Bond defaults in China - Using machine learning to make predictions

Aug 3, 2026

Upload by Skyler

This paper proposes an improved default-prediction model based on machine-learning methods for China’s onshore credit bond market. Conventional risk-assessment tools have limitations, particularly for foreign investors facing transparency and coverage constraints. Using granular financial data on Chinese bond issuers, the model achieves broader coverage than international credit-rating agencies. Results show accuracy above 90% in predicting credit-bond defaults, substantially outperforming Altman’s Z-scores, and supporting early warning and reliable default-risk detection for investors.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Introduction

Market context and investor challenges

Review of related credit-risk modeling approaches

Study contribution and research direction

Received: 23 January 2025 Revised: 24 February 2025 Accepted: 27 February 2025 DOI: 10.1111/irfi.70010 Bond defaults in China: Using machine learning to make predictions Bei Cui 1 | Li Ge 2 | Priscila Grecov 1 1Monash Centre for Financial Studies, Monash Business School, Monash University, Melbourne, Victoria, Australia 2Monash Business School, Monash University, Clayton, Victoria, Australia Correspondence Li Ge, Monash Business School, Monash University, Clayton, VIC 3168, Australia. Email: [li.ge@monash.edu](li.ge@monash.edu) Abstract This paper proposes a superior default-prediction model using machine-learning techniques. Traditional risk-assessment tools have fallen short, especially for foreign investors who face significant transparency issues. Using detailed financial data on Chinese bond issuers, our model provides much broader coverage than international credit-rating agencies offer. We achieve better than 90% accuracy in predicting credit-bond defaults, significantly outperforming Altman's Z-scores. This study not only advances predictive analytics in financial risk management but also serves as an early warning device and reliable default-risk detector for investors aiming to navigate the complexities of the Chinese bond market. KEYW OR DS Chinese bond markets, credit bonds defaults, credit risk assessment, machine learning JE L C LASS IFICAT I O N G1, G22, C53, C58, C45 1 | INTRODUCTION China's onshore bond market is the second largest globally, trailing only the historically dominant United States bond market. Growth in China's onshore bond market in recent years, along with its recent inclusion in global bond indices, has increased its attractiveness to investors globally. Investors have focused particularly on the US$6 trillion Chinese credit bonds market as a strategic avenue for portfolio diversification. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2025 The Author(s) . International Review of Finance published by John Wiley & Sons Australia, Ltd on behalf of International Review of Finance Ltd. International Review of Finance. 2025;25:e70010 . [https://doi.org/10.1111/irfi.70010](https://doi.org/10.1111/irfi.70010) [wileyonlinelibrary.com/journal/irfi](wileyonlinelibrary.com/journal/irfi) 1 of 19 2 of 19 CUI ET AL. Prospective foreign investors have, however, faced significant challenges with credit-risk assessments of Chinese bonds. Defaults on Chinese onshore credit bonds totaled US$17 billion in 2021 and are anticipated to increase in the coming years as the gradual withdrawal of government bailouts of bond issuers continues. Institutional investors have faced a range of hurdles when conducting credit-risk assessments of Chinese bonds. These hurdles include the relatively poor quality of domestic rating-agency assessments, low coverage of Chinese bonds by foreign rating agencies, insufficient research, and distorted market dynamics. To date, most studies related to the Chinese bond market have developed and tested their credit-risk models in the context of the US market. Differences in accounting standards, bankruptcy legislation, and market characteristics between China and the US may, however, result in divergent inferences. Further, existing models, such as the Z-score model of Altman et al. (2017), tend to measure financial stress by predicting afirm's bankruptcy risk—rather than directly measuring the probability that credit bonds default. A few recent studies examine Chinese corporate default risk and the factors that impact default probabilities (such as Liet al., 2022, 2025; Liu & Hu, 2024; Lu et al., 2013; Wang & Ma, 2023), but they focus on firm's distress measures, such as Merton's (1974) distance to defaul

FAQs

What is the main goal of this study?

To develop a superior machine-learning-based model that predicts default risk for Chinese credit bonds and supports investors’ decision-making.

Why are traditional risk tools considered insufficient for foreign investors?

They struggle with transparency and coverage issues in the Chinese bond market, and they often do not measure default probability directly.

How well does the proposed model perform compared with Altman’s Z-scores?

The model achieves better than 90% accuracy in predicting credit-bond defaults, significantly outperforming Altman’s Z-scores.

Aug 3, 2026

Upload by Skyler