Leveraging Risk Adjusted Return on Capital Modeling

Introduction

You're choosing where to put scarce capital, so use Risk-Adjusted Return on Capital (RAROC) to compare returns across different risks and allocate capital only where returns exceed the cost. It matters because RAROC aligns pricing, capital, and strategic choices with true economic risk, letting you price deals, set limits, and avoid hidden losses. The objective is practical: a step-by-step outline to build, use, and govern RAROC models for FY2025 - covering loss modeling, economic capital, hurdle rates, and governance so you can rank business lines and set clear limits. Finance: draft the first-pass RAROC template for FY2025 by Friday, owner Finance. defintely

Key Takeaways

• Use RAROC to compare returns across risks and allocate capital only where returns exceed the cost of capital.
• RAROC = (risk‑adjusted profit) / (economic capital); build it from clean exposures, PD/LGD/EAD, losses, and expense allocation.
• Choose robust capital metrics (VaR vs ES), backtest PDs/LGDs, and avoid stale inputs or mis‑specified correlations.
• Drive pricing, capital allocation, and limits with RAROC; reprice or de‑risk exposures that fall below hurdle rates.
• Govern RAROC as a formal model: validate, document, stress test, and maintain tech pipelines and audit trails.

Leveraging Risk Adjusted Return on Capital Modeling

You're trying to compare returns across lines that carry different risk - use Risk-Adjusted Return on Capital (RAROC) to put returns on the same footing and allocate capital where it earns above the cost. Here's the quick takeaway: RAROC = (risk-adjusted profit) / (economic capital), and you should build it from clean revenue and loss inputs, choose a tail-risk capital metric, and reconcile internal capital to regulatory minima.

What RAROC is - one simple formula and the pieces that go in

One-liner: RAROC = (Risk-adjusted profit) / (Economic capital).

Risk-adjusted profit means operating profit after you remove the expected losses (the average credit, market, and operational losses you plan for) and after allocating recurring expenses and taxes. Economic capital is the buffer you hold for unexpected losses at a chosen confidence level and time horizon.

Steps to compute the numerator and best practices:

• Start with top-line revenue and net interest income.
• Subtract expected credit losses (EL) - use vintage models and overlays.
• Subtract expected market and operational losses (use event histories).
• Allocate operating expenses and tax to the business unit consistently.
• Report risk-adjusted profit pre-capital charge and after tax.

Practical example (2025 fiscal): suppose revenue $200m, expected credit losses $12m, expected market/operational losses $3m, expenses $60m. Risk-adjusted profit = $125m. If economic capital = $1,000m, RAROC = 12.5%. Here's the quick math: 200 - 12 - 3 - 60 = 125; 125 / 1,000 = 0.125.

What this hides: allocation choices (which expenses to charge), tax treatment, and whether you include one-off gains. Make those explicit in documentation and sensitivity tests.

Capital choices - VaR versus Expected Shortfall and implementation steps

One-liner: pick the capital metric that matches how you want to manage tail risk - Value-at-Risk (VaR) measures a percentile loss; Expected Shortfall (ES) measures the average of losses beyond that percentile.

How to choose and implement:

• Decide confidence level and horizon up front (common banking choices: 99% or 99.9% and 1-year horizon).
• Use ES if you care about severity beyond the cutoff - ES captures the tail average and is more conservative.
• Model dependencies (correlations) across borrowers, products, and markets - copulas or factor models help but validate them.
• Calibrate on stressed periods and backtest against realized losses; keep a stressed-window ES if you need regulatory alignment.

Quick example: your portfolio's 99% VaR = $200m but 99% ES = $350m. If you hold capital to VaR you cover the 1% worst-case single loss, but ES tells you average loss in that 1% tail - choose ES if a single catastrophic hit would be intolerable.

Watchouts: defintely avoid underestimating tail dependence, using short sample windows, or ignoring jump risks from market shocks. Backtest and run extreme but plausible scenarios.

Economic capital versus regulatory capital (Basel) - differences and mapping

One-liner: economic capital is your internal estimate of unexpected losses at a chosen confidence; regulatory capital (Basel) is the legally required minimum based on risk-weighted assets (RWAs) and prescribed formulas.

Key distinctions and practical steps to reconcile:

• Define confidence level and horizon for economic capital (e.g., 99.9% one-year UL).
• Compare your economic capital number to the regulatory requirement (Basel's CET1 minimum 4.5% plus 2.5% conservation buffer = 7.0% baseline; add countercyclical or SIFI buffers as applicable).
• Map economic capital to RWAs for a consistent comparison: Required regulatory capital = RWAs × target CET1 ratio.
• If EC < regulatory capital, hold the regulatory amount; if EC > regulatory, consider capital actions or price for the extra capital needed.

