Why Aggregate Trend Can Mislead
An aggregate loss measure changes for two distinct reasons: the cost within a group can change, and the share of business in each group can change. A portfolio may have higher loss costs in every territory but a lower aggregate loss cost if it shifts toward lower-cost territories. The reverse can also occur: a shift toward higher-cost business can make the aggregate trend look much larger than the underlying trends within groups.
This distinction matters when evaluating historical experience. A comparison of aggregate averages alone combines changes in cost with changes in composition. To interpret an observed trend, identify the groups, use weights appropriate to the measure, and compare the observed result with a calculation that holds the group mix constant. This treatment is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
The meaning of a “group” depends on the question. Groups might represent territories, coverage forms, vehicle types, or claim categories. The weights must match the denominator of the measure. For loss cost per exposure, use each group’s share of exposures. For average severity per claim, use each group’s share of claims. If the weights do not correspond to the denominator, the weighted average does not represent the aggregate measure.
Deriving the Weighted-Average and Mix Decomposition
Suppose group \(j\) has \(E_{jt}\) exposures and loss cost \(c_{jt}\) in period \(t\). Its losses are \(E_{jt}c_{jt}\). Therefore, total losses divided by total exposures are
The second equality follows by dividing each term in the numerator by the same total exposure count: \(\frac{E_{jt}c_{jt}}{\sum_k E_{kt}}=\frac{E_{jt}}{\sum_k E_{kt}}c_{jt}=w_{jt}c_{jt}\). Thus, the aggregate loss cost is the exposure-weighted average of group loss costs. For severity, the same reasoning applies with claim counts in place of exposures.
One exact way to separate a change in the aggregate is to use the initial period’s weights for the within-group component and the final period’s group costs for the mix component:
To verify the identity, expand the right-hand side. The first sum is \(\sum_j w_{j0}c_{j1}-\sum_j w_{j0}c_{j0}\); the second is \(\sum_j w_{j1}c_{j1}-\sum_j w_{j0}c_{j1}\). The terms \(\sum_j w_{j0}c_{j1}\) cancel, leaving \(C_1-C_0\). This allocation is exact, but it is not the only possible allocation: using final-period weights for the within-group term gives another exact decomposition. State which convention is being used when reporting components.
Worked Examples
Worked Example: A Shift Toward Higher-Cost Coverage
A hypothetical insurer compares loss costs for two coverage forms. Assume the reported costs are on a consistent basis and use exposure counts as weights. In period 0, Form A has 800 exposures and loss cost $100; Form B has 200 exposures and loss cost $400. In period 1, each form’s loss cost has increased by 10%, but the mix has shifted to 500 exposures of each form. Calculate the observed aggregate trend and the fixed-mix trend. Identify the contribution of the mix shift.
Method. Calculate the aggregate loss cost in each period as an exposure-weighted average. Then compare the period 1 aggregate with the period 1 value at period 0 weights.
In period 0, there are \(800+200=1{,}000\) exposures. The weights are \(w_{A0}=800/1{,}000=0.8\) and \(w_{B0}=200/1{,}000=0.2\). Aggregate losses are \(800(\$100)=\$80{,}000\) for A and \(200(\$400)=\$80{,}000\) for B. Total losses are \(\$160{,}000\), so \(C_0=\$160{,}000/1{,}000=\$160\).
The period 1 group costs are \(c_{A1}=\$100(1.10)=\$110\) and \(c_{B1}=\$400(1.10)=\$440\). Each group has 500 of the 1,000 exposures, so the period 1 weights are 0.5 and 0.5. Aggregate losses are \(500(\$110)=\$55{,}000\) and \(500(\$440)=\$220{,}000\). Thus, \(C_1=\$275{,}000/1{,}000=\$275\). The observed aggregate trend is \(\$275/\$160-1=0.71875\), or 71.875%.
