Policy Implementation and Sustainable Governance in Chinese SOEs: A Study of Mixed-Ownership Reform and ESG Rating Divergence

Abstract

Enterprises fulfilling ESG responsibilities represent a strategic choice to achieve sustainable and high-quality development. Using the sample of A-share state-owned listed companies in China from 2015 to 2022, this study investigates the association of mixed ownership reform (MOR) of state-owned enterprises (SOEs) with ESG rating divergence. The findings reveal that MOR significantly exacerbates ESG rating divergence, particularly in firms with lower equity concentration, smaller scale, those in heavily polluting industries, and those with higher ESG disclosure levels. Robustness checks, including utilizing alternative measurement approaches, lagging sample periods, the propensity score matching (PSM) method, and the difference-in-differences (DID) model, address potential endogeneity issues and confirm the validity of these results. Further analysis demonstrates that MOR increases ESG rating divergence by reducing information transparency and cutting human capital investment, while enhancing ESG disclosure quality mitigates this divergence. These insights advance understanding of the tensions between governance reforms and sustainability metrics in transitional economies, providing a perspective for sustainable corporate governance of enterprises in the background of policy reform.

1. Introduction

As a core agenda of China’s economic system reform, the mixed ownership reform (hereafter referred to as MOR) of state-owned enterprises (SOEs) aims to optimize governance structures and invigorate enterprises by introducing non-state capital. In November 2013, the Third Plenary Session of the 18th CPC Central Committee proposed “vigorously developing the mixed ownership economy.” In September 2015, the State Council’s “Opinions on Developing the Mixed Ownership Economy of State-Owned Enterprises” outlined specific requirements for MOR. The 19th National Congress in October 2017 further clarified the reform direction as “developing the mixed ownership economy,” marking MOR’s transition from pilot exploration to comprehensive deepening. Since 2020, central SOEs have introduced over 900 billion yuan in social capital, significantly amplifying the functional role of state capital.

By the end of 2022, mixed-ownership enterprises accounted for over 70% of SOE subsidiaries. While MOR revitalizes enterprises, divergent governance philosophies among heterogeneous shareholders create multi-tasking demands on boards, posing challenges to the pursuit of dual objectives—corporate social responsibility and economic performance. This raises the central question of this study: How does the mixed ownership reform of SOEs influence ESG rating divergence in China?

The ESG (environmental, social, and governance) framework, an innovation in sustainable development concepts emerging in the 1960s, has reconstructed modern corporate value assessment systems by systematically integrating three-dimensional evaluation dimensions: environmental performance, social responsibility, and governance effectiveness. Under China’s top-level design for achieving its “dual carbon” goals, the developmental philosophy of ESG resonates with the essential requirements of Chinese-style modernization, serving as a critical institutional tool for advancing sustainable development governance. For enterprises as market entities, practicing ESG responsibilities represents a strategic choice for achieving high-quality development [1] ESG performance of enterprises is an important basis for investors, creditors, and even government departments to evaluate the comprehensive development ability of enterprises and then make corresponding decisions [2]. A crucial prerequisite for ESG evaluation, is the accuracy and validity of ESG rating outcomes [3]. Recent studies reveal persistent discrepancies in ESG ratings across institutions. For instance, different rating agencies assign significantly divergent scores to the same company. This phenomenon raises concerns, as relying on a single agency’s results may lead to biased conclusions about a firm’s ESG performance [4,5]. A notable example is Guizhou Moutai (stock code: 600519), which received a “C+” rating from SynTao Green Finance in 2020, while Huazheng Index assigned it an “AA” rating simultaneously, highlighting stark contradictions. Such discrepancies create confusion for investors, complicating their assessment of corporate ESG quality and even casting doubt on a firm’s actual development status [6]. Ref. [7] further demonstrate that ESG rating divergence undermines the risk-mitigation function of ESG information, substantially increasing investors’ information search costs.

Despite the growing importance of ESG ratings, the effect of MOR on ESG rating divergence remains unexplored. This study addresses this gap by investigating how MOR affects ESG rating divergence from 2015 to 2022. Our analysis finds that MOR significantly increases rating divergence, particularly in firms with lower equity concentration, smaller scale, heavier pollution, and higher ESG disclosure levels. Further evidence shows that MOR amplifies divergence by reducing information transparency and human capital investment, while high-quality ESG disclosure can mitigate this effect.

