Handling Sparse Segment Expectations

Why Sparse Segments Need Careful Review

A segment with little premium or few exposure units will often have a small expected loss total. That result can be mathematically correct and still be misleading if it is interpreted as evidence that the segment has little risk. Expected loss calculations combine a volume measure with a selected loss assumption: a small total may reflect small volume, a low selected loss level, or both.

Sparse segments also make the aggregate sensitive to some inputs. A selected expected loss ratio or expected loss cost may be uncertain, and applying it to even a small segment can noticeably change that segment’s own estimate. The aggregate impact depends on the segment’s size relative to the whole portfolio. The appropriate response is to check the basis, weights, and sensitivity—not to change an input merely because the segment is small.

Definition: A sparse segment is a segment with relatively low premium volume or relatively few exposure units for the calculation at hand. The term describes the volume available for the estimate; it does not, by itself, describe the segment’s risk level or establish that a selected assumption is unreliable.

Expected Losses and Their Aggregate

For a premium-based estimate, let $P_i$ be the premium for segment $i$ on the selected premium basis, and let $r_i$ be its selected expected loss ratio on that same basis. The segment expected loss is $P_i r_i$. For an exposure-based estimate, let $E_i$ be the segment’s exposure count and $c_i$ its selected expected loss cost per exposure; the segment expected loss is $E_i c_i$. These calculations treat the selected ratios or costs as inputs. They do not, on their own, estimate how reliable those inputs are.

Key result: When segment estimates cover disjoint parts of the same portfolio and use consistent definitions and valuation bases, the aggregate expected loss is the sum of the segment expected losses. A premium-weighted aggregate expected loss ratio is the sum of expected losses divided by total premium. An exposure-weighted aggregate expected loss cost is the sum of expected losses divided by total exposure.

\widehat{L}_{\mathrm{total}}=\sum_i P_i r_i, \qquad \widehat{r}_{\mathrm{total}}=\frac{\sum_i P_i r_i}{\sum_i P_i}

\widehat{L}_{\mathrm{total}}=\sum_i E_i c_i, \qquad \widehat{c}_{\mathrm{total}}=\frac{\sum_i E_i c_i}{\sum_i E_i}

The first pair applies when the calculation is premium-based; the second applies when it is exposure-based. Do not combine the two formulas without converting the inputs to a common, appropriate basis. In either case, expected losses add across segments by linearity of expectation; independence between segment losses is not required for that addition.

Worked Examples

Worked Example: Premium-Based Estimate With a Very Small Segment

A hypothetical program has three segments. The selected expected loss ratios and premiums are: Segment A, $8,000 of premium at 62%; Segment B, $120,000 at 58%; and Segment C, $2,000 at 130%. Calculate the segment expected losses and the aggregate expected loss ratio. Identify a common incorrect aggregation. This is a deterministic calculation: no interest, mortality, probability distribution, or distribution parameters are assumed.

Method. Apply the premium-based expected loss formula to each segment, then divide the sum of expected losses by the sum of premiums.

For Segment A, the expected loss is $8{,}000(0.62)=\$4{,}960$. For Segment B, it is $120{,}000(0.58)=\$69{,}600$. For Segment C, it is $2{,}000(1.30)=\$2{,}600$.

The total expected loss is $\$4{,}960+\$69{,}600+\$2{,}600=\$77{,}160$. Total premium is $\$8{,}000+\$120{,}000+\$2{,}000=\$130{,}000$. Therefore:

\widehat{r}_{\mathrm{total}} =\frac{77{,}160}{130{,}000} =0.593538\ldots =59.35\%

Check. Applying 59.35% to $130,000 gives approximately $77,155; the small difference from $77,160 is due to rounding the ratio to two decimal places. The unrounded ratio gives exactly $77,160.

