Reinsurance Program Objectives and Constraints

From Program Goals to Tests

A reinsurance program is designed to meet a purpose, not simply to purchase a recovery. An insurer may want to reduce earnings volatility, limit the chance that annual claims exceed available resources, or cap losses in a severe scenario. Those aims become useful in program design only when they are translated into measurable objectives and compared with feasibility constraints.

An objective describes a desired result, such as “keep the probability of net claims above $10 million below 10%.” A constraint is a limit the program must obey, such as “annual premium may not exceed $1.5 million” or “no reinsurer may provide more than $6 million of treaty capacity.” A proposal can meet a risk objective and still be infeasible because it costs too much or exceeds available capacity.

Definition: A reinsurance program objective is a measurable target for the insurer’s financial outcomes after reinsurance. A program constraint is a required limit on the program’s cost, structure, capacity, counterparty exposure, or other practical feature. A proposal is feasible only if it satisfies every applicable hard constraint.

State the outcome being measured, its time period, and its threshold. For example, “reduce severe losses” is not yet a test. “Keep the maximum net loss in the specified scenario set at or below $12 million” is. Similarly, an earnings-stability target needs a defined cost measure and a probability model, not just a preference for “less volatility.”

Choosing Measures That Match the Objective

For a one-year program, let $R$ be annual retained claims and let $P$ be the fixed reinsurance premium. In a simplified model with no other expenses, taxes, or investment income, define net annual claims cost as $C=R+P$. If the premium varies with losses, or if the model includes reinstatement premiums or other contingent costs, those amounts must instead be included in $C$ as random variables.

Earnings stability can be expressed using $\operatorname{Var}(C)$ or the standard deviation $\sqrt{\operatorname{Var}(C)}$. The standard deviation is often easier to interpret in the same units as annual cost. A solvency-protection objective can be stated as a probability that net loss exceeds a specified resource threshold. A peak-loss objective can be stated as a cap on the maximum net loss across an explicitly defined scenario set. These measures answer different questions; one should not be substituted for another without justification.

\text{Stability: } \sqrt{\operatorname{Var}(C)} \leq s_{\max} \qquad \text{Solvency: } P(C>K) \leq \varepsilon \qquad \text{Peak loss: } \max_{j \in \mathcal{S}} C_j \leq M

Here, $s_{\max}$ is a stated volatility limit, $K$ is the loss threshold of interest, $\varepsilon$ is the allowed exceedance probability, and $\mathcal{S}$ is a specified set of scenarios. The final expression is a scenario test, not a claim about every theoretically possible loss unless the scenario set covers all possible outcomes.

Worked Example: Expected Cost and Peak Cost

A hypothetical insurer models annual retained claims, in millions of dollars, as 1 with probability 0.70, 4 with probability 0.20, and 9 with probability 0.10. A proposed treaty has a fixed annual premium of $0.45 million. Ignore expenses and other cash flows. Calculate expected net annual cost and the largest modeled net cost.

Method. Add the fixed premium to retained claims in each outcome, then take the probability-weighted average. Because the premium is fixed, it is added once to the expected retained claims.

The expected retained claims are

$$ E[R]=0.70(1)+0.20(4)+0.10(9)=0.70+0.80+0.90=2.40. $$

Therefore, expected net annual cost is

E[C]=E[R]+P=2.40+0.45=2.85 \text{ million dollars}.

The possible net costs are $1+0.45=1.45$, $4+0.45=4.45$, and $9+0.45=9.45$ million dollars. The largest modeled net cost is therefore $9.45 million. As a check, the probabilities sum to $0.70+0.20+0.10=1.00$, and the largest net cost corresponds to the largest retained-claims outcome.

Common mistake avoided: Adding the premium to the wrong expected claims value can produce an incorrect expected cost. The probability-weighted claims calculation is $2.40 million, so the correct total is $2.85 million, not $2.65 million. The premium is also added to the maximum outcome, giving $9.45 million.