This study makes three primary contributions. First, it enriches the literature on the economic consequences of MOR of state-owned enterprises by examining the relation of MOR to ESG rating divergence from a social responsibility perspective. This provides both theoretical and empirical evidence for resolving debates over the effectiveness of SOE reforms and advancing MOR practices. This exploration is pivotal for advancing sustainable development, as it not only provides robust theoretical and empirical support for evaluating the efficacy of SOE reforms in achieving long-term environmental, social, and economic sustainability but also propels the advancement of MOR practices towards greater alignment with global sustainability standards. Second, it identifies that abundant information disclosure yet low transparency in MOR exacerbates ESG rating divergence, offering practical insights for optimizing ESG implementation during SOE reforms. Specifically, we uncover that while MOR is associated with an abundance of information disclosure, this does not necessarily translate into higher transparency. Instead, the presence of copious yet opaque information can exacerbate ESG rating divergence. This paradoxical situation arises because, in the absence of genuine transparency, the sheer volume of information can overwhelm stakeholders, making it difficult to discern meaningful insights and leading to divergent interpretations of a company’s ESG performance. This discovery offers profound practical implications for refining ESG implementation strategies during SOE reforms. It underscores the necessity of not just increasing the quantity of disclosed information but also enhancing its quality and transparency to truly foster sustainable development. Third, it expands the literature on factors influencing ESG rating discrepancies. By meticulously dissecting the various elements that contribute to divergent ESG ratings, this study sheds light on the complex interplay between corporate governance structures, disclosure practices, and stakeholder perceptions. For investors, these insights are valuable as they navigate the increasingly intricate landscape of sustainable investing. Armed with a clearer understanding of how different factors can skew ESG ratings, investors are better positioned to make rational, informed decisions that align with their sustainability objectives. This not only enhances the effectiveness of their investment strategies but also contributes to the broader goal of driving capital towards companies that genuinely embody sustainable practices.

The remainder of this paper is organized as follows: Section 2 reviews the relevant literature and existing studies. Section 3 presents theoretical analyses and develops hypotheses on how MOR affects ESG rating divergence. Section 4 details the model specification and variable definitions. Section 5 conducts empirical tests to explore the relationship between MOR and ESG rating divergence, including heterogeneity and mechanism analyses. Finally, Section 6 concludes with policy implications and recommendations.

2. Literature Review

With the rise of ESG, research has expanded from examining its economic consequences [8] to its determinants [9], with progressively deeper analysis. In recent years, growing attention has focused on discrepancies in corporate ESG ratings. Current research on ESG rating disagreements primarily explores their causes, contributing factors, and economic consequences. By analyzing data from several rating agencies, it is found that there are discrepancies in ESG ratings among different rating agencies, and such differences can differ in terms of their sample coverage and rating scale [10]. Through difference decomposition and weight estimation, Ref. [11] attribute these differences to inconsistent definitions of corporate sustainable development among rating agencies, differences in rating scopes, and variations in calculation methodologies. Ref. [12] identify heterogeneities in measurement indicators and methods, diversity in information sources, and organizational or commensurability differences among raters as primary drivers of ESG rating variations. Ref. [13] ascribe discrepancies to different indicator weightings, while some scholars link ESG rating differences to political and cultural contexts [14]. While these findings partially explain the causes of ESG rating disagreements, they overlook the fact that ESG rating data originates from enterprises. In [15], using results from U.S. rating agencies, find that greater ESG disclosure actually amplifies rating discrepancies. Refs. [16,17] argue that irregular ESG disclosure practices cause deviations between actual ESG performance and ratings, suggesting that improved information disclosure could reduce such discrepancies. Regarding economic consequences, existing studies investigate the impact of ESG disagreements on corporate financing [18]. Stock liquidity and future returns [19], analyst forecast accuracy [10,20], audit risk [21], corporate social responsibility behaviors [22], and innovation [23].

Research on SOE mixed-ownership reforms reveals effects on financing scale [24], innovation [25], and investment flexibility [26]. Scholars also find a positive correlation between equity integration and cash holdings, with control transfers mitigating excessive cash holdings and over-investment while increasing R&D spending and dividend payments [27]. Appointments of non-state-owned shareholders’ executives to SOEs enhance pay-performance sensitivity, curbing excessive compensation and on-the-job consumption [28] and improving performance [29]. Ref. [30] shows that excessive director appointments by non-state shareholders promote risk-taking in reformed SOEs, with higher appointment ratios linked to elevated risk levels. Overall, the existing SOE reform literature centers on risk-taking, innovation, and performance. Refs. [31,32] address the relationship between SOE reforms and social responsibility. Ref. [33] explore how ownership attributes influence rating disagreements and their causes.

In summary, the existing literature on ESG rating divergence mainly focuses on the analysis of its economic consequences, while the discussion of the factors affecting the divergence remains nascent. Studies related to MOR largely focus on corporate governance impacts, with limited research on whether reforms affect the behavior of the capital market information intermediary, like ESG ratings. This paper identifies the relationship between MOR and ESG rating discrepancies, analyzing whether private capital’s monitoring effects are perceived and recognized by markets, thereby providing a basis for investor decisions. As shown in

Table 1. Comparison with prior studies.

