Scholarlink

<p>Local government implicit debt is a key financing tool for local infrastructure, supporting regional investment through various financing platforms. However, the low short-term default risk and high credit ratings of government implicit debt limit effective credit risk identification, distorting risk pricing. This study employs interpretable machine learning models and post-hoc explainability methods to analyze urban investment bonds (UIBs) credit risk in China from 2015 to 2023. Findings reveal that basic bond characteristics (amount, bond rating) and macroeconomic indicators (monetary policy, market sentiment) are the most influential factors, with regional economic conditions and issuer financial status also playing significant roles. To enhance risk assessment, this study combines the best-performing interpretable EBM model, with the most effective traditional Random Forest model, for a comprehensive predictive analysis. This approach improves both interpretability and accuracy, offering policymakers, investors, and regulators a robust data-driven risk management tool.</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 Credit Default Risk Crisis of Government Implicit Debt: An Interpretable Machine Learning Approach

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Pan Tang,
  • Xiantao Wang,
  • Qingying Han,
  • Rui Dong

摘要

Local government implicit debt is a key financing tool for local infrastructure, supporting regional investment through various financing platforms. However, the low short-term default risk and high credit ratings of government implicit debt limit effective credit risk identification, distorting risk pricing. This study employs interpretable machine learning models and post-hoc explainability methods to analyze urban investment bonds (UIBs) credit risk in China from 2015 to 2023. Findings reveal that basic bond characteristics (amount, bond rating) and macroeconomic indicators (monetary policy, market sentiment) are the most influential factors, with regional economic conditions and issuer financial status also playing significant roles. To enhance risk assessment, this study combines the best-performing interpretable EBM model, with the most effective traditional Random Forest model, for a comprehensive predictive analysis. This approach improves both interpretability and accuracy, offering policymakers, investors, and regulators a robust data-driven risk management tool.