Concrete mapping example: if RWAs = $50bn, regulatory CET1 requirement (including conservation buffer) = 7% → required capital = $3.5bn. If your economic capital estimate for the same risk is $2.8bn, you have a $0.7bn shortfall to meet regulatory minima - you must either increase capital, reduce RWAs, or shrink risky exposures.

Governance tip: maintain an explicit reconciliation table, add a model-risk buffer on top of EC (e.g., 10-25% depending on model complexity), and document assumptions so auditors and regulators can follow your logic.

Immediate next step: Finance & Model Risk - deliver a prototype RAROC for the top 10 portfolios by December 15, 2025; Risk Ops - run two stressed scenarios; Business - set preliminary RAROC floors.

Building a RAROC model: inputs and mechanics

You're building a RAROC model to decide which loans and products actually earn above your cost of capital; use a disciplined inputs-to-output pipeline so decisions are traceable and actionable. Here's the direct takeaway: start with clean exposures and credible loss estimates (PD, LGD, EAD, maturities), convert those into expected losses and economic capital, and report RAROC versus a chosen hurdle (e.g., 12% target ROE) so pricing and allocation follow true economics.

Inputs and data

One-liner: start with clean exposures, PD (probability of default), LGD (loss given default), EAD (exposure at default), and maturities.

Step 1 - clean the exposure ledger. Reconcile loan balances, undrawn commitments, collateral haircuts, and off‑balance items to a single exposure file by origination cohort and current balance date (use fiscal-year-end 2025-09-30 or your firm's year-end). Keep one row per obligor-product-tenor so you can roll up or slice by segment.

Step 2 - estimate PDs. Use vintage performance and survival models (Cox, logistic) with borrower covariates (FICO, LTV, industry, covenants). Benchmark model PDs to external rates (S&P/Moody's sector default tables) and to seasoned internal cohorts. If your model gives a 1.2% PD for a 36‑month unsecured consumer cohort but realized vintage defaults are averaging 2.0%, add an overlay or retrain.

Step 3 - estimate LGD and EAD. Build downturn LGDs (loss given default in stress) rather than cure‑period averages. For committed facilities, calculate EAD using empirical drawdown/conversion factors by vintage or apply regulatory-style CCFs if data is thin. Example quick math: a $10m loan, PD 2.0%, LGD 45%, EAD $10m → expected loss = 0.02 × 0.45 × 10,000,000 = $90,000. Here's the quick math: expected loss scales linearly with PD, LGD, and EAD; what this estimate hides is concentration and tail correlation.

Step 4 - add macro overlays and correlations. Map PD and LGD to macro scenarios (GDP, unemployment, house prices). Use time-varying overlays when forward indicators shift - e.g., add a +50 bps PD overlay when unemployment rises materially. Estimate pairwise and sector correlations from long-run data; if data is short, use conservative priors (higher correlations) to avoid undercapitalizing tail risk.

Best practices and checks:

• Keep raw transaction history for vintages
• Version datasets and document transformations
• Flag manual adjustments and require business justification
• Stitch in macro regressor time series and store scenario IDs

Cost of capital and capital charge

One-liner: pick a hurdle rate (e.g., WACC or target ROE) and allocate a capital charge to each exposure so marginal deals clear the hurdle.

Step 1 - choose the economic capital measure and confidence level. Decide between VaR (value-at-risk) and Expected Shortfall (ES). For lending portfolios the industry leans to tail measures like VaR at 99.9% or ES at 97.5% for severe stress; pick what maps to your ICAAP/board tolerance. VaR gives a single loss percentile; ES averages losses beyond that percentile and better captures tail severity.

Step 2 - compute economic capital per exposure or bucket. Use a portfolio credit model (credit migration, default simulation) to get stand-alone and marginal capital. Example: if simulation gives stand-alone EC of $800,000 for a $10m loan (≈8% of exposure), then capital charge at a 12% hurdle = 0.12 × 800,000 = $96,000.

Step 3 - calculate risk-adjusted profit and RAROC. Define risk-adjusted profit = revenue - expected loss - economic expenses (taxes optional). Then RAROC = risk-adjusted profit / economic capital. Example math: loan revenue = 5% on $10m = $500,000; expected loss = $90,000; operating expense allocation = $60,000; risk-adjusted profit = $350,000; RAROC = 350,000 / 800,000 = 43.8%. Compare that to hurdle 12% - this loan is well above hurdle.

