From Risk Inventory to Model Components
An enterprise risk inventory describes the risks an organization faces; a capital model represents selected exposures through defined calculation components. These are related but not interchangeable lists. Some inventory items can be quantified directly, some are represented within another component, and some are assessed qualitatively or excluded from the quantitative model. A defensible mapping makes each decision visible and reviewable.
The mapping is useful only if a reviewer can follow the chain from an identified risk to its treatment, the component or assessment that addresses it, and the person accountable for the decision. This supports completeness and helps prevent the same loss from being counted twice. It is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
A useful inventory row should be specific enough to support a decision. “Operational risk” is usually too broad to map consistently. A more useful entry might identify payment-processing disruption, the affected activity, and the potential financial consequence. The organization may maintain additional detail, but the mapping should preserve a stable identifier for each row so it can be traced through review and change.
For each item, record at least: its description and identifier; primary disposition; component or qualitative assessment, if applicable; accountable owner; rationale; and any review trigger. The owner is responsible for the decision and its continued relevance, not necessarily for performing every model calculation. Rationale should explain the treatment, rather than simply repeat the risk name.
Choose and Document a Primary Disposition
Use dispositions that distinguish what happens to the risk in the model. In particular, “not a separate component” does not by itself mean “excluded”: the exposure may already be represented elsewhere. The categories below are designed to make that distinction explicit.
| Primary disposition | Meaning | What the record should identify |
|---|---|---|
| Modeled | The exposure is represented quantitatively, either directly or within another component. | The component and the loss exposure it captures; if represented elsewhere, the receiving component and reason. |
| Qualitative | The exposure is considered through a documented non-quantitative assessment rather than a quantitative model component. | The assessment owner, assessment method or record, rationale for this treatment, and any review trigger. |
| Excluded | The exposure is outside the model’s intended representation and is not addressed as a modeled exposure or through the designated qualitative assessment. | The specific exclusion rationale, accountable owner, and circumstances that would prompt reconsideration. |
These are primary dispositions for inventory reconciliation, not a claim that risks are naturally isolated. A single broad risk can involve multiple causes or consequences. If it is split into sub-items, define those sub-items so that their represented losses are distinguishable, or explicitly state how overlap is handled. If several inventory entries feed one component, document how each entry is covered. In either case, avoid treating a list of labels as proof that the underlying exposure has been represented.
An owner and rationale are important even for a risk that is not quantified. An exclusion with no explanation is difficult to distinguish from an oversight. A qualitative treatment with no named assessment owner can become a permanent placeholder. Review triggers—such as a change in product, exposure, or operating arrangement—help keep the mapping current.
A Completeness Check for the Inventory
A basic reconciliation tests whether every inventory row has exactly one primary disposition. Let \(N\) be the number of inventory rows, and let \(n_M\), \(n_Q\), and \(n_E\) be the numbers classified as modeled, qualitative, and excluded, respectively. If the categories are mutually exclusive and exhaustive, then:
This identity is a control on the mapping, not a measure of how much risk is covered. It does not prove that a component is adequate, that an exclusion is justified, or that two components do not overlap. Those questions require review of the row-level descriptions, rationales, and component definitions.
The reason for the check is straightforward: each row must contribute exactly once to the disposition count. A row with no disposition leaves a gap. A row classified in two primary categories violates the classification rule and may conceal inconsistent treatment. Assigning one primary disposition does not prevent a modeled row from being associated with several components when that association is genuinely needed; in that case, specify which losses each component captures so the same loss is not represented repeatedly.
Use a specific description and identifier. Clarify the exposure or consequence at issue when a broad label could refer to several different losses.
Choose modeled, qualitative, or excluded. Do not leave “to be determined” as if it were a final treatment.
For modeled risks, name the receiving component and what it captures. For qualitative risks, identify the assessment and its owner.
Explain why the treatment is appropriate and who will reconsider it if relevant circumstances change.
Confirm that the inventory counts balance, then inspect for vague rows, unsupported exclusions, duplicate coverage, and risks with no accountable owner.