Why a Fixed Premium Does Not Change Cost Variance

A fixed premium increases the level of annual cost but does not create variability in that cost. To see why, suppose $P$ is a constant. By the definition of variance,

\begin{aligned} \operatorname{Var}(C) &=E\left[(C-E[C])^2\right]\\ &=E\left[(R+P-(E[R]+P))^2\right]\\ &=E\left[(R-E[R])^2\right]\\ &=\operatorname{Var}(R). \end{aligned}

The premium cancels when each outcome is measured relative to the mean. This result does not say that a treaty has no effect on volatility: the treaty may change the distribution of retained claims $R$. It says only that adding a fixed amount to every outcome leaves variance unchanged. If the premium depends on losses, that cancellation does not apply.

Worked Example: Testing an Earnings-Stability Target

In a hypothetical program, net annual cost $C$, in millions of dollars, is 2.80 with probability 0.80, 5.80 with probability 0.15, and 10.80 with probability 0.05. These outcomes include a fixed premium of $0.40 million; for example, the corresponding retained claims are 2.40, 5.40, and 10.40. The insurer requires the standard deviation of annual cost to be no more than $2.00 million. Determine whether the target is met.

Method. Calculate the mean, then the probability-weighted squared deviations from that mean. The probabilities sum to $0.80+0.15+0.05=1.00$.

First, the mean is

$$ E[C]=0.80(2.80)+0.15(5.80)+0.05(10.80) =2.24+0.87+0.54=3.65. $$

The squared deviations and weighted contributions are

\begin{aligned} (2.80-3.65)^2&=(-0.85)^2=0.7225, &0.80(0.7225)&=0.5780,\\ (5.80-3.65)^2&=(2.15)^2=4.6225, &0.15(4.6225)&=0.693375,\\ (10.80-3.65)^2&=(7.15)^2=51.1225, &0.05(51.1225)&=2.556125. \end{aligned}

Thus,

\operatorname{Var}(C)=0.5780+0.693375+2.556125=3.8275, \qquad \sqrt{\operatorname{Var}(C)}=\sqrt{3.8275}\approx1.9564.

Rounded to four decimal places, the standard deviation is $1.9564 million. Since $1.9564 \leq 2.00$, the stability target is met. The variance is in squared millions of dollars; the standard deviation returns to millions of dollars.

Common mistake avoided: A large squared deviation in the low-probability outcome must still be weighted by its probability. The last contribution is $0.05(51.1225)=2.556125$, not $0.05$ or $51.1225$. Omitting the probability weight materially changes the variance.

Solvency, Peak Loss, and Feasibility

A solvency test starts by defining the relevant loss and threshold. In a simplified one-year test, $P(C>K)$ is the probability that net claims cost exceeds a specified amount $K$. The test requires a modeled distribution or a set of probability-weighted outcomes. A peak-loss test instead examines the largest outcome among specified scenarios. It is sensitive to which scenarios are included, so the scenario set and loss definition should be stated explicitly.

Program constraints often include several independent limits. A budget constraint limits the premium; a capacity constraint requires purchased treaty limits not to exceed the capacity actually available; and a counterparty constraint limits how much capacity or exposure is assigned to any one reinsurer. Meeting the budget does not imply meeting the capacity or counterparty limits.

Define the objective measure.
Specify the cost or loss, the time horizon, and the scenarios or probability model.

State a threshold.
Write a numerical target, such as a maximum exceedance probability, standard deviation, or scenario loss.

List hard constraints separately.
Record the premium budget, available treaty capacity, and any reinsurer-specific limits.

Test every requirement.
Report each result as pass or fail; do not treat success on one criterion as evidence that all others are satisfied.

Worked Example: Testing a Proposal Against Multiple Constraints

A hypothetical insurer models gross annual claims, in millions of dollars, as 6 with probability 0.60, 12 with probability 0.30, and 20 with probability 0.10. A proposed program has an annual premium of $1.40 million and total treaty capacity of $10 million. The modeled total recoveries in the three outcomes are 0, 3, and 9 million, respectively. The insurer requires premium no greater than $1.50 million, net cost no greater than $11 million in any modeled outcome, and probability of net cost exceeding $10 million no greater than 0.10. Reinsurer R1 can provide at most $6 million of treaty capacity and R2 at most $5 million; the proposal allocates $6 million to R1 and $4 million to R2. Test feasibility.