Common mistake. The unweighted average of the three ratios is $(62\%+58\%+130\%)/3=83.33\%$. Multiplying that by total premium would produce about $108,333, not $77,160. This is wrong because the segments do not have equal premium. The correct aggregate ratio weights each selected ratio by its own premium. Segment C’s high ratio warrants review as an input, but its small premium means it contributes only $2,600 to this calculation.

Worked Example: Exposure-Based Estimate and a Sensitivity Check

A hypothetical equipment policy has two segments. A small segment has 4 exposure units and a selected expected loss cost of $18,000 per unit. A larger segment has 500 units and a selected cost of $4,800 per unit. Calculate the aggregate expected loss and expected loss cost. Then calculate the effect of increasing only the small segment’s selected cost to $27,000. This is a deterministic calculation with no interest, mortality, probability distribution, or distribution parameters.

Method. Multiply exposure by expected loss cost for each segment, sum the resulting losses, and divide by total exposure for the aggregate cost.

At the original selections, the small segment’s expected loss is $4(18{,}000)=\$72{,}000$. The larger segment’s expected loss is $500(4{,}800)=\$2{,}400{,}000$. Thus total expected loss is $2,472,000, and total exposure is $4+500=504$ units.

\widehat{c}_{\mathrm{total}} =\frac{2{,}472{,}000}{504} =\$4{,}904.76\text{ per exposure}

Now increase the small segment’s selected cost to $27,000. Its new expected loss is $4(27{,}000)=\$108{,}000$, an increase of $36,000. The revised total expected loss is $\$108{,}000+\$2{,}400{,}000=\$2{,}508{,}000$. The revised aggregate cost is $2{,}508{,}000/504=\$4{,}976.19$ per exposure.

The aggregate estimate increases by $\$36{,}000/\$2{,}472{,}000=1.456\%$, while the small segment’s expected loss increases by 50%. It accounts for only $4/504=0.794\%$ of exposure, but its selected cost still affects the total. The result is a sensitivity check, not a reason by itself to select either cost.

Common mistake. Using $18,000 as the small segment’s total expected loss would omit the four exposure units. The selected cost is per unit; the correct segment estimate is exposure times cost.

Worked Example: How One Unstable Ratio Can Move the Aggregate

A hypothetical portfolio has three segments. Segment S has $12,000 of premium and a selected expected loss ratio of 150%. Segment M has $108,000 at 60%, and Segment L has $280,000 at 62%. Calculate the initial aggregate expected loss and ratio. Then replace only Segment S’s selected ratio with 75% and measure the change. The calculation is deterministic; no interest, mortality, probability distribution, or distribution parameters are assumed.

Method. Calculate the initial segment losses, then recalculate only the segment affected by the changed selection.

Initially, Segment S contributes $12{,}000(1.50)=\$18{,}000$. Segment M contributes $108{,}000(0.60)=\$64{,}800$. Segment L contributes $280{,}000(0.62)=\$173{,}600$. The total is $256,400 on total premium of $\$12{,}000+\$108{,}000+\$280{,}000=\$400{,}000$. The initial aggregate ratio is $256{,}400/400{,}000=64.10\%$.

At a 75% ratio, Segment S contributes $12{,}000(0.75)=\$9{,}000$, a decrease of $9,000. The revised total expected loss is $\$9{,}000+\$64{,}800+\$173{,}600=\$247{,}400$, and the revised aggregate ratio is $247{,}400/400{,}000=61.85\%$.

The selected ratio for Segment S falls by 75 percentage points, but the aggregate ratio falls by only $9{,}000/400{,}000=2.25$ percentage points. The segment has 3% of total premium and contributes $18,000, or about 7.02% of the initial expected loss. This illustrates why both the input’s sensitivity and its portfolio weight matter.

Check. The change in aggregate expected loss is exactly the change in Segment S’s expected loss, $-\$9{,}000$. Dividing that change by total premium gives $-\$9{,}000/\$400{,}000=-2.25\%$, matching the change from 64.10% to 61.85%.