Worked Applications
Worked Example: Reconciling a Small Insurer’s Inventory
A hypothetical insurer has eight inventory rows. Its model includes insurance, market, credit, and operational components. Assign a primary disposition to each item and verify that the inventory is reconciled. The risk descriptions and treatments below are illustrative, not observations about an actual insurer.
| ID and risk | Disposition and mapping | Owner and rationale |
|---|---|---|
| R1: Property claim severity | Modeled: insurance component | Claims actuary; the component represents the specified claim exposure. |
| R2: Equity-price loss on invested assets | Modeled: market component | Investment risk lead; the component captures the stated asset-price exposure. |
| R3: Default of a corporate bond issuer | Modeled: credit component | Credit risk lead; this is the identified issuer-default exposure. |
| R4: Payment-system interruption loss | Modeled: operational component | Operations risk lead; the component captures the specified interruption loss. |
| R5: Vendor-caused interruption loss | Modeled: operational component | Operations risk lead; it is captured in the same component as R4, with distinct cause tagging and no separate addition of the same interruption loss. |
| R6: Potential change in customer behavior after a new product launch | Qualitative | Product risk owner; the product is not yet operating, so the risk is tracked through a documented launch assessment. |
| R7: Possible adverse reputation effect from a future service change | Qualitative | Communications risk owner; the effect is assessed qualitatively because the proposed change has not been approved. |
| R8: Earthquake loss at a location the insurer no longer owns or insures | Excluded | Property risk owner; the stated exposure is absent from the defined inventory boundary, with reconsideration required if the exposure returns. |
Step 1: Count modeled rows. R1 through R5 are modeled, so \(n_M=5\). R5 is not a new component: its rationale says that the interruption loss is represented in the operational component and must not be added again as a second copy.
Step 2: Count qualitative and excluded rows. R6 and R7 are qualitative, so \(n_Q=2\). R8 is excluded, so \(n_E=1\).
Step 3: Reconcile. The total is \(5+2+1=8\), equal to the eight inventory rows. The count check passes. The mapping also identifies each owner and gives a treatment-specific rationale; an owner or rationale still needs review if its underlying assumptions change.
Method and common mistake avoided: Assign one disposition per row, then reconcile the counts. Do not count R5 twice merely because the same component covers two inventory causes; component coverage is not an additional inventory row.
Worked Example: Preventing Duplicate Representation of an Event
A hypothetical insurer identifies a cyber intrusion that interrupts a claims platform. For a specified stress scenario, the inventory records two distinct consequences: $1,200,000 of unauthorized transfers and $800,000 of claims-processing interruption loss. The cyber component is defined to capture unauthorized transfers; the operational component captures interruption losses, whether caused internally or by a cyber event. Determine the mapping and check the represented loss total for this scenario.
Step 1: Separate the consequences. The two recorded amounts refer to different loss categories: transfers of $1,200,000 and interruption of $800,000. The mapping should preserve that distinction rather than assign the entire event amount to both components.
Step 2: Assign component coverage. Map the unauthorized transfers to the cyber component and the interruption loss to the operational component. Record the cyber intrusion as a shared cause, so the operational component’s treatment of the interruption is visible.
Step 3: Add only the distinct losses. The represented total for the stated scenario is \( \$1{,}200{,}000+\$800{,}000=\$2{,}000{,}000 \). The check is that the two terms are the two distinct consequences stated in the problem.
Common mistake and correction: If each component is assigned the entire $2,000,000 event loss, the resulting component entries sum to $4,000,000. That is wrong because the same transfer and interruption losses have each been counted twice. Assign the transfer loss once to cyber and the interruption loss once to operational. If a component definition instead includes the whole event loss, the mapping must state that explicitly and prevent the other component from representing the same loss.
Method: Map loss consequences, not just event names. A shared cause can appear in the descriptions of several risks, but each loss included in the model needs a clear representation that does not duplicate another component’s coverage.
Worked Example: Documenting Qualitative Treatment and Exclusion
A hypothetical regional insurer reviews three inventory items. Its documented screening rule says that a risk with an estimated annual loss below $100,000 may receive qualitative treatment if the estimate is sufficiently reliable and no separate severe-event concern has been identified. The rule does not require exclusion. A rare-event exposure with an unreliable estimate is to receive qualitative review rather than being dismissed on the basis of that estimate. Apply the rule, assign ownership, and give a review trigger.
