Expected Cost As The Actuarial Benchmark

Why an Expected-Cost Benchmark Matters

A price decision needs a clear measure of the costs associated with the business that might be written at that price. For insurance, the central cost component is expected claims. Relevant expenses must also be included: some arise for each policy written, some vary with premium, and others are incurred regardless of the price choice. Combining these components carefully provides an actuarial benchmark for comparing alternatives.

The benchmark is not, by itself, a recommended price or an optimization result. It is a cost measure built from the assumptions and decision horizon in the problem. Expected total cost may change when a price changes because the number or mix of policies changes, even when the expected cost per policy is unchanged. Conversely, a fixed expense does not become price-sensitive merely because it is allocated across policies for reporting.

Definition: For a specified price decision and time horizon, the expected-cost benchmark is the expected amount of claims and relevant expenses associated with that decision. It includes costs that vary with a sale or with the price charged, and includes a fixed cost only when that cost is incurred within the stated horizon. A cost is fixed with respect to the decision if its amount does not change across the alternatives being compared.

Building the Cost Components

For one bound policy, expected claims can be calculated directly from the claim model. For example, if a policy has claim probability $r$, and the claim amount conditional on a claim has expected value $m$, the expected claim cost is $rm$. For a claim count with expected value $E[N]$ and identically distributed claim severities with mean $E[Z]$, independence of count and severities gives $E[\sum_{k=1}^{N}Z_k]=E[N]E[Z]$. If that independence assumption does not hold, the appropriate joint model must be used instead.

The expense side depends on how the expense is incurred. A per-policy issuance cost is incurred when a policy is written, so expected total issuance expense depends on the expected number written. A commission defined as a percentage of premium depends on both the premium and the number of policies. A fixed system cost that is incurred under every alternative contributes the same amount to each alternative’s total cost. A cost that is triggered only if a sales threshold is reached is not fixed across alternatives unless the threshold outcome is the same under all of them.

The relevant distinction is not whether an accountant labels an expense “fixed” or “variable” in a general budget. It is whether the amount changes with the particular decision being analyzed, over the stated horizon. Likewise, a cost may be fixed in the short run but change over a longer horizon. State the horizon before deciding how to treat it.

Worked Example: Expected Cost for Two Price Alternatives

A hypothetical insurer considers two prices for a one-year coverage offer to 80 prospects. At the lower price, 75% of Group A prospects and 25% of Group B prospects are expected to bind. At the higher price, the corresponding probabilities are 50% and 12.5%. The expected annual claim costs per bound policy are $400 for Group A and $1,000 for Group B. Each bound policy also generates $100 of variable servicing expense. A fixed $8,000 system expense is incurred under either alternative. Calculate expected total cost at each price. Assume no discounting; claim costs are one-year expected amounts, with no mortality or other time-value adjustment.

Method. First calculate expected bound policies by group. Then multiply each group’s expected count by its expected claims plus per-policy expense, and add the fixed expense once.

At the lower price, Group A has $80(0.75)=60$ expected bound policies, and Group B has $80(0.25)=20$. Their total is $60+20=80$ expected policies.

For Group A, expected claims and servicing expense per bound policy are $$400+$100=$500$. Its expected total cost is $60($500)=$30,000$. For Group B, the corresponding per-policy cost is $$1,000+$100=$1,100$, giving $20($1,100)=$22,000$. Including fixed expense, the lower-price total is $$30,000+$22,000+$8,000=$60,000$.

At the higher price, expected bound counts are $80(0.50)=40$ in Group A and $80(0.125)=10$ in Group B. Group A contributes $40($500)=$20,000$, and Group B contributes $10($1,100)=$11,000$. Therefore expected total cost is $$20,000+$11,000+$8,000=$39,000$.

Check. The fixed expense is $8,000 in both totals. The remaining amounts, $52,000 and $31,000, equal the expected claims plus servicing costs for the policies expected to bind. A common error is to use 80 policies at both prices, despite the different binding probabilities.

