Beyond the Rating: How Disagreement Among ESG Agencies Affects Bond Credit Spreads

Abstract

Based on data from Chinese corporate bonds issued between 2014 and 2023, this study examines how ESG rating disagreement affects credit spreads. The results indicate that such disagreement significantly increases spreads through financial risk and information asymmetry channels, though this effect is mitigated by higher bond ratings. The impact is more pronounced in developed regions, highly marketized areas, less polluted and less competitive industries, non-Big Four audited firms, small enterprises, and state-owned enterprises. Increases in credit spreads are mainly driven by environmental and social rating disagreements, with the governance dimension playing a limited role.

1. Introduction

Global sustainable development has accelerated improvements in the ESG (Environmental, Social, and Governance) institutional framework. The European Union’s (EU) Sustainable Finance Disclosure Regulation took effect in March 2021, and the International Sustainability Standards Board (ISSB) released its first sustainability disclosure guidelines in June 2023. These initiatives have strengthened the standardization of ESG ratings. As ESG investment concepts spread across global capital markets, corporate ESG performance increasingly shapes bond credit spreads (Lian et al. 2023; Jiang et al. 2023). Strong ESG practices help firms lower compliance risks, stabilize operations, and raise investor confidence, which ultimately narrows credit spreads.

However, the absence of a unified ESG framework and standardized methods creates systematic differences among rating agencies. Variations in indicators, weights, and data processing drive persistent rating disagreement (Berg et al. 2022). This disagreement is reshaping traditional bond pricing rules. From an information perspective, rating discrepancies increase market asymmetry, making it harder for investors to identify firms’ true ESG risks (Arevalo et al. 2024). From a pricing perspective, such discrepancies distort risk assessments and lead to biased bond pricing (Yin et al. 2025). Therefore, studying how ESG rating disagreement influences credit spreads deepens traditional pricing theory and expands its boundaries. It also offers regulators guidance to enhance disclosure and helps investors refine portfolio strategies, carrying both theoretical and practical significance.

As the largest developing country, we have selected a sample of listed firms in China for this study, for several reasons. Firstly, China provides a valuable research sample. China integrates ESG factors into its financial system and advances sustainable development through the “dual-carbon” goal. Policy innovations, such as the Guidelines on Green Finance for the Banking and Insurance Industry issued in June 2022, strengthen institutional support and expand the ESG financial market. Secondly, according to the China Bond Market Development Report released by the National Association of Financial Market Institutional Investors in 2024, China’s bond market ranks second globally in terms of scale. In 2024, bond issuance reached 79.62 trillion yuan, while trading volume amounted to 2735.44 trillion yuan. Its scale and activity provide abundant data to study credit spreads. In addition, China’s ESG investment has expanded rapidly. According to the Green Finance International Research Institute of Central University of Finance and Economics, by the end of 2023, ESG public funds reached 586, with a total scale exceeding RMB 540 billion. More prominently, China’s green bond market also leads globally. Data from the Climate Bonds Initiative (CBI) shows that it accounted for 15% of global issuance and ranked first worldwide for two consecutive years between 2016 and 2023. These developments show that ESG factors increasingly shape China’s financial market and underscore China’s contribution to global green finance. Studying China’s bond market allows us to reveal how ESG rating disagreement influences credit spreads.

We first explore the association between ESG rating disagreement and corporate bond credit spreads. The analysis reveals a significant positive correlation after controlling for firm and bond level variables. This relationship remains when we apply instrumental variables to address endogeneity. To verify the robustness of the findings, we conduct regression tests by changing the ESG rating disagreement calculation method, changing the ESG rating disagreement sample organization, and replacing the sample range. The results consistently show a robust association between ESG rating disagreement and corporate bond credit spreads.

We further explore the mechanism by which ESG rating disagreement affects corporate bond credit spreads. One key channel is financial risk. Disagreement in ESG ratings of the same firm by rating agencies is a form of uncertainty in the information environment and a warning signal of financial risks hidden in the corporate, which will likely translate into real financial risks in the future. On the other hand, if rating agencies disagree significantly on the ESG performance of the same company, it indicates that the market is unable to recognize and capture information about the long-term sustainability of the firm. It raises information costs and prompts investors to price cautiously. We therefore identify information asymmetry as another transmission path. It is worth noting that ESG ratings are not the only reference, corporate bond ratings can provide a more accurate measure of bond default probability and a more unified framework, which can compensate for the information gap of ESG rating differences, and weaken the positive impact of ESG rating disagreement on corporate bond credit spreads.

Third, we systematically analyze the heterogeneous impact of ESG rating disagreement on corporate bond credit spreads at the micro, meso and macro levels. At the micro level, ESG rating disagreement has a greater impact on non-Big Four audited, small, and state-owned firms; at the meso level, it is more pronounced in non-heavy polluting and low-competition industries; and at the macro level, the effect intensifies in developed and highly marketized regions. In terms of the internal sub-dimensions of ESG ratings, their impacts are also differentiated, as evidenced by the fact that disagreement in the environmental and social dimensions has a significant positive impact on credit spreads, while governance disagreement is relatively limited.

This study achieves a significant breakthrough in existing research on ESG rating disagreement. While prior studies have confirmed the impact of ESG factors on bond pricing from multiple dimensions (Berg et al. 2022; Christensen et al. 2022; Brandon et al. 2022), establishing the role of non-financial information in credit spread formation, existing literature exhibits notable limitations: On one hand, existing literature predominantly focuses on the impact of individual ESG dimensions or overall ESG ratings on bond credit spreads, with insufficient research on the relationship between ESG rating disagreement and bond credit spreads. This study identifies the relationship between overall ESG rating disagreement and corporate bond credit spreads, and systematically analyzes disagreement across the environmental, social, and governance dimensions, providing more granular evidence for understanding the applicability of ESG pricing mechanisms. On the other hand, existing research has yet to reach consensus on the specific mechanisms through which ESG rating disagreement affect credit spreads, and discussions on heterogeneity factors such as region, industry, and firm characteristics remain insufficient. This paper advances the field by constructing a dual-channel mechanism model, identifying key moderating factors, and establishing a three-dimensional heterogeneity analysis framework. It systematically dissects the transmission pathways, moderating factors, and boundary conditions through which ESG rating disagreement influence bond pricing, thereby deepening the field’s progression from correlational identification to causal mechanism analysis.

The marginal contributions of our study to the existing literature are as follows: first, this study empirically examines the impact of ESG rating disagreement and the three sub-dimensions of Environmental (E), Social (S) and Governance (G) on corporate bond credit spreads by using data on Chinese listed companies and ESG rating data. We further the mechanism from the two channels of corporate financial risk and information asymmetry, to reveal the impact of ESG rating disagreement on corporate bond credit spreads. Second, this study extends the research perspective to the important role of debt ratings, and constructs a moderating effect model to investigate how debt ratings, which have a more standardized evaluation system, can modulate the impact of ESG rating disagreement on corporate bond credit spreads. Finally, we break through the traditional research framework and construct a “micro-medium-macro” dimension to comprehensively analyze the impact of ESG ratings disagreement on corporate bond credit spreads at different levels of companies, industries, and regions. Overall, this paper aims to systematically analyze the correlation mechanism between ESG ratings disagreement and corporate bond credit spreads, which enriches the existing literature in terms of theoretical depth, data application, and practical value.

The subsequent sections of this paper are organized as follows: Section 2 provides a literature review of existing studies; Section 3 explains the impact of ESG rating disagreement on corporate bond credit spreads and puts forward research hypotheses; Section 4 outlines the empirical design and constructs the analytical model; Section 5 presents the empirical findings; Section 6 discusses the theoretical contributions, practical implications, limitations, and future research directions of this study; and Section 7 provides a general conclusion of the paper.

2. Literature Review

2.1. Related Studies on Bond Credit Spreads

Bond credit spread is the most direct reflection of bond financing cost, which is the risk compensation for investors in bond price. In the field of credit bond spread research, a systematic theoretical structure has been established. Over decades, scholars developed structural model, simplified model, and hybrid model, which complement one another in methodology and application.

Merton’s (1974) structural model marked a milestone. He reconstructed corporate debt as a contingent claim on asset value, organically linked financial, equity market and bankruptcy law factors. The model constructed a corporate spread function with leverage, asset volatility, and debt maturity as the important indexes, clarifying the value of a corporation affects default risk was explained from a micro point of view. Jarrow and Turnbull (1995) proposed a simplified model that treated default as exogenous. It used market prices to estimate default probability and recovery rates. The model’s main advantage was avoiding the complex process of enterprise value modeling, which improved applicability. The hybrid model of Duffie and Lando (2001) is an extension of the former. The introduction of the information asymmetry assumption is an inheritance of the advantages of the structural model in studying the individual characteristics of the firms, and at the same time, it also draws on the improvement of the simplified model in approaching the characteristics of the market, creating a more comprehensive model for credit risk analysis. The development of the above theories, from the initial creation of the structural model, the specificity of the simplified model, and the overall innovation and integration of the hybrid model, has promoted scholars’ understanding of credit risk. These models also offer financial institutions effective tools for credit risk management.

