2015年-EBA欧洲银行管理局_EBA_results_from_the_2014_Low_Default_portfolio_28LDP2920exercise_59页_2mb

Summary of the EBA Report on the 2014 Low Default Portfolio (LDP) Exercise

Core Content

This report presents the findings of the first supervisory benchmarking study under Article 78 of the Capital Requirements Directive (CRD) on the internal models used to calculate risk-weighted assets (RWAs) for low-default portfolios (LDPs) in large EU institutions. The LDPs include exposures to sovereigns, institutions, and large corporates, which are characterised by low default rates. The study aimed to identify discrepancies in internal models and support Competent Authorities (CAs) in assessing these models.

Main Findings

• Participation: 41 institutions from 14 EU countries participated in the study, with 36 also having participated in the 2013 LDP exercise.
• Data Collection: Data was collected as of 30 June 2014, focusing on quantitative figures. Qualitative data was gathered through interviews with nine institutions.
• Benchmarking Tool: The EBA calculated benchmarks for IRB parameters and provided detailed feedback to CAs, which were used to monitor internal models and identify potential underestimations.
• Key Drivers of GC Variability:
  • 75% of the variability in global charge (GC) levels was attributed to the proportion of defaulted exposures and portfolio mix.
  • For large corporate portfolios, 40% of GC differences were due to defaulted exposures, while the remaining 60% were influenced by bank-specific factors.
• Impact of PD and LGD:
  • For the AIRB approach, the LGD effect had a more significant impact on RW differences than the PD effect.
  • Six out of nine institutions showed that real LGD parameters led to lower RW than benchmark LGD parameters.
  • Two out of six institutions showed that real PD parameters led to lower RW than benchmark PD parameters.
• Comparative Analysis:
  • The dispersion of GC was similar between the Standardised Approach (SA) and IRB for sovereign and institutions portfolios, but higher for large corporate portfolios.
  • The SA data quality was inconsistent, which may affect the dispersion results.
• Collateral and Facility Type:
  • Collateral information and facility type breakdowns were important in explaining RW variability, though data quality issues limited the depth of analysis.
  • A hypothetical LGD based on senior unsecured facilities was used to isolate the impact of collateral and deal structures.
• Regulatory Approaches:
  • The Advanced IRB (AIRB) approach was more widely used than the Foundation IRB (FIRB) approach, especially in large corporate portfolios.
  • Differences in regulatory approach (AIRB vs. FIRB) significantly affected RWs, particularly for defaulted exposures.

Key Insights

• Data Quality Constraints:
  • The study faced challenges due to the absence of reporting requirements and incomplete or poor-quality submissions.
  • Legal Entity Identifier (LEI) mapping was not fully integrated in institutions' IT systems, leading to potential mismatches.
• Supervisory Implications:
  • CAs used the benchmarks to assess internal models and identified issues with some institutions.
  • The results confirmed earlier findings on LDP variability and supported policy options for improving IRB approaches.
  • Feedback on benchmark parameters is planned to help institutions improve data quality in future exercises.

Portfolio Composition

• Regulatory Approach Usage:
  • 30 institutions used the IRB approach for the Institutions portfolio.
  • 20 institutions used the IRB approach for the Large Corporate portfolio.
  • 14 institutions used the IRB approach for the Sovereign portfolio.
• Portfolio Mix:
  • The average portfolio of participating institutions consists of 43% large corporate, 19% institutions, and 38% sovereign exposures.
  • Some institutions focused on specific portfolio types, which affects the overall representativeness of the study.
• Representativeness:
  • The LDPs may not fully represent the total IRB portfolios of individual institutions.
  • Sovereign exposures, in particular, may constitute a small portion of the total IRB credit portfolios for some institutions.

Future Directions

• Enhanced Data Collection:
  • Future benchmarking exercises will benefit from improved data collection, especially in areas like collateral valuation and LGD estimation.
• Continued Supervisory Benchmarking:
  • Annual benchmarking exercises will be conducted starting in 2016, following the implementation of the ITS on supervisory benchmarking.
• Policy and Regulatory Development:
  • The EBA has outlined policy implications and potential regulatory measures to enhance the functioning of internal models.
  • A workshop with participating institutions is planned to improve future benchmarking exercises.

Conclusion

The 2014 LDP exercise served as a pilot for future annual benchmarking under the EBA's technical standards. It provided valuable insights into the variability of RWAs and highlighted the importance of improving data quality and consistency. The findings support ongoing regulatory discussions and will be used to refine the IRB approach and supervisory practices.