Operational steps and allocation rules:

• Decide whether capital is economic (internal) only or blended with regulatory minima
• Allocate capital by marginal contribution (preferred) or pro rata exposure if portfolio sims not feasible
• Define expense allocations and tax treatment consistently across products
• Document the chosen hurdle: target ROE (12%), WACC (8-10%), or internal hurdle

Validation and watchouts

One-liner: backtest PDs and losses, benchmark models, and run out-of-sample checks so RAROC is a reliable decision tool and not a spreadsheet illusion.

Validation steps:

• Backtest PDs versus realized default rates by vintage and cohort; produce two-year and lifetime hit-rate tables
• Backtest LGD and EAD with realized recoveries and drawdown behavior; compare downturn LGD in stress years (e.g., 2008/2020 analogs)
• Run out-of-sample performance tests: train on pre-2022 data, test on 2022-2025 to capture late-cycle changes
• Use sensitivity analysis to compute elasticities: ΔRAROC/ΔPD, ΔRAROC/ΔLGD, ΔRAROC/Δhurdle

Key watchouts and fixes:

Stale PDs - defintely avoid using old scorecards without recalibration. If vintages drift, add time-based overlays or retrain quarterly for fast-moving portfolios (quarterly) and semi-annually for stable ones.

Mis-specified correlations - underestimating correlation flattens capital. If your historical sample is short, assume higher correlation (stress calibration) and document the rationale. Run a sensitivity sweep: increase correlations by +0.1 and report capital delta.

Hidden underwriting adjustments - manual pricing overrides, undocumented collateral rehypothecation, or portfolio transfers will bias model inputs. Detect these by monitoring acceptance rates, average pricing vs model price, and pre/post underwriting PD drift. Require any manual override to create an audit trail and corrective overlay.

Governance for model risk:

• Assign model owner, independent validator, and an approval committee
• Validate code, seed randomness, and version outputs; keep reproducible pipelines
• Set validation cadence: quarterly light checks for high-impact models, annual full validation
• Keep fall-back rules: if model fails, apply conservative overlays or hard RAROC floors

Leveraging RAROC for decisions: pricing, allocation, and limits

You're deciding price changes, where to put scarce capital, and which products to shrink - and you need one consistent metric to drive those choices. RAROC (risk-adjusted return on capital) should be your decision engine: it aligns pricing with the economic capital each exposure consumes so you fund what beats the hurdle and fix what doesn't.

Pricing: add a capital charge so marginal business meets hurdle RAROC

One-liner: let RAROC determine the incremental price needed so marginal deals meet your hurdle.

Steps to implement pricing with RAROC:

• Measure all cash flows: contract revenue, expected losses (PD×LGD), direct expenses, and fees.
• Allocate economic capital (EC) to the exposure using your capital model (e.g., VaR or ES at chosen confidence).
• Compute current RAROC = (revenue - expected loss - expenses) / EC.
• Compare to hurdle RAROC (your target return on capital). If current < hurdle, compute required price uplift: required profit = hurdle × EC; price uplift = (required profit - current profit) spread over exposure or term.

Worked example (prototype calculation): revenue = $120, expected loss = $6, expenses = $14 → profit = $100. Economic capital = $800 → RAROC = 100/800 = 12.5%. If your hurdle is 15%, you need profit = 0.15×800 = $120, so raise price to generate an extra $20 (about 2.0 percentage points on a $1,000 annualized exposure). Here's the quick math: additional profit needed / exposure = rate uplift.

Practical checks and best practices:

• Price by marginal RAROC for incremental deals, not by book averages.
• Pass-through capital charge as an explicit line on internal pricing sheets.
• Use tenor-adjusted capital (longer tenor → more EC) and show annualized bps impacts.
• Stress-test price elasticity: if customers churn at small increases, quantify churn vs RAROC benefit.

Allocation: rank product lines and customers by RAROC for capital deployment

One-liner: treat RAROC like ROI - rank, fund, and cap until your capital budget is deployed to highest-return risk-adjusted uses.

Concrete allocation process:

• Calculate RAROC for each product line, business unit, and customer segment on a consistent EC basis.
• Standardize assumptions (hurdle, confidence level, LGD) so comparisons are apples-to-apples.
• Rank buckets by RAROC; apply a liquidity/strategic overlay (e.g., growth or market share rules).
• Allocate capital top-down: fund highest RAROC buckets until aggregate EC target is reached, then stop or reserve contingent capacity.

