Expense Allocation and Cross-Subsidies

Why Expense Allocation Can Shift Costs Between Classes

A rate provision can collect enough total expense to cover an insurer’s expenses and still distribute that expense unevenly among classes. The issue is not necessarily the total amount collected; it is whether the allocation basis assigns costs to classes in a way that reflects what drives those costs. When an allocation basis and a cost driver differ, one class may be assigned more than its driver-based share while another is assigned less. That redistribution is a cross-subsidy, measured relative to the chosen cost-driver benchmark.

For example, a shared processing expense might be driven mainly by the number of policies serviced, while the rate provision is allocated in proportion to exposures. If classes have different average policies per exposure, those two bases need not produce the same class assignments. The comparison is a tool for diagnosing a distributional effect; it does not, by itself, prove that the selected basis is inappropriate. The cost-driver assumption must be supported by how the expense is incurred.

Definition: Let $E$ be a shared expense to be allocated across classes. An allocation basis assigns amounts $A_j$ to class $j$, with $\sum_j A_j=E$. A cross-subsidy, relative to a specified cost-driver benchmark $B_j$, is the difference $T_j=A_j-B_j$. If $T_j>0$, class $j$ is assigned more than its benchmark share; if $T_j<0$, it is assigned less. When both allocations distribute the same total expense, $\sum_j T_j=0$.

The benchmark is not automatically a uniquely correct allocation. It is an explicit comparison based on a stated assumption about the expense’s cause. A class’s benchmark responsibility may be estimated from policy counts, exposures, premium, activity counts, or another relevant measure. The key question is whether the chosen measure tracks the expense being allocated.

A General Comparison Framework

Suppose the shared expense is assumed to vary in proportion to a measurable driver $z_j$ for class $j$. Assume the driver is measured consistently across classes, the total driver amount is $Z=\sum_j z_j>0$, and the expense is proportional to the driver over the period being studied. Then the benchmark cost per unit of driver is $c=E/Z$, and class $j$’s benchmark assignment is $B_j=cz_j$.

B_j=E\frac{z_j}{\sum_k z_k}

If the actual rate allocation uses a different basis $w_j$, with $W=\sum_j w_j>0$, then $A_j=Ew_j/W$. The signed transfer is the allocated amount minus the benchmark amount. A positive result indicates an over-allocation relative to the driver; a negative result indicates an under-allocation.

T_j=A_j-B_j =E\left(\frac{w_j}{W}-\frac{z_j}{Z}\right)

This comparison has an important conservation property. Since both sets of shares sum to one, the positive and negative differences offset across classes.

\sum_j T_j =E\left(\sum_j\frac{w_j}{W}-\sum_j\frac{z_j}{Z}\right) =E(1-1)=0

This does not mean the insurer has no expense shortfall or excess. It means only that reallocating the same total $E$ among classes redistributes responsibility without changing the total assigned. To express the class effect as an expense amount per exposure, divide $T_j$ by that class’s exposures. A positive amount is an additional assigned expense per exposure relative to the benchmark; a negative amount is a reduction.

Worked Class-Level Comparisons

Worked Example: Policy-Driven Service Costs Allocated by Exposure

A hypothetical insurer has two classes. Class A has 100 policies and 100 exposures; Class B has 50 policies and 100 exposures. Shared service expense is $9,000. Assume the expense is driven by policies serviced, but the proposed allocation assigns expense in proportion to exposures. Calculate the benchmark, the proposed allocation, and each class’s transfer.

Method and assumptions. Use policy count as the cost-driver benchmark, so each policy is assumed to generate the same service cost. Use exposure count for the proposed allocation. Both allocation methods must distribute the full $9,000.

There are $100+50=150$ policies. The benchmark cost per policy is $9{,}000/150=\$60$. Therefore, the benchmark assignments are:

Class A: $100\times \$60=\$6{,}000$. Class B: $50\times \$60=\$3{,}000$. These sum to $9,000.

There are $100+100=200$ exposures. The proposed allocation per exposure is $9{,}000/200=\$45$. The proposed assignments are $4,500 to A and $4,500 to B.

The transfers are proposed allocation minus benchmark. For A, $T_A=4{,}500-6{,}000=-\$1{,}500$. For B, $T_B=4{,}500-3{,}000=+\$1{,}500$. Per exposure, A is assigned $-1{,}500/100=-\$15$ relative to the benchmark; B is assigned $1{,}500/100=+\$15$.

Check and trap avoided. The transfers sum to zero, and each allocation method totals $9,000. A common mistake is to conclude that the classes should receive equal allocations because they have equal exposures. That is the proposed basis, not the stated cost driver: under the policy-count assumption, A has twice as many policies as B and therefore twice the benchmark expense.

Worked Example: Premium-Driven Commission Expense Allocated by Exposure

A hypothetical book has 80 exposures in Class C with $400,000 of premium and 120 exposures in Class D with $800,000 of premium. A shared acquisition expense of $36,000 is assumed to be a 3% commission on premium. A proposed rate allocation assigns it by exposure. Find the benchmark assignments, proposed assignments, and transfers.

Method and assumptions. Treat premium as the driver because the expense is assumed to equal 3% of premium. For the proposed allocation, use the class exposure shares. This example assumes the stated premium totals are on a consistent basis.

The benchmark follows directly from the assumed commission rate. For C, $0.03\times \$400{,}000=\$12{,}000$. For D, $0.03\times \$800{,}000=\$24{,}000$. The benchmark totals $12{,}000+24{,}000=\$36{,}000$.

