IFRS 9 Expected Credit Loss (ECL) Modelling

Day One

Overview

• IFRS 9 Stage 1, 2 and 3 loans – classification, expected credit loss (ECL) provisioning and interest income recognition

• ECL formula

• ECL as probability-weighted average ECL over macroeconomic scenarios

• Motivation for IFRS 9 loan loss provisioning standard vs its predecessor (IAS 39)

• Cliff effect in ECL provisioning as loans migrate from Stage 1 to Stage 2 – going from 12 month ECL to lifetime ECL

• Case Study 1 – Illustrative sensitivity of ECL provisioning for a range of Stage 1 loans migrating to Stage 2

• Importance of Significant Increase in Credit Risk (SICR) threshold. Setting SICR threshold based on current lifetime probability of default (PD) as a multiple of lifetime PD at loan origination

• Case Study 2 – Illustrative sensitivity of ECL provisioning for Stage 2 loans to changes in SICR threshold

• Interaction of IFRS 9 ECL provisioning with regulatory capital (Internal Ratings Based approach, IRB) – treatment of expected provisions shortfall and surplus (IFRS 9 ECL vs 12 month through-the-cycle ECL within IRB) within regulatory capital

• Case Study 3 – IFRS 9 ECL provisioning during COVID

Modelling 12 month Point-In-Time (PIT) & Lifetime PD – Method 1 – Markov Chain

• Transition matrices to set out transition probabilities between different rating grades over 12 months – continuous time homogeneous method

• Case Study 4 – Transition matrix PD model – illustrated with corporate loan portfolio

• Case Study 5 – Logit and probit PD models – illustrated with corporate loan portfolio

• Case Study 6 – Scoring PD model – illustrated with retail loan portfolio

• Matrix multiplication of transition matrices combined with macroeconomic vector to arrive at lifetime PD

• Case Study 7 – Using Markov Chain to estimate lifetime PD on a hitherto Stage 1 corporate loan, test for SICR threshold and estimate ECL depending on whether loan remains Stage 1 or has migrated to Stage 2

Modelling Lifetime PD – Method 2 – Parametric Survival Regression (Weibull)

• Survival time of borrower or segment of borrowers linked to macroeconomic covariates according to Weibull distribution. Lifetime PD derived from survival time

• Case Study 8 – Using Weibull model to estimate lifetime PD on hitherto Stage 1 retail loan portfolio, test for SICR threshold and estimate ECL depending on whether loans remain Stage 1 or have migrated to Stage 2

Modelling Lifetime PD – Other Methods

• Vasicek single factor model

• Forward intensity model on distance to default

• Pluto Tasche

Modelling 12m PIT LGD & Lifetime LGDs

• Workout LGD

• Market LGD

• Case Study 9 – Modelling 12m PIT LGD and LGDs over lifetime

IFRS 9 ECL Modelling For Off-Balance Sheet Facilities

• Revolving credit facilities (RCF)

• Credit guarantees

• Letters of credit (LC)

• Credit conversion factors and exposure at default (EAD)

• Case Study 10 – Example IFRS 9 ECL modelling for RCF, credit guarantee and LC