关键词: 公司债, 违约风险, Logit模型, ROC曲线

Abstract: The Chinese corporate bond market has experienced rapid development since the issuance of the first corporate bond in 2007, becoming the world’s second largest credit bond market after the United States. In 2014, “11 Chaori Bond” defaulted, breaking the rigid payment mechanism for Chinese corporate bonds. As of the end of 2020, a total of 590 corporate bonds have defaulted in China, with a total default amount of 520 billion yuan. This motivates us to investigate the determinants of default risk of Chinese corporate bond.
This paper selects 5160 corporate bonds listed on the Shanghai and Shenzhen Stock Exchanges from 2008 to 2020, including 590 defaulted bonds. All sample data are sourced from Wind Financial Database. The number of defaulted bonds increased from 6 in 2014 to 155 in 2020. This paper studies the determinants of default risk from the two aspects: the ownership and the industry. Companies owned by state-owned enterprises issue more bonds for public infrastructure construction or national strategic purposes, with lower default probability than private enterprises. Meanwhile, we find that manufacturing and comprehensive industries have the highest number of defaults of all industries.
Macroeconomic state also has a significant impact on debt repayment. When the economy is in a downward cycle, it often leads to declining corporate performance, limited revenue generation, and difficulty in paying back debts on schedule. If the economy improves, rising social demand will lead to better corporate performance and reduce the risk of default. Areas with higher economic development levels have better market environment and development prospects for companies, and we believe that the economic level and default risk in the region are negatively correlated. Many scholars have analyzed the factors of corporate finance and bond characteristics, and this paper takes corporate financial indicators and bond features as control variables.
The empirical results show that in terms of corporate finance factors, net asset growth rate, operating income growth rate, net profit margin, inventory turnover ratio, total asset turnover ratio, and cash ratio are negatively correlated to corporate bond defaults, indicating that deteriorating corporate finances increase the risk of default. Both the ownership and industry are significantly associated with corporate bond defaults. Credit rating is not significantly related to default, indicating that the quality of China’s credit rating market still needs to be evaluated, and the credit rating agencies in China cannot effectively measure the default risk.
This paper calculates the default probability of corporate bonds and performs ROC curve testing on the model. The area under the ROC curve reaches 0.9608, indicating that the model can accurately predict the default of Chinese corporate bond. The average marginal effect test shows that when the corporate belongs to a state-owned enterprise, its bond default probability decreases by 2.36%. Manufacturing and comprehensive industries have the highest default rates, followed by wholesale and retail trade, construction, information technology, transportation and storage, and other sectors. From the perspective of the economic environment, if the GDP increases by 1 trillion yuan per year, the issuer’s corporate bond default probability decreases by 0.8%.
The contributions of this paper have two folds. First, the existing literature predicts the default risk mainly based on financial conditions and macroeconomic state. This paper incorporates the ownership and industry factors to study the default risk of Chinese corporate bonds. Second, unlike previous research conclusions, this paper finds that credit ratings and maturity periods have no significant impact on Chinese corporate bond default risk. Chinese corporate bonds generally have high credit ratings, and rating adjustments are relatively lagging, making credit ratings unable to effectively measure credit risk.

Key words: corporate bond, credit risk, Logit model, ROC curve