Implementation Plan¶

This document provides the detailed implementation plan including acceptance test scenarios, contract definitions, and implementation order.

Overview¶

The implementation follows a test-first approach with phases:

1. Phase 1: User Acceptance Tests - Define expected inputs and outputs
2. Phase 2: Process Contracts - Establish interfaces between components
3. Phase 3: Implementation - Build components in dependency order
4. Phase 4: Acceptance Testing - Validate against expected outputs

Regulatory Framework Prioritisation: 1. Phase A: CRR (Basel 3.0) - Current UK implementation (effective until 31 Dec 2026) 2. Phase B: Basel 3.1 - Future UK implementation (effective from 1 Jan 2027 per PRA PS1/26)

CRR Acceptance Test Scenarios¶

Each scenario defines specific inputs and expected outputs with hand-calculated values based on current CRR rules (EU 575/2013 as onshored into UK law).

Scenario Group CRR-A: Standardised Approach (SA)¶

Scenario Group CRR-B: Foundation IRB (F-IRB)¶

Note: F-IRB only applies to wholesale exposures (corporate, institution, sovereign). Retail exposures must use A-IRB or Standardised Approach.

Scenario Group CRR-C: Advanced IRB (A-IRB)¶

Note: CRR A-IRB has NO LGD floors (unlike Basel 3.1 which has 25% unsecured floor).

Scenario Group CRR-D: Credit Risk Mitigation (CRM)¶

Scenario Group CRR-E: Specialised Lending (Slotting)¶

CRR Slotting Risk Weights:

Scenario Group CRR-F: Supporting Factors¶

CRR SME Supporting Factor - Tiered Approach (CRR2 Art. 501): - Tier 1: Exposures up to €2.5m (£2.2m): factor of 0.7619 (23.81% RWA reduction) - Tier 2: Exposures above €2.5m (£2.2m): factor of 0.85 (15% RWA reduction) - Formula: factor = [min(E, threshold) × 0.7619 + max(E - threshold, 0) × 0.85] / E

Scenario Group CRR-G: Provisions & Impairments¶

Scenario Group CRR-H: Complex/Combined¶

Scenario Group CRR-I: Defaulted Exposures¶

Defaulted Exposure Rules: - F-IRB: K = 0, RWA = 0 (regulatory capital requirement is zero for defaulted) - A-IRB: K = max(0, LGD - BEEL) where BEEL is best estimate of expected loss - Expected Loss = LGD × EAD for all defaulted exposures - CRR 1.06 scaling factor applies to corporate defaulted RWA

Basel 3.1 Acceptance Test Scenarios¶

These scenarios test the Basel 3.1 implementation per PRA PS1/26, effective 1 Jan 2027.

Scenario Group B31-A: Standardised Approach (SA)¶

Scenario Group B31-F: Output Floor¶

Output Floor Phase-In:

Key CRR Implementation Details¶

1. 1.06 Scaling Factor: Applies to ALL IRB RWA under CRR (removed in Basel 3.1)

   CRR:      RWA = K × 12.5 × 1.06 × EAD × MA
   Basel 3.1: RWA = K × 12.5 × EAD × MA
   

2. SME Firm Size Adjustment: R_adjusted = R - 0.04 × (1 - (max(S, 5) - 5) / 45) for turnover < EUR 50m

3. PD Floor: Single 0.03% floor for all exposure classes (Basel 3.1 has differentiated floors)

4. F-IRB LGDs: 45% unsecured senior, 75% subordinated

5. Maturity: Floor 1 year, Cap 5 years

Test Data Fixtures¶

Test datasets cover all key regulatory scenarios:

tests/fixtures/
├── counterparty/
│   ├── sovereign.py              # UK Govt, US Govt, Brazil, Argentina, etc.
│   ├── institution.py            # Banks (Barclays, HSBC, JPMorgan), CCPs
│   ├── corporate.py              # Large corp, SME, unrated, org hierarchy groups
│   ├── retail.py                 # Individuals, mortgages, SME retail, lending groups
│   └── specialised_lending.py    # Project finance, IPRE, HVCRE, object finance
├── exposures/
│   ├── facilities.py             # RCFs, term facilities, mortgages, hierarchy test
│   ├── loans.py                  # Drawn exposures for all acceptance scenarios
│   ├── contingents.py            # Off-balance sheet (LCs, guarantees, commitments)
│   └── facility_mapping.py       # Facility-to-loan/contingent relationships
├── collateral/
│   └── collateral.py             # Cash, bonds, equity, real estate, receivables
├── guarantee/
│   └── guarantee.py              # Sovereign, bank, corporate guarantees
├── provision/
│   └── provision.py              # SCRA/GCRA, IFRS9 stages 1-3
├── ratings/
│   └── ratings.py                # External (S&P/Moody's) and internal ratings
└── mapping/
    ├── org_mapping.py            # Parent-subsidiary relationships
    └── lending_mapping.py        # Retail lending group connections

CRR Regulatory Parameters¶

SA Risk Weights¶

Sovereigns (Art. 114)¶

Institutions (Art. 120-121) - ECRA¶

Corporates (Art. 122)¶

Real Estate (Art. 125-126)¶

Residential (Art. 125): - LTV ≤80%: 35% - LTV >80%: Split treatment (35% on 80%, 75% on excess)

Commercial (Art. 126): - LTV ≤50%: 50% (where rental income >1.5x interest) - Otherwise: 100%

CCFs (Art. 111)¶

Supervisory Haircuts (Art. 224)¶

F-IRB LGDs¶

Next Steps¶

Completed Fixture Work¶

The majority of fixture data has been completed, enabling 71 of 74 acceptance tests to pass:

• [x] SA scenario exposures with risk weights (CRR-A: 12/14 pass, 2 skip)
• [x] A-IRB exposures with own estimates (CRR-C: 6/7 pass, 1 skip)
• [x] CRM scenario exposures with collateral/guarantees (CRR-D: 9/9 pass)
• [x] Specialised lending exposures with slotting categories (CRR-E: 9/9 pass)
• [x] Supporting factor scenario exposures (CRR-F: 15/15 pass)
• [x] Provision scenario exposures with EL data (CRR-G: 7/7 pass)
• [x] Complex/combined scenario exposures (CRR-H: 4/4 pass)
• [x] Defaulted exposure scenarios with F-IRB and A-IRB treatment (CRR-I: 9/9 pass)

Remaining Fixture Work¶

• [ ] Commercial RE low LTV fixture data (CRR-A7)
• [ ] Off-balance sheet commitment CCF fixture data (CRR-A8)
• [ ] Specialised lending A-IRB fixture data (CRR-C3)

Basel 3.1 Extension¶

• [ ] Basel 3.1 expected outputs
• [ ] Basel 3.1 acceptance tests
• [ ] Updated risk weight tables (LTV-based real estate)
• [ ] Differentiated PD floors
• [ ] A-IRB LGD floors
• [ ] Output floor phase-in validation

Related Documentation¶

• Testing Guide - How to run tests
• Framework Comparison - CRR vs Basel 3.1 differences