptions and inputs used to construct it. Aggregation preserves those selections while expressing them at portfolio level; it does not make incompatible segment assumptions compatible.

Check Your Understanding

For each problem, calculate the requested portfolio quantity from the segment data rather than taking an unweighted average.

A hypothetical portfolio has Segment A premium of $400,000 at an expected loss ratio of 55%, and Segment B premium of $600,000 at 80%. Find aggregate expected loss and the aggregate expected loss ratio. (Answer: expected loss $700,000; aggregate ratio 70%)
A hypothetical policy group has 500 exposure units at an expected loss cost of $240 and 1,500 units at $360. Find total expected loss and aggregate loss cost per unit. (Answer: expected loss $660,000; aggregate loss cost $330 per unit)
Three hypothetical segments have premiums of $2,000,000, $1,000,000, and $500,000, with expected loss ratios of 50%, 75%, and 90%, respectively. Find aggregate expected loss and ratio. (Answer: expected loss $2,200,000; aggregate ratio 62.86%)
Three hypothetical segments have expected losses of $125,000, $300,000, and $210,000, and corresponding premiums of $250,000, $600,000, and $350,000. Find the portfolio expected loss ratio. (Answer: 52.92%)
A hypothetical portfolio has 4,000 low-risk exposure units at an expected loss cost of $275 and 1,000 high-risk units at $610. A revised mix has 3,000 low-risk and 2,000 high-risk units, with the same total exposure. Calculate expected loss under each mix, the change in expected loss, and each aggregate loss cost. (Answer: original expected loss $1,710,000 and cost $342; revised expected loss $2,045,000 and cost $409; increase $335,000)

What You'll Learn