At the initial mix, the period 1 costs would produce \(C_1^{\mathrm{fixed}}=0.8(\$110)+0.2(\$440)=\$88+\$88=\$176\). The fixed-mix trend is \(\$176/\$160-1=10\%\), matching the 10% increase in both group costs.
For the exact decomposition, the within-group change is \(0.8(\$110-\$100)+0.2(\$440-\$400)=\$8+\$8=\$16\). The mix component is \((0.5-0.8)(\$110)+(0.5-0.2)(\$440)=-\$33+\$132=\$99\). Together, \(\$16+\$99=\$115=C_1-C_0\). The calculations check: the group costs rose 10%, while the shift toward Form B increased the aggregate by an additional \(\$99\) per exposure under this decomposition.
Common mistake avoided: Applying 10% to the original aggregate and concluding that the observed aggregate must be \(\$176\). That is the fixed-mix value, not the actual period 1 value, because the exposure shares changed.
Worked Example: A Mix Shift Reverses the Aggregate Trend
A hypothetical property portfolio is divided into standard and high-hazard territories. In period 0, 25% of exposures are standard, with loss cost $200, and 75% are high-hazard, with loss cost $1,000. In period 1, both territory costs have increased by 5%, but the exposure shares have changed to 75% standard and 25% high-hazard. Find the observed and fixed-mix trends, then reconcile the aggregate change.
Method. Use exposure shares to form each period’s aggregate loss cost, and use the exact decomposition with period 0 weights for the within-group component.
The period 0 aggregate is \(C_0=0.25(\$200)+0.75(\$1{,}000)=\$50+\$750=\$800\). Period 1 costs are \(\$200(1.05)=\$210\) and \(\$1{,}000(1.05)=\$1{,}050\). The period 1 aggregate is \(C_1=0.75(\$210)+0.25(\$1{,}050)=\$157.50+\$262.50=\$420\). Therefore, the observed trend is \(\$420/\$800-1=-0.475\), or -47.5%.
Holding the original mix fixed gives \(C_1^{\mathrm{fixed}}=0.25(\$210)+0.75(\$1{,}050)=\$52.50+\$787.50=\$840\). The fixed-mix trend is \(\$840/\$800-1=5\%\).
The within-group component is \(0.25(\$210-\$200)+0.75(\$1{,}050-\$1{,}000)=\$2.50+\$37.50=\$40\). The mix component is \((0.75-0.25)(\$210)+(0.25-0.75)(\$1{,}050)=\$105-\$525=-\$420\). Their sum, \(\$40-\$420=-\$380\), equals \(C_1-C_0=\$420-\$800=-\$380\). The aggregate decreased despite a 5% increase in each territory’s loss cost because the portfolio shifted toward the lower-cost territory.
Common mistake avoided: Calling the negative aggregate trend evidence that territory costs improved. The group-level data show the opposite; the decline is caused by the composition change.
Worked Example: Use Claim Shares for Average Severity
A hypothetical liability portfolio has minor and major claims. In period 0, there are 900 minor claims averaging $1,000 and 100 major claims averaging $9,000. In period 1, each type’s average severity rises by 8%; the claim mix becomes 800 minor and 200 major claims. Calculate the average severity in both periods and compare the observed trend with the fixed-mix trend. Assume these are claim counts and average severities for the stated periods.
Method. Average severity is total losses divided by total claims, so its weights are shares of claims, not shares of exposures.
In period 0, losses are \(900(\$1{,}000)=\$900{,}000\) for minor claims and \(100(\$9{,}000)=\$900{,}000\) for major claims. Total losses are \(\$1{,}800{,}000\) over 1,000 claims, giving average severity \(C_0=\$1{,}800\). The claim shares are 0.9 minor and 0.1 major.
Period 1 severities are \(\$1{,}000(1.08)=\$1{,}080\) and \(\$9{,}000(1.08)=\$9{,}720\). Losses are \(800(\$1{,}080)=\$864{,}000\) and \(200(\$9{,}720)=\$1{,}944{,}000\). Total losses are \(\$2{,}808{,}000\) over 1,000 claims, so \(C_1=\$2{,}808\). The observed average-severity trend is \(\$2{,}808/\$1{,}800-1=56\%\).
