EBA欧洲银行-B1Hahnenstein2C20Greve-Stress-Testing-the-Credit-Risk-of-Mortgage-Loans_19页_1mb

报告摘要

Summary of "Stress Testing the Credit Risk of Mortgage Loans: The Relationship between Portfolio-LGD and the Loan-to-Value Distribution"

Core Content

This document explores the relationship between the Loan-to-Value (LTV) distribution and the Loss Given Default (LGD) in mortgage loan portfolios, with a focus on how stressed LGD can be more accurately assessed in the context of regulatory stress testing. The paper addresses a key challenge in the banking sector: the information asymmetry between banks and regulators, where banks have access to detailed loan-level data, while regulators rely on aggregated or peer-based data.

Main Viewpoints

LTV as a Key Factor: The LTV ratio, defined as the loan exposure divided by the collateral value, is a critical determinant of LGD. It is identified as the primary driver of realized loss rates in empirical studies and is a predominant input in LGD models.
LTV Distribution Importance: The paper emphasizes that the distribution of LTV ratios within a portfolio significantly affects the sensitivity of LGD to stress scenarios. Two banks with similar average LTV ratios can exhibit very different LGD responses due to the dispersion in their LTV distributions.
Regulatory Challenges: The EBA currently only requires banks to report average LTV figures, which limits the ability of regulators to accurately assess stressed LGD. This lack of detailed data creates an information asymmetry, making it difficult to compare portfolios fairly.
Proposed Solution: The paper suggests using a parametric characterization of the LTV distribution, specifically assuming a Beta distribution, to derive a formula that can estimate the stressed portfolio LGD without access to loan-level data. This approach provides a "rule-of-thumb" for benchmarking and stress testing, allowing for more consistent and comparable assessments across institutions.

Key Information

LTV and LGD Relationship

LTV ratio is defined as $ LTV_i = L_i / C_i $, where $ L_i $ is the loan amount and $ C_i $ is the collateral value.
LGD is calculated as $ LGD_i = \max[0, (LTV_i - RR_i) / LTV_i] $, where $ RR_i $ is the recovery rate for the loan.
Portfolio LGD is the weighted average of individual LGDs, calculated as $ LGD_P = \sum_{i=1}^n LGD_i \cdot \frac{L_i}{\sum_{i=1}^n L_i} $.

Stress Testing and LTV Distributions

The paper presents a two-bank example showing that even with similar average LTVs, the stress sensitivity of LGD can differ significantly based on the dispersion of the LTV distribution.
The Beta distribution is used to model LTV ratios, with parameters $ p $ and $ q $ estimated via Maximum-Likelihood-Estimation from loan-level data.
The formula for portfolio LGD under Beta distribution is:

$$
LGD_P = 1 - F(RR, p, q) - RR \cdot \frac{p + q - 1}{p - 1} \cdot \left(1 - F(RR, p - 1, q)\right)
$$

Where $ F(u, p, q) $ is the cumulative distribution function of the Beta distribution.

Policy and Practical Implications

Transparency is identified as a key requirement for a sound LGD benchmarking framework. Regulators and banks should have access to consistent and comparable LTV distributions.
The proposed approximation method allows for the estimation of stressed LGD using only the shape parameters of the LTV distribution, thus reducing the need for detailed loan-level data in stress testing.
The paper recommends incorporating this approach into future regulatory stress tests and benchmarking exercises to improve the accuracy and consistency of risk assessments.

Future Research

Empirical analysis could begin with the LTV buckets of RMBS/CMBS pools and their associated loss data.
The approximation approach could be extended to other risk parameters (e.g., Probability of Default) and portfolio types (e.g., corporate loans).

Conclusion

The paper concludes that understanding and modeling the LTV distribution is essential for accurate stressed LGD estimation. By introducing a Beta distribution-based approximation, it offers a practical solution to the information asymmetry in regulatory benchmarking. This approach supports a more transparent and consistent risk assessment framework for mortgage loan portfolios.