Exposure Mix as a Source of Territory Bias

Why Exposure Mix Matters

A territory’s aggregate losses depend on both the risks insured there and the losses associated with those risks. If one territory has more exposure in a class with higher expected losses, its total loss per exposure can be higher even when the corresponding risks have the same costs in both territories. An unadjusted territorial comparison can therefore confuse who is insured with where the risk is located.

This issue arises when territories differ in the mix of classification classes, coverages, limits, deductibles, or other rating characteristics. The key diagnostic is to compare aggregate results with results for reasonably comparable groups. This tutorial focuses on understanding the source of the distortion; it does not set out a procedure for producing a mix-adjusted territorial indication.

Definition: An exposure cell is a group of exposures with the same relevant rating characteristics for a particular comparison, such as class and limit. The exposure mix of a territory is the distribution of its exposures among those cells. For territory $t$ and cell $k$, let $e_{tk}$ be the exposure, $L_{tk}$ the loss assigned to that exposure, $c_{tk}=L_{tk}/e_{tk}$ the cell loss per exposure, and $w_{tk}=e_{tk}/\sum_j e_{tj}$ the cell’s share of the territory’s exposure.

The exposure unit must be appropriate to the data and comparison. For example, if policies have different effective durations, earned car-years may be more comparable than policy counts. Losses and exposures must also be measured on a consistent basis across territories. Differences in development, valuation, or the definition of an exposure can create additional distortions that are not exposure-mix effects.

How the Aggregate Comparison Combines Mix and Cell Results

For a territory with positive total exposure, its aggregate loss per exposure is total losses divided by total exposures. Because each cell’s losses equal its exposure multiplied by its cell loss per exposure, the aggregate is a weighted average of cell results.

c_t=\frac{\sum_k L_{tk}}{\sum_k e_{tk}} =\frac{\sum_k e_{tk}c_{tk}}{\sum_k e_{tk}} =\sum_k w_{tk}c_{tk}.

The weights are exposure shares: they are nonnegative and sum to one. Thus, an aggregate can change because the cell loss amounts change, because the cell exposure shares change, or both. A territorial comparison that reports only the aggregate does not by itself distinguish these explanations.

Why Different Mixes Can Create Different Aggregates

Suppose two territories have the same loss per exposure in every corresponding cell. Their aggregate results can still differ if their exposure weights differ. The reason follows directly from the weighted-average formula: each territory gives a different weight to the same set of cell results.

Identify comparable cells.
Specify the rating characteristics to hold alike in the comparison, such as class and coverage limit.

Find each cell result.
For each territory and cell, divide the cell losses by the corresponding exposure.

Inspect exposure shares.
Determine whether the territories assign substantially different weights to the cells.

Interpret the aggregate cautiously.
If the cell results are alike but the weights differ, the aggregate difference can arise from mix rather than a territory effect within those cells.

This is a diagnostic, not proof that territory has no effect. Cells may omit relevant characteristics, contain limited experience, or combine risks that are not truly comparable. Conversely, an apparent aggregate difference may remain important even if mix contributes to it. The point is that an aggregate comparison alone cannot identify the source.

Worked Example: Class Mix Produces a Higher Aggregate

Assume a hypothetical insurer has two risk classes with the same cell loss costs in two territories: $200 per exposure for Class L and $800 per exposure for Class H. Territory North has 800 Class L exposures and 200 Class H exposures. Territory South has 300 Class L exposures and 700 Class H exposures. All losses are on the same valuation basis, and each exposure has the same meaning.

Method. Calculate each territory’s aggregate loss per exposure as the exposure-weighted average of the two class loss costs. This method makes the class weights explicit rather than treating the territories’ overall averages as directly comparable.

For North, total exposure is $800+200=1{,}000$. Its Class L weight is $800/1{,}000=0.80$, and its Class H weight is $200/1{,}000=0.20$. The cell losses are $800\times \$200=\$160{,}000$ and $200\times \$800=\$160{,}000$, so total losses are $\$320{,}000$. The aggregate is:

c_{\text{North}}=\frac{\$320{,}000}{1{,}000} =0.80(\$200)+0.20(\$800)=\$320.