The proposed allocation uses 200 exposures in total. Its expense per exposure is $\$36{,}000/200=\$180$. Thus C is assigned $80\times \$180=\$14{,}400$, while D is assigned $120\times \$180=\$21{,}600$. These amounts also total $36,000.

For C, $T_C=14{,}400-12{,}000=+\$2{,}400$, or $\$2{,}400/80=\$30$ per exposure. For D, $T_D=21{,}600-24{,}000=-\$2{,}400$, or $-\$2{,}400/120=-\$20$ per exposure.

Check and trap avoided. The dollar transfers offset. A common error is to apply the $180 proposed expense per exposure and call that the benchmark cost. It is only the result of the proposed basis. Under the stated 3% driver assumption, the benchmark costs are instead $12,000 and $24,000.

Worked Example: Inspection Costs Allocated by Exposure

A hypothetical insurer incurs $40 for each property inspection. Class E has 60 exposures and 90 inspections; Class F has 40 exposures and 30 inspections. The shared inspection expense is allocated by exposure. Calculate the total expense, the driver-based benchmark, the proposed allocation, and the transfers.

Method and assumptions. Treat the number of inspections as the driver, with the same $40 cost per inspection in both classes. The proposed allocation uses exposures. Assume the inspection counts are complete for the expense period.

There are $90+30=120$ inspections, so total expense is $120\times \$40=\$4{,}800$. Benchmark assignments are E: $90\times \$40=\$3{,}600$, and F: $30\times \$40=\$1{,}200$.

There are $60+40=100$ exposures. Allocation by exposure assigns $\$4{,}800/100=\$48$ per exposure. Thus E receives $60\times \$48=\$2{,}880$, while F receives $40\times \$48=\$1{,}920$.

The transfers are $T_E=2{,}880-3{,}600=-\$720$, or $-\$720/60=-\$12$ per exposure; and $T_F=1{,}920-1{,}200=+\$720$, or $+\$720/40=+\$18$ per exposure.

Check and trap avoided. Both allocation methods total $4,800, and the transfers offset. Do not infer driver responsibility from exposure share: E has 60% of exposures but 75% of inspections, so its exposure-based assignment is below its inspection-based benchmark.

Interpreting the Results Carefully

A per-exposure transfer helps translate the allocation comparison into a rate-level quantity, but it is not automatically the change to the final premium for every policy. It measures the difference in allocated expense per exposure under the specified comparison. Other components of a rate and the exposure basis used for a policy still matter.

Likewise, the benchmark depends on the cost-driver assumption. If inspection costs include a fixed component that does not vary with the number of inspections, assigning the entire expense in proportion to inspections may not be appropriate. If classes have materially different service requirements per policy, a simple policy count may be an imperfect benchmark. The calculation identifies the consequence of the assumptions; it does not establish those assumptions as fact.

State the expense and its period.
Identify the shared amount being distributed and ensure the class data refer to a consistent period and scope.

Name the proposed basis and the driver assumption.
For example, distinguish allocation by exposure from a benchmark assumed to vary with policy count or activity count.

Calculate both class assignments.
Confirm that each method distributes the same total expense before interpreting class differences.

Compute and interpret the signed transfer.
Subtract benchmark responsibility from proposed allocation. Positive means more assigned than the benchmark; negative means less.

A frequent exam trap is to call any difference between class allocations a cross-subsidy without naming the benchmark. A difference is meaningful only relative to a stated cost driver or other defensible allocation standard. Another trap is to report which class is subsidizing another without checking signs: with $T_j=A_j-B_j$, a positive value means the class pays more under the proposal than under the benchmark, not less.

Key takeaway: A cross-subsidy is a redistribution relative to an explicit benchmark, not a statement that total expenses are misstated. Compare the proposed allocation with a cost-driver-based assignment, verify that both totals reconcile, and interpret each class’s signed difference in context.

Check Your Understanding

For each problem, calculate the driver-based benchmark and compare it with the stated proposed allocation. Treat all scenarios as hypothetical.

A $10,000 shared expense is assumed to be driven by policy count. Class A has 80 policies and 160 exposures; Class B has 120 policies and 120 exposures. The proposed allocation is by exposure. Find each class’s transfer. (Answer: A is over-allocated by $1,714.29; B is under-allocated by $1,714.29.)
A hypothetical $18,000 commission expense is 10% of premium. Class A has $60,000 of premium and 60 exposures; Class B has $120,000 of premium and 60 exposures. The proposed allocation is by exposure. Find each transfer and the transfer per exposure. (Answer: A is over-allocated by $3,000, or $50 per exposure; B is under-allocated by $3,000, or $50 per exposure.)
A hypothetical service cost is $25 per claim. Class A has 90 claims and 60 exposures; Class B has 30 claims and 40 exposures. Total expense is allocated by exposure. Calculate the signed transfer for each class and check the total. (Answer: A is under-allocated by $450; B is over-allocated by $450; total transfer is $0.)
A $8,400 shared expense is assumed to be $70 per policy. Class A has 40 policies and $30,000 of premium; Class B has 80 policies and $90,000 of premium. The proposed allocation is by premium. Calculate both transfers. (Answer: A is under-allocated by $700; B is over-allocated by $700.)
Two allocation methods both assign a total shared expense of $24,000. Relative to a stated benchmark, Class R has a transfer of $+\$1,800$ and Class S has a transfer of $-\$1,200$. There are no other classes. Is the comparison internally consistent? If not, what transfer is required for S? (Answer: No; the transfers must sum to zero, so Class S must have a transfer of $-\$1,800$.)

What You'll Learn