In the existing research, the influencing factors of bond credit spreads can be mainly categorized into two levels: the internal factors of the debt-issuing enterprises as well as the external factors. At the firm level, factors such as leverage ratio, information environment, equity structure, credit rating, and nature of property rights can have a profound impact on bond credit spreads. Ericsson et al. (2009) show that higher leverage raises credit risk premiums. When a firm has greater information transparency (Yu 2005) or higher quality of accounting information (Bharath et al. 2008), the market’s suspicion of default risk is reduced, and the financing cost of bonds decreases. Bond spreads are lower when founder family shareholding is high; however, too high a founder family shareholding ratio may give rise to agency conflicts between shareholders and creditors, which in turn pushes up bond credit spreads (Anderson et al. 2003). Meanwhile, firms with higher government shareholdings face greater political risk, which raises bond financing costs, though economic conditions may alter this effect (Borisova et al. 2015). Bond spreads are lower when analysts’ forecasts for debt-issuing firms are more positive (Mansi et al. 2009). It is worth noting that a higher share of institutional investors leads to more objective ratings, while greater individual participation increases the chance of inflated ratings (Hirth 2014). Sánchez-Ballesta and García-Meca (2011) show that government ownership lowers bond issuance costs, suggesting implicit guarantees that reduce corporate debt expenses.

At the macro level, various external market factors also have an impact on bond credit spreads, including the risk-free rate, industry, political uncertainty, and investor sentiment. Lower risk-free rates (Longstaff and Schwartz 1995) and higher Treasury yields (Duffee 1998) raise financing costs. Industry conditions further influence spreads, with energy, basic materials, and communications firms facing distinct cost structures (Garay et al. 2019). Policy uncertainty significantly raises financing costs, especially in policy-sensitive sectors (Kaviani et al. 2020). Investor sentiment also plays a key role: optimism narrows spreads, while pessimism widens them (Liu and Zhang 2025).

With the development of financial markets and the evolution of research paradigms, scholars increasingly recognize the limits of traditional frameworks and the growing role of non-financial factors in credit risk pricing (Apergis et al. 2022). In this context, environmental, social, and corporate governance (ESG) factors have received extensive attention as an emerging research perspective. ESG performance influences credit spreads through multiple channels. First, green bond issuance often triggers positive market reactions (Wang et al. 2020; Tang and Zhang 2020), which is conducive to increasing shareholder value (Flammer 2021). In terms of green premium, studies also show evidence of a green premium: green bonds carry lower risk premiums than conventional bonds (Gianfrate and Peri 2019; Zerbib 2019). Strong corporate governance further reduces agency costs of creditor capital. Secondly, some studies show that strong corporate governance can strengthen the degree of corporate disclosure, reduce the information asymmetry between firms and investors (Cormier et al. 2011; Klock et al. 2005). It can also reduce the risk of default (Bhojraj and Sengupta 2003), and narrow corporate bond credit spreads. In addition, corporate social responsibility (CSR) reduces firms’ cost of capital, particularly equity financing (Dhaliwal et al. 2011; El Ghoul et al. 2011). Firms that fulfill their social responsibility better will signal commitment, strengthen their social influence, attract the attention of investors, and reduce the degree of corporate information asymmetry (Richardson and Welker 2001; Gelb and Strawser 2001). Numerous studies have found that active fulfillment of social responsibility can help firms significantly reduce various unsystematic risks (Chowdhury et al. 2021; Clancey-Shang and Fu 2024; Goss and Roberts 2011; Luo and Bhattacharya 2009; Sun and Cui 2014). The quality of CSR disclosure is significantly and negatively related to the cost of bond financing, and this negative relationship is more pronounced in firms with weak corporate governance and in regions with weak environmental governance (Gong et al. 2018).

2.2. Related Research on ESG Rating

Academic research on ESG ratings has developed in stages. Early studies focused on the correlation mechanism between ESG ratings, corporate operations, and market performance, highlighting the multifaceted value embedded in strong ESG practices. Dhaliwal et al. (2011) confirmed through their analysis that good ESG practices can effectively cut down the cost of corporate financing. Garcia and Orsato (2020) found that it has an uplifting effect on corporate valuation. In addition, improved ESG performance aids firms in better coping with external challenges and risks (Albuquerque et al. 2019; Mirza et al. 2025). Moreover, ESG performance is associated with investment efficiency (Gao et al. 2021), corporate innovation (Yang et al. 2024; Liu et al. 2024b), and firm value (Wang et al. 2022), as well as with corporate governance (Wang et al. 2022). These positive impacts are mainly realized through the paths of improving stakeholder relationships, reducing the chances of negative events, and attracting long-term stable investments. Other scholars focus on ESG influencing factors such as financial flexibility, regional regulatory environments, and green credit policy pressures (Li et al. 2025). Together with the economic consequences of ESG performance, these enrich the research framework on ESG factors.

In recent years, the accelerating trend of ESG integration in the global bond market and the strengthening of ESG disclosure requirements by regulators have drawn scholarly attention to how ESG factors affect credit spreads. Studies have shown that listed companies with higher ESG performance have lower bond credit spreads (Roggi et al. 2024; Razak et al. 2023). ESG performance reduces spreads by lowering financial risk, improving transparency, and cutting agency costs of debt (Jiang et al. 2023). This effect is more pronounced in non-state-owned enterprises and in regions with a higher degree of marketization (Lian et al. 2023). At the same time, however, other scholars have come to the opposite conclusion, finding that the relationship between firms’ environmental and social performance and bond spreads is negligible (Amiraslani et al. 2025).

With the increasing richness and diversity of ESG assessment systems, the issue of rating disagreement has gradually emerged. Academic attention has gradually shifted to its underlying causes and implications. The lack of consensus on the measurement and interpretation of ESG characteristics has led to differences in ESG ratings across third-party rating agencies for the same company (Brandon et al. 2021; Chatterji et al. 2016; Serafeim and Yoon 2023). Berg et al. (2022) suggest that differences in assessment methodological differences are a key factor contributing to rating disagreement in a number of ways, including the design of the indicator system, data collection methods, and the development of scoring rules. Christensen et al. (2022) focuses on the importance of information transparency, arguing that the quality of corporate disclosure directly shapes the consistency of rating results. Brandon et al. (2022) find the potential influence of business interest factors and the flexibility of some assessment organizations in the implementation of standards, which in turn affects the objectivity of ratings. These factors are intertwined and work together to ultimately cause significant differences in ESG rating results.

Existing research on ESG rating disagreement focuses specifically on its impact on key areas such as stock returns (Wang et al. 2024b; Avramov et al. 2022; Anselmi and Petrella 2025), corporate investment efficiency (Lin et al. 2025), green innovation (Geng et al. 2024), and credit risk (Porenta and Rant 2025). In contrast, Zhang et al. (2024a) reveal the role of ESG rating disagreement on the cost of debt financing, finding that only firms with consistently high ESG ratings benefit from cost losses and larger amounts of debt financing. Some scholars have also found that corporate financing constraints mediate the relationship between ESG rating disagreement and the cost of debt financing (Yang and Deng 2025).

2.3. Literature Review Commentary

Current research on the relationship between ESG ratings and credit spreads has preliminarily confirmed the impact of non-financial factors on bond pricing, but still faces key limitations. Firstly, most studies adopt a single-dimensional focus. Existing literature mostly focuses on the impact of a single dimension of E or S or G on credit spreads. Part of the literature examines the impact of ESG composite ratings on the cost of corporate debt financing, while there is limited exploration of the impact of ESG rating disagreement on credit spreads. Secondly, the research on rating disagreement remains limited. Although the economic consequences of ESG rating disagreement have gradually received attention, scholars have not yet reached consensus on the path of its effect on credit spreads, nor have they examined sub-dimension disagreement in depth. Thirdly, heterogeneity analysis is weak. Although a few studies have addressed the relationship between ESG rating disagreement and credit spreads, they rarely explored the effect of heterogeneity across different regional scopes, industry attributes, and corporate characteristics, which limits the external validity of the findings.

This study focuses on the relationship between ESG rating disagreement and corporate bond credit spreads. By constructing a comprehensive research framework that incorporates mediating pathways, moderating effects, multidimensional heterogeneity analysis, and disaggregated disagreement regressions across E, S, and G dimensions, it deepens the understanding of the role of non-financial information disputes in bond pricing. This approach effectively enriches the relevant research field and provides empirical evidence for practical application.