For South, total exposure is $300+700=1{,}000$. Its Class L weight is $0.30$, and its Class H weight is $0.70$. The cell losses are $300\times \$200=\$60{,}000$ and $700\times \$800=\$560{,}000$, for total losses of $\$620{,}000$. Therefore:

c_{\text{South}}=\frac{\$620{,}000}{1{,}000} =0.30(\$200)+0.70(\$800)=\$620.

South’s aggregate exceeds North’s by $\$620-\$320=\$300$ per exposure. Yet the loss cost is identical between territories within each class: $\$200$ in Class L and $\$800$ in Class H. The aggregate difference is entirely consistent with South having a larger share of its exposure in the higher-cost class. Check: the weighted averages reproduce the total losses divided by total exposures in both territories. The common mistake avoided is to interpret South’s higher aggregate as evidence that comparable risks cost more there.

Mix Differences Beyond Class

The same reasoning applies when the cells are defined by coverage, limit, or deductible. For example, if higher limits are associated with higher expected payments, a territory with a larger share of high-limit exposure may have a higher aggregate loss per policy, even if the corresponding limit groups have the same loss costs across territories. Deductibles can also change the amount paid per claim. Comparing portfolios with different deductible distributions without accounting for that difference can obscure what the aggregate says about location.

The appropriate cell definition depends on the question. If class and limit are both relevant, comparing only by class may leave a limit-mix difference inside each class. If many rating characteristics are combined, the resulting cells may become sparse. This creates a practical trade-off: cells should distinguish material differences in risk or coverage, but they also need enough exposure and loss experience to support useful comparisons.

Worked Example: Coverage-Limit Mix Changes the Aggregate

Consider a hypothetical liability product with two limits. Assume the loss per policy exposure is $\$180$ at the lower limit and $\$300$ at the higher limit in both territories. Territory East has 300 lower-limit and 100 higher-limit exposures. Territory West has 50 lower-limit and 150 higher-limit exposures. The losses are measured consistently, and there are no other differences in the example.

Method. Calculate losses within each limit group, then divide total losses by total exposure. This isolates the arithmetic effect of giving different exposure weights to the two limit groups.

East has $300+100=400$ exposures. Its lower-limit losses are $300\times \$180=\$54{,}000$, and its higher-limit losses are $100\times \$300=\$30{,}000$. Total losses are $\$84{,}000$, so its aggregate is $\$84{,}000/400=\$210$ per exposure. Equivalently, the weights are $300/400=0.75$ and $100/400=0.25$, giving $0.75(\$180)+0.25(\$300)=\$210$.

West has $50+150=200$ exposures. Its lower-limit losses are $50\times \$180=\$9{,}000$, and its higher-limit losses are $150\times \$300=\$45{,}000$. Total losses are $\$54{,}000$, so its aggregate is $\$54{,}000/200=\$270$ per exposure. Its weights are $0.25$ and $0.75$, yielding $0.25(\$180)+0.75(\$300)=\$270$.

West’s aggregate is $\$60$ higher per exposure, while the loss cost at each limit is the same in both territories. The difference follows from West’s larger high-limit share. Check: the calculated aggregates agree with the respective total losses divided by total exposures. The common mistake avoided is comparing the two overall averages as though both territories had the same limit distribution.

Aggregate Rankings Can Reverse Within Cells

A mix distortion can do more than exaggerate a difference: it can reverse the direction of an aggregate comparison. This is possible when one territory has a lower aggregate only because it has more exposure in lower-cost cells, even though its cell results are higher in every corresponding group. Such a reversal is a warning that the aggregate ranking does not describe the within-cell pattern.

Worked Example: A Reversal Between Cell and Aggregate Results

Assume two territories, Ridge and Harbor, and two classes. For Class L, the loss per exposure is $\$110$ in Ridge and $\$100$ in Harbor. For Class H, it is $\$510$ in Ridge and $\$500$ in Harbor. Ridge has 900 Class L and 100 Class H exposures; Harbor has 100 Class L and 900 Class H exposures. Assume consistent exposure and loss bases.

Method. First compare the territories within each class. Then compute the aggregates from cell losses. This two-part calculation shows whether the overall ordering agrees with the cell-level ordering.

Within Class L, Ridge exceeds Harbor by $\$110-\$100=\$10$ per exposure. Within Class H, Ridge also exceeds Harbor by $\$510-\$500=\$10$ per exposure.

