2017年-EBA欧洲银行管理局_EBA_Report_results_from_the_2016_high_default_portfolio_exercise_March_2017_69页_4mb

Summary of EBA Report: Results from the 2016 High Default Portfolios (HDP) Exercise

Core Content

This report presents the results of the 2016 supervisory benchmarking exercise conducted by the European Banking Authority (EBA) on high default portfolios (HDPs), including residential mortgage, SME retail, SME corporate, and corporate-other portfolios. The study aims to assess the variability in risk-weighted assets (RWA) and identify the drivers of such variability across EU banks.

The reference date for the data is 31 December 2015, with 114 institutions from 17 EU countries participating. The analysis is based on data reported at the highest level of consolidation, ensuring consistency and avoiding double-counting. A total of 99 institutions were included after data cleansing.

Main Findings

1. Top-Down Approach

• The EBA used a top-down approach to quantify the proportion of GC (global charge) variability that can be explained by key drivers such as default status, counterparty country, and portfolio mix.
• The average GC increased to 75% (from 67% in the 2013 HDP report), and GC variability ranges from 8% to 293%, indicating a higher dispersion compared to previous studies.
• The average RW per institution ranges from 7% to 129%, with a simple average of 37.3%.
• 82% of GC variability is explained by the drivers of heterogeneity, which includes the proportion of defaulted exposures, non-EU exposures, and portfolio mix. This is slightly higher than the 78% from the 2013 report.
• The remaining 18% of variability is attributed to underlying credit risk, modeling assumptions, and supervisory practices.

2. Cross-Sectional Approach

• This approach examines the distribution of risk parameters and portfolios, focusing on interquartile ranges (IQRs).
• SME corporate and corporate-other portfolios show significant variability in RWs across EU countries.
• The country of the counterparty is a major driver of RW variability, with banks in EU countries experiencing stressed macroeconomic conditions showing higher average RWs.
• FIRB and AIRB approaches differ in their estimation of risk parameters:
  • LGDs under AIRB are generally lower than under FIRB.
  • PDs under FIRB are smaller than under AIRB.
• These differences may be due to the lower LGDs and CCFs under AIRB, which could lead to more conservative PD estimates under FIRB.

3. Outturn (Backtesting) Approach

• This method compares observed default rates and loss rates with estimated PDs and LGDs.
• Banks generally exhibit conservative estimates, with observed values often lower than estimated ones.
• However, some banks show observed values above estimated PDs and LGDs, indicating potential underestimations in their models.
• The country of the reporting bank and its counterparties is a key factor in RW variability, influenced by both underlying risk and supervisory practices.
• The findings are subject to data quality constraints, and the impact of using observed defaults instead of PD estimates on the CET1 ratio is limited, decreasing by only 17 bps on average.

4. Competent Authorities' Assessments

• CAs assessed the quality of internal models for each institution and found that most deviations from EU benchmarks were justified and not significant.
• Residential mortgages were identified as a critical portfolio to monitor due to their importance and potential impact on RWA.
• Corporate-other and SME corporate portfolios showed the highest potential for underestimations, with no immediate justification for such deviations.
• Internal validations by banks did not identify most potential underestimations, suggesting the need for supervisory actions and comprehensive analysis.
• CAs noted that some underestimations were identified in advance, particularly for SME corporate and corporate-other portfolios.

Key Information

• The study uses two main indicators: average RW and average GC.
• Standard deviation, interquartile range, and maximum vs. minimum distance are used to quantify variability.
• The EBA provides feedback to institutions to improve data quality and model accuracy.
• The report highlights the importance of supervisory knowledge and country-specific circumstances in interpreting results.
• The 2016 HDP exercise applies the new framework from the ITS and RTS published in March 2015, marking a shift in supervisory benchmarking methodology.

Methodology and Limitations

• The EBA used three main approaches: top-down, cross-sectional (distribution analysis), and outturn (backtesting).
• The top-down approach does not account for the partial use of the Standardised Approach (SA), limiting direct comparisons with earlier reports.
• The cross-sectional analysis identifies outliers and extreme values for key parameters, such as PD and LGD.
• The outturn approach provides insights into model accuracy and conservatism but has limitations in capturing full variability due to differences in data interpretation.
• Data quality issues are noted throughout the report, emphasizing the need for caution in interpreting results.

Conclusion

The 2016 HDP exercise provides valuable insights into the variability of RWA and the factors influencing it. While the results show a general trend of increased GC and RW variability, the majority of this can be attributed to measurable factors such as default status and portfolio mix. The EBA encourages ongoing improvements in data quality and model accuracy, and the findings support the development of a regular benchmarking process for internal models.