3. Theoretical Analysis and Research Hypotheses

3.1. Impact of ESG Rating Disagreement on Corporate Bond Credit Spreads

In the relationship between ESG performance and credit spreads, ESG rating disagreement is a relatively new research topic, which will affect the credit spreads of corporate bonds from a variety of channels. When rating agencies disagree sharply, capital market participants will be highly concerned about it and make more conservative decisions. ESG rating disagreement essentially reflects the market’s ambiguous expectations of the sustainable development of the enterprise, which in turn affects creditor risk pricing.

From the perspective of information asymmetry, weak ESG disclosure standards make it difficult to assess true sustainability when ratings diverge. Due to information uncertainty, creditors will demand risk compensation to guard against risk. According to signaling theory, investors can easily capture the information implied by rating disagreement in an efficient market. If the ESG ratings of major rating agencies diverge significantly, the market may interpret it as poor disclosure, “greenwash” or even weak governance. Such negative information will amplify investors’ risk expectations, which will in turn react to bond pricing.

According to stakeholder theory, large differences in different ESG ratings reflect serious problems in environmental, social, or corporate governance. For example, if a rating agency gives a lower rating to a company on the basis of negative environmental information, while another rating agency gives a higher rating to a company on the basis of positive information provided by the company, this indicates that the company has deeper problems in ESG management. According to reputation theory, in the era of prevalent ESG investment, rating disagreement will weaken the reputation of the company’s responsible investment and reduce the company’s attractiveness and reputation in the bond market. The damaged reputation will increase the cost of corporate financing, leading to wider credit spreads.

From the aforementioned analysis, we propose the following research hypothesis:

H1: 

ESG rating disagreement has a positive effect on corporate bond credit spreads.

3.2. The Mediating Role of Financial Risk

The financial risk channel plays an important role in how ESG rating disagreement affects bond credit spreads. Disagreement in ESG ratings of the same company is a form of uncertainty in the information environment and a warning signal of financial risks hidden in ESG, which will likely be transformed into real financial risks in the future. For example, rating agencies may identify environmental compliance risks from non-public data, or confused by a company’s ESG publicity gimmick. Such disagreement signals that the company may need to incur environmental penalties, litigation damages, or pay financial costs in the future, which ultimately translate into greater financial risk.

In terms of the mechanism of action, divergent ESG ratings raise corporate financing constraints. If the capital market disagrees on corporate ESG ratings, corporate ESG rating disagreement triggers financial institutions and investors to question information about the company, leading to higher financing constraints or risk premiums. Financing constraints raise the cost of corporate financing and lower the level of corporate liquidity reserves, making it difficult for corporations to cope with market volatility and uncertainty. Rating disagreement can also affect a corporation’s business network relationships. Customers may turn to competitors due to corporate ESG ratings, and suppliers may tighten their payment terms, affecting the corporation’s cash flow from operating activities and intensifying financial pressure.

At the firm level, differences in rating agencies’ ratings of corporate governance levels often imply that there are board structure problems, flawed incentives, or poor investor protection. If there are problems with a company’s internal governance, it can lead to inefficient decisions, oversight failures, and even financial irregularities in the company. Companies with poor corporate governance face higher risks of investment losses and asset erosion. Investors are bound to demand higher bond yields to compensate for the risks once they find them in the rating discrepancies.

From a capital market perspective, ESG rating disagreement can also limit corporate refinancing. Rating disagreement in bond issuance is one of the main risks associated with ratings, and for companies, rating disagreement faces tougher requirements when issuing rolled-over debt or issuing new bonds. When market sentiment fluctuates, companies with large rating disagreement are likely to experience a sharp reduction in financing channels, and the company’s refinancing difficulties will aggravate the company’s financial risk.

Given the preceding analysis, the study proposes the following hypotheses:

H2: 

Through the mediating role of financial risk, ESG rating disagreement positively effects on corporate bond credit spreads.

3.3. The Mediating Role of Information Asymmetry

The information asymmetry amplification effect is another transmission path that drives ESG rating disagreement leading to corporate bond credit spreads. Significant rating differences indicates that the market is unable to recognize and capture information about the long-term sustainable development of the company. For debt investors with conflicting ratings, it is difficult to discern the true state of a company’s ESG performance, and they can only increase the valuation risk premium to hedge against the risks associated with uncertainty. The variability in rating results may come from differences in rating methodologies, data sources, and indicator weights, but market participants cannot distinguish whether this variability is caused by differences in valuation systems or actual ESG differences. Increased information processing costs prompt investors to price prudently.

From the perspective of market effectiveness, ESG rating disagreement alters the quality of information disclosure. ESG ratings should guide investors in assessing a company’s sustainable growth, yet disagreement often creates market confusion. For listed companies with insufficient ESG disclosure, rating disagreement may mask risks and make it more difficult for investors to identify key factors. And a deteriorating information environment naturally leads creditors to demand higher income returns to hedge against information risk. Valuation differences can also cause changes in the investment behavior of institutional investors, who will tend to increase the cost of independent research and in turn affect bond pricing. Meanwhile, ESG rating disagreement can also cause differences in investors’ perception of bond risk, and lower trading activity and lower liquidity can result in higher yields.

From a firm behavior perspective, ESG rating disagreement is, in a sense, a matter of firm disclosure behavior. Some companies may selectively disclose to different rating agencies or else deliberately hide some negative ESG indicators. Such non-transparent disclosure behavior may help a company win higher ratings from some rating agencies, but it may create a larger credibility crisis once disagreement becomes visible. Investors, suspecting information manipulation, demand higher risk premiums, which widen credit spreads.

Based on the above analysis, this paper proposes the following hypothesis:

H3: 

Through the mediating role of information asymmetry, ESG rating disagreement positively effects on corporate bond credit spreads.

3.4. The Moderating Role of Debt Ratings

ESG rating disagreement influences corporate bond spreads through several channels, with debt ratings playing a clear moderating role. Debt rating refers to the assessment of the default probability of a bond issuer by a professional organization, which is a more accurate and direct measure to assess default risk. To a certain extent, it can mitigate the risks arising from ESG rating disagreement. A better debt rating for a debtor represents a better ability to repay debt and a solid financial position, reducing the impact of ESG rating disagreement on credit spreads.

In terms of information quality, debt rating evaluation standards are unified, and risk assessment is more accurate. Compared with ESG ratings, the debt rating evaluation system is more unified and can make up for the information gap of ESG rating differences. Since debt rating is a measure of bond default risk and recovery at maturity, when investors obtain information on debt rating and ESG rating at the same time, they are more inclined to use debt rating as a basis for price formation. Higher debt rating means lower default risk, which can reduce the influence of ESG rating disagreement on credit spreads.

At the investment practices level, debt ratings may carry more weight in pricing. For institutional investors, debt ratings are usually prioritized for asset allocation before non-financial factors such as ESG. It is worth noting that higher debt rating require higher disclosure requirements and earlier rating disclosure, which can enhance transparency and avoid the risk of misjudgment due to the inconsistency of ESG ratings to a certain extent.

And in terms of the risk pricing structure, generally speaking, if the debt rating adequately reflects the company’s credit risk level, the room for excess risk premium implied by ESG rating disagreement is actually very limited. Especially for AAA companies with the highest debt ratings, the market is sufficiently confident in its judgment of their solvency. As a result, ESG rating disagreement has only a minimal effect on market outcomes.

Based on the above analysis, this paper proposes the following hypothesis:

H4: 

Debt ratings can weaken the positive impact of ESG rating disagreement on corporate bond credit spreads.

4. Research Design

4.1. Data Source

This paper utilizes data on corporate bonds issued by Chinese listed companies and ESG rating data from 2014–20231 to empirically test the impact of ESG rating disagreement on corporate bond credit spreads. Listed company-related characteristics data, financial data, and bond-related data come from CSMAR (China Stock Market & Accounting Research Database) database, and ESG rating data come from six mainstream rating agencies: Wind, China Securities Index, Bloomberg, Runling Global, Susallwave, and SynTao Green Finance.

Debt issued by listed companies mainly includes corporate bonds, enterprise bonds, and other types. This paper chooses corporate bonds for the study mainly because of their high degree of marketization, transparent information disclosure, and strong sample representativeness, which can more accurately reflect the impact of ESG ratings on credit risk. Compared with the implicit government guarantee of enterprise bonds and the special terms of convertible bonds, the pricing of corporate bonds more purely reflects the credit status of the debt issuer, which is an ideal sample for studying ESG factors.

To ensure the reliability of the research design, this study matches each year’s ESG rating data with that year’s credit spreads, thus more accurately capturing the real-time impact of ESG information on bond pricing. In terms of ESG rating data, considering the differences in the rating frequency and update time of the six ESG rating agencies, this study selects the final rating data released by each rating agency in each year. For credit spreads, it measures the difference between the bond’s year-end yield to maturity and that of a Treasury bond with the same remaining maturity. This ensures that all ESG rating data are strictly earlier than the point in time when credit spreads are calculated. The treatment effectively avoids forward-looking bias while maintaining the contemporaneous correlation between ESG ratings and credit spreads as much as possible. In this way, we satisfy the chronological requirement of causality and also fully incorporate the latest assessment results of the agencies, which improves the reliability of our research conclusions.