Expected Cost When Purchase Is Uncertain

When a prospect may or may not purchase, it is useful to distinguish the cost conditional on a sale from the cost expected before the purchase outcome is known. Let $I_j(p)$ equal 1 if prospect $j$ buys at price $p$, and 0 otherwise. Let $q_j(p)=P(I_j(p)=1)$. If $C_j$ is expected claims per bound policy and $e_j(p)$ is the expense incurred per bound policy, then the cost from that prospect is zero if no sale occurs and $C_j+e_j(p)$ if a sale occurs. A fixed cost $F$ is added if it is incurred regardless of the purchase outcome.

E[T(p)] = F+\sum_j q_j(p)\bigl(C_j+e_j(p)\bigr)

This expression does not require purchase outcomes for different prospects to be independent. Linearity of expectation allows the expected cost contributions to be added without an independence assumption. The expression does require that $C_j+e_j(p)$ represent the appropriate expected cost conditional on that prospect buying. If the mix of purchasers changes with price, the conditional expected claims cost may also change by price; in that case, use the appropriate $C_j(p)$ or segment-specific costs.

Specify the decision and horizon.
Identify the prices being compared and the period over which costs are counted.

Estimate conditional costs.
Calculate expected claims and per-sale expenses for the relevant risks or segments.

Apply expected sales.
Multiply each conditional cost by the expected number of policies written, or sum prospect-level purchase probabilities.

Add decision-invariant costs once.
Include a fixed amount that is incurred under each alternative; do not multiply it by the number of policies unless the cost definition itself is per policy.

Why the Expected-Cost Formula Holds

The formula follows by writing the cost as a fixed amount plus the cost contribution from each potential sale. For prospect $j$, the sale indicator $I_j(p)$ is 1 with probability $q_j(p)$ and 0 otherwise. If the expected conditional cost is $C_j+e_j(p)$, the expected cost contribution is the purchase probability times that conditional amount.

E\left[I_j(p)\bigl(C_j+e_j(p)\bigr)\right] =E[I_j(p)]\bigl(C_j+e_j(p)\bigr) =q_j(p)\bigl(C_j+e_j(p)\bigr).

Adding these contributions and the fixed amount gives the total expected-cost formula. If the conditional cost is itself random, the same argument applies using its conditional expectation given a sale: $E[I_j(p)K_j]=P(I_j(p)=1)E[K_j\mid I_j(p)=1]$. This is why the relevant claim estimate should describe the risks actually expected to be written, not automatically the entire prospect pool.

Worked Example: Converting Frequency and Severity to Expected Cost

A hypothetical portfolio has 300 prospects. At a selected price, each prospect has a 0.60 probability of buying, and a bound policy has a 0.08 probability of one claim during the year. Conditional on a claim, severity is $2,000 or $3,000 with equal probability. Each bound policy incurs $45 of servicing expense, and a fixed $12,000 platform expense is incurred regardless of sales. Find expected total cost. Assume the claim event and severity model are the same for all bound policies, with no discounting or mortality.

Method. Calculate expected severity, then expected claims per bound policy, then expected bound count and portfolio cost.

The conditional mean severity is $(0.5)($2,000)+(0.5)($3,000)=$2,500$. Expected claims per bound policy are $(0.08)($2,500)=$200$. Thus expected claims plus servicing expense per bound policy are $$200+$45=$245$.

Expected policies bound are $300(0.60)=180$. Expected claims and servicing expense for these policies are $180($245)=$44,100$. Adding the fixed expense gives expected total cost of $$44,100+$12,000=$56,100$.

Check. The claims component is $180(0.08)($2,500)=$36,000$; servicing expense is $180($45)=$8,100$. These sum to $44,100 before fixed expense. The common mistake of applying the 0.08 claim probability to all 300 prospects would count claims for people who are not expected to buy.

Expenses That Depend on Premium

Some expenses are per policy but do not depend on the premium amount; others are explicitly calculated as a percentage of premium. Both are relevant costs, but their behavior differs across price alternatives. If the commission rate is $c$, the premium is $p$, and $n(p)$ policies are expected to bind, expected commission is $n(p)cp$. It changes when either the expected number of policies or the premium changes.