Example allocation table (simplified):

• Product A: RAROC 18% - fund fully.
• Product B: RAROC 10% - restrict new originations.
• Product C: RAROC 14% - marginal funding, requires efficiency program.

Operational tips:

• Translate RAROC into actionable KPIs for front-line sales (bps price floors, target deal sizes).
• Use allocation rounds quarterly with clear cut-offs; avoid continuous ad-hoc rebalancing that confuses businesses.
• Keep a small strategic carve-out (e.g., 10-15% of EC) for market-making or strategic initiatives even if RAROC is below hurdle.

Limits and strategy: set minimum RAROC floors and reprice or shrink low-RAROC exposures

One-liner: define clear RAROC floors and required actions so low-return risks get fixed fast.

How to set and operationalize RAROC floors:

• Pick a floor tied to capital cost and strategic needs - typical floor range: 10-15% depending on target ROE.
• Tag exposures below the floor as remediation candidates and require a documented action plan within a fixed window (e.g., 60-90 days).
• Action options: reprice (raise spread), reduce exposure (limit/line cut), improve underwriting, or exit.

Enforcement mechanics and governance:

• Automate exceptions reporting: monthly list of accounts/products below floor with owner and due date.
• Require business owners to show three levers: price, volume, or cost reduction, and expected RAROC after action.
• Escalate persistent underperformers to a capital allocation committee for final decision (reprice, cap, or close).

Sample trigger policy:

• Below floor: business must submit remediation plan within 30 days.
• No adequate plan: reduce new origination by 50% within 90 days.
• Still below floor after 180 days: exit or transfer to special workout.

What this estimate hides: RAROC depends on EC model inputs and correlations - if you defintely under-estimate capital, floors look artificially achievable. Maintain overlays and human review.

Immediate next step: Modeling/Finance build first-pass RAROC for top 10 portfolios by December 15, 2025; Risk Ops run two adverse stress scenarios and Business set RAROC floors. Owner: Finance/Model Risk to deliver prototype; Business to confirm floors.

Stress testing, sensitivity, and scenario analysis

one-liner: test RAROC under realistic and adverse macro states to find fragility

You're running RAROC today, but you need to know where it breaks under stress so you can act quickly. Run baseline and adverse states that map to revenue, expected loss, expenses, and economic capital, then compare RAROC paths over a chosen horizon (typically 1-3 years).

Start simple: pick a baseline and two stress levels (adverse and severe). Translate each macro shock into changes in PD (probability of default), LGD (loss given default), EAD (exposure at default), and correlations. Run forward P&L, expected losses, and capital needs at monthly or quarterly cadence.

Here's the quick math to check you built the pipeline: project incremental expected loss = ΔPD × LGD × EAD; update economic capital if your model links capital to stressed VaR/ES; recompute RAROC = (Revenue - Expected Loss - Expenses - capital charge) / Economic Capital. What this estimate hides: non-linear capital add-ons and second-order business responses from pricing and origination.

scenarios: GDP shock, unemployment rise, LGD inflation, correlation spikes

Design scenarios that are realistic, policy-relevant, and plausibly severe. Example canonical scenarios to include:

• GDP downside: -3% over 12 months (adverse), -6% (severe)
• Unemployment shock: baseline ++3ppt (adverse), ++6ppt (severe)
• LGD inflation: ++10% relative increase in unsecured LGD
• Correlation spike: add +0.15 to asset correlations for credit portfolios

Map each macro input to risk-driver deltas using empirical overlays or internal econometric models. For example, convert GDP shock to PD changes with a GDP-PD sensitivity vector by vintage, or use unemployment as the driver for consumer PDs and house-price shocks for mortgage LGDs.

Operational steps: (1) lock scenario narratives and numeric paths by date; (2) run driver-models to produce PD/LGD/EAD/time series per portfolio; (3) feed those into the RAROC engine to produce quarterly RAROC curves; (4) compare to baseline and regulatory thresholds.

sensitivity and triggers: compute elasticities of RAROC to PD, LGD, and capital cost, and set action thresholds

Compute sensitivities with simple derivatives and a small shock approach. Use a one-factor approximation first, then full revaluation. Formulaic starter: dRAROC/dPD ≈ -(d(Expected Loss)/dPD) / Economic Capital, where d(Expected Loss)/dPD = LGD × EAD.

