Pricing Is a Decision Under Uncertainty
A price is not adequate merely because it covers an expected claim cost in a spreadsheet. The price may affect how many customers buy, which customers choose to buy, how competitors respond, and whether operations can deliver the product at the assumed cost. A pricing decision should therefore connect the proposed price to those uncertainties and specify how the decision will be reviewed as experience emerges.
The analysis here concerns a hypothetical product and uses contribution as a decision measure. It does not replace applicable actuarial pricing standards, governance, or other required analyses. The aim is to make assumptions visible: identify which uncertainties can change the pricing conclusion, test their effects, and define what evidence would require a response. This decision discipline is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Let \(p_s\) be the premium per policy in scenario \(s\), \(N_s\) the number of policies, \(C_s\) the expected claim cost per policy, \(E_s\) the variable expense per policy, and \(F_s\) the fixed cost. For a one-period analysis, assume all quantities are measured over the same period and on a consistent basis. Scenario contribution is
If scenario \(s\) has probability \(q_s\), with \(\sum_s q_s=1\), the expected contribution is \(\sum_s q_s\Pi_s\). The scenarios should represent coherent outcomes: for example, a change in price might affect both volume and the risk characteristics of customers who accept the offer.
Why Volume and Cost Must Be Considered Together
A tempting shortcut is to multiply average volume by average margin per policy. That shortcut is valid only under suitable conditions, such as when volume and unit cost are independent and the relevant averages are formed consistently. When higher volume occurs in a scenario with higher claim costs, using separate averages can give the wrong expected contribution.
The reason follows directly from the scenario formula. Expected variable contribution is \(E[N(p-C-E)]\), not automatically \(E[N](p-E[C]-E[E])\). For a fixed premium, the claim-cost component includes \(E[NC]\). In general,
This identity follows from the definition \(\operatorname{Cov}(N,C)=E[NC]-E[N]E[C]\). Unless the covariance is zero, multiplying the two averages omits their dependence. Scenario-by-scenario calculation makes that dependence explicit.
Worked Example: Comparing Demand and Risk-Mix Scenarios
A hypothetical insurer is considering an annual premium of $1,200 for a small commercial policy. Its one-year planning model has two scenarios. The normal scenario has probability 0.70, 400 policies, expected claims of $760 per policy, and variable expenses of $140 per policy. The adverse selection scenario has probability 0.30, 300 policies, expected claims of $900 per policy, and variable expenses of $150 per policy. Fixed launch costs are $80,000 in either scenario. Calculate expected contribution and assess the result against a hypothetical approval rule requiring expected contribution of at least $15,000 and no scenario worse than a $40,000 loss.
Method. Calculate contribution in each scenario before weighting by its probability. This preserves the modeled relationship between volume and claim cost.
In the normal scenario, contribution per policy is \(1,200-760-140=300\) dollars. Total contribution before fixed costs is \(400(300)=120,000\) dollars, so
In the adverse selection scenario, contribution per policy is \(1,200-900-150=150\) dollars. Total contribution before fixed costs is \(300(150)=45,000\) dollars, so
Weight the scenario results by their probabilities:
The expected contribution exceeds the $15,000 hurdle, and the worst modeled scenario is a $35,000 loss, within the $40,000 limit. The proposal meets this hypothetical rule. That is not a guarantee of a favorable outcome: the conclusion depends on the scenarios, probabilities, and cost estimates being suitable for the decision.
Common mistake and correction. A shortcut is to use average volume \(0.70(400)+0.30(300)=370\), average claim cost \(0.70(760)+0.30(900)=802\), and average variable expense \(0.70(140)+0.30(150)=143\). That gives \(370(1,200-802-143)-80,000=\$14,350\), not the scenario-weighted result. The shortcut combines averages as though volume and unit costs could be multiplied independently. Calculate the contribution in each joint scenario, then apply its probability.
Check. The probability-weighted contributions are $28,000 and −$10,500; their sum is $17,500, consistent with the approval-rule comparison.
Test the Price Against Customer and Competitor Responses
The price and customer response should be evaluated together. A higher price can improve contribution per policy but reduce the number of policies sold. The effect on the mix of buyers can also matter: the customers who remain may not have the same expected costs as those who leave. Competitive reactions can further change sales or require a different market position.