In terms of sample selection, we restrict bond issuance to the same year as listing. This rule firstly ensures the observability of the complete duration of the bond; secondly, it can to a certain extent eliminate the time mismatch between ESG ratings and market pricing caused by inter-annual issuance; and lastly, it can enhance the homogeneity of the samples in order to strengthen the comparability of the results. After matching bond-level, firm-level, and ESG rating data, we process the samples as follows. First, we remove ST and ST* corporate samples. Second, we exclude corporate samples in the financial industry. Third, we drop corporate samples with seriously missing data on major variables. Fourth, we remove firms covered by fewer than two rating agencies in a year to compute ESG rating disagreement. Fifth, we exclude floating rate bond samples. Finally, we winsorize continuous variables at the 1% and 99% quantiles. After the above processing, the final sample includes 871 corporate bonds issued by 190 listed firms over ten years.2

4.2. Variable Definitions

Dependent Variable: Credit Spread (CS). Referring to the study of Lian et al. (2023), we measure corporate bond credit spreads as the difference between the yield to maturity on the annual closing date and the yield to maturity of treasury bonds with the same remaining maturity. For the missing yields to maturity of treasury bonds for a given year, they are calculated using interpolation. Unlike previous studies, which only use key maturities such as 5-, 7-, 10-, 15-, and 20-year treasury data to estimate the missing maturities by the interpolation method, we match maturities exactly using the full treasury yield curve from the CSMAR database, which spans 0 to 40 years. This approach improves the accuracy of our credit spread measure.

Explanatory Variable: ESG Rating Disagreement (ESG_DIS). Considering the mainstream of ESG rating agencies and the availability of data, we finally select the mean of the standard deviation of ratings between two agencies of six rating agencies (Wind, China Securities Index, Bloomberg, Runling Global, Susallwave, and SynTao Green Finance) to measure the ESG rating disagreement. The specific construction methodology is as follows:

Wind ESG ratings range from CCC to AAA, but the sample in this paper includes only six grades, from CCC to AA. We assign these grades a numeric value from 1 to 6, from low to high.

The CSI ESG ratings have nine grades from C to AAA, while the CSI ratings of the sample in this paper contain seven grades from C to A. In this paper, these seven grades are assigned a numeric value from 1 to 7 in ascending order.

Bloomberg adopts a scoring system to evaluate the ESG performance of enterprises, with a full score of 100, and the higher the score, the better the ESG performance of the enterprise. The Bloomberg rating score of this sample only includes the score range from 10.74 to 73.53, so we divide the score by ten and then rounds it up, only retaining the whole number, and finally obtains the Bloomberg rating results of seven grades from 1 to 7.

Runling Global ESG ratings have seven grades from CCC to AAA, while the Runling Global ratings in the sample of this paper contain only five grades from CCC to A. In this paper, We assign these grades a numeric value from 1 to 5, from low to high.

Susallwave ESG ratings are divided into C to AAA, a total of nine basic grades, which are detailed into 19 enhancement grades. The Susallwave ratings of the sample in this paper contain only CC to AA, a total of seven basic grades, so we assign these seven grades in ascending order as 1 to 7.

The SynTao Green Finance ratings are divided into ten levels of D, C−, C, C+, B−, B, B+, A−, A, A+, while the samples in this paper contain only six levels of C, C+, B−, B, B+, A−. Consequently, we assign these six levels in ascending order from 1 to 6 in this paper.

After assigning the original ESG scores from various institutions, this study measures the ESG rating disagreement with reference to the method proposed by Avramov et al. (2022). First, standardization is conducted to eliminate differences in the dimension and distribution of scores across institutions. The original ESG scores adopt three different scales, namely 1–5, 1–6, and 1–7. Direct comparison of these scores lacks a unified benchmark. Therefore, we use the cross-sectional percentile ranking method for standardization. In each year, the original scores of all sample companies given by each institution are converted into percentile rankings within the institution’s own score sequence. This process maps scores of different scales to the interval [0, 1], ensuring direct comparability of scores across institutions. Second, for a single company rated by multiple institutions in a specific year, we first identify all possible pairwise institution combinations. Then, we calculate the standard deviation of the standardized percentile rankings between the two institutions in each combination, which is used to measure the rating difference between each pair of institutions. Finally, the arithmetic mean of the standard deviations of all pairwise institution combinations for the company is computed, and this mean is defined as the ESG rating disagreement of the company in the corresponding year. It should be noted that companies rated by only one institution are excluded from the research sample, as inter-institutional differences cannot be calculated for them.

Mediating Variables: Financial risk and information asymmetry play important roles in how ESG rating disagreement affects bond credit spreads.

Measurement of Financial Risk (ZSCORE). The Z-value is used as a proxy for financial risk (Altman 1968). ZSCORE = 1.2X₁ + 1.4X₂ + 3.3X₃ + 0.6X₄ + 0.999X₅; where X₁ is the ratio of working capital to total assets, which is used to reflect the liquidity of the assets and the size of the characteristics; X₂ is the ratio of retained earnings to total assets, which reflects the cumulative profitability of the company; X₃ indicates the ratio of EBIT to total assets, which measures the profitability of assets; X₄ is the ratio of market value of equity to the book value of total liabilities, which can reflect the financial structure of the enterprise and the relative relationship between owner’s equity and creditor’s equity, and is used to assess the solvency; X₅ is the ratio of operating income to total assets, which is used to measure the efficiency of corporate asset turnover and the ability to utilize assets.

Measurement of Information Asymmetry (FERROR). Following Leukhardt et al. (2022), we use analyst surplus forecast bias as a proxy for information asymmetry. This variable helps capture the degree of misalignment between analysts’ expectations and actual outcomes. Analysts’ forecast samples are screened as follows: first, remove the samples with missing actual EPS and forecast dates; second, remove the records with forecast dates later than the publication date of the annual report; and finally, for multiple forecasts of the same company by the same analyst in the same year, only the last forecast is retained. The formula for analyst surplus forecast bias is as follows. FEPS is the average analyst surplus forecast and AEPS is the actual company surplus.

$$FERROR={\displaystyle \frac{\left|\left.FEPS-AEPS\right|\right.}{\left|\left.AEPS\right|\right.}}$$

Moderating Variable: Debt Rating (RATING). The credit rating at the time of bond issuance. Combined with the sample data, we code AAA as 4, AA+ as 3, and AA as 2.

Control Variables: With reference to studies such as Lian et al. (2023), He et al. (2025), and Cang and Li (2024), we select the following three levels of variables as control variables.

The bond characteristics variables include coupon rate, resalability, callability and remaining maturity. Coupon rate (CP), which is the annual interest paid by the issuer divided by the bond’s face value. Resaleability (PUT), a dummy variable, is taken as 1 if the bond contains an investor resale clause, otherwise it is taken as 0. Callability (CALL), a dummy variable, is taken as 1 if the bond contains an issuer redemption clause, otherwise it is taken as 0. Residual maturity (MATURITY), the remaining number of years from the calculation date to the maturity date of the bond. In other words, it is the remaining life of the bond from the date of calculation to the maturity date.

Firm characteristics variables include whether to disclose the internal control evaluation report, the proportion of independent directors, and the age of the firm. DISCLOSURE, a dummy variable, equals 1 if the firm discloses the internal control evaluation report and 0 otherwise. INDEP, the ratio of the number of independent directors to the total number of board members. AGE, the natural logarithm of the difference between the observation year and the listing year plus one.

Firm financial variables include asset-liability ratio, Tobin’s Q, and operating profitability. The asset-liability ratio (DA) is the ratio of total liabilities to total assets. Tobin’s Q (TQA) is the ratio of the market value of the firm to its total assets. The operating profit margin (OPM) is the ratio of operating profit to operating revenue.

4.3. Descriptive Statistics

Descriptive statistics are analyzed for the explained variable, explanatory variable, control variables, mediator variables, and moderator variable in this paper.

Table 1. Summary of Descriptive Statistics of Variables.

