Log in to the Sina Finance app and search for "Information Disclosure" to view more assessment ratings. From October 20 to 22, on the banks of the Huangpu River in Shanghai, the 2026 Sustainable Global Leaders Conference will return once again. Nobel Prize and Turing Award winners will speak on-site about the future and trends of technology, and the internet-famous professor who "rescues" people from anxiety and the era of involution, Michael McFarlane, will hold an on-site dialogue with Xiang Biao!
Limited spots are available, first come, first served. Let us join together for the appointment with sustainability! [Click to register] Author: Zhang Ling, Deputy General Manager of the Risk Management Department of Huaxia Wealth Management Co., Ltd. This article was originally published in Bond, September 2026 issue.
Abstract
At present, credit risk in the bond market is generally converging, but structural tail risks remain challenging. Forward-looking identification of default risk is crucial for implementing regulatory requirements, ensuring the smooth operation of asset management products, and protecting investors. Based on the situation and requirements faced by bond credit risk monitoring and early warning, this article sorts out the general patterns of credit risk evolution, constructs a four-dimensional monitoring framework covering industry, fundamentals, related-party transmission, and individual bonds, explains the thinking for strengthening the early warning system from the five levels of governance, data, indicators, methods, and application, and finally proposes an intelligent transformation path of "artificial intelligence + monitoring and early warning," providing reference for asset management institutions to improve forward-looking prevention and control of bond credit risk.
Keywords
Asset management, bond credit risk, monitoring and early warning, artificial intelligence.
Credit risk is one of the main risks faced by bond investment. Accurate identification and effective management of credit risk are conducive to the smooth operation of asset management products and fully protecting investor interests. In a broad sense, credit risk includes both default risk and valuation volatility risk caused by deterioration in credit quality. As credit assets, bonds show the characteristics of relatively certain returns and uncertain losses. Once default occurs, disposal is difficult, and from practical experience, the average recovery rate is relatively low. Therefore, early risk monitoring, early warning, and resolution play an important role in bond investment operations. Since valuation volatility risk mainly appears alongside the process of credit deterioration, this article focuses on research into monitoring and early warning of default risk.
The Situation and Requirements Faced by Risk Monitoring and Early Warning
(1) The objective need to respond to a complex and changeable risk situation. The 15th Five-Year Plan points out that China's development is in a period where strategic opportunities and risks and challenges coexist, and uncertain and unpredictable factors are increasing. From the perspective of the domestic bond market, overall credit risk is at a low level and shows structural differentiation. Regulation guides the rating industry to return to its origins, focus on risk identification, and strictly control the generalization of high-grade ratings, which is conducive to promoting more reasonable and accurate bond credit stratification. Under the disturbance of external variables and uncertainty, tail risks in certain fields may continue to be exposed. For asset management institutions, if bonds held in asset management products default, it may shock investor confidence and product operations, falling into a vicious cycle of "bond default - negative public opinion fermentation - valuation adjustment - net value decline - product redemption - further asset selling." Monitoring and early warning can move the risk management threshold forward, break the transmission chain of the vicious cycle, and contain risk in the bud.
(2) The inevitable requirement of complying with regulatory provisions and fulfilling fiduciary responsibilities. In 2018, the Guiding Opinions on Regulating Asset Management Business of Financial Institutions issued by the People's Bank of China and other departments stipulated that financial institutions need to fulfill due diligence obligations for the interests of principals. In addition, asset management institutions such as public funds, bank wealth management, and private funds should meet corresponding requirements in aspects such as investable bond grades, credit risk monitoring, valuation of defaulted bonds, emergency handling of default events, and information disclosure. This is mainly reflected in the following aspects. First, the product's investable scope. After a bond defaults, its rating and valuation usually decline accordingly, and it may no longer comply with the product's originally agreed investment scope, facing disposal pressure. Second, product liquidity. Bonds that cannot be transferred or traded after the issuer defaults will be included in liquidity-restricted assets. As liquidity-restricted assets increase, product liquidity management faces challenges. Third, product concentration. Regulation sets concentration requirements for money market funds, cash management wealth management products, and other investments in low-credit-rating assets. When default occurs, relevant products may breach concentration limits and be forced to adjust holdings, thereby indirectly affecting product operations. Fourth, information disclosure. The Administrative Measures for Information Disclosure of Asset Management Products of Banking and Insurance Institutions, effective from September 1, 2026, stipulates that for bond defaults involving a single asset whose post-look-through holding proportion is 10% or more, asset management products should disclose relevant information to investors within the prescribed time limit.
