Why the Allocation Base Matters
An expense total does not, by itself, determine how much of that expense belongs to each class of business. The allocation base determines the distribution. If the base does not reflect the activity that generates the cost, the allocation can shift costs between classes even when the total allocated expense is correct.
For example, a charge that is assessed as a percentage of premium has a natural premium base. A cost incurred once for each policy may be better represented by policy count. These are not interchangeable choices: two classes can have similar premium but very different policy counts, or similar policy counts but different exposure volume. The selection question is therefore not simply, “Which data are available?” It is, “Which measurable quantity best represents the driver of this expense?”
The base is a method for distributing an expense, not necessarily a statement that every dollar of the expense is literally caused by that base. In practice, an actuary may use a proxy because direct measurement is unavailable. The proxy should be chosen deliberately and reviewed for whether it gives a reasonable distribution.
A General Allocation Rule
Suppose an expense pool \(C\) is to be allocated among \(n\) classes. Let \(b_j\) be the amount of the chosen base for class \(j\), and let \(B\) be the total base across all classes. If the expense is allocated in proportion to the base, class \(j\)'s allocation is:
This rule assumes that expense is reasonably proportional to the selected base over the segments being considered. In other words, each base unit receives the same share of the pool. If some costs are incurred once per policy while others vary with premium, combining them into one pool and allocating all of it on only one base may distort the distribution. Where the components have different drivers, separate pools or a documented proxy may be more appropriate.
How to Select a Base
Start with the expense’s operational description: what event, transaction, quantity, or financial measure makes the expense increase? Then identify a base that measures that driver consistently across the classes being compared. A useful first-pass guide is:
| Possible base | Best fit when cost is driven by | Question to check |
|---|---|---|
| Premium | A charge that varies with premium dollars, such as a percentage-based assessment | Does the expense actually rise with premium, or is premium merely convenient to observe? |
| Exposure | The amount of insured risk or coverage provided over the applicable period | Does the exposure measure represent the relevant coverage and time at risk? |
| Policy count | Activities performed once per policy or policy transaction | Are policies reasonably similar in the work or cost generated per policy? |
| Losses | An expense that varies with the amount of losses, when that relationship is supported | Which loss measure is relevant, and is it on a consistent basis across classes? |
A base should be defined precisely before it is applied. “Exposure” might mean insured vehicle-years, building-years, or another measure appropriate to the coverage; these quantities are not automatically comparable. “Policy count” also needs a clear counting rule: for example, whether the unit is a policy in force at a specified date or a policy transaction during a period. Similarly, a loss base requires a stated loss measure and valuation basis.
Consider the cost driver across the entire pool, not just one class. A policy-processing cost may be approximately the same for every policy, but that approximation can fail if policies differ greatly in the number of transactions or the amount of service required. A premium base can be easy to obtain yet still be a poor proxy for work performed. Conversely, an expense may have a genuine premium-related mechanism, making premium the more defensible base even if another quantity is also correlated with it.
Worked Applications
Worked Example: A Premium-Related Assessment
A hypothetical insurer is allocating a $90,000 assessment that is charged in proportion to written premium. Class A has $1,500,000 of premium and Class B has $1,000,000. A review shows that the classes have 3,000 and 3,000 exposure units, respectively. Select the more suitable base and allocate the pool. State the mistake avoided.
Method. Because the assessment is charged in proportion to premium, use premium as the base. Total premium is:
Class A's premium share is \(1{,}500{,}000/2{,}500{,}000=0.60\), so its allocation is \(0.60\times \$90{,}000=\$54{,}000\). Class B's share is \(1{,}000{,}000/2{,}500{,}000=0.40\), so its allocation is \(0.40\times \$90{,}000=\$36{,}000\).
The check is \(\$54{,}000+\$36{,}000=\$90{,}000\). If exposure units were used instead, the equal counts would assign $45,000 to each class. That calculation is arithmetically valid but inconsistent with the stated premium-proportional assessment. Common mistake avoided: choosing a convenient available base instead of the driver specified by how the expense is charged.
Worked Example: A Cost Incurred Per Policy
A hypothetical policy administration pool of $84,000 covers work that is performed once for each policy issued. Class C has 1,200 policies and Class D has 800. Their exposure totals are 900 and 1,100 exposure units. Assuming comparable work per policy, allocate the pool using the appropriate base.
Method. The expense is incurred per policy, so policy count is the more direct base. The total count is \(1{,}200+800=2{,}000\) policies. Class C has a share of \(1{,}200/2{,}000=0.60\), giving \(0.60\times \$84{,}000=\$50{,}400\). Class D has a share of \(800/2{,}000=0.40\), giving \(0.40\times \$84{,}000=\$33{,}600\).