Ridge has 1,000 exposures. Its Class L losses are $900\times \$110=\$99{,}000$, and its Class H losses are $100\times \$510=\$51{,}000$. Total losses are $\$150{,}000$, so its aggregate is $\$150$ per exposure.

Harbor also has 1,000 exposures. Its Class L losses are $100\times \$100=\$10{,}000$, and its Class H losses are $900\times \$500=\$450{,}000$. Total losses are $\$460{,}000$, so its aggregate is $\$460$ per exposure.

The aggregate is much lower in Ridge, even though Ridge has a higher loss per exposure in each class. Ridge’s 90% Class L mix contrasts with Harbor’s 10% Class L mix, and Class L has much lower costs than Class H. Check: the totals give $\$150{,}000/1{,}000=\$150$ and $\$460{,}000/1{,}000=\$460$; both within-class comparisons favor Harbor. The common mistake avoided is concluding from the aggregate ranking that Ridge has lower costs for comparable risks.

Common Mistakes and Interpretation

Using policy counts when exposure durations differ. A policy in force for a fraction of a year should not automatically receive the same weight as a full-year exposure. Use a consistent, appropriate exposure measure.
Comparing unlike cells. A class-only comparison may still mix different limits or deductibles. The cells need to reflect the characteristics material to the question.
Assuming mix explains every difference. Different weights can explain an aggregate gap when cell results are alike, but they do not establish that all other territory effects are absent.
Over-interpreting sparse cells. A small cell can have a volatile observed loss per exposure. A dramatic cell difference may reflect limited experience rather than a stable pattern.
Ignoring the loss basis. Different treatment of valuation, development, or included coverages can make cell comparisons misleading even when the exposure mix is described correctly.

The practical conclusion is narrow but important: an unadjusted aggregate territorial comparison combines within-cell experience and exposure composition. Before interpreting a difference as a territory difference, examine whether the territories insure comparable mixes and whether the comparison is based on consistent exposures and losses. That diagnosis identifies the issue; choosing and applying a method to place territories on a common mix is a separate step.

Check Your Understanding

For each hypothetical situation, calculate the requested aggregate and identify what the mix comparison does and does not establish.

Two territories have cell costs of $\$240$ for Class A and $\$720$ for Class B. North has 700 A and 300 B exposures; South has 400 A and 600 B exposures. Calculate both aggregates and their difference. (Answer: North $\$384$ per exposure; South $\$528$; South exceeds North by $\$144$, with different class mixes despite identical cell costs.)
A product has a $\$90$ loss per exposure at a higher deductible and $\$210$ at a lower deductible in both territories. Cedar has 800 higher-deductible and 200 lower-deductible exposures; Lake has 300 higher-deductible and 700 lower-deductible exposures. Calculate each aggregate. (Answer: Cedar $0.80(\$90)+0.20(\$210)=\$114$; Lake $0.30(\$90)+0.70(\$210)=\$174$, a $\$60$ difference attributable to the specified mix.)
Within Class A, Territory X has a loss cost of $\$160$, and Territory Y has $\$260$. Each territory has 500 exposures. X has 400 A and 100 B exposures; Y has 100 A and 400 B exposures. In both territories, Class B costs $\$360$. Calculate the two aggregates and state whether the aggregate ranking matches the Class A comparison. (Answer: X has losses $400(\$160)+100(\$360)=\$100{,}000$, or $\$200$ per exposure; Y has $100(\$260)+400(\$360)=\$170{,}000$, or $\$340$. X is lower in aggregate and in Class A, so the rankings match, though the different mixes affect the aggregate gap.)
Territory P has Class L cost $\$105$ and Class H cost $\$405$; Territory Q has costs $\$100$ and $\$400$, respectively. P has 950 L and 50 H exposures; Q has 50 L and 950 H exposures. Calculate both aggregate costs, then compare the within-class and aggregate rankings. (Answer: P has $[950(\$105)+50(\$405)]/1{,}000=\$120$; Q has $[50(\$100)+950(\$400)]/1{,}000=\$385$. P is higher in both classes but lower in aggregate, a mix-driven reversal.)
Two territories have exactly the same loss cost in every rating cell, but the territories have different exposure shares among those cells. What can be concluded from a difference in their aggregate loss costs, and what cannot be concluded from that difference alone? (Answer: The aggregate difference is consistent with the different exposure weights applied to the common cell costs; by itself, it does not show that comparable risks have different costs because of territory.)

What You'll Learn