(3) An effective means of forward-looking prevention of bond default risk. The scale and number of defaults in the bond market over the years have shown certain cyclical and contagious characteristics (see Figure 1). From 2018 to 2021, affected by a combination of various factors, bond default events increased significantly. With the gradual clearing of risks and the continuous improvement of default disposal and risk resolution mechanisms, bond market default risk has visibly converged in recent years, but structural and hidden challenges in risk performance still exist, placing higher demands on the coverage breadth and response speed of monitoring and early warning. Risk monitoring and early warning is one part of full-process risk management and an important part of bond duration management. Faced with the above challenges, it is necessary to continuously broaden the monitoring scope, optimize monitoring tools and methods, and capture abnormal signals in a timely manner.
(4) The practical driver of adapting to the development of artificial intelligence (AI) to improve early warning quality and efficiency. At present, the development of artificial intelligence has entered a new stage, evolving from a simple question-and-answer tool into an intelligent assistant and productivity tool. The Boston Consulting Group's 2026 Global Asset Management Report shows that artificial intelligence is accelerating industry transformation, playing roles in reducing costs, expanding research coverage, and rapidly carrying out personalized investment plan preparation on a large scale. Traditional credit risk analysis based mainly on fundamental research cannot track a large number of bond issuers in a timely manner. The industry needs to rely on frontier technologies represented by large models and financial agents to build an all-weather risk early warning system, provide a powerful engine for monitoring and early warning work, and greatly shorten response time and improve work quality and efficiency through real-time scanning and deep tracking of the entire market.
Risk Monitoring and Early Warning Framework and Methods
(1) The general evolution process of credit risk. The exposure of an issuer's credit risk is the result of the joint action and dynamic evolution of multiple risk factors, generally going through three stages: "generation and evolution of risk factors - risk trigger point - substantive default." Initially, several risk factors appear and accumulate within the issuer itself. Subsequently, some factors further deteriorate. Combined with sudden shocks such as negative market public opinion and legal litigation, the issuer's own risk is ultimately transmitted to investors, and the quality of bond assets deteriorates, such as extension of bond principal and interest, rating downgrades, and so on. From the perspective of risk early warning, risk monitoring indicators are the external manifestation of the issuer's credit risk. In practice, monitoring and early warning work needs to adhere to a combination of top-down and bottom-up approaches, specifically manifested as risk indicator monitoring, quantitative model monitoring, and so on.