The allocated amounts total \(\$50{,}400+\$33{,}600=\$84{,}000\). The implied average allocation is \(\$84{,}000/2{,}000=\$42\) per policy; multiplying by 1,200 and 800 gives the same class amounts. Using exposure units instead would allocate 45% to Class C and 55% to Class D, or $37,800 and $46,200. That would assign more of a per-policy cost to the class with fewer policies merely because it has more exposure units. Common mistake avoided: treating a policy count and an exposure count as equivalent units.
Worked Example: An Expense Linked to Exposure Volume
A hypothetical inspection program has a $60,000 expense pool. The program’s service records indicate that inspections are scheduled in proportion to insured vehicle-years, the selected exposure measure. Class E has 9,000 vehicle-years and Class F has 3,000. Allocate the pool and explain why policy count alone could misstate the distribution.
Method. The stated driver is vehicle-year volume, so use exposure. The total is \(9{,}000+3{,}000=12{,}000\) vehicle-years. Class E's share is \(9{,}000/12{,}000=0.75\), giving \(0.75\times \$60{,}000=\$45{,}000\). Class F's share is \(3{,}000/12{,}000=0.25\), giving \(0.25\times \$60{,}000=\$15{,}000\).
The allocations add to $60,000. If the two classes had the same number of policies but different vehicle-year totals, a policy-count allocation would split the pool equally despite the program scheduling inspections in proportion to vehicle-years. The exposure base is suitable here because it measures the specified activity, not because exposure is always preferable to policy count. Common mistake avoided: assuming one policy represents the same quantity of insured exposure in every class.
Worked Example: A Loss-Related Expense Pool
A hypothetical vendor agreement charges a variable service expense in proportion to expected loss dollars. The expense pool to distribute is $48,000. Class G has $800,000 of expected losses and Class H has $400,000. Allocate the pool, and identify a condition that would make a loss base questionable.
Method. The stated agreement links the expense to expected loss dollars, so use those losses as the base. Total expected losses are \(\$800{,}000+\$400{,}000=\$1{,}200{,}000\). Class G's share is \(800{,}000/1{,}200{,}000=2/3\), yielding \((2/3)\times \$48{,}000=\$32{,}000\). Class H's share is \(400{,}000/1{,}200{,}000=1/3\), yielding \((1/3)\times \$48{,}000=\$16{,}000\).
The check is \(\$32{,}000+\$16{,}000=\$48{,}000\). A loss base would be questionable if the expense were actually incurred once per claim regardless of claim size: claim count might then represent the driver more directly. Common mistake avoided: assuming that all claim-related expenses should be allocated by loss dollars without checking what makes the expense vary.
Practical Checks and Common Traps
Before finalizing a base, test both its conceptual fit and its measurement. A strong allocation process asks what generates the expense, whether the chosen quantity captures that activity, and whether the data are defined consistently across classes. If a base is only a proxy, document why it is reasonable and whether a better measure is available.
- Do not confuse correlation with a cost driver. Premium and exposure may move together in some data, but that does not show that a premium-related expense is caused by exposure, or vice versa.
- Do not combine unlike costs without checking their drivers. If a pool contains policy-level and premium-level components, one base may distribute one component poorly. Separate components when the information supports doing so.
- Do not mix units. A count of policies, a count of transactions, and exposure-years measure different things. State the unit and period for each base.
- Do not treat conservation as validation. An allocation that sums exactly to the expense pool can still be based on the wrong driver.
- Use a consistent loss basis. If losses are the selected base, specify which loss measure is used and ensure it is measured comparably for all segments. A change in the loss definition can change the allocation even if the expense pool is unchanged.
Check Your Understanding
For each problem, identify the most suitable base from the information given and calculate the allocations.
- A hypothetical $36,000 charge is assessed as a fixed percentage of premium. Two classes have premiums of $400,000 and $600,000. What is the allocation to each class? (Answer: $14,400 and $21,600, respectively.)
- A hypothetical $84,000 expense is incurred in proportion to exposure volume. Two classes have 3,500 and 7,000 exposure units. Allocate the pool. (Answer: $28,000 and $56,000, respectively.)
- A hypothetical $45,000 processing pool is incurred once per policy, with comparable processing effort per policy. Class A has 900 policies and Class B has 600. Allocate the pool. (Answer: $27,000 to Class A and $18,000 to Class B.)
- A hypothetical $120,000 expense varies in proportion to expected losses. Two classes have $1,500,000 and $2,500,000 of expected losses. Allocate the pool and name one circumstance in which policy count might be a better base. (Answer: $45,000 and $75,000, respectively; policy count may be better if the expense is incurred per claim or policy independently of loss size.)
- A hypothetical insurer has two expense pools: $50,000 for work done once per policy and $100,000 for a premium-proportional charge. Class X has 600 policies and $900,000 of premium; Class Y has 400 policies and $2,100,000 of premium. Allocate each pool using its matching base, then give each class’s combined allocation. (Answer: policy pool—$30,000 to X and $20,000 to Y; premium pool—$30,000 to X and $70,000 to Y; combined—$60,000 to X and $90,000 to Y.)