2023-06-08-国际清算银行-信贷损失率洞察_全球数据库_31页_657kb

Summary of "Insights into Credit Loss Rates: A Global Database"

Core Content

This paper introduces a comprehensive global database of credit loss rates, aiming to address the significant data gap in reliable economy-level credit risk information. The database provides time series of both actual and forward-looking credit loss metrics, which are essential for financial stability analysis. The authors build upon the work of Hardy and Schmieder (2020), combining different sources of data to estimate credit loss rates for a wide range of jurisdictions.

Main Purpose

The primary goal of the database is to offer a public good by providing accessible, standardized, and updated credit loss data. This information helps in assessing credit risk and supports scenario analyses based on GDP trajectories. It is particularly useful for understanding how credit losses evolve in response to macroeconomic conditions and financial cycles.

Key Concepts

• Credit Loss Rates: These are measures of potential losses due to borrower default, and are used in financial risk management and regulatory frameworks.
• Forward-looking Credit Loss Rates: These are estimates based on models and macroeconomic indicators, rather than historical data. They include:
  • Merton-model/market-implied credit loss rates: Derived from firm-level PDs and macro-implied LGDs.
  • Macro-implied credit loss rates: Based on projected GDP trajectories and calibrated using historical relationships from past banking crises.
• Contemporaneous Realized Credit Loss Rates: These reflect actual losses observed in the current period, including:
  • Impairment rates: Calculated from net impairment charges.
  • Charge-off rates: Reflect actual loan defaults and write-offs.
  • GDP-implied realized credit loss rates: Use observed GDP trajectories to estimate losses.
• Implied Credit Loss Rates from NPL Stock Data: Derived from nonperforming loan (NPL) ratios and recovery rates, which are used to estimate cumulative losses over time.

Main Viewpoints

• Credit risk has been a major contributor to financial crises, including the Great Financial Crisis (GFC) and the ongoing effects of the pandemic.
• Despite the importance of credit loss data, such information is not systematically available to the public, creating a significant challenge for financial stability analysis.
• The "COVID-19 bankruptcy gap" refers to the discrepancy between expected and actual bankruptcies, attributed to the support measures that delayed insolvencies.
• The GDP elasticity of credit losses is a critical factor in estimating credit loss rates, with different coefficients for advanced economies (AEs) and emerging market economies (EMEs).
• Forward-looking credit loss rates are crucial for anticipating future losses and stress testing, especially in the context of macroeconomic downturns and financial cycles.
• Macroprudential and microprudential approaches are both used to estimate credit loss rates, with the former focusing on system-wide risks and the latter on individual institutions.

Key Information

• Data Sources:
  • NUS-CRI PD data: Provides firm-level probability of default (PD) estimates for non-financial corporations.
  • World Bank LGD data: Offers loss given default (LGD) estimates and time to resolve insolvencies.
  • BankFocus: Provides bank-level data on impairment charges, charge-offs, and balance sheet information.
  • IMF FSI: Offers economy-level NPL ratios and GDP data.
• Metrics:
  • Forward-looking credit loss estimates: Include Merton model/market-implied and GDP-implied rates.
  • Contemporaneous realized credit loss rates: Include impairment, charge-off, and GDP-implied rates.
  • Implied credit loss rates from NPL stock data: Based on NPL ratios and LGD data.
  • Stock of credit losses: Reflects the total amount of defaulted loans and NPL stocks.
• Methodology:
  • The authors use a combination of microprudential and macroprudential data to estimate credit loss rates at the economy level.
  • They adjust for missing values, outliers, and ensure comparability by capping extreme losses or removing them from the dataset.
  • The database is updated regularly as new source data becomes available.
• Scenario Analysis Tool:
  • Users can run simple scenario analyses using GDP forecasts to estimate potential future credit losses.
  • The dashboard allows for easy access to data and tailored analysis based on jurisdiction, time period, and credit loss metric.
• Limitations:
  • Some data sources are not consistently available or are confidential.
  • Market-implied credit loss rates are limited to public firms.
  • The GDP elasticity model may over- or underestimate actual credit losses depending on the calibration period.

Conclusion

The authors emphasize the importance of having reliable, economy-level credit loss data for financial stability analysis. They present a database that combines actual and forward-looking credit loss metrics, making it a valuable resource for policymakers, regulators, and financial institutions. The data is updated continuously and made publicly accessible through an interactive dashboard, supporting both current analysis and future scenario modeling. The work highlights the need for improved data transparency and the role of macroeconomic indicators in forecasting credit risk.