IFRS 9 SICR Quantitative Calibration: Relative Lifetime PD Ratios vs Absolute Basis-Point Thresholds

How large a credit deterioration moves a performing loan from Stage 1 (12-Month ECL) into Stage 2 (Lifetime ECL)? Under IFRS 9 paragraphs 5.5.9 and B5.5.1–B5.5.24, Significant Increase in Credit Risk (SICR) is explicitly a relative origination-anchored test—comparing the remaining-lifetime Probability of Default (PD) at the reporting date against the forward remaining-lifetime PD that was expected for that exact residual tenor back on Day 1!

Why You MUST Compare Remaining-Lifetime PD Against Forward Origination Remaining-Lifetime PD!#

Spot the fatal mathematical bug that ECB SSM, EBA, and UK PRA targeted review auditors flag immediately inside our IFRS 9 & ASC 326 CECL cornerstone guide:

• Suppose on Day 0, a bank originates a healthy 10-year loan with cumulative 10-year Lifetime PD(0 -> 10) = 8.0%. Seven years pass cleanly (t = 7), so only 3 years remain (t = 7 -> 10). Even if the borrower’s annual hazard rate tripled in Year 7 (0.8%/yr -> 2.4%/yr), its 3-year remaining-lifetime PD (~7.0%) is numerically smaller than its original 10-year Day-0 Lifetime PD (8.0%) simply because 7 years of time rolled off!
• IFRS 9.B5.5.11 Rule: Never compare PD_Remaining(7 -> 10 | Info_t=7) against PD_Full(0 -> 10 | Info_t=0)! You MUST compare PD_Remaining(7 -> 10 | Info_t=7) against the conditional forward survival-adjusted PD_Orig_Forward(7 -> 10 | Info_t=0) (~2.4%)—instantly catching the 7.0% / 2.4% = 2.92x relative surge!

Why Pure Relative Ratios (2.0x) Fail at Both Ultra-Low and High PDs: The Hybrid Threshold Formula#

Testing a flat Relative Ratio >= 2.0x (100% PD increase) alone breaks down at both ends of the credit spectrum across BNP Paribas, Bank of Montreal, and ANZ Bank:

1. Ultra-Prime Noise Trap (PD = 3 bps -> 7 bps): If an AA corporate’s remaining PD ticks from 0.03% (3 bps) to 0.07% (7 bps), the relative ratio is 2.33x (> 2.0x)—falsely dumping a pristine AA- borrower into Stage 2 on micro market noise!
2. Sub-Prime Insensitivity Trap (PD = 12.0% -> 22.0%): Conversely, if a high-yield B-rated borrower deteriorates by +1,000 bps from 12.0% to 22.0% PD, its relative ratio is only 1.83x (< 2.0x)—missing a massive +10 percentage point credit plunge!

Institutional Hybrid SICR Standard: Quants pair a Grade-Stratified Relative Ratio (R_trigger) with an Additive Minimum Basis-Point Floor AND an Absolute PD Surge Cap: SICR_Quant <=> ( PD_Rem_t / PD_Rem_0 >= R_grade AND PD_Rem_t - PD_Rem_0 >= Floor_bps ) OR ( PD_Rem_t - PD_Rem_0 >= Cap_bps ) (e.g., require >= 2.0x–2.5x AND >= +30 to +50 bps change on low-PD grades, OR any absolute jump >= +200 to +300 bps annualized on sub-IG grades! Note: IASB / EBA prohibit using an Absolute Bps trigger ALONE* without relative calibration!)*

Three-Layer SICR Waterfall: Quantitative PD + Qualitative Watchlist + 30-DPD Backstop#

Audit-grade Stage 1/2/3 engines layer three complementary gates:

• Layer 1 — Quantitative Forward-Looking PD Test (accounts for 75%–85% of timely Stage 2 inflows).
• Layer 2 — Qualitative Early-Warning Triggers: Inclusion on Credit Risk Watchlist / Early Warning List, Performing Forbearance / Restructuring, adverse sector/geopolitical Collective Assessment transfer (ECB "Stage 2 collective sector overlays"), or LTV breach.
• Layer 3 — 30 Days Past Due (30 DPD) Backstop: Catches the residual <5%–10% of unmodeled retail/SME delinquencies.

Backtesting SICR Discrimination: Type I vs Type II Error & Pre-Default Capture Rate#

Model validation teams (under SR 11-7 / ECB TRIM / ISA 540 audit rules) optimize SICR cutoffs by measuring the Default Prior-to-Stage-3 Capture Rate: what percentage of loans entering Stage 3 Default in Year T were already flagged in Stage 2 at least 6 to 12 months prior (Target >= 70%–80%), balanced against the False-Positive Cure Rate (Type I Error)!

Using 12-Month PD as a Permissible Proxy (IFRS 9.B5.5.13)#

IFRS 9.B5.5.13 allows testing PD_12m changes as a proxy for PD_Lifetime only for non-back-loaded amortizing loans where empirical validation proves 12m and Lifetime term structures move monotonically across the IFRS 9, CECL & Fair Value Accounting topic hub.