How do we aggregate model risk across the enterprise?

Question

AcadiFi · Accepted Answer

Enterprise model risk aggregation is inherently qualitative but must be structured. Banks use a combination of five approaches: 1. Findings-based aggregation. Count and weight open validation findings by severity (high/medium/low) and age. An increase in high-severity open findings signals growing exposure. Roll up by business line, model type, and region. 2. Tier-weighted exposure. Aggregate the number of Tier 1 / Tier 2 / Tier 3 models with issues, weighted by tier importance. A single Tier 1 issue may outweigh 10 Tier 3 issues. 3. Limitation and override tracking. Every model in production has documented limitations. Count instances where users operate outside stated bounds, where overrides occur, or where management actions (conservative overlays) are applied. 4. Quantified model reserves. Some banks set capital-like reserves against specific models known to have weaknesses — an explicit dollar amount recorded as 'model risk buffer' in capital. Aggregated reserves are a direct enterprise measure. 5. Qualitative heat map. Board-facing dashboard colored by business line and model category, indicating overall health. Supported by narrative from the CMO. ```mermaid flowchart TD A[Model
Inventory] --> B[Findings] A --> C[Tier
Distribution] A --> D[Overrides] A --> E[Reserves] B --> F[Heat
Map] C --> F D --> F E --> F F --> G[Board] ``` A concrete example: Cypress Ridge Financial's quarterly board pack includes: (1) total models by tier (1,180); (2) open findings by severity (4 high, 22 medium, 58 low); (3) limitation breaches in the quarter (3 instances); (4) model risk reserve balance ($125M); and (5) a 5x5 heat map across business lines and model categories with one red cell flagged: 'Wealth Management — ML-based next-best-offer model, interpretability limitation pending remediation.' For FRM Part II exam: know that enterprise model risk lacks a single quantitative measure (unlike VaR for market risk), but structured aggregation through findings, tiers, overrides, reserves, and heat maps gives the board a coherent view. Regulators (OCC, Fed) specifically look for this in horizontal reviews.

How do we aggregate model risk across the enterprise?

Master Part II with our FRM Course

Related Questions

Practice Questions