Example quick calculation (illustrative): assume Revenue = 15, Expenses = 3, EAD = 100, PD = 2%, LGD = 45%, Economic Capital = 40 (dollars). Baseline expected loss = 0.02×0.45×100 = 0.9. Capital charge at 12% hurdle = 0.12×40 = 4.8. Baseline RAROC = (15 - 0.9 - 3 - 4.8) / 40 = 15.75%.

Raise PD to 3%: EL = 1.35, RAROC ≈ 14.625%. Delta = -1.125 percentage points. Raise LGD to 55%: EL = 1.1, RAROC ≈ 15.25%. Increase hurdle to 15%: capital charge = 6.0, RAROC ≈ 12.75%. Here's the quick math on elasticity: Elasticity = (ΔRAROC / RAROC) / (Δdriver / driver).

Use those elasticities to prioritize actions. If a 1pp PD increase reduces RAROC by more than 20% relative to baseline, flag the exposure as fragile and require business remediation.

Define concrete triggers and actions tied to breaches. Suggested rule set:

• RAROC < hurdle (example: 12%) → require repricing within 10 business days
• RAROC drop > 25% vs baseline → suspend new originations and present remediation plan within 5 business days
• Stress scenario causes capital shortfall > 10% of regulatory CET1 floor → Finance model capital raise within 30 days
• Correlation spike increases aggregated economic capital > 15% → business to de-risk top 3 counterparties

Govern triggers with clear owners and SLAs: Risk Ops runs monthly stress runs, Model Risk approves scenario design quarterly, Finance translates model outputs to capital plans, and Business executes repricing/de-risk actions. Keep an action log and automated alerts so nobody searches for the reason after losses hit - defintely avoid manual slow chains.

Immediate action: Risk Ops and Finance build and run two stress scenarios for the top 10 portfolios and deliver RAROC sensitivity tables by December 15, 2025. Owner: Finance/Model Risk to produce the prototype; Business to prepare repricing playbooks.

Governance, implementation, and model risk

You're putting RAROC into production and need it to behave like a financial-control tool, not a spreadsheet toy. Direct takeaway: treat RAROC as a regulated model-document, validate, and govern so decisions rest on accountable outputs.

Governance and model life-cycle

One-liner: treat RAROC as a regulated model-document, validate, and govern.

Assign clear ownership. Have a Model Owner in Finance (builds model), an Independent Validator in Model Risk (challenges assumptions), and a Business Sponsor (uses outputs). Hold a Governance Committee that meets at least quarterly to approve material changes and exceptions.

• Set review cadences: full validation every 12 months, change-control reviews for code/data changes, and monthly performance monitoring for top exposures.
• Document everything: model spec, data dictionary, assumptions, version history, validation report, and decision logs.
• Create an audit trail: time-stamped runs, user IDs for changes, and signed approval for production runs.
• Define KPIs and KRIs (key risk indicators): actual default vs expected default, realised loss vs modelled loss, RAROC drift, and capital utilisation.

Here's the quick math: show monthly realised losses vs expected losses and flag breaches beyond a threshold (see Limits). What this estimate hides: initial governance work often takes 6-12 weeks to operationalize for a top portfolio.

Technology, reproducibility, and data pipelines

One-liner: use scalable data pipelines, reproducible code, and versioned model outputs.

Build for repeatability. Pipeline design should guarantee data lineage from source to RAROC output and allow fast re-runs for scenario tests. Use tooling that forces versioning and testing so your model runs are auditable.

• Data: enforce a canonical exposure table with snapshots, timestamps, and upstream lineage.
• Code: store model code in version control, require unit tests, and use CI/CD to push only approved changes to production.
• Environments: separate dev, staging, and prod; require peer review and validator sign-off before promotion.
• Reproducibility: capture random seeds, package versions, and config files with each run so outputs are recreatable.
• Monitoring: deploy dashboards showing run-to-run deltas, latency, and data-quality failures with alerts.

Practical stack choices: use orchestration for scheduled runs, row-level lineage tools for data QA, and secure artifact stores for model outputs. Start small: pilot the pipeline on top 3 portfolios, then scale to the top 10 by December 15, 2025.

Regulatory mapping and model risk limits

One-liner: map RAROC outputs to ICAAP, CCAR, and internal capital plans and maintain overlays with human review.

Map outputs into capital governance. For banks, feed stressed RAROC runs into your ICAAP documentation and use RAROC as an input to CCAR-style capital planning exercises. Keep a clear mapping table from each RAROC line item (expected loss, economic capital, risk charge) to the corresponding regulatory concept.