Research Theme	Representative Studies	Core Focus	Contribution of This Study
Mixed-Ownership Reform (MOR)	[26,27,29,30,31]	Effects on governance risk, performance, and social responsibility	Extends the MOR literature to sustainability evaluation, focusing on ESG divergence
ESG Rating Divergence	[1,11,15,18]	Causes and economic consequences of divergence	Investigates MOR as a novel determinant of ESG divergence
Integrated Research Gap	—	Lack of analysis connecting ownership reform and ESG inconsistency	Provides first evidence of MOR’s relationship with ESG rating divergence through transparency and human capital pathways

, we systematically compare our study with prior literature on both MOR-related research and ESG rating divergence studies.

3. Hypotheses Development

Current research indicates that non-state-owned directors’ governance exhibits a pursuit of short-term interests, meaning that such governance increases the level of related-party transactions between state-owned enterprises (SOEs) and non-state-owned shareholders [34]. When related-party transactions serve as a tool for major shareholders to tunnel listed companies, listed companies often utilize earnings management to conceal such behaviors to avoid significant impacts on corporate performance [35]. According to the principal-agent theory and information asymmetry theory, after SOEs undergo mixed-ownership reform, non-state-owned shareholders, driven by the goal of maximizing their own interests, may induce profit-driven managerial behaviors, increase the enterprise’s risk-taking level [30], raise the probability of earnings management, lead to reduced information transparency, and ultimately prevent ESG rating agencies from obtaining highly consistent information. Based on the perspective of subjective information interpretation differences proposed by Feng et al. [1], the more information disclosed, the greater the differentiation in disclosed information among companies, and the larger the discrepancies between information sets. Compared to enterprises with a single ownership nature, after state-owned listed companies introduce private capital, shareholder information, equity change information, operating data, personnel information, etc., will all change, and various financial and non-financial information will become more complex. Especially in situations where information transparency is insufficient, this will lead to more subjective interpretations by third-party rating agencies, thereby increasing rating discrepancies.

The mixed-ownership reform of SOEs has a “governance effect” [31]. From the perspective of governance effects, the profit-seeking attribute of non-state-owned shareholders enables them to fully leverage their governance advantages and enhance the overall governance level of SOEs. This requires SOEs to make full use of information disclosure to convey positive corporate information to the outside world, enhance the role of corporate financial performance [26,36], improve internal control quality [37], and further enhance the quality of corporate social responsibility information disclosure, thereby reducing rating discrepancies [32,38]. According to the information asymmetry mitigation perspective proposed by Feng et al. [1], when listed companies have high-quality information disclosure, the degree of information disclosure can provide effective incremental information to the market, alleviate information asymmetry, and, thereby, reduce opinion discrepancies among rating agencies. After state-owned listed companies introduce private capital, they will have stronger incentives to engage in information disclosure and improve disclosure quality, sending positive signals to investors and capital markets. This information facilitates horizontal comparisons by rating agencies, thereby reducing ESG rating discrepancies.

During the process of mixed-ownership reform, discrepancies in ESG practices have become increasingly prominent: on the one hand, there is a natural tension between the profit-seeking nature of non-state-owned shareholders and the policy-oriented social responsibilities undertaken by SOEs; on the other hand, asymmetries in governance mechanisms, differences in marketization processes, and heterogeneities in control hierarchies have further exacerbated conflicts among ESG objectives. Therefore, we believe that subjective information interpretation holds true in the relationships between MOR and ESG rating discrepancies. This paper proposes the following hypotheses:

H₁: 

Mixed-ownership reform (MOR) of state-owned enterprises is positively associated with ESG rating divergence.

We present theoretical framework of our study in Figure 1. Based on information asymmetry and pricipal-agent theory, we argue that the mixed-ownership reform of state-owned enterprises (SOEs) affects ESG disagreement by influencing information transparency and human capital investment. The arrows in Figure 1. represent the underlying mechanisms of association.

4. Research Design

4.1. Sample Selection and Data Sources

This study focuses on A-share state-owned listed companies in China from 2015 to 2022, representing the population of enterprises most directly affected by the country’s mixed-ownership reform (MOR) initiatives. This population includes all firms with a controlling state ownership status as identified in the CSMAR Database. The population size of A-share state-owned listed firms during this period is approximately 1500 enterprises, resulting in 103,645 firm-year observations after screening.

The sampling frame is constructed from multiple authoritative databases. MOR data and financial indicators are obtained from the CSMAR Database, while ESG rating information is drawn from six major agencies—Huazheng, RKS, SynTao Green Finance, Wind, FTSE Russell, and Menglang. These institutions are chosen because they are the most widely used and represent both domestic and international rating perspectives, thereby ensuring comprehensive coverage and comparability. To ensure the validity of the divergence measure, we include only firms that received ratings from at least three agencies [38] within the same year.