(2) Main monitoring dimensions. 1. Industry risk. The production and operation status, intensity of competition, financial characteristics, and level of default risk of an enterprise are to a large extent affected by its industry, and a combined "industry + entity" analysis is needed. First, single-industry risk. The default characteristics of bonds in various industries are shown in Figure 2. From the overcapacity industry risk from 2015 to 2016, to the real estate industry risk after 2021, different industries have large differences in development cycles, policy environments, and industry prosperity, leading to significant differentiation in risk exposure. Second, industrial chain risk. Within the same industrial chain, the risk of core enterprises may spill over to upstream and downstream. From the perspective of the industrial chain, the risk transmission path can be monitored and predicted, such as risks in the real estate industrial chain including real estate development, building materials, and construction. 2. Fundamental risk. The issuer's fundamentals include company operating risk, management risk, financial risk, and so on. In addition, attention to external ratings and outlooks, negative public opinion information, regulatory penalties, and legal litigation is also one of the important ways to reveal credit risk. With the in-depth advancement of the national "dual carbon" strategy, asset management institutions should promptly integrate environmental, social, and corporate governance (ESG) evaluations into the bond credit risk management process and carry out monitoring of negative ESG events. It should be noted that the credit risk of different bonds of the same entity differs, such as ordinary credit bonds, guaranteed bonds, and asset-backed securities. For this, further analysis can be conducted in combination with individual bonds. 3. Related-party transmission risk. First, parent-subsidiary related risk. This mainly refers to credit events occurring at a parent company or subsidiary in a group enterprise causing risk contagion within the group. Second, guarantee-related risk. If a bond issuer defaults, the guarantor needs to fulfill compensatory obligations. If this is combined with difficulties in the guarantor's operations or an excessively large amount of external guarantees, the risk may worsen. Third, cross-default clauses. Due to the existence of cross-default clauses, once a risk event triggers cross-default, it may bring a series of chain reactions. Fourth, after a single bond defaults, investor expectations and risk preferences change accordingly, leading to market risk repricing. This negative sentiment often spreads along the path of "same industry, same characteristics, same region," causing large fluctuations in the prices of related bonds and even affecting subsequent primary market bond issuance. 4. Individual bond risk. First, the manifestations of credit risk are more diverse. In recent years, in responding to bond risk, issuers have adopted various methods such as extending performance grace periods, adjusting installment repayment plans, and lowering bond redemption interest rates. Evaluation of such risk manifestations needs to be strengthened. Second, cross-border investment risk. As China's bond market steadily achieves high-level two-way opening, in the process of domestic funds investing in dim sum bonds, Chinese dollar bonds, and other bonds, the credit risk influencing factors and manifestations of these bonds differ somewhat from domestic bonds. Institutions can carry out targeted monitoring, such as paying attention to rating actions by international rating agencies and changes in dollar bond spreads, to promptly identify the transmission chain of cross-border credit.
(3) Monitoring methods. 1. Combine regular monitoring with real-time monitoring. Resources should be reasonably allocated and monitoring frequency set according to the risk size and change trend of the monitored objects. In addition, for specific matters or scenarios, real-time monitoring can be adopted to discover risk signals in a timely manner. 2. Combine daily monitoring with special monitoring. On the basis of consolidating normalized risk monitoring, an event-driven perspective can be introduced to carry out special monitoring and screening for credit risks derived from major industry risks and market emergencies. 3. Combine indicator monitoring with model monitoring. Indicator monitoring is the foundation of risk monitoring. Relying on the full-process electronic operation of technology systems, it can automatically monitor in batches and issue early warnings, realizing closed-loop management of bond credit risk identification, monitoring, early warning, disposal, and feedback. Model early warning is gradually being applied, such as machine learning-based models, logistic regression-based statistical models, and the KMV model based on option pricing. Quantitative models and monitoring indicators complement each other's advantages and jointly build a more three-dimensional risk monitoring and early warning network.
Ideas for Strengthening the Construction of the Risk Monitoring and Early Warning System
From existing research on bond credit risk monitoring and early warning, it mainly includes enterprise financial risk early warning, machine learning modeling early warning, ESG factor risk early warning, specific industry risk early warning, and cross-market risk spillovers. There is a relative lack of systematic research on bond credit risk monitoring and early warning from the investor perspective. Therefore, in order to adapt to the new situation and new requirements, it is necessary to promote transformation and upgrading around risk monitoring concepts, data, indicators, methods, applications, and other aspects, and systematically improve the core capabilities of credit risk monitoring and early warning.
(1) Governance level. First, institutional construction. Corresponding management systems should be formulated, responsibilities and division of labor in each link should be standardized, and a full-process work closed loop should be formed. In situations where there are assessment disagreements over the severity and scope of impact of risk, a collective deliberation mechanism should be established. Second, tiered management. Monitored risk information should be graded according to the degree of risk harm, urgency, and so on, with different risk levels corresponding to differentiated management requirements. For example, differentiated management should be implemented in risk response measures, feedback time limits, and so on. Third, review and optimization. By retrospectively reviewing the operating effect of the monitoring and early warning process, early warning rules, early warning processes, and disposal plans should be optimized and improved.