Variables	Obs	SD	Min	Median	Max
CS	871	1.194	−0.106	1.117	5.105
ESG_DIS	871	0.097	0.000	0.170	0.544
CP	871	1.339	2.450	3.940	7.900
PUT	871	0.488	0.000	1.000	1.000
CALL	871	0.305	0.000	0.000	1.000
MATURITY	871	1.669	1.114	4.210	9.694
DISCLOSURE	871	0.235	0.000	1.000	1.000
INDEP	871	0.061	0.308	0.364	0.583
AGE	871	0.503	0.693	2.996	3.434
DA	871	0.131	0.275	0.668	0.866
TQA	871	0.403	0.754	0.995	3.534
OPM	871	0.370	−2.036	0.109	1.174
ZSCORE	871	0.964	0.248	1.087	6.107
FERROR	751	1.406	0.003	0.208	9.935
RATING	792	0.647	2.000	4.000	4.000

presents the descriptive statistics for all variables used in this study. In our samples, the maximum value of credit spread is 5.105, the minimum value is −0.106, the mean value is 1.515, and the standard deviation is 1.194, which indicates that there is a large difference in the credit spread of different corporate bonds. ESG rating disagreement has a mean of 0.175 and a standard deviation of 0.097. Its values range between 0 (minimum) and 0.544 (maximum), which indicates that there is also a significant difference in the disagreement of ESG ratings among different companies. The mean (standard deviation) of Z-value is 1.408 (0.964). The maximum and minimum values of Z-value are 6.107 and 0.248, respectively. The maximum value of analysts’ surplus prediction bias is 9.935, the minimum value is 0.003, the mean value is 0.565, and the standard deviation is 1.406, which indicates that there is a right-skewed distribution of financial risk and information asymmetry among the sample firms. Debt ratings has a mean of 3.595 and a standard deviation of 0.647. Its values range between 2 (minimum) and 4 (maximum), which indicates that the credit qualification level of the sample bonds is generally high. In terms of the comprehensive set of control variables, all indicators show reasonable distribution characteristics.

4.4. Regression Design

This study analyzes the impact of ESG rating disagreement on corporate bond credit spreads using a high-dimensional fixed effects linear regression model. To eliminate potential confounding factors, we controlled for firm, year, industry, and province fixed effects. This approach accounted for unobservable variables including firm-specific characteristics that remain constant over time, macroeconomic cycles and policy shocks, industry-specific risks, and business model differences, as well as regional development levels and regulatory environments. Considering the clustering characteristics of multiple bonds within the same corporate year, we cluster standard errors at the firm level to ensure the validity of statistical inference. Given that ESG data from the six rating agencies are released one after another during the year, and this study uses the yield to maturity on the annual closing date of the bond to calculate credit spreads, the lagged one-period methodology leads to a timing mismatch problem. To address this issue, we adopts a more precise time-matching methodology—for each rating agency’s ESG data, we match each agency’s ESG ratings with credit spreads calculated from yields on December 31 of the same year. This approach ensures that all ESG ratings data are released before the point in time when credit spreads are calculated. It can also capture their cumulative effect on pricing and avoid potential estimation bias from mismatched years.

To examine the impact of ESG rating disagreement on corporate bond credit spreads, this paper constructs the following estimation model to test H1:

$$\begin{gathered} {CS}_{i,k,t}={\beta }_0+{\beta }_1{ESG\_DIS}_{i,t}+X+\sum Stkcd+\sum Year+\sum Industry \\ +\sum Province+{\epsilon }_{i,k,t} \end{gathered}$$

(1)

The dependent variable is ${CS}_{i,k,t}$ and the explanatory variable is ${ESG\_DIS}_{i,t}$. The subscripts i, k and t represent firm, bond, and year, respectively. X is a series of control variables, and ΣStkcd, ΣYear, ΣIndustry, and ΣProvince represent the firm, year, industry, and province fixed effects, respectively. ${\epsilon }_{i,k,t}$ represents the random error term.

Construct the following model to test H2 and H3:

$$\begin{gathered} {CS}_{i,k,t}={\beta }_0+{\beta }_1{ESG\_DIS}_{i,t}+X+\sum Stkcd+\sum Year+\sum Industry \\ +\sum Province+{\epsilon }_{i,k,t} \end{gathered}$$

(2)

$$\begin{gathered} {Mediator}_{i,t}={\beta }_0+{\beta }_1{ESG\_DIS}_{i,t}+X+\sum Stkcd+\sum Year+\sum Industry \\ +\sum Province+{\epsilon }_{i,k,t} \end{gathered}$$

(3)

${Mediator}_{i,t}$ represents the mediator variable ZSCORE/FERROR and the other variables are the same as above.

Construct the following model to test H4:

$$\begin{gathered} {CS}_{i,k,t}={\beta }_0+{\beta }_1{ESG\_DIS}_{i,t}+{\beta }_2{RATING}_{i,k,t}+{\beta }_3{ESG\_DIS}_{i,t} \times {RATING}_{i,k,t}+X \\ +\sum Stkcd+\sum Year+\sum Industry+\sum Province+{\epsilon }_{i,k,t} \end{gathered}$$

(4)

${RATING}_{i,k,t}$ represents the debt rating of firm and other variables are as above.

5. Empirical Analysis and Results

5.1. Basic Regression

Table 2. The impact of ESG rating disagreement on corporate bond credit spreads.

	(1)	(2)	(3)
Variables	CS	CS	CS
ESG_DIS	0.501 ***	0.494 ***	0.508 ***
	(2.75)	(2.75)	(2.89)
CP	0.855 ***	0.856 ***	0.856 ***
	(21.04)	(21.12)	(21.48)
PUT	−0.081	−0.081	−0.086
	(−1.45)	(−1.45)	(−1.57)
CALL	−0.112 *	−0.111 *	−0.103 *
	(−1.94)	(−1.95)	(−1.85)
MATURITY	−0.054 ***	−0.054 ***	−0.054 ***
	(−4.81)	(−4.83)	(−4.87)
DISCLOSURE		−0.017	−0.007
		(−0.28)	(−0.12)
INDEP		0.161	0.144
		(0.27)	(0.24)
AGE		0.021	0.038
		(0.09)	(0.16)
DA			0.394
			(0.81)
TQA			−0.024
			(−0.21)
OPM			0.009
			(0.23)
Constant	−2.026 ***	−2.134 ***	−2.412 ***
	(−10.63)	(−2.91)	(−2.83)
Stkcd FE	YES	YES	YES
Year FE	YES	YES	YES
Industry FE	YES	YES	YES
Province FE	YES	YES	YES
N	871	871	871
Adj. R2	0.925	0.925	0.924

*** and * indicate significant levels at 1% and 10%, respectively, with t-values in parentheses.

presents the estimation results examining the relationship between ESG rating disagreement and corporate bond credit spreads. Column (1) reports the regression results with only bond characteristic control variables. After progressively incorporating firm characteristic and financial control variables, column (3) shows that the regression coefficient of ESG rating disagreement is 0.508, significant at the 1% level. This indicates a statistically significant positive association between ESG rating disagreement and corporate bond credit spreads, thus supporting Hypothesis 1. Regarding control variables, the coupon rate shows a significantly positive coefficient at the 1% level, callability demonstrates a significantly negative coefficient at the 10% level, and remaining maturity exhibits a significantly negative coefficient at the 1% level. Thus, Hypothesis 1 of this study is confirmed.

5.2. Robustness Tests

This section supports the regression relationship between ESG rating disagreement and corporate bond credit spreads using multiple methodologies. First, we change the calculation of ESG rating disagreement ESG rating discrepancy sample institution and sample range to re-test the relationship. Second, we employ instrumental variable techniques to address endogeneity concerns. These robustness tests aim to ensure the reliability and consistency of our findings.

5.2.1. Changes in the Calculation of ESG Rating Disagreement

Table 3. Robustness test—changes in the calculation of ESG rating disagreement.

	Overall Standard Deviation (1)	Average Absolute Error (2)
Variables	CS	CS
ESG_DIS	0.381 **	0.127 **
	(2.38)	(2.38)
Controls	YES	YES
Stkcd FE	YES	YES
Year FE	YES	YES
Industry FE	ES	YES
Province FE	YES	YES
N	871	871
Adj. R2	0.924	0.924

** indicates significant levels at 5%, with t-values in parentheses.

demonstrates the robustness test based on the change in methodology for calculating ESG rating disagreement. The above regressions use the mean of two-by-two standard deviations across the ESG ratings of the six agencies to measure ESG rating disagreement. In order to further confirm the positive effect of ESG rating disagreement on corporate bond credit spreads, we also apply using two alternative measures: the overall standard deviation of the ESG ratings of the six agencies (Christensen et al. 2022) and the mean absolute error (Dhaliwal et al. 2011) to calculate ESG rating disagreement and regress it. As shown in Column (1), the regression coefficient of ESG rating disagreement measured by the overall standard deviation of ESG ratings on corporate bond credit spreads is 0.381, which is significantly positive at the 5% significance level. Column (2) shows that the mean absolute error produces a coefficient of 0.127, also positive and significant at the 5% level. The above results both remain largely consistent with the benchmark regression results, confirming the robustness of the positive relationship between ESG rating disagreement and corporate bond credit spreads.

5.2.2. ESG Rating Discrepancy Sample Institution Change

In constructing the initial model of ESG rating disagreement indicator, this paper mainly selects Bloomberg as a representative agency for foreign ESG ratings and combines five domestic mainstream ESG rating agencies, in order to capture methodological differences between Chinese and foreign rating agencies and their impact on credit spreads. However, considering potential biases in the assessment frameworks, data sources, and coverage of different ESG rating agencies, the analysis further incorporates Russell and Ming Sheng, two international authoritative ESG rating agencies, to ensure the robustness of the study’s conclusions. The regression results are shown by

Table 4. Robustness Test—ESG Rating Agency Change.