The rationale for selecting the 2015–2022 period is twofold. First, 2015 marks the State Council’s release of the Opinions on Developing the Mixed Ownership Economy of State-Owned Enterprises, which established the institutional framework and officially launched the MOR process nationwide. Second, 2015 is the first year in which multiple ESG rating agencies began consistent coverage of A-share firms, enabling reliable cross-agency comparison. The selected time frame thus aligns with both the policy implementation and data availability required for this study.

To improve data reliability, we exclude firms labeled as ST/*ST, firms in the financial sector, and observations with missing key variables. Continuous variables are winsorized at the 1% and 99% levels to mitigate the impact of outliers.

4.2. Model Design

This study conducts OLS regression to investigate the relationship between mixed ownership reform (MOR) of state-owned enterprises and ESG rating divergence. The baseline regression model is specified as follows:

$${ESG\text{\_}div}_{it}={\beta }_0+{\beta }_1{Mix}_{it}+{\beta }_2{X}_{it}+{\phi }_{it}+{\epsilon }_{it}$$

Among them, i and t represent listed companies and years, respectively. The dependent variable $ESG$ is the divergence in ESG ratings for state-owned listed company i provided by rating agencies in year t. Drawing on the approaches of [38,39], this paper selects six ESG rating agencies: Huazheng, RKS, SynTao Green Finance, Wind, FTSE Russell, and Menglang. We include these six rating providers because they represent the major players in the ESG rating market in China, and their ratings are widely used by practitioners as well as in a growing number of academic studies [1,12,38]. Since different rating agencies differ in their scoring methods and indicator settings, we list the rating methods of the six rating agencies selected in this paper. As shown in

Table 2. The indicator system of rating agencies.

Rating Agencies	ESG Ratings Methodology
Huazheng	3 first-tier pillars, 16 second-tier themes, 44 third-tier key issues, 80 fourth-tier indicators, and 300+ underlying data points
FTSE Russell	11 areas, more than 300 indicators
SynTao Green Finance	14 core topics, 51 industry models, and 200 ESG indicators
Menglang	6 first-tier, 30+ second-tier, and 90+ third-tier indicators covering 6 different dimensions
RKS	56 industries 26 topics, 100+ indicators
Wind	500+ management practice indicators, 1200+ risk indicators

, there are significant differences in the rating indicators among these six rating agencies. Furthermore, we follow the approach of [11,40] to analyze descriptive statistics on the ESG rating scores of full samples. As presented in

Table 3. Descriptive statistics of the ESG ratings for the six rating agencies.

Panel A: Aggregate ESG Score
Rating agencies	Median		Mean	Standard Dev
Huazheng	4.44		4.55	1.16
FTSE Russell	2.4		2.66	1.10
SynTao Green Finance	5		5.41	1.08
Menglang	4.74		4.70	1.83
RKS	1.98		2.20	1.05
Wind	5.83		5.93	0.85
Panel B: Correlations between ESG ratings
	Huazheng	FTSE Russell	SynTao Green Finance	Menglang	RKS
Huazheng	1
FTSE Russell	0.31	1
SynTao Green Finance	0.32	0.65	1
Menglang	0.43	0.55	0.50	1
RKS	0.41	0.64	0.62	0.62	1
Wind	0.32	0.51	0.59	0.55	0.56
Average	0.51

, Panel A, the ESG rating medians and means of Huazheng and Menglang, FTSE Russell and RKS, and SynTao Green Finance and Wind are relatively close. Table 3, Panel B, shows that correlations at the ESG level are on average 0.51 and range from 0.31 to 0.65. In sum, these six selected agencies exhibit both differences and correlations in their ratings, which lay the foundation for our subsequent analysis.

Due to variations in scoring ranges among these agencies, comparable processing is necessary before calculating the ESG rating divergence. Currently, there are two main methods in academic research for this purpose. The first method involves unifying all rating scores into nine or ten levels and then using the standard deviation to represent the ESG rating divergence. The second method involves standardizing and ranking each rating before calculating the standard deviation to measure the ESG rating divergence. This paper follows [38] by adopting the first method to calculate the divergence of corporate ESG ratings using the standard deviation, with the maximum ESG rating score uniformly set to 10 points. Specifically, the Huazheng ESG rating system has nine levels, with scores ranging from 1 to 9, which are then multiplied by 10/9 to obtain standardized data. The FTSE Russell ESG rating ranges from 0 to 5 points, which are multiplied by 2 to achieve standardization. The Wind rating system has a maximum score of 10 points, so no further standardization is required. The SynTao Green Finance ESG rating system has ten levels, from A+ to D, with D rated as 1 and A+ rated as 10. Menglang’s ESG data has nine basic levels, further divided into 19 enhanced levels. The ratings, from strongest to weakest, are AAA, AA+, AA, AA−, A+, A, A−, BBB+, BBB, BBB−, BB+, BB, BB−, B+, B, B−, CCC, CC, and C. These are assigned scores from 19 down to 1, which are then multiplied by 10/19 for standardization. The RKS ESG rating system has a maximum score of 10 points, so no further standardization is required. If a rating agency provides multiple ratings for a company within a year, the average value is taken as the original data.

