1. Executive Summary
This document defines the methodology for measuring Interest Rate Risk in the Banking Book (IRRBB) for an Indian commercial bank. It covers:
- Economic Value of Equity (EVE) measurement under standardized rate shocks
- Net Interest Income (NII) sensitivity over a 1-year horizon
- Behavioural modelling for 5 product categories with non-contractual cashflows
- Data architecture for sourcing, processing, and computing risk metrics
- Reporting framework for management and regulatory submission
The approach is aligned with RBI Master Direction on IRRBB (2023) and Basel Committee BCBS 368.
2. Regulatory Background
2.1 RBI Requirements
The Reserve Bank of India requires all scheduled commercial banks to:
- Measure IRRBB using both EVE and NII approaches
- Apply standardized interest rate shock scenarios
- Incorporate behavioural assumptions for NMD, prepayable products, and term deposits
- Report ΔEVE as percentage of Tier 1 capital
- Report ΔNII as percentage of Net Interest Income
- Maintain internal limits and escalation triggers
2.2 Basel BCBS 368 Framework
- Banks must capture IRRBB comprehensively across all banking book positions
- Both EVE and NII perspectives required (complementary measures)
- Behavioural assumptions must be documented and regularly reviewed
- Standardized framework prescribes 6 interest rate shock scenarios
- Outlier test: ΔEVE > 15% of Tier 1 capital triggers supervisory action
2.3 Indian Context
- INR yield curve (RBI DBIE / FBIL published zero-coupon rates)
- MCLR/EBLR linked floating-rate products
- Savings account rate regulation (floor at 2.70%)
- SLR/G-Sec portfolio treatment
- CRR as non-interest-bearing reserve
3. Scope & Definitions
| Term | Definition |
|---|
| EVE | Economic Value of Equity = PV(Assets) − PV(Liabilities) |
| ΔEVE | Change in EVE under a shocked yield curve vs base curve |
| NII | Net Interest Income = Interest Income − Interest Expense |
| ΔNII | Change in NII under shocked rates over 1-year horizon |
| NMD | Non-Maturity Deposit — no contractual maturity (savings, current) |
| AIRO | Automatic Interest Rate Option — embedded optionality |
| CPR | Conditional Prepayment Rate |
| Core Deposits | Stable portion of NMD balances unlikely to be withdrawn |
| Volatile Deposits | Portion of NMD sensitive to rate changes |
4. Model Inventory
4.1 Product-to-Model Mapping
| Product | Cashflow Type | Model Required | Complexity |
|---|
| G-Sec (fixed coupon) | Contractual | None (known CFs) | Low |
| Corporate loans (MCLR/EBLR) | Repricing | Repricing schedule | Low |
| Retail home loans (fixed) | Contractual + behavioural | CPR / Prepayment | Medium |
| Retail savings accounts | No contractual maturity | NMD behavioural | High |
| Current accounts | No contractual maturity | NMD behavioural | High |
| Retail term deposits | Contractual + behavioural | Premature withdrawal | Medium |
| Wholesale term deposits | Contractual + optionality | AIRO (Hull-White) | High |
| Borrowings (fixed) | Contractual | None | Low |
4.2 Behavioural Model Inventory
| Model | Products | Methodology |
|---|
| BM-01: Retail NMD | Savings, Current accounts | Core/volatile decomposition, decay function |
| BM-02: Wholesale NMD | Large institutional deposits | Concentration-adjusted decay |
| BM-03: Term Deposit | Retail FDs | Rate-incentive prepayment model |
| BM-04: Fixed-Rate Loans | Home loans (fixed) | CPR based on refinancing incentive |
| BM-05: AIRO | Wholesale term deposits | Hull-White, Bermudan swaption |
5. Data Requirements
5.1 Source Systems
| Source | Data | Frequency | Granularity |
|---|
| Core Banking (CBS) | Account positions, balances, rates, dates | Daily | Account-level |
| Treasury System | G-Sec, money-market, swaps | Daily | Deal-level |
| General Ledger | GL balances for validation | Monthly | GL-code level |
| RBI DBIE / FBIL | Government security yield data | Daily | Tenor-level |
5.2 Key Data Fields — Deposits
| Field | Purpose |
|---|
| Account ID | Unique identifier |
| Product Code | Maps to risk category |
| Outstanding Balance | Current position |
| Interest Rate (booked) | Current paying rate |
| Rate Type | Fixed / Floating / Administered |
| Maturity Date | For term products |
| Customer Segment | Retail / Wholesale / Institutional |
6. EVE Methodology
6.1 Concept
EVE = Present Value of all future asset cashflows − Present Value of all future liability cashflows
ΔEVE = EVE(shocked curve) − EVE(base curve)
This measures how the bank's economic net worth changes when interest rates move.