• Stress tests: run at least a baseline and an adverse macro stress; capture impacts on PD, LGD, correlation, and capital.
• Triggers: set action thresholds-example trigger: if realized defaults exceed model PD by more than 30% (or > 2 percentage points absolute) open a model review and stop new originations into that cohort.
• Overlays: require model overlays when data is thin, recent vintage shifts occur, or correlation assumptions break; document overlay rationale and exit criteria.
• Model risk appetite: define quantitative limits (e.g., max portfolio weight for unvalidated models, min RAROC floor) and enforce at the business-line level.
• Human review: mandate senior sign-off for pricing changes driven by RAROC adjustments and for any replay of model outputs used in capital plans.

Validation steps: backtest losses quarterly, benchmark PDs to external data annually, and run an out-of-sample stress once per year. Defintely avoid letting stale PDs or mis-specified correlations silently erode capital adequacy.

Immediate action: Finance/Model Risk to deliver a prototype RAROC for the top 10 portfolios by December 15, 2025; Risk Ops to run two stress scenarios and produce a one-page governance checklist for the Governance Committee.

Conclusion and next step

You want RAROC live for the biggest exposures fast so capital and pricing reflect true economic risk - start small, prove it, and scale. Below are the practical steps, timelines, and owners to make that happen by December 15, 2025.

Operationalize RAROC quickly for top exposures, then expand iteratively

One-liner: stand up a lean RAROC prototype for the top exposures, validate it, then broaden coverage.

Start with the portfolios that drive most risk - aim to cover portfolios that together represent at least 70 percent of your risk-weighted assets (RWA) or funded exposure in the 2025 fiscal year. That keeps work focused and impact high.

Concrete steps:

• Map: list top 10 portfolios by RWA/exposure and data owners.
• Scope: limit initial metrics to PD (probability of default), LGD (loss given default), EAD (exposure at default), term, fees, and direct expenses.
• Prototype: deliver one RAROC computation per portfolio (pre-tax and post-tax) with capital at a chosen percentile (VaR or ES).
• Validate: backtest with last 5 fiscal years of loss vintages and one out-of-sample year.

Best practices: use monthly granularity where possible, require minimum sample sizes of 500 default observations for PD calibration or apply conservative overlays if samples are smaller; version outputs and keep a one-line change log. What this estimate hides: if data cleanup is heavy, plan +4 weeks.

Immediate action: Modeling and Finance build first-pass RAROC for top 10 portfolios by the deadline

One-liner: deliver a repeatable prototype across the top 10 portfolios by December 15, 2025.

Project plan (high-level, for planning approval):

• Team: 3 modelers, 2 data engineers, 1 product lead, and 0.5 business SME.
• Timeline: 6 weeks development, 2 weeks validation, 1 week executive sign-off.
• Deliverables: pipeline-ready data extract, RAROC engine (PD, LGD, EAD, capital), documented assumptions, and a dashboard with portfolio ranks.

Practical guidance: build modular code with parameter files so you can flip capital method (VaR vs ES) and tax/ROE assumptions quickly. Use a conservative default hurdle of 12 percent post-tax RAROC for pricing tests and flag any exposure below that for the business to review. Keep the first iteration transparent - no black-box ensembles yet.

Owner: Finance/Model Risk deliver prototype; Risk Ops run two stress scenarios; Business set RAROC floors

One-liner: assign clear owners, deliverables, and triggers so RAROC moves from model to decision.

Owner matrix and deadlines:

• Finance / Modeling & Model Risk: build and document prototype; deliver by December 15, 2025.
• Risk Operations: execute two stress scenarios (baseline stress and adverse stress) against the prototype within 5 business days after delivery.
• Business leaders: set RAROC floors by portfolio and product line within 10 business days after stress results, and decide repricing or run-off actions.

Stress scenarios to run (practical defaults):

• Baseline stress: GDP decline -3 percent, unemployment +200 bps, LGD +50 bps.
• Adverse stress: GDP decline -6 percent, unemployment +400 bps, LGD +150 bps, correlation spike to 0.6.

Triggers and actions:

• If RAROC falls below hurdle by 300 bps, require immediate repricing or reduced origination.
• If capital need increases > 20 percent under adverse, trigger funding/capital plan review.
• Maintain a model overlay allowance of 25-50 bps on PD/LGD for model risk until multiple backtests validate performance - defintely document overlays.

Governance notes: require a one-page model charter, quarterly review cadence, and a version-controlled repository for inputs and outputs. Next action: Finance/Model Risk to schedule kickoff this week and confirm data extracts for the top 10 portfolios.