The independent variable is mixed ownership reform (${Mix}_{it}$) of state-owned enterprises. Following the existing literature [41,42], we use two metrics to capture MOR:

Dummy Variable ($Mix\text{\_}dum$), defined as whether non-state shareholders hold more than 10% of shares among the top ten shareholders. The 10% threshold is chosen because China’s Company Law stipulates that shareholders holding ≥10% of shares have the right to request an extraordinary general meeting, signifying a qualitative shift in decision-making power.

Continuous Variable ($Mix\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$): Calculated as the proportion of non-state shares among the top ten shareholders.

Following [1,26,32], we select a series of control variables and account for year and industry fixed effects to mitigate confounding factors, represent industry and year fixed effects, and include a random error term. The standard errors of estimates are clustered at the firm level. Variable definitions are provided in

Table 4. Definition of primary variables.

	Variable Name	Variable Abbreviation	Variable Definition
Dependent variable	ESG rating divergence	$\mathrm{E}\mathrm{S}\mathrm{G}\text{\_}\mathrm{d}\mathrm{i}\mathrm{v}$	The degree of divergence in ESG ratings is measured by the standard deviation of scores from six rating agencies: Huazheng, RKS, SynTao Green Finance, Wind, FTSE Russell, and Menglang
Independent variable	MOR	Mix_dum	The dummy variable is defined as whether the proportion of non-state shareholders among the top ten shareholders exceeds 10% (coded as 1 if true, 0 otherwise).
		Mix_ratio	The sum of the proportion of shares held by non-state shareholders among the top ten shareholders.
Controlled variables	Firm size	Size	The natural logarithm of total assets
	Debt-to-Asset Ratio	Lev	Debt-to-asset ratio
	Return on Total Assets	ROA	Net income/total assets
	Return on Equity	ROE	Net income/total shareholders’ equity
	Cash Flow Level	Cashflow	Operating cash flow/sales
	Sales growth rate	Growth	Revenue change rate compared to last year
	The proportion of independent directors	Indep	The number of independent directors/the number of board directors
	Top ten ratio	Top10	The shareholding ratio of the top ten shareholders
	Market-to-book ratio	BM	Market-to-book ratio
	TobinQ	TobinQ	TobinQ
	Years Since Listing	ListAge	Years since listing
	Turnover of total capital	ATO	Net sales/Average total assets
	Board size	Bsize	The natural logarithm of total directors

.

5. Results and Discussion

5.1. Descriptive Statistics

Table 5. Summary Statistics.

Variable	N	Mean	SD	Min	p25	p50	p75	Max
Esg_div	10,365	0.930	0.820	0	0	0.840	1.640	3.130
Mix_dum	10,365	0.570	0.500	0	0	1	1	1
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$	10,365	21.82	20.60	0.940	5.130	13.28	34.29	78.61
Size	10,365	22.93	1.420	19.92	21.92	22.79	23.78	27.16
Lev	10,365	0.490	0.200	0.0700	0.340	0.500	0.640	0.950
ROA	10,365	0.0300	0.0600	−0.330	0.0100	0.0300	0.0500	0.220
ROE	10,365	0.0400	0.160	−1.030	0.0200	0.0600	0.110	0.470
Cashflow	10,365	0.0500	0.0700	−0.170	0.0100	0.0500	0.0800	0.280
Growth	10,365	0.140	0.470	−0.700	−0.0500	0.0700	0.220	5.220
Indep	10,365	0.380	0.0600	0.300	0.330	0.360	0.420	0.600
Top10	10,365	0.570	0.150	0.220	0.460	0.570	0.680	0.920
BM	10,365	1.640	1.910	0.0500	0.530	1.010	1.960	14.25
TobinQ	10,365	1.900	1.460	0.770	1.080	1.410	2.110	15.40
ListAge	10,365	2.750	0.550	1.100	2.480	2.940	3.140	3.430
ATO	10,365	0.620	0.480	0.0400	0.310	0.510	0.780	3.380
Bsize	10,365	0.770	0.0900	0.480	0.730	0.790	0.790	1