6.2 Computation Steps
Step 1: Identify all banking book positions
Step 2: Generate cashflows (contractual + behavioural)
Step 3: Map cashflows to time buckets (O/N, 1M, 3M, 6M, 1Y, 2Y, 3Y, 5Y, 7Y, 10Y, 15Y, 20Y, 30Y+)
Step 4: Obtain base yield curve (INR zero-coupon rates from FBIL)
Step 5: Discount all cashflows: PV = Σ CF(t) / (1 + r(t))^t
Step 6: EVE(base) = PV(assets) − PV(liabilities)
Step 7: For each shock scenario:
a. Apply shock to base curve
b. Re-discount all cashflows
c. ΔEVE = EVE(shocked) − EVE(base)
Step 8: Report ΔEVE as % of Tier 1 capital
6.3 EVE Treatment by Product
| Product | Cashflow Treatment |
|---|
| Fixed-rate loans | Scheduled principal + interest to maturity |
| Floating-rate loans | Principal at next reset date (repricing assumption) |
| G-Sec (fixed coupon) | Coupon + principal at maturity |
| NMD (savings/current) | Behavioural cashflows from NMD model |
| Term deposits | Contractual CFs adjusted by prepayment model |
| Wholesale TD | Option-adjusted CFs from AIRO model |
7. NII Methodology
7.1 Concept
ΔNII = NII(shocked rates) − NII(base rates) over 1-year horizon.
This captures near-term earnings risk from rate movements.
7.2 Key Assumptions
- Horizon: 1 year (12 months forward)
- Balance sheet: Static (constant balance sheet for regulatory reporting)
- Repricing: Products reprice at scheduled reset dates within the year
- Pass-through: Administered rates (savings) move partially and with lag
7.3 Pass-Through Assumptions
| Product | Rate Behaviour | Pass-Through |
|---|
| MCLR-linked loans | Reprices at reset | 100% at reset |
| EBLR-linked loans | Reprices immediately | 100% immediate |
| Fixed-rate loans | No repricing | 0% within 1Y |
| Savings accounts | Administered | 30–50% partial, lagged |
| Term deposits (existing) | Locked in | 0% for existing |
| Wholesale deposits | Negotiated | 80–100% |
8. Rate Shock Scenarios
| # | Scenario | Short End | Long End | Description |
|---|
| 1 | Parallel Up | +200 bps | +200 bps | Entire curve shifts up uniformly |
| 2 | Parallel Down | −200 bps | −200 bps | Entire curve shifts down (floor at 0) |
| 3 | Steepener | −100 bps | +100 bps | Short rates fall, long rates rise |
| 4 | Flattener | +100 bps | −100 bps | Short rates rise, long rates fall |
| 5 | Short Rate Up | +300 bps | 0 bps | Only short end rises (graduated) |
| 6 | Short Rate Down | −300 bps | 0 bps | Only short end falls (graduated) |
9. Behavioural Models Overview
Why behavioural models are needed: Contractual cashflows assume all products behave exactly as their contract states. In reality, NMD deposits have no maturity, term deposits can be withdrawn early, and fixed-rate loans can be prepaid. Behavioural models convert contractual positions into realistic expected cashflows.
10. Retail NMD Model
10.1 Methodology: Core/Volatile Decomposition
- Historical analysis: Collect 36+ months of monthly aggregate savings balances
- Core identification: Statistical decomposition — core = stable floor (typically 68–75% for Indian savings)
- Volatile balance: Total − Core (placed in O/N bucket)
- Decay function: Core decays exponentially: Balance(t) = Core × e−λt
- Repricing: Savings rate adjusts by β × ΔMarket_Rate with lag L months
- Cashflow generation: Bucketed using decay: CF(bucket) = Core × [Surv(start) − Surv(end)]
10.2 Illustrative Parameters
| Parameter | Savings | Current |
|---|
| Core % | 72% | 55% |
| Decay rate (λ) | 0.15 p.a. | 0.25 p.a. |
| Half-life | 4.6 years | 2.8 years |
| Pass-through (β) | 40% | N/A (zero rate) |
| Repricing lag | 4 months | N/A |
| Max effective maturity | 7 years | 5 years |
11. Term Deposit Prepayment Model
11.1 Methodology
Depositors may withdraw before contractual maturity. Likelihood depends on rate differential between current market rate and booked rate, penalty, remaining tenor, and deposit size.
CPR Model: CPR = Base_CPR + Sensitivity × max(0, Rate_Differential − Penalty)
11.2 Illustrative Parameters
| Parameter | Value |
|---|
| Base CPR (no incentive) | 3–5% p.a. |
| Sensitivity per 100bps | 2–4% p.a. additional |
| Penalty (typical) | 1% of deposit rate |
12. AIRO — Hull-White Model
12.1 Concept
Wholesale term deposits carry an implicit early-withdrawal option (AIRO). Valued using Hull-White one-factor short-rate model.