	Addition of Russell (1)	Addition of Russell and Ming Sheng (2)
Variable	CS	CS
ESG_DIS	0.508 ***	0.516 ***
	(2.82)	(2.90)
Controls	YES	YES
Stkcd FE	YES	YES
Year FE	YES	YES
Industry FE	YES	YES
Province FE	YES	YES
N	871	871
Adj. R2	0.924	0.924

*** indicates significant levels at 1%, with t-values in parentheses.

. From Column (1), after adding Russell, the regression coefficient of ESG rating disagreement on corporate bond credit spreads is 0.508, which is significantly positive at 1% significance level. In Column (2), after adding both Russell and Ming Sheng, the regression coefficient of ESG rating disagreement on corporate bond credit spreads rises slightly to 0.516, still significant at the 1% level. These results align with the benchmark regression and confirm that ESG rating disagreement consistently exerts a significant effect on credit spreads across different agency configurations.

5.2.3. Replacement of Sample Range

In the basic regression, some listed companies may issue multiple bonds in the same year, resulting in several bond observations for the same firm in the sample. Firms with frequent issuance may be over-represented in the sample, causing the estimates to be biased toward the characteristics of such firms. In addition, the credit spreads of multiple bonds issued by the same firm in a given year may be affected by the same unobserved factors, leading to correlation in the error term. To exclude the potential interference of the number of bond issues on the results, this paper randomly retains one bond as a proxy for multiple bonds per year for each listed firm in the robustness test, ensuring that each stkcd-year portfolio contributes only one observation (Chen et al. 2007).

Random screening ensures that each firm has equal weight in each year to avoid a few firms dominating the regression results. Retaining a single bond also eliminates intra-group correlation and satisfies the independence assumption of the regression model.

Table 5. Robustness Test—Alternative Sample Ranges.

	Single Bond Two-by-Two Standard Deviation Mean (1)	Overall Standard Deviation of Single Bond (2)
Variable	CS	CS
ESG_DIS	0.585 ***	0.454 **
	(2.84)	(2.84) (2.38)
Controls	YES	YES
Stkcd FE	YES	YES
Year FE	YES	YES
Industry FE	YES	YES
Province FE	YES	YES
N	508	508
Adj. R2	0.915	0.914

*** and ** ndicate significant levels at 1% and 5%, respectively, with t-values in parentheses.

reports the regression results. In Column (1), the regression coefficient of ESG rating disagreement on credit spreads, measured by the mean of the two-by-two standard deviations of ratings, equals 0.585, which is significantly positive at 1% significance level and consistent with the benchmark regression results. In Column (2), the regression coefficient of ESG rating disagreement on credit spreads, measured by the overall standard deviation of ratings, is 0.454, which is significantly at 5% significance level positive and is also generally consistent with the benchmark regression results. Together, these results reinforce the reliability of the study’s findings.

5.2.4. Endogeneity Treatments

When empirically analyzing the effect of ESG rating disagreement on corporate bond credit spreads, the model may encounter endogeneity that biases estimates. The first concern is the reverse causation problem: firms with higher credit spreads may reduce ESG disclosures or practices due to increased financing constraints, thus widening rating agencies’ disagreement on their ESG performance. The second concern is the omitted variable problem: unobservable variables may exist that simultaneously affect both ESG rating disagreement and credit spreads despite controlling for correlated variables and fixed effects. And the third is the measurement error: ESG rating disagreement construction relies on limited agency coverage, which may lead to disagreement indicators underestimating the true differences if some agencies do not cover low-transparency firms.

To mitigate the problems of reverse causation, omitted variables, and measurement error, this study adopts an instrumental variable approach following Lian et al. (2023). We use the mean ESG rating disagreement of other firms in the same province and industry as an instrumental variable for ESG rating disagreement, ensuring both relevance and exogeneity. In terms of relevance, firms in the same region and industry face similar ESG regulatory policies, resource conditions, and social expectations, which link their ESG performance and rating disagreement. In terms of exogeneity, the mean ESG disagreement of other firms in the same industry in the same province only indirectly influences the ESG disagreement of the target firms in terms of credit spreads, which is not directly affected by the individual credit risk of the target firms, thus satisfying the exclusivity constraint. It will not be directly affected by the individual credit risk of the target firm, which satisfies the exclusivity constraint.

Table 6. Endogeneity Test: 2SLS.

	First Stage (1)	Second Stage (2)
Variables	ESG_DIS	CS
ESG_DIS		0.589 *
		(1.79)
ivESG_DIS	−2.000 ***
	(−6.55)
Controls	YES	YES
Stkcd FE	YES	YES
Year FE	YES	YES
Industry FE	YES	YES
Province FE	YES	YES
N	853	853
Adj. R2	0.654	0.618

*** and * indicate significant levels at 1% and 10%, respectively, with t-values in parentheses.

reports the endogeneity test results. In the first stage, the coefficients of instrumental variables on credit spreads are significantly positive at the 1% level. In the second stage, the coefficients of ESG rating disagreement on credit spreads remain significantly positive at the 10% level. The test of non-identifiability shows that the Kleibergen-Paap rk LM statistic is 44.365, with a p-value less than 0.01, which rejects the hypothesis of “under-identifiability”. The weak identification test reports a Cragg-Donald Wald F-statistic of 298.517 and a Kleibergen-Paap rk Wald F-statistic of 43.022, both exceeding the Stock-Yogo critical value of 16.38 at the 10% level. These results reject weak instruments and confirm the validity of the instrument choice. After adopting the instrumental variable method to mitigate the endogeneity problem, the effect of ESG rating disagreement on corporate bond credit spreads is basically consistent with the benchmark regression results, indicating that the benchmark model possesses robust estimation efficiency.

5.3. The Mechanism of ESG Rating Disagreement’s Impact on Corporate Bond Credit Spreads

We have shown that ESG rating disagreement is positively related to corporate bond credit spreads and this relationship is likely to be causal. What is more important is how ESG rating disagreement affects corporate bond credit spreads. In Section 3, we have argued that information asymmetry and financial risk might serve as important channels through which ESG rating disagreement affects corporate bond credit spreads. To formally test these propositions, in this section, we empirically examine the aforementioned potential channels of influence. The results of the mediation effect test are shown in

Table 7. Channels: the mediating effect of financial risk and information asymmetry.

	(1)	(2)	(3)
Variable	CS	ZSCORE	FERROR
ESG_DIS	0.508 ***	−0.296 **	1.935 **
	(2.89)	(−2.05)	(2.26)
Controls	YES	YES	YES
Stkcd FE	YES	YES	YES
Year FE	YES	YES	YES
Industry FE	YES	YES	YES
Province FE	YES	YES	YES
N	871	871	738
Adj. R2	0.924	0.971	0.430

*** and ** indicate significant levels at 1% and 5%, respectively, with t-values in parentheses.

.

Column (1) shows that ESG rating disagreement significantly increases corporate bond credit spreads. In Column (2), it can be seen that the regression coefficient of ESG rating disagreement on Z-value is −0.296, which is significantly negative at 5% significance level. It indicates that the higher the ESG rating disagreement, the higher the corporate financial risk. Firms’ financial risk directly raises the probability of default, prompting investors to demand a higher risk premium to compensate for the potential default risk, which raises the credit spread of bonds (Lin et al. 2025; Dai et al. 2024). In addition, financially weaker firms also face lower bond liquidity, market makers demand higher bid-ask spreads, and corporate bond credit spreads increase (Friewald et al. 2012; Dick-Nielsen and Rossi 2019). Therefore, Hypothesis 2 of this study is supported.

Column (3) shows that ESG rating disagreement raises analysts’ surplus forecast bias, with a coefficient of 1.935 significant at the 5% level. This result indicates that the higher the ESG rating disagreement, the higher the information asymmetry. Due to higher information asymmetry, creditors will demand higher risk premiums to compensate for the assessment difficulties and potential default risks caused by information opacity, thus driving up credit spreads (Lu et al. 2010). Therefore, Hypothesis 3 of this study is verified.

5.4. The Moderating Roles of Debt Ratings

Table 8. The moderating effect of debt ratings.