reports the descriptive statistical results for the main variables. As can be seen from Table 5, the higher the degree of mixed reform in state-owned enterprises, the greater the divergence of ESG, which increases the Mix_dum and Mix_ratio values in the minimum, 25%, 50%, and 75% quantiles and maximum, and the corresponding ESG divergence also increases in the same direction. It preliminarily shows that with the increase of the degree of mixed reform, the difference in ESG rating will become more and larger. In the sample, we took Yunnan Baiyao (stock code: 000538) as an example. The company completed the mixed reform in April 2017 and completed the overall listing in 2019. The proportion of non-state-owned shareholders in the top 10 shareholders was 20.67%, 21.14%, and 44.63%, respectively, from 2017 to 2019. The ESG ratings of the six ESG rating agencies were 0, 1, and 2, respectively. It can be seen that the reform has a significant relationship to ESG divergence.

5.2. Baseline Results

Table 6. The relationship of MOR and ESG divergence.

Variables	Esgdiv	Esgdiv	Esgdiv	Esgdiv
	(1)	(2)	(3)	(4)
Mix_dum	0.077 ***	0.046 ***
	(3.91)	(2.69)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$			0.001 ***	0.001 ***
			(2.78)	(3.13)
Size		0.207 ***		0.210 ***
		(20.40)		(20.73)
Lev		−0.095 *		−0.099 *
		(−1.69)		(−1.77)
ROA		−0.882 ***		−0.896 ***
		(−3.34)		(−3.40)
ROE		−0.024		−0.021
		(−0.26)		(−0.23)
Cashflow		0.107		0.103
		(0.97)		(0.94)
Growth		−0.046 ***		−0.048 ***
		(−4.13)		(−4.27)
Indep		−0.047		−0.028
		(−0.29)		(−0.17)
Top10		0.029		−0.010
		(0.45)		(−0.16)
BM		−0.054 ***		−0.053 ***
		(−7.68)		(−7.64)
TobinQ		0.050 ***		0.051 ***
		(10.14)		(10.34)
ListAge		0.018		0.017
		(1.00)		(0.98)
ATO		−0.007		−0.006
		(−0.34)		(−0.27)
Bsize		0.011		0.037
		(0.11)		(0.34)
Constant	0.318 **	−4.363 ***	0.332 **	−4.432 ***
	(2.09)	(−17.09)	(2.25)	(−17.42)
N	10,365	10,365	10,365	10,365
R2	0.491	0.546	0.490	0.547
Controls	No	Yes	No	Yes
Industry/year	Yes	Yes	Yes	Yes

The t-statistics are reported in parentheses on robust standard errors clustered at the firm and year level. *, ** and *** designate statistical significance at the 10%, 5%, and 1% level, respectively.

reports the regression results of the relation of mixed reform of state-owned enterprises to ESG ratings. No control variables are included in columns (1) and (3) of Table 3, and no control variables are included in columns (2) and (4). Columns (1), (2), (3), and (4), respectively, make a regression for Mix_dum and Mix_ratio, two measurement indicators indicating the MOR of state-owned enterprises, and all the results control the fixed effect of industry and year. From the results in Table 6, the regression coefficients of both Mix_dum and Mix_ratio were significantly positive. This result shows that the higher the degree of MOR of state-owned enterprises, the greater the differences between ESG rating agencies, thus supporting Hypothesis 1.

5.3. Robustness Checks

We performed the robustness test by replacing the metrics and alleviating the endogeneity problem. First, we replace the explanatory variables for mixed change diversity (Mix_diversity). According to the method of [31], we classify the nature of the top ten shareholders as state-owned shareholders, private shareholders, natural persons and family shareholders, institutional investors, foreign shareholders, and other shareholders. The number of the top ten shareholders classified by nature is used as an indicator to measure the diversity of mixed reform. The results of the regression after replacement of the explanatory variables are reported in

Table 7. Robustness checks.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
VARIABLES	Mix_Diversity	Esgdiv4	Esgdiv4	L. Mix	L. Mix	PSM	PSM	DID
Mix_dum_post								0.050 *
								(1.70)
Post								−0.006
								(−0.20)
Mix_diversity	0.021 **
	(2.28)
Mix_dum		0.060 ***		0.066 ***		0.038 **
		(3.54)		(3.55)		(2.17)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$			0.001 ***		0.002 ***		0.001 ***
			(3.20)		(3.54)		(2.60)
Constant	−4.403 ***	−1.137 ***	−1.211 ***	−4.296 ***	−4.386 ***	−4.583 ***	−4.625 ***	−4.363 ***
	(−17.31)	(−4.27)	(−4.55)	(−15.55)	(−15.95)	(−17.05)	(−17.23)	(−17.09)
Controls	YES	YES	YES	YES	YES	YES	YES	YES
Observations	10,365	10,365	10,365	8572	8572	8458	8458	10,365
R-squared	0.546	0.420	0.420	0.503	0.504	0.546	0.547	0.546
industry/year	YES	YES	YES	YES	YES	YES	YES	YES