Common mistake. Changing all segments’ ratios when only Segment S’s selection changes would overstate the effect. Hold other inputs constant when measuring the sensitivity of one assumption.

Why the Weighted Formula Holds

The aggregate formula follows directly from the segment definitions. For each premium-based segment, its expected loss is $P_i r_i$. Adding those expected losses gives $\sum_i P_i r_i$. The aggregate ratio is defined as total expected loss divided by total premium, so substitution gives the weighted formula:

\widehat{r}_{\mathrm{total}} =\frac{\widehat{L}_{\mathrm{total}}}{P_{\mathrm{total}}} =\frac{\sum_i P_i r_i}{\sum_i P_i}

If total premium is positive, this is a weighted average of the segment ratios, with weights $P_i/\sum_i P_i$. Those weights sum to one. Consequently, the aggregate ratio lies between the smallest and largest segment ratios when all premiums are nonnegative and the total is positive. The same argument applies to expected loss costs with exposure as the weight. A zero-premium or zero-exposure segment contributes no loss through these formulas; it cannot be assigned a meaningful weight for the corresponding aggregate ratio.

Practical Checks and Common Traps

Sparse-segment review is primarily about whether the calculation is being interpreted and assembled correctly. A low expected loss total is not, by itself, evidence of a low loss level per unit. Likewise, a high segment ratio or cost is not automatically an error; it may be a valid selection, but its basis and influence should be checked.

Separate volume from level. Review the segment’s premium or exposure count alongside its selected ratio or cost. A small total can result from low volume even when the selected per-unit loss level is high.
Reconcile the aggregate. Confirm that segment premiums or exposures sum to the stated portfolio total and that segment expected losses sum to the aggregate estimate.
Check the basis. A ratio must match its premium basis, and an expected loss cost must match its exposure definition and unit. Avoid combining unlike bases without an explicit conversion.
Test influential assumptions. Recalculate the total after changing one questionable input while holding the others fixed. Report both the segment-level and aggregate-level change.
Do not treat pooling as automatic. Combining segments may be useful only when their definitions, measurement bases, and risk characteristics support that treatment. Small volume alone does not establish that segments should be combined.

These checks matter because an aggregate can conceal a concentrated input effect: a high selected ratio on a small premium may have a modest portfolio impact, while a modest change to a larger segment can move the total more. Evaluate the segment calculation and its aggregate weight together. This tutorial is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Check Your Understanding

For each problem, calculate on the stated basis and keep the segment weights explicit.

A hypothetical book has Segment A premium of $15,000 at a 72% expected loss ratio and Segment B premium of $85,000 at 64%. Find each expected loss and the aggregate expected loss ratio. (Answer: A $10,800; B $54,400; aggregate $65,200 and 65.20%.)
A small segment has 6 exposure units at $12,500 expected loss cost per unit. A large segment has 240 units at $3,750 per unit. Find total expected loss and aggregate expected loss cost, rounded to the nearest cent. (Answer: $975,000 total and $3,963.41 per exposure.)
A portfolio has $5,000 premium at an 120% expected loss ratio and $95,000 at 58%. Calculate the aggregate expected loss ratio. Explain briefly why the simple average of the two ratios is not the aggregate. (Answer: $61,100 expected loss; 61.10%; the simple average ignores unequal premium weights.)
A $200,000-premium portfolio has one small segment with $2,000 premium at a 90% expected loss ratio and all remaining premium at 61%. If the small segment’s ratio changes to 140%, calculate the change in aggregate expected loss and in aggregate expected loss ratio. (Answer: expected loss increases by $1,000; aggregate ratio increases by 0.50 percentage points, from 61.29% to 61.79%.)
Three segments have premiums of $10,000, $40,000, and $150,000, with selected expected loss ratios of 80%, 65%, and 60%, respectively. Calculate each segment expected loss, total expected loss, and the aggregate ratio. (Answer: $8,000, $26,000, and $90,000; total $124,000; aggregate ratio 62.00%.)

What You'll Learn