	(1)	(2)	AA&AA+ (3)	AAA (4)
Variables	CS	CS	CS	CS
ESG_DIS	0.508 ***	2.026 ***	0.698 *	0.171
	(2.89)	(2.65)	(1.73)	(0.68)
RATING × ESG_DIS		−0.444 *
		(−1.96)
RATING		0.182 **
		(2.49)
Controls	YES	YES	YES	YES
Stkcd FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
N	871	779	229	519
Adj. R2	0.924	0.929	0.904	0.900

***, **, and * indicate significant levels at 1%, 5%, and 10%, respectively, with t-values in parentheses.

reports the moderating role of debt ratings in the effect of ESG rating disagreement on corporate bond credit spreads. From Column (2), the coefficient of ESG rating disagreement on credit spreads after considering debt ratings is 2.026 and significantly positive at the 1% level, consistent with the benchmark regression. The interaction term between debt ratings and ESG rating disagreement equals −0.444, significantly negative at the 10% level, indicating that debt ratings weaken the positive impact of ESG rating disagreement on credit spreads. Hypothesis 4 of this study is verified.

In order to verify this relationship in depth, we divide the sample into two groups by median rating. Bonds rated AA+ and AA form the RATING < 4 group, while AAA bonds form the RATING = 4 group. From Column (3), the regression coefficient of ESG rating disagreement on credit spreads is 0.698 for non-AAA rated bonds, which is significantly positive at the 10% level. In contrast, Column (4) shows a coefficient of 0.171 for AAA bonds, but it is not statistically significant, which corroborates the above moderating effect of debt ratings.

5.5. Heterogeneity Analysis

5.5.1. Heterogeneous Analysis Based on Regional Level

Based on the regional development where the firms are located, we divide firms into Beijing–Shanghai–Guangzhou and non-Beijing–Shanghai–Guangzhou groups.

Table 9. Heterogeneous Analysis Results Based on Regional Level.

	Beijing-Shanghai-Guangzhou (1)	Non-Beijing-Shanghai-Guangzhou (2)	High Marketization (3)	Low Marketization (4)
Variables	CS	CS	CS	CS
ESG_DIS	0.557 ***	0.486	0.748 ***	0.374
	(2.94)	(1.42)	(2.78)	(1.21)
Controls	YES	YES	YES	YES
Stkcd FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
N	484	385	418	426
Adj. R2	0.880	0.923	0.920	0.911

*** indicates significant levels at 1%, with t-values in parentheses.

reports subgroup results. As shown in columns (1) and (2), ESG rating disagreement significantly increases credit spreads at the 1% level in the more developed Beijing–Shanghai–Guangzhou region but shows no significant effect elsewhere.

Additionally, we classify provinces into high and low marketability groups using the median of their marketization index. The data of marketization index is obtained from China Marketization Index (1997–2019). And then we project values for 2020–2023 by using the historical average growth rate of all years. As shown in columns (3) and (4), for provinces with higher marketization, the regression coefficient of ESG rating disagreement on corporate bond credit spreads is significantly positive at the 1% significance level, while for provinces with lower marketization, the effect of ESG rating disagreement is insignificant. This result further strengthens the results of the earlier heterogeneity analysis.

5.5.2. Heterogeneous Analysis Based on Industry Level

The study classify firms by the pollution level of their industries into heavily polluting and non-heavily polluting groups based on the degree of environmental pollution caused by the production activities of the industries in which the sample enterprises are located. The identification of heavily polluting industries is mainly based on the Guidelines for Environmental Information Disclosure of Listed Companies formulated by the Ministry of Environmental Protection in 2010 and the Guidelines for Industry Classification of Listed Companies revised by the China Securities Regulatory Commission in 2012. It mainly covers 16 heavily polluting industries including coal, chemicals, and textiles. The regression analysis results in columns (1) and (2) of

Table 10. Heterogeneous Analysis Results Based on Industry Level.

	Non-Heavily Polluting (1)	Heavily Polluting (2)	Low Industry Competition (3)	High Industry Competition (4)
Variables	CS	CS	CS	CS
ESG_DIS	0.556 **	0.768	0.784 ***	0.167
	(2.50)	(1.46)	(3.20)	(0.35)
Controls	YES	YES	YES	YES
Stkcd FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
N	585	285	417	416
Adj. R2	0.931	0.902	0.933	0.910

*** and ** indicate significant levels at 1% and 5%, respectively, with t-values in parentheses.

indicate that ESG rating disagreement significantly increases corporate credit spreads in non-heavily polluting industries at the 5% level, but has no significant effect in heavily polluting industries.

In addition, we use the median of the industry Lerner index of the sample firms to separate high-competition and low-competition industries. The higher the industry Lerner index, the lower the degree of competition in the industry and the stronger market monopoly power of enterprises. The regression analysis results in columns (3) and (4) of Table 10 indicate that the regression coefficient of ESG rating disagreement on corporate bond credit spreads is significantly positive at the 1% level in low-competition industries, while the effect of ESG rating disagreement is not significant in highly competitive industries.

5.5.3. Heterogeneous Analysis Based on Firm Level

Firstly, the sample firms are divided into two dimensions based on whether their auditors are from the Big Four accounting firms. The regression analysis results in columns (1) and (2) of

Table 11. Heterogeneous Analysis Results Based on Firm Level.

	Non-Big Four Audited (1)	Big Four Audited (2)	Small Scale (3)	Large Scale (4)	SOEs (5)	Non-SOEs (6)
Variables	CS	CS	CS	CS	CS	CS
ESG_DIS	0.393 **	0.309	0.393 **	0.295	0.294 **	0.874
	(2.17)	(0.75)	(3.85)	(0.65)	(2.11)	(1.57)
Controls	YES	YES	YES	YES	YES	YES
Stkcd FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
N	595	272	426	427	644	224
Adj. R2	0.924	0.910	0.922	0.917	0.892	0.891

** indicates significant levels at 5%, with t-values in parentheses.

indicate that the regression coefficient of ESG rating disagreement on corporate bond credit spreads is significantly positive at the 5% level for the sample of firms audited by non-Big Four accounting firms, while the effect of ESG rating disagreement is insignificant for firms audited by Big Four.

Secondly, the median firm size of the sample firms is used as the criterion to divide the two dimensions of small and large firm size. The regression results in columns (3) and (4) reveal a significantly positive relationship between ESG rating disagreement and credit spreads for small firms, with a 1% significance level. This indicates that the market is more inclined to regard ESG rating disagreement of small firms as an important risk signal. For large firms, however, ESG rating disagreement does not have a significant effect on credit spreads of corporate bonds.

Lastly, we classify the sample firms into state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) according to the different nature of corporate ownership. The regression analysis results in columns (5) and (6) of Table 11 indicate that in the sample of state-owned enterprises, the regression coefficient of ESG rating disagreement on credit spreads is significantly positive at the 5% level, indicating that the market will incorporate the ESG rating disagreement of these enterprises into the credit risk assessment. Whereas, in non-state-owned enterprises, ESG rating disagreement does not have a significant impact on corporate bond credit spreads.

5.6. The Heterogeneous Impact of ESG Sub-Dimensions Rating Disagreement

In the previous benchmark regression, this paper examines the overall impact of ESG composite rating disagreement on corporate bond credit spreads and finds that the widening of ESG rating disagreement significantly increases credit spreads. In order to deeply examine the heterogeneity of the impact of ESG disagreement, this paper further explores the impact of rating disagreement on credit spreads in the environmental, social, and corporate governance dimensions separately. This approach identifies which dimension drives stronger effects on credit spreads. In this paper, we construct rating disagreement indicators using the ESG rating data from Wind, CSI, Bloomberg and Runling Global. The initial data source contains ESG ratings of six agencies. We exclude Susallwave due to the lack of environmental, social, and corporate governance segmentation data as it only provides comprehensive ESG ratings. Meanwhile, there are structural changes in the scoring system of SynTao Green Finance, which shifted from simple summation to weighted scoring, leading to the inconsistent scoring standards of the segmentation dimensions. It is also excluded to ensure data comparability.

Table 12. Heterogeneous Impact of ESG Sub-Dimension Rating Disagreement.

	(1)	(2)	(3)
Variable	CS	CS	CS
E_DIS	0.334 **
	(2.31)
S_DIS		0.327 *
		(1.84)
G_DIS			−0.166
			(−1.11)
Controls	YES	YES	YES
Stkcd FE	YES	YES	YES
Year FE	YES	YES	YES
Industry FE	YES	YES	YES
Province FE	YES	YES	YES
N	864	864	864
Adj. R2	0.925	0.925	0.925

**, and * indicate significant levels at 5% and 10%, respectively, with t-values in parentheses.

shows the results of the sub-dimension regression. As shown in Column (1), the regression coefficient of environmental rating disagreement on credit spread is 0.334, which is significant and positive at 5% level. Column (2) shows that social rating disagreement raises spreads by 0.327, significant at the 10% level. And as shown in column (3), the corporate governance rating disagreement does not have any significant effect on credit spread. Taken together, these findings indicate that investors pay more attention to corporate environmental compliance risk and social responsibility effectiveness, but assign limited weight to governance controversies when pricing credit risk.

