From Analysis to a Defensible Decision
A risk transfer analysis is not complete when it produces a capital-relief estimate. A decision-maker needs to know what the organization is trying to achieve, which assumptions drive the estimates, whether the alternatives meet the relevant constraints, and why one feasible action is preferred. A recommendation that reports only the largest relief figure can be misleading: that action may cost more, leave an unacceptable stress exposure, or depend on a transfer mechanism that performs poorly in an adverse scenario.
This tutorial develops a practical way to turn quantitative results into a decision. The examples use hypothetical amounts in millions of dollars. Their figures are illustrative, not observations about any actual insurer or market. The methods are aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Begin by separating hard constraints from preferences. A hard constraint is a condition an action must satisfy, such as a minimum capital reduction, maximum annual cost, or maximum residual stress loss. A preference helps choose among actions that satisfy the constraints, such as lower cost, more stable protection, or less dependence on a single counterparty. This distinction prevents a favorable score on one measure from concealing a failure on a non-negotiable requirement.
Define Comparable Measures Before Ranking
State the reference position and the post-action position on the same basis. For example, let \(K_0\) be modeled capital before an action and \(K_1\) modeled capital after it. Define modeled capital relief as the decrease in that measure:
Likewise, if \(L_0\) is the loss under a specified stress before the action and \(L_1\) is the corresponding residual loss after it, the stress-loss reduction is \(L_0-L_1\). These measures answer different questions. Capital relief describes a change in a specified capital measure; stress-loss reduction describes a change in a specified stress outcome. Neither should be substituted for the other.
The subtraction is justified by measuring the same quantity in the same units under the same scenario or model basis. If \(K_0\) and \(K_1\) are both in millions of dollars and use the same capital definition, their difference is the amount by which modeled capital decreases. Comparing a pre-action rating-agency measure with a post-action regulatory measure would not support that interpretation. The comparison must first be made consistent.
Similarly, state what “cost” includes and over what period it is measured. A total annual cost might include premiums or fees and specified operating or financing expenses. If retained losses or a risk charge are included, say so. Do not compare an all-in annual cost for one alternative with a quoted premium alone for another. Record any material features not captured in the numerical measures, such as collateral terms, counterparty exposure, liquidity timing, or basis risk.
Specify the target, measurement basis, time horizon, and any limits that an acceptable action must meet.
Use consistent definitions for capital, stress losses, cost, and the scenarios used to measure them.
Identify which alternatives satisfy every hard constraint. Do not allow a strong result on one measure to offset a failed hard limit unless the decision rule explicitly permits that trade.
Vary important assumptions, identify changes in feasibility or ranking, and state why the selected action remains preferable—or what condition would change the recommendation.
Worked Example: Apply Hard Limits Before Choosing
A hypothetical insurer compares a treaty with retaining the risk. The modeled rating capital before the treaty is $160 million and after it is $116 million. A specified stress loss is $130 million before and $94 million after the treaty. The treaty’s total annual cost, measured on the same stated basis as the alternatives, is $12.10 million. The decision criteria require at least $40 million of capital relief, at least $40 million of stress-loss reduction, and annual cost no greater than $12 million. Is the treaty feasible?
Method. Calculate each measure separately, then test all three constraints. This avoids treating capital relief as if it were stress-loss reduction.
Capital relief. Subtract post-treaty capital from pre-treaty capital:
The capital-relief test is passed because \(44 \ge 40\).
Stress-loss reduction. Subtract the residual stress loss from the pre-treaty stress loss:
The stress test is failed because \(36 < 40\).
Cost. The annual cost is $12.10 million, which exceeds the $12 million limit by \(12.10-12.00=0.10\) million dollars, or $100,000.
Decision. No. The treaty is not feasible under the stated criteria: it passes the capital-relief test but fails both the stress-loss and cost tests. The common mistake is to see $44 million of capital relief and conclude that the treaty meets the risk-reduction requirement. The stress calculation shows that it does not.
Rank Only the Alternatives That Meet the Requirements
Once infeasible alternatives have been identified, make the preference among feasible alternatives explicit. If the objective is to minimize annual cost subject to minimum capital and stress reductions, the least-cost feasible action is preferred on the supplied measures. If decision-makers instead require a margin above those minimums, that additional requirement must be stated; it should not be introduced after reviewing the results to justify a preferred option.
A ranking is conditional on its assumptions. In particular, modeled relief may depend on how the action is recognized in a capital model, and stress-loss reduction may depend on whether the transfer responds to the actual loss. The recommendation should identify assumptions that could materially change feasibility or ranking. It should also distinguish quantified results from qualitative considerations instead of disguising a judgment as a calculation.
