Scholarlink

<p>Loss Given Default (LGD) modelling for residential mortgages presents persistent challenges in bank risk management. This study proposes a three-stage decomposition framework for LGD prediction, aligned with operational collection processes: pre-collateral disposition, collateral disposition, and post-collateral disposition recovery. Using the Freddie Mac Single-Family Loan-Level dataset (1999–2020), we demonstrate that this approach significantly outperforms traditional single-component models, including OLS regression, two-step selection models, and random forest techniques in out-of-time predictions. The decomposition reveals previously obscured stage-specific drivers and their varying influences across recovery stages. Our findings contribute to both theoretical understanding of recovery processes and practical applications in risk management, capital allocation, and loss provisioning.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting loss severities for residential mortgage loans: a decomposition approach

  • 期刊论文

  • 发表日期：

    2025年11月4日

查看全文
  • Justin Tang,
  • Hung Xuan Do,
  • David Tripe,
  • David Woods

摘要

Loss Given Default (LGD) modelling for residential mortgages presents persistent challenges in bank risk management. This study proposes a three-stage decomposition framework for LGD prediction, aligned with operational collection processes: pre-collateral disposition, collateral disposition, and post-collateral disposition recovery. Using the Freddie Mac Single-Family Loan-Level dataset (1999–2020), we demonstrate that this approach significantly outperforms traditional single-component models, including OLS regression, two-step selection models, and random forest techniques in out-of-time predictions. The decomposition reveals previously obscured stage-specific drivers and their varying influences across recovery stages. Our findings contribute to both theoretical understanding of recovery processes and practical applications in risk management, capital allocation, and loss provisioning.