(2) Data level. Data is the foundation of monitoring and early warning work. The ability to collect, govern, integrate, mine, and analyze data will directly determine the early warning effect. First, data scope. Integrate risk data from various internal systems of the organization, continuously expand external risk data sources, and obtain large amounts of high-quality data in a timely manner at controllable cost. In the data collection process, properly handle issues of data legality and compliance. Second, data integration. On the basis of coordinating endogenous business data and third-party market data, further release the factor value of derived data. Third, data quality. Establish a data quality assessment system, carry out regular governance, and improve data authenticity and completeness. Fourth, build an integrated data foundation. Implement centralized data management and form a data asset repository integrating data governance, data mining, and data analysis.
(3) Indicator level. Monitoring indicators are the main carrier of risk control work. Relying on various quantitative indicators, they can depict the absolute degree and changing trend of credit risk. First, establish a unified monitoring indicator system. Establish a standardized and normalized management mechanism for the indicator system and implement dynamic optimization. Second, improve the multi-dimensional indicator library and threshold management. Enrich monitoring dimensions, reasonably set indicator thresholds, actively learn from market practice (such as the credit bond early warning function already built on the Wind terminal), and continuously accumulate and transform practical experience in risk management to continuously expand the monitoring indicator library. Third, pay attention to the forward-looking nature of market price signals. Some studies point out that implied ratings can in most cases react before external debt ratings, so abnormal changes such as bond implied ratings and secondary market price fluctuations can be tracked to capture early risk.
(4) Method level. First, deepen the integrated application of frontier technologies. Through new technologies, potential risks that were previously difficult to monitor online or costly to manage can be transformed into early warning indicators that can be implemented, monitored, and controlled. In terms of unstructured data, methods such as text mining, word cloud analysis, and quantitative modeling of negative public opinion can be used to enhance the early warning value of alternative data. Second, conduct linked monitoring across risk categories. Different types of risks, such as credit risk, market risk, and liquidity risk, have mutual transformation and contagion effects. For example, when secondary market liquidity dries up, the issuer's financing cost rises sharply or refinancing is blocked, which may ultimately lead to credit risk. Third, strengthen the construction of a risk case library, summarize the causes and evolution patterns of risk, and provide support for risk monitoring and early warning.
(5) Application level. First, risk response and resolution. The period between the emergence of bond risk signals and substantive default is a "window period" for taking risk resolution measures. Active response is needed to strive to exit before negative market information breaks out intensively and bond prices fall sharply, avoiding a series of shocks after substantive default. Second, assist investment decisions. The results of monitoring and early warning can be better applied to policy improvement, customer access, review and approval, post-investment management, and other links, placing equal emphasis on preventing risk and promoting development. Third, promote management improvement. Relying on monitoring and early warning clues, draw inferences about other cases from one instance, analyze possible weak links, and improve the level of refined management.
Advancing the Intelligent Transformation of Monitoring and Early Warning
Monitoring and early warning is a necessary component of the credit risk management process. With the in-depth development of a new round of technological revolution and industrial transformation, as well as the deep integration of emerging technologies such as AI, big data, and cloud computing with the asset management industry, asset management institutions need to actively adapt to changes in industry trends, take "AI + monitoring and early warning" as the goal, strengthen top-level design and optimize governance structures, systematically formulate implementation plans, reshape risk monitoring and early warning models, optimize data management, clarify role positioning, and build a new paradigm of intelligent risk monitoring and early warning based on AI technology.
(1) Build a bond risk management knowledge base. Under the artificial intelligence application model, the knowledge base is the "foundation" of the monitoring and early warning system, and to a certain extent determines the usability of artificial intelligence applications and the interpretability of early warning results. First, establish a bond risk knowledge management system covering the entire process of knowledge collection, processing, review, storage, and updating, and clarify management norms and operating mechanisms. Second, unify and manage diverse knowledge such as regulatory provisions, management systems, bond issuance and duration documents, knowledge accumulated in internal systems, and expert experience to form an underlying knowledge base. Through the systematic organization of massive explicit knowledge, the internalization and transformation of external knowledge, and the explicit precipitation of individual tacit knowledge, realize the accumulation, preservation, and sharing of bond risk management knowledge, and provide a solid knowledge foundation for artificial intelligence retrieval and invocation.