Worked Example: Choose Among Feasible Alternatives
A hypothetical insurer has modeled capital of $240 million before any action. It requires at least $40 million of capital relief, a reduction of at least $40 million in a specified stress loss, and annual total cost no greater than $15 million. The stress loss before action is $190 million. Alternative R is reinsurance; alternative C is a cat bond; and alternative P is a portfolio change. Their modeled post-action results are:
| Alternative | Post-action capital | Residual stress loss | Annual total cost |
|---|---|---|---|
| R | $188 million | $142 million | $14.20 million |
| C | $196 million | $138 million | $15.60 million |
| P | $198 million | $147 million | $12.80 million |
Assume all costs use the same annual basis, and the stated capital and stress figures already reflect the modeled effects of each action. The objective is to minimize annual total cost among alternatives that meet all three constraints.
Step 1: Calculate capital relief.
Each alternative meets the $40 million minimum capital-relief requirement.
Step 2: Calculate stress-loss reduction.
Each also meets the $40 million stress-reduction requirement.
Step 3: Apply the cost limit. R costs $14.20 million and P costs $12.80 million, so both meet the $15 million limit. C costs $15.60 million, so C fails. Its stronger stress reduction does not remove its cost-limit failure.
Step 4: Choose among feasible actions. Of R and P, P has the lower annual cost: \(14.20-12.80=1.40\) million dollars less. Therefore, under the stated objective and measures, P is preferred. R provides $10 million more capital relief than P (\(52-42=10\)) and $5 million more stress-loss reduction (\(48-43=5\)), but costs $1.40 million more annually. If the organization requires a larger buffer above the minimum reductions, that preference could change; no such buffer is part of the stated criteria. The common mistake is to select C solely because its stress-loss reduction is largest, without checking that C exceeds the hard cost limit.
Sensitivity and the Strength of the Recommendation
A recommendation should test assumptions that could plausibly change the result. A sensitivity is especially decision-relevant when it can turn a feasible option into an infeasible one, or reverse the ranking of feasible options. Examples include weaker-than-expected risk reduction, a higher price, lower capital recognition, delayed cash flows, or a stress event that exposes basis risk. The sensitivity should change an identified assumption while keeping the comparison interpretable; changing several unrelated assumptions at once may make the cause of a changed result unclear.
For a simple proportional sensitivity, suppose only a fraction \(q\) of the modeled relief and stress reduction is achieved, while the pre-action measures and cost remain unchanged. Then the adjusted relief is \(q\) times modeled relief, and the adjusted stress reduction is \(q\) times modeled stress reduction. This is not a universal model of transfer performance: it is an explicit simplifying assumption for the sensitivity. A real analysis should use a scenario that reflects the risk being tested.
Worked Example: Test Whether a Preferred Action Is Robust
Use the alternatives and criteria from the preceding example. P was preferred on base assumptions, with $42 million of capital relief, $43 million of stress-loss reduction, and $12.80 million annual cost. R had $52 million of capital relief, $48 million of stress-loss reduction, and $14.20 million annual cost. Now test a hypothetical sensitivity in which only 85% of each alternative’s modeled capital relief and stress-loss reduction is achieved. Costs do not change. Does P remain feasible, and which alternative meets the constraints under this sensitivity?
Method and assumption. Multiply each modeled reduction by \(q=0.85\). Keep the $240 million pre-action capital, $190 million pre-action stress loss, and cost figures fixed. This isolates the effect of weaker risk reduction.
Alternative P. Its adjusted capital relief is \(42 \times 0.85=35.70\) million dollars, giving adjusted post-action capital of \(240-35.70=204.30\) million dollars. Its adjusted stress-loss reduction is \(43 \times 0.85=36.55\) million dollars, giving residual stress loss of \(190-36.55=153.45\) million dollars. Its unchanged cost is $12.80 million. P meets the cost limit but fails both reduction requirements because \(35.70 < 40\) and \(36.55 < 40\).
Alternative R. Its adjusted capital relief is \(52 \times 0.85=44.20\) million dollars, giving adjusted post-action capital of \(240-44.20=195.80\) million dollars. Its adjusted stress-loss reduction is \(48 \times 0.85=40.80\) million dollars, giving residual stress loss of \(190-40.80=149.20\) million dollars. Its unchanged cost is $14.20 million. R passes all three requirements: \(44.20 \ge 40\), \(40.80 \ge 40\), and \(14.20 \le 15\).
