2014年-EBA欧洲银行管理局_20140611_Fourth_interim_report_on_the_consistency_of_risk-weighted_asset_31页_1mb

Summary of the Fourth Report on the Consistency of Risk Weighted Assets in Residential Mortgage Portfolios

Core Content

This report presents the findings of the second phase of an analysis on the consistency of risk-weighted assets (RWA) in residential mortgage portfolios across the European Union (EU). It focuses on the impact of drill-down variables such as Loan-to-Value at Origination (LTVO), Indexed Loan-to-Value (ILTV), Debt-to-Service at Origination (DTSO), Loan-to-Income at Origination (LTIO), and Credit Risk Mitigation (CRMO) on RWA variability.

Main Findings

1. Use of Drill-Down Variables

• Most Common Variables: Indexed loan-to-value (ILTV) and credit risk mitigation (CRMO) are the most frequently used variables in risk parameter estimations, with ILTV used in 58% of the cases.
• Less Common Variables: DTSO and LTIO are used less frequently, primarily in PD models.
• Country-Specific Usage: Different countries show distinct preferences:
  • ILTV is widely used in the Czech Republic, Portugal, UK, Spain, and Ireland.
  • Other credit risk mitigants are more relevant in the Netherlands and France.
  • DTSO is more used in Italy and Belgium.

2. Definitions and Variability

• Inconsistent Definitions: Banks use varying definitions for the same variables, often influenced by country-specific practices and internal methodologies.
• Numerator and Denominator Variations: For LTVO, some banks include prior liens or other credit risk mitigants in the numerator or denominator.
• ILTV Complexity: ILTV includes different indexing frequencies (quarterly, semi-annually, yearly) and methods, such as external indicators or internal models.

3. Quantitative Analysis

• Correlation and Variability:
  • A correlation of above 80% was observed for most variables.
  • ILTV shows the highest variability, with 65% of observations having a standard deviation greater than 25% relative to the mean.
• Price Effect vs. Bucket Mix Effect:
  • The price effect (RW level relative to the benchmark) is more significant than the bucket mix effect (distribution of EAD across buckets).
  • The level effect is more important than the sensitivity effect in the EU sample, though both are present.
  • The sensitivity effect is generally negative, meaning RWs tend to be lower than the benchmark for similar ILTV buckets.

4. Impact of Vintage and Portfolio Composition

• Vintage Analysis: The level of LTVO and RW is closely linked, with portfolio composition by vintage playing a role in explaining RW variation.
• Portfolio Mix Effect: The composition of the portfolio across different buckets influences RW levels, but this effect varies by bank and country.

5. Country-Specific Patterns

• CRMO Influence: In France, the Netherlands, and Belgium, the presence of other credit risk mitigants (such as government guarantees) leads to lower RWs than the portfolio average.
• Lower RW Sensitivity: These country-specific mitigants may explain the reduced sensitivity of RWs to the value of financed real estate.

6. Limitations

• Data Granularity: The use of common and predefined buckets limits the ability to capture detailed variations.
• Missing Data: Some variables, such as DTSO and LTIO, have a high percentage of missing data, indicating recent data collection challenges.
• Indirect Influence: The indirect role of variables in credit assessment may affect RWs, but their direct use in PD/LGD models does not consistently explain RW differences.

Key Variables and Their Roles

Summary of Findings

• Positive Correlation: Drill-down variables are positively correlated with RWs, but the sensitivity to these variables is not always a clear driver of variation.
• ILTV Dominance: ILTV is the most significant variable influencing RW variation.
• Country-Specific Adjustments: Certain countries exhibit unique patterns where other mitigants reduce RW sensitivity.
• Vintage and Portfolio Mix: Vintage analysis and portfolio composition play a role in explaining RW differences.
• No Clear Discrimination: The direct use of variables in PD/LGD models does not significantly differentiate RWs across banks.

Conclusion

The report highlights the complexity and variability in the application of drill-down variables across EU banks, emphasizing the need for more granular data and consistent definitions. While ILTV and CRMO are the most influential variables, the overall impact of these variables on RWs is not uniform and depends on both the bank's internal practices and country-specific conditions.