12.2 Short-Rate Dynamics
dr(t) = [θ(t) − α·r(t)] dt + σ dW(t)
Where:
r(t) = instantaneous short rate
α = mean reversion speed
θ(t) = time-dependent drift (fits current term structure)
σ = short-rate volatility
dW(t) = Wiener process
12.3 Calibration
- Estimate α: OLS regression on discretized rate changes using historical 91-day T-Bill data
- Estimate σ: Standard deviation of regression residuals, annualized
- Calibrate θ(t): Fit to current INR zero-coupon curve (RBI DBIE / FBIL)
12.4 Option Valuation
- Early withdrawal modeled as Bermudan receiver swaption
- Exercise dates: Monthly
- Strike: Booked rate + penalty spread
- Valuation: Monte Carlo with Longstaff-Schwartz exercise boundary
- Output: Option-adjusted cashflows for EVE/NII
13. Architecture & Data Flow
┌─────────────────────────────────────────────────────────────┐
│ SOURCE SYSTEMS │
│ CBS │ Treasury │ GL │ Market Data (RBI) │
└────┬─────┴──────┬───────┴──────┬──────┴───────────┬──────────┘
│ │ │ │
▼ ▼ ▼ ▼
┌─────────────────────────────────────────────────────────────┐
│ ETL LAYER │
│ Extract → Cleanse → Map products → Generate cashflows │
│ → Apply behavioural models → Construct yield curves │
└────────────────────────────┬────────────────────────────────┘
▼
┌─────────────────────────────────────────────────────────────┐
│ MODEL ENGINE │
│ NMD model → TD prepay → Loan CPR → AIRO/Hull-White │
│ → Cashflow bucketing → Discounting → EVE/NII computation │
└────────────────────────────┬────────────────────────────────┘
▼
┌─────────────────────────────────────────────────────────────┐
│ REPORTING LAYER │
│ EVE by scenario → NII by scenario → ΔEVE/Tier1 │
│ → Product contribution → Limit monitoring → Reg submission │
└─────────────────────────────────────────────────────────────┘
14. ETL & Performance Considerations
| Challenge | Solution |
|---|
| Repeated expensive table joins | Pre-join instrument × product_map × cashflow once, reuse downstream |
| Range-based bucket classification | Replace with format/lookup-based classification (single-pass) |
| Repeated UNION ALL processing | Single-pipeline architecture — each record processed once |
| Sequential bucket-by-bucket loops | Vectorized computation — all buckets simultaneously |
| Data volume (20M+ rows) | Minimize I/O, reduce table scans, parallel by product |
Result: Architecture redesign achieved approximately 70% reduction in overnight batch run-time with zero change to final computation outputs.
15. Validation Approach
| Validation Type | Method |
|---|
| Conceptual soundness | Methodology review against BCBS 368 / RBI standards |
| Data integrity | Reconcile model inputs with GL / source balances |
| Implementation correctness | Independent recalculation on sample positions |
| Back-testing | Apply model to historical period, compare predicted vs actual |
| Sensitivity analysis | Vary parameters ±10–20%, measure output impact |
| Mass balance | Sum of bucketed cashflows = total portfolio balance |
16. Implementation Considerations
Phased Approach
| Phase | Duration | Deliverables |
|---|
| Phase 1 | 3–4 months | Data sourcing, basic EVE/NII, simple NMD model, regulatory submission |
| Phase 2 | 2–3 months | All 5 behavioural models, Hull-White calibration, enhanced reporting |
| Phase 3 | Ongoing | Automation, performance optimisation, quarterly recalibration |
Common Challenges in Indian Banks
| Challenge | Mitigation |
|---|
| Product-code proliferation (300+ codes) | Master mapping table, validate with business |
| Missing data fields | Phased approach: start with available, enhance over time |
| No historical behavioural data | Conservative assumptions initially, refine with data |
| Stakeholder alignment | Joint workshops, documented decisions |
| Regulatory timeline pressure | Phased delivery with early regulatory submission capability |
17. Governance & Recalibration
- All behavioural model parameters approved by ALCO / Risk Committee
- Model documentation maintained and version-controlled
- Annual independent validation (or more frequent if material changes)
- Change management process for parameter updates
- Audit trail for all model runs
| Model | Recalibration | Ad-Hoc Trigger |
|---|
| NMD (Retail) | Quarterly | Balance shift >15% |
| NMD (Wholesale) | Quarterly | Large depositor entry/exit |
| TD Prepayment | Semi-annual | Rate regime change |
| Loan CPR | Semi-annual | Major rate cut/hike |
| AIRO (Hull-White) | Semi-annual | Yield curve shape change |