Why Scenario Governance Matters
A stress scenario can be analytically sophisticated and still produce a misleading decision if no one is accountable for its assumptions, if a claimed management action cannot be implemented, or if approval occurs without independent scrutiny. Scenario governance is the set of responsibilities, controls, records, and review procedures that make a scenario’s purpose, construction, use, and limitations transparent.
Governance does not mean that every scenario must pass through an identical committee or that challenge must always produce a different estimate. The process should be proportionate to the scenario’s materiality, uncertainty, and intended use. A scenario used to inform a significant capital decision generally warrants more formal review and evidence than an exploratory sensitivity used for discussion. The objective is a result that decision-makers can understand, question, reproduce, and update.
Roles and Decision Rights
Good governance separates responsibilities that can otherwise become blurred. A scenario owner coordinates its purpose, scope, and lifecycle. Subject-matter experts supply and explain evidence in their areas. Analysts implement the scenario and produce traceable results. An approver decides whether the scenario is fit for its stated use, considering the review and any unresolved limitations. Independent challengers test the rationale rather than merely checking that calculations were performed. These roles can be arranged differently across organizations, but accountability for each function should be explicit.
A challenger need not build a competing model or reject every assumption. The challenge should be sufficiently independent to ask whether the scenario is coherent, material assumptions are supported, important dependencies or model limitations are visible, and management actions are credible. If a material objection remains unresolved, the decision-maker should see both the objection and its potential effect; approval should not silently convert uncertainty into certainty.
| Role | Core responsibility | Evidence of completion |
|---|---|---|
| Scenario owner | Define purpose, scope, accountable contributors, and review timetable | Named owner and approved scope |
| Analyst or model team | Implement assumptions and produce reproducible results | Versioned inputs, methods, and outputs |
| Independent challenger | Question rationale, evidence, limitations, and action feasibility | Challenge log with responses and dispositions |
| Approver | Accept, conditionally accept, defer, or reject use for a stated purpose | Recorded decision, conditions, and escalation |
Independence is functional, not just a title. A reviewer who reports to the scenario owner may still offer useful technical scrutiny, but the organization should consider whether that relationship limits the reviewer’s ability to escalate disagreement. Conversely, separation alone is not enough: a reviewer must have access to relevant material, suitable expertise, and adequate time.
A Controlled Governance Cycle
A practical cycle begins by defining the decision the scenario is intended to inform. The owner then identifies the affected exposures, time horizon, key assumptions, contributing functions, and approval route. Before results are used, the organization reviews the construction and evidence, documents challenges and responses, and records an approval decision with any conditions. After approval, monitoring and refresh controls keep the scenario relevant. This cycle does not replace technical stress analysis; it controls how that analysis is owned, reviewed, and used.
State the intended decision, scenario horizon, covered risks, owner, contributors, challenger, and approver. Identify materiality criteria and the evidence required for this use.
Record assumptions, data sources, model versions, limitations, and proposed management actions. Apply change control so that material edits can be distinguished from the reviewed version.
Test whether assumptions are supported and internally consistent, and whether actions are feasible within the scenario’s timing and constraints. Record each material issue, its owner, response, and disposition.
Record who decided, what version was reviewed, what use is permitted, and any conditions or unresolved limitations. Escalate material disagreements rather than hiding them in a summary.
Set review dates and event-driven triggers. Reassess the scenario when exposures, evidence, models, or the decision context change materially.
Assumptions, Actions, and the Audit Trail
An assumption register should make it possible to identify what the scenario assumes and why. For each material assumption, record its definition, value or range, source, rationale, owner, relevant horizon, and known uncertainty. Where evidence is weak, the record should say so and explain how the limitation is considered in interpreting results. A scenario need not pretend that uncertain assumptions are precisely known.
Management action claims require particular care. An action should not be credited solely because it is theoretically available. Governance should ask who has authority to take it, what operational steps and approvals it requires, how quickly it can take effect, whether counterparties or markets would cooperate under stress, and whether it conflicts with other actions or constraints. The record should distinguish an action assumed in the scenario from an action approved, funded, or otherwise demonstrably executable. An action’s modeled benefit is not proof of its feasibility.
Documentation should allow another qualified reviewer to reconstruct the decision and understand its limitations. At a minimum, retain the scenario’s purpose and scope, version identifier, data and assumption sources, methods and model versions, results, challenge comments, responses, approval record, and refresh triggers. A change log should show what changed, when, who authorized the change, and whether the change required renewed review. Retaining only a final number or presentation is not enough to reproduce the reasoning behind it.