(2) Explore and promote the application of financial agents. Financial agents are undergoing a leap from single-point auxiliary tools to "job-level digital twins." In industry practice, some financial institutions have actively explored and studied a matrix of "digital employees" covering multiple functional roles, and have even evolved toward agent cluster collaboration. By introducing agents, automated generation of data governance and analysis reports can be achieved, the operating cost of the risk control middle office can be reduced, and the transformation of risk control operating models can be promoted. In mechanism design, the entire process of risk monitoring and early warning can be broken down into different links, and agents with different skills can be configured to automatically complete the full-process work of risk information search, risk analysis, judgment on triggering risk compliance matters, early warning grading, and early warning push. Through a continuous feedback loop, agents can be iteratively optimized to improve their adaptability and reliability in various monitoring and early warning work scenarios.
(3) Strengthen full-lifecycle management of the model system. Risk monitoring and early warning is gradually shifting from experience-driven to data-driven. In this process, large models, machine learning models, and statistical models are combined, each with its own emphasis and complementing one another, playing roles from different dimensions. First, use retrieval-augmented generation (RAG) technology to make up for shortcomings that may exist when directly invoking general large models in risk monitoring and early warning, such as outdated knowledge, content hallucination, and poor traceability. Second, advance the construction of quantitative early warning models, taking into account both early warning coverage and precision, and enhance the predictive ability of models. Third, implement full-lifecycle management of models, covering development, validation, deployment, operation, and exit, establish a manual verification mechanism for model screening results and a training feedback update mechanism, and continuously improve model early warning accuracy in practice.
(4) Innovate a human-machine collaborative monitoring and early warning model. Although the role of artificial intelligence is increasingly important, the subjective initiative of risk managers still plays a unique role, and higher requirements are also placed on human-machine collaboration. A future-oriented intelligent monitoring and early warning system urgently needs to build a new model of collaborative cooperation among risk managers, digital employees, and financial agents to improve work quality and efficiency. First, skill updating. Risk managers need to strengthen training in structured problem decomposition and the application of intelligent tools, quickly master new tools, and break complex problems into subtasks that AI can understand and execute. Second, reshaping functional positioning. After being freed from transactional work such as searching for information, compiling statistics, and writing reports, risk managers should focus more on risk judgment, risk response, and other matters. Third, at key points such as automatic operation of agents and output of quantitative model results, manual review links should be embedded when necessary to ensure the reliability of the risk management system.
(5) Strengthen risk management in artificial intelligence applications. While unleashing productivity, artificial intelligence also brings risk challenges such as model hallucination, loss of control over agent behavior, and algorithmic bias. The relevant provisions of the Guiding Opinions on the Safe Development and Application of Artificial Intelligence in the Banking and Insurance Industries should be strictly implemented, the artificial intelligence risk governance framework should be improved, and the boundaries of rights and responsibilities of each responsible entity in the risk monitoring and early warning system should be clarified. Establish and improve data security management and technical protection systems, and enhance the data security awareness of all staff. Strengthen data privacy management and effectively safeguard the legitimate rights and interests of consumers. Strengthen business continuity guarantees, improve emergency management mechanisms, prevent information technology outsourcing risks, and ensure that intelligent applications are robust and reliable. (The views in this article do not represent the opinions of the author's institution.)
References
[1] Wang Zhenhan, Song Siyue. Constructing a Credit Risk Identification Model through Implied Rating Spreads [J]. Bond, 2022(1). DOI: 10.3969/j.issn.2095-3585.2022.01.016. [2] Zhang Ling. Research on Innovating the Credit Risk Early Warning and Monitoring System for Farmer Loans - Based on a Big Data Perspective [J]. Rural Finance Research, 2017(4). [3] Zhang Ling. Investment Risks and Implications of Chinese Dollar Bonds [J]. Financial Expo (Wealth), 2020(6).
Comments