Item A: Minor equipment damage. The estimated annual loss is $60,000, the estimate is considered reliable, and no separate severe-event concern has been identified. Since \( \$60{,}000 < \$100{,}000 \), assign qualitative treatment under the stated rule. The facilities risk owner is accountable for the assessment. Reconsider the disposition if the estimate reaches the threshold or the exposure changes materially.
Item B: Rare regional flood exposure. The estimated annual loss is $20,000, but the estimate is unreliable and the item is identified as a rare-event concern. The numerical estimate alone does not satisfy the screening rule’s conditions. Assign qualitative treatment, with the catastrophe risk owner accountable for documenting the assessment. Reconsider if better information becomes available or the exposure changes.
Item C: Coverage for a product the insurer has never offered and has no current plan to offer. The risk owner confirms that there is no current exposure within the inventory boundary. Assign exclusion, not qualitative treatment. Record the evidence for the absence of exposure and require reconsideration if the product is proposed or offered.
Step-by-step check: Item A meets both conditions for the stated threshold-based qualitative route. Item B fails the reliability condition and triggers the explicit rare-event provision, so it remains qualitative. Item C has no current exposure, so its rationale is boundary-based absence, not low estimated loss.
Common mistake avoided: Treating every estimate below $100,000 as an automatic exclusion would misread the rule. It authorizes qualitative treatment only when the estimate is reliable and no separate severe-event concern exists; it does not say to ignore the exposure.
Review the Mapping, Not Just the Counts
A balanced count is necessary but not sufficient. A review should challenge whether the inventory description is specific, the component actually captures the stated exposure, and the rationale matches the disposition. In particular, examine the following:
- Unmapped or vague items. “Other risk” and broad labels can conceal distinct exposures that require different treatment.
- Duplicate coverage. Two components may appear to capture the same consequence, especially when different risk causes lead to the same loss.
- Unsupported exclusions. “Not modeled” is not a rationale. State why the exposure is outside the model representation and what change would prompt reconsideration.
- Unowned assessments. Qualitative treatment still requires an accountable person and a record of how the risk is considered.
- Stale decisions. A treatment can become inappropriate after a change in products, operations, or exposure. Use a review trigger that is concrete enough to prompt action.
The purpose is not to force every risk into a quantitative component. It is to ensure that the chosen treatment is explicit, justified, and traceable. The mapping therefore serves both as a bridge from inventory to model and as a record of where the model intentionally does not provide a quantitative representation.
Check Your Understanding
For each problem, identify the mapping decision and check the relevant coverage or reconciliation. All scenarios are hypothetical.
- A four-row inventory has two modeled risks, one qualitative risk, and one excluded risk. A reviewer discovers that one modeled row has no named component, while the other is mapped to the credit component. Does the count reconcile, and what additional issue remains? (Answer: The count reconciles as \(4=2+1+1\), but the unmapped modeled row still lacks traceable component coverage.)
- A hypothetical event causes $500,000 of direct fraud loss and $300,000 of service interruption loss. The fraud component captures only direct fraud; the operational component captures interruption. What is the represented total if the losses are distinct, and what should the mapping record? (Answer: $800,000; map $500,000 to fraud and $300,000 to operational, and record the shared event cause without duplicating either consequence.)
- A risk has a $70,000 annual loss estimate under a $100,000 qualitative-screening threshold. The estimate is unreliable, and the organization’s rule requires a separate severe-event concern with unreliable estimates to receive qualitative review. Is exclusion supported? (Answer: No; assign qualitative treatment, document the unreliable estimate and severe-event concern, and name an assessment owner.)
- An inventory has 12 rows: 7 modeled, 3 qualitative, and 2 excluded. One modeled row is assigned to two components, but both component definitions include the same loss consequence. What does the count check show, and what must be resolved? (Answer: The count check passes because \(12=7+3+2\); the overlapping component coverage must be clarified so the same consequence is not represented twice.)
- A risk is excluded because a hypothetical product is not currently offered. The product is approved for launch, but no policies have yet been written. What should happen to the mapping? (Answer: Apply the documented review trigger and reconsider the exclusion; record the current status and decide whether the prospective exposure is modeled, qualitative, or remains excluded with a justified rationale.)