Why the Capital Reconciliation Must Be Clear
Governance is strengthened when the reported result can be reconciled to its components using an explicit sign convention. Suppose \(C_0\) is capital surplus before stress, \(L\) is the gross adverse stress impact, \(A\) is the supported benefit of management actions, and \(I\) is the signed effect of interactions not already included in \(L\) or \(A\). A negative \(I\) is adverse; a positive \(I\) is favorable. Under this convention, stressed surplus is:
This identity follows by applying each change to the starting amount in sequence: subtract the gross adverse impact, add the supported action benefit, and then add the signed interaction effect. It is valid only if the components use consistent units and horizons and do not double-count effects. If an interaction is already embedded in the gross loss, it must not be subtracted again. If an action is unsupported or unavailable within the modeled horizon, it should not be treated as a supported benefit merely because it appears in a model output.
Worked Example: Challenging an Action Before Approval
A hypothetical insurer reports capital surplus in millions of dollars. The starting surplus is $80 million; the gross scenario impact is an adverse $52 million; and the scenario owner claims management actions will add $9 million. A separate interaction effect is an adverse $2 million and is not included in the gross impact or action estimate. The $9 million claim consists of a $5 million asset sale expected by day 3 and a $4 million expense reduction that can begin by day 30. The scenario’s decision horizon is one year, but the stressed liquidity need requiring the asset sale arises on day 2. An independent reviewer finds no evidence that the sale can settle by day 2. Assume the expense reduction is authorized and its $4 million benefit is supported within the one-year horizon.
Method. Reconcile surplus using the stated sign convention, then credit only the action supported for the modeled timing and constraints. The scenario owner’s arithmetic before challenge is:
Challenge. The proposed asset sale is not supported as an available response to a day-2 liquidity need because the evidence does not establish settlement by then. The $5 million benefit is therefore excluded. The $4 million expense reduction is supported by day 30, within the one-year horizon, so the supported action benefit is \(A=4\).
Revised result. With \(C_0=80\), \(L=52\), \(A=4\), and \(I=-2\):
The revised surplus is $30 million, which is $5 million below the owner’s reported $35 million because the unsupported part of the action claim is excluded. Check: the net change from the starting $80 million is \(-52+4-2=-50\) million, giving \(80-50=30\) million. The governance decision is to document the timing objection and approve only the supported action benefit, or defer approval if the intended use requires the asset sale to be available.
Common mistake avoided: treating an action’s modeled dollar estimate as available regardless of timing. The arithmetic \(80-52+9-2=35\) is internally correct, but it is not an appropriate approved result when $5 million of the action cannot be supported for the scenario’s stated liquidity need.
Independent Challenge in Practice
Effective challenge is specific and evidenced. Rather than asking only whether an assumption is “reasonable,” a reviewer can ask what evidence supports it, whether the evidence matches the scenario horizon, what changes if the assumption is varied, and which decision depends on it. Challenge can also test consistency across assumptions: for example, whether a scenario simultaneously assumes constrained markets and rapid execution of a transaction that depends on market access. A finding should lead to a response, not simply be filed.
Worked Example: Recording an Assumption Challenge
A hypothetical insurer’s scenario has a starting capital surplus of $120 million, gross adverse impacts of $68 million, a supported management action benefit of $12 million, and an adverse interaction effect of $5 million. One component of the gross impact is a $40 million loss based on a stated demand-severity assumption. The analyst confirms that increasing this component by 10% increases the loss by 10%, with all other components fixed. An independent challenger asks for the sensitivity to be shown alongside the base result. Assume the change is a sensitivity only, not a replacement assumption.
Method. First calculate the base result, then calculate the specified sensitivity while holding the other components fixed. In the base case:
The challenged component increases by \(0.10 \times 40=4\) million dollars, so the revised gross adverse impact for the sensitivity is \(68+4=72\) million dollars. The other inputs remain unchanged:
Check. The surplus falls by \(59-55=4\) million dollars, exactly the increase in gross adverse impact. The owner should record the sensitivity, its fixed-input assumption, and whether it changes the decision. The challenger has tested decision sensitivity; the challenger has not, by that calculation alone, established that the higher demand-severity assumption is the best estimate.
Common mistake avoided: presenting the sensitivity value as though it were the approved central scenario, or changing other inputs without saying so. A sensitivity is interpretable only when its changed and held-fixed assumptions are clear.
