EBA欧洲银行-Session-3-Slides-F.-McCann2C20E.-Gaffney_20页_743kb

Summary of "Towards an IFRS9-ready probability of default framework"

Core Content

This document presents a framework for assessing the probability of default (PD) for Irish residential mortgages in alignment with IFRS 9 requirements. The authors, Fergal McCann and Edward Gaffney from the Financial Stability Division (FSD) of the Central Bank of Ireland, focus on the implications of IFRS 9 for identifying a stock of performing mortgage balances that are subject to stricter provisioning requirements (Stage 2 assets). The framework is based on a Markov multi-state model using loan-level panel data.

Main Objectives and Findings

• Objective: To develop a PD framework suitable for IFRS 9 compliance and to analyze the impact of IFRS 9 on the classification of mortgage loans into different stages.
• Key Finding: Irish mortgage PDs mostly remain above origination levels, resulting in a significant stock of performing mortgage balances classified as Stage 2 due to high PDs. These cannot be identified using traditional methods such as forbearance or early arrears.

Irish Residential Property and Mortgage Markets

• Property Price Bubble: There was a significant 12-month growth in Irish residential property prices during the bubble period.
• Deleveraging: Mortgage lending declined significantly from Q1 2005 to Q1 2017, indicating a period of deleveraging.
• Arrears Trends: There was a notable increase in mortgage arrears, especially during the financial crisis.

Central Bank of Ireland's PD Model

• The model forecasts one-year PDs for Irish residential mortgages using loan-level data.
• It was re-estimated in 2016 based on an earlier study from 2014.
• The model uses a quarterly unbalanced panel from 2009 to 2015.
• It has been applied in various contexts, including bank stress tests and macroprudential policy evaluation.

Markov Multi-State Model

• The model uses time-dependent covariates to estimate transition probabilities between performing and default states.
• It incorporates variables such as current loan-to-value (CLTV), local unemployment rate, depth of arrears, and borrower type.
• The model's coefficients are challenging to interpret due to their logarithmic nature.

PD Coefficients and Interpretation

• A baseline loan with mean values of dummies and medians of quantitative variables has a PD of 0.57%.
• The PD changes significantly when certain parameters are varied:
  • Never modified → modified: PD increases from 0.47% to 2.26%
  • Non-BTL → BTL: PD increases from 0.52% to 0.90%
  • Fixed → SVR: PD increases from 0.38% to 0.71%
  • Fixed → Tracker: PD increases from 0.38% to 0.50%
  • One-loan → Multi-loan: PD increases from 0.56% to 0.59%
  • Change in instalment: PD increases from 0.57% to 0.61%
  • House price misalignment: PD increases from 0.57% to 0.64%

IFRS 9 and Stage Classification

• Under IFRS 9, loans with a significant increase in credit risk since origination must be classified as Stage 2, based on lifetime PD rather than one-year PD.
• The authors classify loans into IFRS 9 stages using the following rules:
  • Stage 1: Performing loans not in Stage 2
  • Stage 2: Performing loans in arrears (31-90 days), performing forborne, or with a material increase in PD (CPD ≥ 3 * OPD)
  • Stage 3: Non-performing loans
• The PD-based share of Stage 2 assets is estimated to be around 50%.

IFRS 9 and Economic Cycle

• IFRS 9 can be pro-cyclical, as many loans issued during the property price bubble (2004-08) are now in Stage 2 due to higher PDs.
• Since end-2015, property prices have risen by 19%, which may reduce the Stage 2 share among performing loans.

Remaining Challenges

• Estimating Lifetime PD: This is more complex for long-maturity portfolios like mortgages.
• Forecast Uncertainty: The authors emphasize the need for models that are transparent and analytically useful.
• Model Simplification: The use of a 2x2 transition matrix to model a fundamentally 3x3 process is a limitation.
• Future Dynamics: Forecasting the movement of balances between stages remains an ongoing challenge.

References

• Cox, D.R. 1972. "Regression Models and Life-Tables." Journal of the Royal Statistical Society Series B (Methodological) 34 (2), 187-220.
• Gaffney, Edward, Robert Kelly and Fergal McCann. 2014. "A transitions-based framework for estimating expected credit losses." Research Technical Paper 16/RT/14, Central Bank of Ireland.
• Jackson, Christopher. 2011. Multi-State Models for Panel Data: The msm Package for R. Journal of Statistical Software 38: (8)
• Joyce, John, and Fergal McCann. 2016. "Model-based estimates of the resilience of mortgages at origination." Economic Letters 09/EL/16, Central Bank of Ireland.
• Kang, Heedon, and Fergal McCann. 2016. "Simulation analyses of probabilities of default in household sector." In Ireland Financial Sector Assessment Program Technical Note – Nonbank Sector Stability Analyses, 43-50. Washington, D.C.: International Monetary Fund.
• Kelly, Robert, and Terence O'Malley. 2016. "The good, the bad and the impaired: A credit risk model of the Irish mortgage market." Journal of Financial Stability 22(C), pages 1-9.