The t-statistics are reported in parentheses on robust standard errors clustered at the firm level. *, ** and *** designate statistical significance at the 10%, 5%, and 1% level, respectively.

column (1), and the coefficient of Mix_diversity is significantly positive, consistent with the results of the principal regression in Table 6. Second, the dependent variables are replaced because RKS Global only rated the companies within CSI 800, with few data and is eliminated. Moreover, because FTSE Russell belongs to a foreign rating agency, the six rating agencies are changed to 4 as the divergence indicators of ESG rating after replacement. The regression results after replacing the explained variables are reported in Table 7, columns (2)–(3). The coefficient of Mix_dum and Mix_ Ratio is significantly positive, consistent with the results of the main regression in Table 6. Third, given the potential endogeneity issue that may exist between the MOR and discrepancies in ESG ratings, we first adopt the approach of lagging the independent variable by one period. Table 7 columns (4)–(5) presents the regression results obtained by repeating the same regression as in Table 6, but with the independent variable lagged by one period. The coefficients of both Mix_dum and $Mix\text{\_} ratio$ are positive and statistically significant, indicating that the main regression results are robust. Fourth, considering the potential sample selection issue that may arise in the MOR of state-owned enterprises, we further employ the propensity score matching (PSM) method. Specifically, SOEs participating in MOR are designated as the treatment group, while other SOEs serve as the control group. The control variables from the original model are utilized as matching covariates, and kernel matching is applied. Regression analysis is then conducted using the PSM-matched samples. The PSM test results are shown in Table 7, columns (6)–(7), indicating that the original main conclusions remain unchanged after accounting for potential sample selection issues. Fifth, we adopt a multi-period difference-in-differences (DID) model, where SOEs participating in MOR are designated as the treatment group, and the year of participation, as well as subsequent years, are marked as post. The results of the DID analysis, as presented in Table 7, column (8), reveal that the interaction term between treatment and post is positive and statistically significant. This suggests that, compared to the period before participating in MOR, the ESG rating discrepancies of SOEs significantly increase after participating in the reform. Accordingly, we conclude that the core results presented in Table 6 are robust to alternative measurements of variables and to endogeneity issues.

5.4. Heterogeneity Test

Table 8. Heterogeneity test.

Panel A	(1)	(2)	(3)	(4)
	Higher Top10	Lower Top10	Higher Top10	Lower Top10
Mix_dum	0.039	0.046 **
	(1.13)	(2.36)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$			0.001	0.002 ***
			(1.30)	(3.18)
Constant	−4.585 ***	−4.260 ***	−4.685 ***	−4.303 ***
	(−10.38)	(−13.47)	(−10.50)	(−13.63)
Controls	YES	YES	YES	YES
Observations	2592	7773	2592	7773
R-squared	0.542	0.558	0.542	0.559
industry/year	YES	YES	YES	YES
Panel B	(1)	(2)	(3)	(4)
	Larger Firm	Smaller Firm	Larger Firm	Smaller Firm
Mix_dum	−0.017	0.081 ***
	(−0.72)	(3.61)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$			0.000	0.002 ***
			(0.31)	(4.11)
Constant	−5.495 ***	−1.872 ***	−5.477 ***	−1.935 ***
	(−11.87)	(−3.46)	(−11.90)	(−3.60)
Controls	YES	YES	YES	YES
Observations	4897	5468	4897	5468
R-squared	0.575	0.548	0.575	0.549
industry/year	YES	YES	YES	YES
	(1)	(2)	(3)	(4)
Panel C	Polluted = 1	Polluted = 0	Polluted = 1	Polluted = 0
Mix_dum	0.079 ***	0.034
	(2.74)	(1.63)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$			0.002 **	0.001 **
			(2.20)	(2.53)
Constant	−3.527 ***	−4.692 ***	−3.616 ***	−4.762 ***
	(−8.96)	(−15.94)	(−9.07)	(−16.24)
Controls	YES	YES	YES	YES
Observations	3175	7190	3175	7190
R-squared	0.566	0.540	0.566	0.540
industry/year	YES	YES	YES	YES
	(1)	(2)	(3)	(4)
Panel D	ESG_dis = 1	ESG_dis = 0	ESG_dis = 1	ESG_dis = 0
Mix_dum	0.064 *	−0.014
	(1.95)	(−0.50)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$			0.002 *	0.001
			(1.89)	(0.90)
Constant	−2.764 ***	−5.452 ***	−2.902 ***	−5.427 ***
	(−5.53)	(−12.14)	(−5.85)	(−12.22)
Controls	YES	YES	YES	YES
Observations	3175	7190	3175	7190
R-squared	0.566	0.540	0.566	0.540
industry/year	YES	YES	YES	YES

presents the results of the heterogeneity analysis. We mainly discuss the heterogeneity effect of state-owned enterprises from four aspects of equity concentration, company size, industry of the company, and ESG information disclosure degree.