6. Discussion

6.1. Theoretical Contributions

This study systematically examines the impact of ESG rating disagreement on corporate bond credit spreads, based on bond and ESG rating data from Chinese listed companies between 2014 and 2023. Empirical results indicate that ESG rating disagreement significantly widen corporate bond credit spreads, with this conclusion remaining robust after a series of stability tests. This aligns with Zhang et al.’s (2023) findings that ESG rating disagreement elevate corporate debt financing costs. From the perspective of non-financial information uncertainty, this study provides new empirical evidence for understanding pricing mechanisms in China’s bond market. It reveals that ESG rating disagreement primarily affect credit spreads through two channels: exacerbating financial risk and increasing information asymmetry. This finding strongly resonates with the conclusions of Zhang et al. (2024b) and Jiang et al. (2023). While prior studies confirm that favorable ESG ratings narrow spreads by mitigating these risks, this research provides a reverse validation of this mechanism, demonstrating that conflicting ESG information widens spreads by exacerbating the same risk pathways. Furthermore, bond ratings effectively attenuate the positive impact of ESG rating disagreement on credit spreads. This aligns with the intrinsic logic of Wang et al.’s (2024c) finding that divergence effects are more pronounced in “unsecured bonds.” Together, these studies indicate that when uncertainty surrounds non-financial information like ESG ratings, investors increasingly rely on traditional financial metrics such as credit ratings for risk assessment. This reflects the risk-hedging function of standardized credit information in mitigating disputes over non-financial data.

This study employs heterogeneity analysis to reveal that the impact of ESG rating disagreement on credit spreads varies significantly across different scenarios, deepening our understanding of non-financial information pricing mechanisms. At the regional level, economically developed and highly marketized regions exhibit greater sensitivity to ESG rating disagreement, consistent with the findings of Lian et al. (2023). This indicates that mature institutional environments and rational investors not only recognize the value of ESG but also more accurately assess the risks stemming from its uncertainties. At the industry level, ESG rating disagreement trigger stronger market reactions among non-heavily polluting industries compared to heavily polluting sectors where environmental risks are already fully priced in. This finding cross-validates Wang et al.’s (2024a) discovery that “low-carbon companies are more sensitive to ESG disagreement” in equity markets. Furthermore, low-competition industries exhibit heightened sensitivity to ESG rating disagreement. Within these sectors, stable market positions prompt investors to focus on long-term corporate development, where ESG rating divergences are perceived as critical risk signals indicating governance deficiencies, thereby triggering significant market penalties. Conversely, in highly competitive industries, intense market competition functions as an effective external governance mechanism (Giroud and Mueller 2011), where credit risk is primarily driven by short-term financial performance, significantly diluting the impact of non-financial information like ESG rating disagreement on credit spreads. At the firm level, non-Big Four audited firms, small-scale enterprises, and state-owned enterprises are more affected by these disagreements. According to information asymmetry theory, non-Big Four audit firms and small-scale enterprises, due to their inherent information transparency deficits (Liu et al. 2024a), are more susceptible to information frictions caused by ESG rating disagreement. This finding appears to contradict Wang et al.’s (2024c) conclusion that non-SOEs exhibit greater sensitivity to rating divergences in credit spreads. However, it may reveal a deeper mechanism: markets hold higher ESG expectations for SOEs. When rating divergences occur, they trigger stronger “disappointment of expectations,” leading to elevated risk premiums. Sub-dimensional analysis indicates that in bond markets, rating disagreement in the Environmental (E) and Social (S) dimensions significantly widen credit spreads, while the Governance (G) dimension shows no significant impact. This contrasts sharply with equity markets, where governance (G) disagreement predominantly exert a significant negative effect on stock returns (Wang et al. 2024b). This divergence reflects differing investor priorities across markets: bond investors focus more on direct regulatory and reputational risks stemming from environmental and social factors, which directly impact debt security; equity investors prioritize governance structures’ influence on long-term corporate value. Evidence from both markets collectively demonstrates that the economic consequences of ESG dimensions exhibit systematic differentiation based on investor attributes and market characteristics.

This study’s exploration of ESG rating disagreement in China’s bond market yields core findings with significant implications for global markets. The mechanism through which ESG disagreement amplifies credit spreads by exacerbating financial risks and information asymmetry reflects the market’s fundamental response to uncertainty surrounding non-financial information. This mechanism may manifest more prominently in mature ESG investment markets like Europe and the United States, where investors exhibit heightened sensitivity to rating quality and disclosure systems are more robust. The pronounced influence of environmental and social dimensions holds reference value for emerging market economies at similar developmental stages. Meanwhile, the particular sensitivity of state-owned enterprises to ESG disagreement provides a crucial case study for examining corporate ESG behavior across different ownership structures. The cross-market applicability of these findings requires consideration of each market’s foundational conditions. Developed markets, with their stronger corporate governance traditions, may exhibit more pronounced impacts from governance-related disagreement. Emerging markets, whose rating systems, and investor structures are closer to China’s, are likely to benefit from the applicability of this research.

6.2. Practical Implications

This study expands the domain of ESG rating disagreement and credit spreads. In conclusion, this paper proposes the following policy recommendations for corporations, investors, rating agencies, and regulators. At the corporate level, improving the environmental data collection and disclosure system is the key to improving ESG management, especially focusing on strengthening the disclosure quality of environmental and social dimensions. At the same time, it is indispensable to build an ESG rating monitoring mechanism, and enterprises need to compare the rating results of different agencies on a regular basis to accurately optimize the disclosure of information. At the investor level, ESG ratings should be deeply integrated into the credit analysis system. When making investment decisions, it is important to consider the absolute value of ESG ratings and pay more attention to the consistency of the rating results. In addition, investors can exercise their shareholder rights to push companies to improve the quality of ESG disclosure and help establish uniform disclosure standards for the industry. In the process of improving the quality of ESG ratings, it is important for rating agencies to enhance the transparency of their methodologies. Strengthening industry collaboration, reaching consensus on key indicator definitions and measurement methods, and developing a unified rating framework and minimum disclosure requirements can effectively reduce disagreement. For major disagreement points, rating agencies should provide special explanations, and they can also develop disagreement warning products to provide investors with multi-dimensional analysis services. Regulators can start from system construction to promote the standardization of the market. Introduce ESG disclosure policies for the bond market, strictly standardize the disclosure requirements for key indicators of environmental and social dimensions, and implement stricter supervision of high-risk industries. Establish a filing management system for rating agencies, requiring regular submission of methodology changes and rating quality reports to ensure the orderly operation of the market.

6.3. Limitations and Future Research Directions

While yielding significant findings, this study also has certain limitations. First, it is based entirely on data from China’s bond market. Although this facilitates in-depth analysis of ESG’s operational mechanisms within emerging market contexts, the universality of its conclusions warrants caution. China’s unique institutional backdrop, investor structure, and regulatory environment may cause the impact mechanisms of ESG rating disagreement to exhibit characteristics distinct from other markets. Additionally, while we included six major rating agencies, differences in their coverage, methodologies, and update frequencies may introduce selection bias that affects the robustness of our findings. Finally, despite adopting internationally recognized measurement approaches, quantifying ESG rating disagreement remains methodologically challenging, with varying standardization techniques potentially yielding divergent results.

Future research directions can systematically expand upon these limitations. First, promote cross-country comparative studies by constructing a unified analytical framework to compare the impact mechanisms of ESG disagreement under different market systems. This would identify key contextual factors influencing ESG pricing and establish more universally applicable theoretical explanations. Second, expand the coverage of rating agencies, deepen investigations into the root causes of methodological differences, and establish dynamic evaluation mechanisms to track the evolution of rating systems. Third, research design could leverage policy changes to create natural experiments or incorporate cutting-edge methods like machine learning to capture nonlinear relationships between ESG disagreement and bond pricing.

7. Conclusions

This study systematically examines the impact of ESG rating disagreement on corporate bond credit spreads using data from Chinese listed companies’ bonds between 2014 and 2023. The findings reveal the following: First, ESG rating disagreement significantly widen corporate bond credit spreads, with financial risk and information asymmetry serving as key transmission channels, while bond ratings effectively mitigate this negative effect. Second, this impact exhibits significant heterogeneity. It is more pronounced in economically developed regions and areas with higher marketization levels; it is more evident in non-heavily polluting and low-competition industries; and it has a greater impact on non-Big Four audited, small-scale, and state-owned enterprises. Furthermore, rating disagreement in the environmental and social dimensions exert a significantly stronger influence on credit spreads than those in the governance dimension.

The theoretical contribution of this study lies in constructing a dual-mechanism model and a multidimensional heterogeneity analysis framework. This reveals the specific pathways and boundary conditions through which ESG rating disagreement influence bond pricing, deepening our understanding of the role of non-financial information in credit pricing. In practical terms, the findings provide empirical evidence for investors to identify ESG-related risks, for enterprises to optimize information disclosure, for rating agencies to refine evaluation systems, and for regulators to formulate standardized policies. This research holds significant reference value for promoting the healthy development of China’s green bond market.