Approval, Refresh, and Escalation
Approval should identify the scenario version and its permitted use. A scenario may be acceptable for exploratory discussion but not sufficiently supported for a material capital decision. Approval can therefore be conditional, limited to a specific use, or deferred pending evidence. The record should name the decision-maker, date, conditions, and person responsible for satisfying each condition. Material issues that remain unresolved should be visible to the people relying on the result.
Refresh should be based on both a timetable and events. Periodic review helps prevent a scenario from becoming stale even when no obvious change has been reported. Event-driven triggers can include a material change in exposure, a new or revised model, changed data or external evidence, an assumption moving outside a defined range, a management action becoming unavailable, or a changed decision purpose. A trigger does not automatically mean the scenario is invalid; it means the owner must assess whether it remains fit for use and document the decision.
Worked Example: Refreshing a Scenario After Exposure Growth
A hypothetical enterprise has a $100 million starting capital surplus. Its approved scenario has gross adverse components of $30 million, $25 million, and $20 million, for a total gross impact of $75 million. A supported action benefit is $8 million, and the adverse interaction effect is $3 million. A later portfolio acquisition increases the exposure represented by the third component by 50%. Assume the other components, action benefit, interaction, and starting surplus are unchanged for this comparison. The scenario owner’s governance policy requires review after a material exposure change.
Method. Establish the approved baseline, then update the affected component and recalculate the result with the other stated inputs held fixed. The original gross impact is:
The third component increases by \(0.50 \times 20=10\) million dollars, becoming \(20+10=30\) million dollars. The updated gross impact is \(30+25+30=85\) million dollars. Before the exposure change, stressed surplus was:
After the exposure change, the fixed-input comparison is:
Check. The gross impact rises by $10 million, so stressed surplus falls from $30 million to $20 million, also a $10 million change. The acquisition meets the stated review trigger. The owner should document the changed exposure, produce a new version, reassess whether the assumptions and action remain appropriate, obtain the required challenge, and secure approval before relying on the updated result for a decision.
Common mistake avoided: continuing to use the approved version without review because the scenario name and method have not changed. A material exposure change can make a previously approved result unsuitable even when the calculation method is unchanged.
Common Governance Traps
- Confusing review with approval. A reviewer’s comments do not establish that an authorized decision-maker accepted the scenario for a particular use.
- Using an untracked final edit. A small change to a material assumption can alter results; preserve versions and identify which one was challenged and approved.
- Crediting actions without constraints. Record timing, authority, dependencies, and operational feasibility rather than relying on an unqualified benefit estimate.
- Closing challenge without disposition. Record whether an issue was accepted, rejected with reasons, or left unresolved and escalated.
- Refreshing only on a calendar. Periodic review should be supplemented by event-driven triggers for material changes.
A disciplined governance process makes scenario results more useful because it clarifies who is accountable, what evidence supports the result, what remains uncertain, and when the result must be reconsidered. These controls are aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Check Your Understanding
For each question, distinguish the governance decision from the arithmetic or evidence that supports it.
- A hypothetical scenario has a $70 million starting surplus, $46 million of gross adverse impact, a supported $6 million action benefit, and a $3 million adverse interaction not included elsewhere. Calculate stressed surplus and identify one item the approval record should specify. (Answer: $27 million; for example, the approved scenario version and its permitted use.)
- A scenario owner claims a $10 million action benefit. Evidence supports $7 million by the scenario horizon, while the remaining $3 million depends on an approval that has not been obtained. What action benefit should be credited pending approval, and what should be documented? (Answer: $7 million; document the unsupported approval dependency and the challenge disposition.)
- A gross stress impact is $50 million. A challenger requests a sensitivity increasing one component of $16 million by 25%, holding all else fixed. By how much does the gross impact increase, and what must the report clarify? (Answer: $4 million; identify the changed component and that other inputs are held fixed.)
- A scenario’s gross components are $18 million, $22 million, and $10 million. An exposure change increases only the third component by 30%. What is the updated gross impact, and what governance response is appropriate if policy requires review after material exposure changes? (Answer: $53 million; trigger a documented fitness-for-use review and required challenge and approval before relying on the updated scenario.)
- An independent reviewer identifies a material disagreement about whether a management action can occur within the stress horizon. The owner responds, “The model includes it, so the assumption is approved.” Explain the governance flaw and the appropriate response. (Answer: Model inclusion is not evidence of feasibility or approval; assess and document timing and supporting evidence, record the challenge disposition, and escalate or exclude the unsupported action as appropriate.)