Worked Example: Separating Commission From Fixed Expense

A hypothetical insurer compares two annual prices for a coverage option. At $600 it expects 100 policies to bind; at $800 it expects 60. Expected claims are $420 per bound policy, issue expense is $35 per bound policy, and commission is 10% of premium. A fixed expense of $9,000 is incurred in either case. Find expected total cost under each alternative. Treat claim amounts as one-year expected costs, with no discounting or mortality adjustment.

At $600, commission per policy is $0.10($600)=$60$. Total expected variable cost per policy is $$420+$35+$60=$515$. For 100 expected policies, this is $100($515)=$51,500$. Adding $9,000 fixed expense gives $60,500.

At $800, commission per policy is $0.10($800)=$80$. Total expected variable cost per policy is $$420+$35+$80=$535$. For 60 expected policies, this is $60($535)=$32,100$. Adding fixed expense gives $41,100.

Check. At $600 the commission total is $100($60)=$6,000$; at $800 it is $60($80)=$4,800$. Each total includes claims, issue expense, commission, and the same fixed $9,000. Excluding commission because it is “an expense loading” rather than a claim would understate the cost benchmark.

Fixed Costs, Allocations, and Common Traps

A fixed cost may be allocated to policies for financial reporting or unit-cost analysis. That allocation is not necessarily the cost that changes with a price decision. If fixed expense $F$ is spread evenly over $n$ expected policies, the allocated amount per policy is $F/n$. When $n$ changes, the allocation changes even though the total $F$ does not. Keep the total-cost calculation and any per-policy allocation conceptually separate.

A useful test is to ask: Would this amount change if the decision changed, holding the stated assumptions and horizon fixed? If not, it is fixed with respect to that decision. If it changes with policy count, premium, coverage, or another modeled outcome, include that dependence. Do not automatically classify an expense as fixed simply because it is paid monthly, or as variable simply because it is reported per policy.

Wrong: Divide a fixed $18,000 expense by 100 expected policies and then treat the resulting $180 as an additional cost for every policy in every alternative, even when the expected count changes.
Correction: Include $18,000 once in each alternative’s total expected cost if it is incurred regardless of that alternative. If reporting per-policy cost, allocate afterward using that alternative’s expected count.
Wrong: Calculate expected claims for all prospects and add them to the expected cost even though only a fraction is expected to purchase.
Correction: Use the expected number of bound policies or prospect-level purchase probabilities, and use expected claims conditional on binding.

Key takeaway: The expected-cost benchmark is expected claims plus all relevant sale- or price-sensitive expenses, together with fixed costs that are incurred over the decision horizon. Apply purchase probabilities to costs conditional on sale, and include decision-invariant fixed amounts once rather than confusing them with per-policy costs.

Check Your Understanding

For each problem, calculate the expected cost requested. Treat stated fixed expenses as incurred under every alternative unless the problem says otherwise.

A hypothetical offer is presented to 125 prospects. Each has a 0.40 probability of binding. Expected claims per bound policy are $300, variable expense is $20 per policy, and fixed expense is $2,500. Find expected total cost. (Answer: $18,500)
There are 200 prospects, each with a 0.25 purchase probability. For a bound policy, claim probability is 0.10 and conditional claim severity is $1,800. Variable servicing expense is $30 per bound policy, and fixed expense is $4,000. Find expected total cost. (Answer: $14,500)
A hypothetical price is $500. Of 80 prospects, 50% are expected to bind. Expected claims per bound policy are $240, issue expense is $18, and commission is 5% of premium. Fixed expense is $6,000. Find expected total cost. (Answer: $17,320)
A portfolio has 150 Group A and 100 Group B prospects. At the selected price, binding probabilities are 0.40 for A and 0.20 for B. Expected claims per bound policy are $250 for A and $700 for B; per-policy expenses are $25 for A and $40 for B. Fixed expense is $3,000. Find expected total cost. (Answer: $34,300)
At alternative 1, 100 policies are expected to bind; at alternative 2, 60 are expected to bind. Expected claims plus per-policy variable expense are $360 for either alternative. Fixed expense is $18,000 under both. Find each total expected cost and the allocated fixed expense per expected policy. (Answer: Alternative 1, total $54,000 and allocation $180 per policy; alternative 2, total $39,600 and allocation $300 per policy.)

What You'll Learn