Method. Subtract the stipulated recovery from gross claims in each outcome, add the fixed premium, and test each risk objective and hard limit separately.

The three net costs are

\begin{aligned} C_1&=6-0+1.40=7.40,\\ C_2&=12-3+1.40=10.40,\\ C_3&=20-9+1.40=12.40. \end{aligned}

The peak modeled net cost is $12.40 million, which exceeds the $11 million limit. The probability that cost exceeds $10 million is $0.30+0.10=0.40$, because both $10.40 million and $12.40 million exceed the threshold. This is greater than the permitted 0.10.

The premium test passes because $1.40 \leq 1.50$. The capacity allocations also pass: R1 provides $6 million, equal to its maximum, and R2 provides $4 million, below its $5 million maximum. Total capacity is $6+4=10$ million, matching the proposal. Nevertheless, the proposal is not feasible because it fails both stated risk objectives. Passing the financial and capacity constraints alone is not enough.

Common mistake avoided: The exceedance probability uses a strict “exceeds $10 million” test. Both $10.40 million and $12.40 million qualify, so the probability is 0.40, not 0.10. Also, the peak test uses the full net cost, including premium.

Common Design and Exam Traps

Using a vague objective. “Improve stability” is not testable until the cost measure, time horizon, and target are specified.
Confusing average with protection in a severe outcome. A lower expected cost does not establish a lower peak cost or a smaller probability of exceeding a threshold.
Mixing units or periods. Compare annual premiums with annual limits and keep dollars, thousands, or millions consistent throughout the calculation.
Ignoring probabilities in a solvency test. Identify every modeled outcome above the threshold and add its probability.
Combining a fixed premium incorrectly. It increases expected cost and each outcome’s cost, but it does not change variance when it is truly fixed.
Stopping after one constraint passes. A program can satisfy its budget and available capacity while failing a risk objective or a counterparty limit.

Key takeaway: Translate each program goal into a defined measure and numerical target, then test it alongside—not instead of—the budget, capacity, and counterparty constraints. A proposal is acceptable only when every required test is satisfied under the stated model and assumptions. This framework is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Check Your Understanding

For each hypothetical problem, identify the relevant measure and check the stated target or constraint.

A treaty has a fixed annual premium of $0.50 million. Retained claims are $2 million with probability 0.50, $6 million with probability 0.30, and $10 million with probability 0.20. Calculate expected net annual cost and the maximum modeled net cost. (Answer: Expected cost $5.30 million; maximum cost $10.50 million.)
Net annual cost, in millions, is 1 with probability 0.50, 4 with probability 0.30, and 7 with probability 0.20. Calculate its mean and standard deviation, then test a requirement that standard deviation be no more than $2.40 million. (Answer: Mean $3.10 million; standard deviation approximately $2.3431 million; target passes.)
After reinsurance, net claims are $5 million with probability 0.70, $11 million with probability 0.20, and $16 million with probability 0.10. The solvency target is $P(\text{net claims}>10)\leq0.25$. Does it pass? (Answer: Exceedance probability $0.20+0.10=0.30$; target fails.)
A proposal costs $1.10 million, against a $1.20 million budget, and needs $9 million of capacity. Reinsurer A can provide at most $5 million and Reinsurer B at most $4 million. The counterparty limit for A is $4.50 million. Does the proposal satisfy all stated constraints if it assigns $5 million to A and $4 million to B? (Answer: No; budget and total capacity pass, but A’s $5 million allocation exceeds its $4.50 million counterparty limit.)
A program has net cost of $8 million with probability 0.80 and $13 million with probability 0.20. The peak-loss target is at most $12 million, and the probability of cost exceeding $10 million must be no more than 0.15. Which tests pass? (Answer: Neither; peak cost is $13 million and exceedance probability is 0.20.)

What You'll Learn