A demand estimate is a model, not a promise. It should be used within the range and conditions for which it was developed. Where a pricing recommendation depends on a particular response, compare the recommendation with plausible alternatives and identify what evidence would show that the response assumption is failing.
Worked Example: Price Choice Under a Competitive Stress
A hypothetical insurer estimates annual demand for a product using \(N(p)=900-0.5p\), where \(p\) is the annual premium in dollars and \(N(p)\) is expected policy count within the proposed price range. Expected claims are $720 per policy, variable expenses are $130 per policy, and fixed costs are $60,000 per year. Compare premiums of $1,000 and $1,100 under the base model. Then test the $1,100 price under a combined competitive and claims stress: a competitor response reduces expected volume by 15%, and expected claims per policy rise by 8%. Assume variable expenses and fixed costs do not change under the stress.
Method. Calculate expected volume and contribution under the base assumptions, then change only the assumptions specified in the stress. The demand model is assumed applicable to these two prices; it should not be extrapolated beyond its proposed range without support.
At a $1,000 premium, expected volume is \(900-0.5(1,000)=400\) policies. Unit contribution before fixed costs is \(1,000-720-130=150\) dollars. Annual contribution is \(400(150)-60,000=\$0\).
At a $1,100 premium, expected volume is \(900-0.5(1,100)=350\) policies. Unit contribution is \(1,100-720-130=250\) dollars. Annual contribution is \(350(250)-60,000=\$27,500\). Under the base model, the $1,100 price has the higher contribution.
Under the stress, expected volume is \(350(1-0.15)=297.5\) policies. The stressed claim cost is \(720(1+0.08)=777.60\) dollars per policy. Unit contribution is therefore \(1,100-777.60-130=\$192.40\). Stressed annual contribution is
The price ranks better under the base model but produces a loss under the specified combined stress. This result does not establish that the stress will occur; it shows that the recommendation is sensitive to the assumptions. An approval could therefore be conditional on testing competitor response and early claim experience, with a predefined review if observed results depart materially from the assumptions.
Common mistake avoided. Choosing $1,100 solely because its base contribution is $27,500 ignores that competition can reduce volume while a changed risk mix can increase claims. The correction is to state and calculate the joint stress rather than treating the volume and cost assumptions as unrelated.
Check. In the stress, unit contribution is $192.40 and expected volume is 297.5; their product is $57,239, which is $2,761 below fixed costs.
Make Approval Conditions Operational
Pricing approval can be more useful when it specifies not only the price, but also the conditions under which the price is acceptable. Conditions should connect to an assumption that materially affects the result. For example, if the business case depends on an implementation cost estimate, an approval gate can require a confirmed vendor quote or limit the initial commitment until costs are verified. If it depends on customer take-up, a limited launch can provide evidence before broader expansion.
A useful approval condition is observable, assigned to a decision owner, and linked to an action. “Monitor costs” is difficult to act on. “Reconcile actual implementation costs to the approved estimate before expanding the launch; return for review if the approved cost allowance is exceeded” specifies an evidence source, timing, and response.
List the customer, cost, competitor, and execution assumptions that can change whether the price meets the decision criteria.
Use consistent periods and definitions; vary the assumptions that could plausibly move together as well as those that can move separately.
State any prerequisite, launch limit, or threshold required before approval or expansion.
Specify who will compare experience with assumptions, when the comparison occurs, and what decision follows a material variance.
Worked Example: Setting an Execution Gate and Review Threshold
A hypothetical product is priced at $1,500 per annual policy. The plan assumes 600 policies, expected claims of $900 per policy, variable administration and acquisition expenses of $180 per policy, and fixed implementation costs of $90,000. A launch review also tests a combined adverse execution case: only 480 policies are issued, fixed implementation costs rise by $45,000, and expected claims per policy rise by 10%. Variable expenses per policy remain $180. Calculate base and stressed contribution and determine the minimum policy count for nonnegative contribution under the stressed unit costs and fixed costs.
Method. First calculate unit contribution and total contribution under each set of assumptions. For the stressed break-even count, divide stressed fixed costs by stressed contribution per policy; because policies are whole units, round the minimum count up.