At the original claim mix, period 1 average severity is \(0.9(\$1{,}080)+0.1(\$9{,}720)=\$972+\$972=\$1{,}944\), an 8% increase from \(\$1{,}800\). The difference between actual and fixed-mix period 1 severity is \(\$2{,}808-\$1{,}944=\$864\), reflecting the change in claim shares.
Common mistake avoided: Weighting the severity figures by exposure shares. The denominator for average severity is claims, so the appropriate weights are claim shares. If the target were loss cost per exposure instead, the relevant weights and calculation would need to reflect exposures and claim frequency.
Practical Interpretation and Exam Traps
An observed aggregate trend is a valid description of the aggregate measure as recorded, but it does not by itself identify the trend in comparable business. The fixed-mix comparison provides one useful diagnostic: it applies the later group costs to the earlier group weights. It does not claim that the initial mix is inherently preferable, or that a particular mix shift is accidental. It isolates the effect of changing group costs under a clearly stated reference mix.
The chosen groups also matter. If broad territory averages conceal different coverage distributions within each territory, a territory-only comparison may still contain a mix effect. Grouping should be sufficiently relevant to the question and supported by credible, consistently defined data. Conversely, dividing data into many small groups can make individual results unstable. Interpretation should therefore include the group definitions, the denominator used to form weights, and the reference mix used for any fixed-mix calculation.
State whether the analysis concerns loss cost per exposure, average severity per claim, or another aggregate measure.
Use exposure shares for loss cost per exposure and claim shares for average severity. Confirm that weights sum to 1 in each period.
Compute each period’s actual aggregate, then apply the later group costs to the stated reference-period weights.
Explain whether composition amplifies, reduces, or reverses the change suggested by within-group costs.
A common exam trap is to multiply a single aggregate by a trend factor when the question supplies group-specific costs and changing group shares. That shortcut assumes the mix is unchanged. Another is to compare percentages without checking the dollar or loss-cost levels: a large percentage change in a small group may have little effect on the total, while a modest change in a heavily weighted group may matter substantially. Always verify the weighted totals and, when decomposing, check that the components sum to the observed aggregate change.
Check Your Understanding
For each problem, identify the relevant weights and distinguish the observed aggregate from the fixed-mix comparison.
- A hypothetical insurer’s low-cost territory has loss cost $120 and high-cost territory has loss cost $600 in period 0. Exposure shares are 70% and 30%. In period 1, each cost rises 10%, while the shares become 40% and 60%. What is the observed aggregate trend? (Answer: 70%)
- Two coverage groups have period 0 loss costs of $300 and $900, with exposure shares 60% and 40%. In period 1, both costs rise 4%, while the shares become 80% and 20%. Calculate the observed aggregate trend and the fixed-mix trend. (Answer: observed trend -19.11%; fixed-mix trend 4%)
- Average severity is $800 for mild claims and $4,800 for severe claims. In period 0, 85% of claims are mild and 15% are severe. In period 1, both severities rise 6%, but only 10% of claims are severe. Find the observed average-severity trend and the fixed-mix trend. (Answer: observed trend -9.14%; fixed-mix trend 6%)
- Group A and Group B have loss costs of $200 and $800 in period 0, with equal exposure shares. In period 1, their costs are $220 and $880, and their exposure shares are 30% and 70%. Using initial weights for the within-group component, calculate the observed aggregate trend and the mix component in dollars per exposure. (Answer: observed trend 36.4%; mix component $132 per exposure)
- A hypothetical portfolio has 1,000 exposures in each of two groups. Group A has 100 claims with average severity $1,000; Group B has 500 claims with average severity $500. What is the portfolio average severity? (Answer: $583.33 per claim)