Conclusion. P is not feasible under the sensitivity, whereas R remains feasible. If this sensitivity is considered decision-relevant, the recommendation should change from P to R, or state that P is preferred only if the base assumptions are accepted and R is the more robust feasible alternative. The common mistake is to apply the 85% factor to post-action capital rather than to the modeled reduction. Under the stated assumption, the adjustment applies to relief; subtract the adjusted relief from the unchanged pre-action capital.
Document the Trade-Off and the Conditions
A concise recommendation can be organized so that a reviewer can reconstruct the reasoning without relying on unstated judgment. It should identify the action and decision objective; state the key modeling and cost assumptions; show the principal before-and-after measures; identify constraints passed and failed; summarize the most decision-relevant sensitivity; and explain the trade-off behind the preference. If the recommendation depends on a condition, such as a minimum level of modeled relief or a maximum price, make that condition explicit.
For example, the recommendation might prefer a lower-cost portfolio change on base assumptions, while noting that its relief falls below the required minimum when only 85% of modeled reductions are achieved. If the decision-maker considers that sensitivity plausible, the more expensive alternative that remains feasible may be preferable. This is not inconsistency: it is a conditional conclusion based on a changed assumption. The analysis should also record limitations, such as unmodeled execution risk or an assumed transfer response, rather than implying that the numerical comparison captures every consequence.
Common Recommendation Errors
- Reporting gross relief as the decision result. Use the specified post-action measure and explain any material effects included or excluded from it.
- Confusing capital relief with stress reduction. Calculate each against its own consistent before-and-after measure.
- Ranking before screening. First identify failures of hard constraints; then compare the feasible alternatives according to the stated objective.
- Comparing inconsistent costs. State the cost basis, period, and included components for every alternative.
- Calling a base-case result robust without testing it. Show whether plausible sensitivities change feasibility, ranking, or the conditions attached to the recommendation.
- Hiding judgment in a score. Explain the priorities and trade-offs directly, particularly where a qualitative consideration affects the choice.
Check Your Understanding
For each hypothetical decision, apply the stated criteria and show the key calculations before checking the answer.
- A treaty reduces modeled capital from $210 million to $164 million and a specified stress loss from $175 million to $131 million. Its annual cost is $9.50 million. Criteria require capital relief of at least $40 million, stress-loss reduction of at least $45 million, and cost no greater than $10 million. Is it feasible? (Answer: No. Capital relief is $46 million and cost is within the limit, but stress-loss reduction is \(175-131=$44\) million, below $45 million.)
- Before action, modeled capital is $300 million and stress loss is $220 million. Criteria require at least $55 million capital relief, at least $50 million stress-loss reduction, and annual cost at most $18 million. A has post-action capital of $240 million, residual stress loss of $164 million, and cost of $17.50 million. B has corresponding values of $238 million, $170 million, and $18.20 million. Which alternative is feasible? (Answer: A only. A provides $60 million relief and $56 million stress reduction and costs $17.50 million. B provides $62 million relief and $50 million stress reduction but exceeds the cost limit.)
- In the preceding problem, suppose only 80% of A’s modeled capital relief and stress-loss reduction is achieved, with pre-action amounts and cost unchanged. Does A still meet the requirements? (Answer: No. Adjusted relief is \(60 \times 0.80=$48\) million and adjusted stress reduction is \(56 \times 0.80=$44.80\) million; both are below their required amounts.)
- A hypothetical transfer has modeled post-action capital of $142 million from a $180 million baseline. A $9 million counterparty adjustment and a $6 million collateral-related capital addition must also be included. Before action, stress loss is $120 million; modeled gross recovery under stress is $30 million, but a basis-risk reduction of $8 million and a counterparty nonpayment reduction of $5 million apply. Requirements are at least $25 million capital relief and $20 million stress-loss reduction. Is the transfer feasible on these adjusted figures? (Answer: No. Adjusted post-action capital is \(142+9+6=$157\) million, so relief is \(180-157=$23\) million. Net stress-loss reduction is \(30-8-5=$17\) million. Both are below the requirements.)
- Two actions each satisfy all base-case limits. Action M costs $13.60 million annually and gives $48 million of capital relief and $46 million of stress reduction. Action N costs $12.90 million and gives $43 million of capital relief and $42 million of stress reduction. The minimum reductions are $40 million each and cost is capped at $15 million. The objective is lowest cost, but a sensitivity retains only 90% of each modeled reduction. Which action is preferred on base assumptions, and which remains feasible under the sensitivity? (Answer: N is preferred on base assumptions because it is feasible and costs $0.70 million less. Under sensitivity, M has $43.20 million relief and $41.40 million stress reduction and remains feasible; N has $38.70 million relief and $37.80 million stress reduction and fails both reduction limits.)