In the base case, unit contribution is \(1,500-900-180=\$420\). Contribution before fixed costs is \(600(420)=\$252,000\), so base contribution is \(252,000-90,000=\$162,000\).
In the combined adverse case, expected claims per policy are \(900(1.10)=\$990\), and unit contribution is \(1,500-990-180=\$330\). Fixed costs are \(90,000+45,000=\$135,000\). At 480 policies, contribution is \(480(330)-135,000=158,400-135,000=\$23,400\).
For nonnegative contribution under the stressed unit contribution, the policy count must satisfy \(330N-135,000\geq 0\). Thus \(N\geq135,000/330=409.09\ldots\). The minimum whole-number count is 410 policies. At 410, contribution is \(410(330)-135,000=\$300\); at 409, it is \(409(330)-135,000=-\$30\). This verifies the threshold.
A proportionate approval condition could require confirmation of the implementation-cost allowance before full launch and a review of take-up and claims during the initial period. If experience indicates that the assumptions are unlikely to support at least 410 policies under the stressed cost basis, the decision owner should reassess expansion or the price rather than treating the base-case contribution as assured.
Check. The base and stressed contributions are positive at $162,000 and $23,400. Under the stated stress costs, 410 policies is the smallest whole-number volume that avoids a loss.
Review the Assumptions After Launch
A pricing review should compare like with like. For example, actual policies issued over six months should not be compared directly with an annual volume assumption without adjusting for the period and any seasonal pattern. Costs should use consistent definitions: claim payments, claim estimates, and expenses may be recorded on different bases. A difference that appears favorable at an early date can reverse as claims develop or expenses emerge.
Useful review measures follow the drivers of the original decision. They can include take-up or policy count, observed customer and risk mix, claims per policy, variable expense per policy, competitor actions, and implementation cost. The review should distinguish ordinary fluctuation from a change that challenges the original model, without treating a single observation as conclusive.
- Reconcile the basis. Match the observed period, exposure, and cost definitions to those used in the pricing analysis.
- Compare drivers, not only total profit. A result near plan can conceal offsetting errors, such as lower volume and lower claims cost.
- Use pre-agreed triggers. Set materiality thresholds or decision dates before results are known, where practicable.
- Specify a response. Possible actions include further investigation, a pricing reassessment, a limit on new business, or a return to the approval authority.
The central pitfall is treating a forecast as a fact. A defensible pricing decision explains how the price performs under its stated assumptions, which assumptions matter most, what conditions support approval, and how new evidence can change the decision. The aim is not to eliminate uncertainty, but to make its consequences visible and manageable.
Check Your Understanding
For each hypothetical case, calculate the requested result using the assumptions stated. Keep scenario outcomes together when volume and cost assumptions differ.
- A product has a $900 premium and fixed costs of $20,000. In scenario A, probability 0.50, volume is 200, claims are $500 per policy, and variable expenses are $100 per policy. In scenario B, probability 0.50, volume is 120, claims are $700 per policy, and variable expenses are $100 per policy. What is expected contribution? (Answer: $16,000)
- A hypothetical policy has a $900 premium, $100 variable expense per policy, and $25,000 fixed costs. With probability 0.60, volume is 300 and expected claims are $600 per policy; with probability 0.40, volume is 100 and expected claims are $1,000 per policy. What is expected contribution? (Answer: $3,000)
- Expected annual demand is \(N(p)=500-0.2p\), where \(p\) is the annual premium in dollars, for the two prices considered. Expected claims are $550 per policy, variable expenses are $0 per policy, and fixed costs are $30,000. Calculate base contribution at $1,000 and $1,200 and identify which price has the higher modeled contribution. (Answer: $105,000 at $1,000; $139,000 at $1,200; $1,200 has the higher modeled contribution)
- A hypothetical product has a $1,400 premium, 250 policies, $850 expected claims per policy, $150 variable expenses per policy, and $50,000 fixed costs. Recalculate contribution if expected claims rise by 12% and all other assumptions remain unchanged. (Answer: $24,500)
- A product has an $800 premium, expected volume of 400, claims of $520 per policy, variable expenses of $80 per policy, and fixed costs of $60,000. In a launch stress, volume is 15% lower and fixed costs are $12,000 higher; unit claims and expenses are unchanged. What is stressed contribution? (Answer: −$4,000)