• Equity concentration degree

The research of Jin [39] shows that equity concentration on the performance of corporate social responsibility has a significant positive effect. Because major shareholders hold large stakes in the company, they have strong incentives and sufficient influence to choose whether to disclose environmental information with high quality. On the contrary, a decentralized equity structure will weaken the effectiveness of internal control, reduce the transparency of corporate governance, and make it difficult to unify the quality and standards of ESG information disclosure, which may lead to greater differences in ESG ratings. We use the quartile of equity concentration (Top10, the sum of the shareholding ratios of the top ten shareholders) to group the sample. Those greater than the third quartile are regarded as the group with relatively higher equity concentration, and those below the third quartile are regarded as the group with lower equity concentration. Table 8 shows that the coefficients in columns (2) and (4) are both significant, and the coefficients in columns (1) and (3) are not significant, indicating that compared with enterprises with high equity concentration, the MOR of state-owned enterprises has a more significant positive relation to ESG rating divergence in enterprises with low equity concentration.

2.

Firm size

According to whether the asset size of listed companies is greater than the industry average [1], we divide our samples into large and small enterprises, and the heterogeneity of enterprise size between the MOR of state-owned enterprises and ESG rating divergence is further analyzed. Compared with larger enterprises, small companies have a relatively low degree of regulatory compliance and market attention, and the transparency of information disclosure may be lower. As a result, third-party organizations rely more on private information and have more subjective interpretations, leading to greater rating differences. Table 8, Panel B, shows that the coefficients in columns (2) and (4) are positive and significant, and the coefficients in columns (1) and (3) are insignificant, indicating that compared with large enterprises, the positive relation to small-scale SOE mixed reform on ESG rating differences is more significant.

3.

Industry

Environmental information disclosure by enterprises in heavily polluting industries faces problems such as inconsistent disclosure content, insufficient detail in disclosure, failure to disclose industry-specific environmental information, non-standardized disclosure methods, and the absence of independent third-party verification in disclosure reports. Ref. [43], which makes the comparability of information disclosure poor. Therefore, compared with non-heavy pollution industries, the ESG information disclosure quality of heavy pollution enterprises is lower, more difficult to collect and analyze information, and more likely to produce differences of opinions. We refer to [1]. According to the Ministry of Environmental Verification Industry Classification Management List of listed companies and the listed company environmental information disclosure guide, combined with the CSRC-issued 2012 listed companies’ industry classification guidelines, we determined the 16 heavy pollution industries. We define heavy pollution industry as polluted equaling 1, and 0 equaling otherwise. The results of the heterogeneity analysis are shown in Table 8, Panel C. The coefficients shown in columns (1) and (3) are positive and significant, while the coefficient in column (2) is not significant, but the coefficient in column (4) is significant. The coefficient of 0.002 in column 3 is greater than that of 0.001 in column (4), and we further find that there are significant differences between the coefficients of groups through the inter-group coefficient difference test (p-value of 0.000), indicating that the positive relationship between MOR of state-owned enterprises in heavily polluting industries and ESG rating disagreements is more significant.

4.

Degree of ESG disclosure

Bloomberg, one of the world’s most influential ESG rating agencies, will score the ESG disclosure of listed companies to measure the amount of ESG information, but it does not measure the quality of the ESG information disclosure. Following the method of [1], the degree of ESG disclosure of the company is measured by the number of ESG information disclosures and is divided into groups with higher disclosure (ESG_dis = 1) and lower disclosure (ESG_dis = 0). Table 8, Panel D, shows the results of the heterogeneity analysis. The coefficients in column (1) and column (3) are positive, and the coefficients in column (2) and column (4) are not significant, which indicates that the larger the number of ESG disclosures, the greater the ESG divergence, consistent with the inference of the previous Hypothesis 1.

5.5. Further Analysis

We have empirically verified that MOR exacerbates ESG rating divergence. MOR will lead to diversified ownership structures. Non-state shareholders may focus more on short-term economic benefits, while state-owned shareholders consider long-term goals such as social responsibility. So, it is difficult for enterprises to reach a consensus on the content and focus of ESG information disclosure. The goal of maximizing non-state shareholders’ own interests leads to enterprises’ profit-seeking behaviors [35], increasing corporate risks [30] and information opacity. Consequently, different ESG rating agencies find it hard to obtain accurate and consistent information and, thus, exacerbate ESG discrepancies. We proposed that MOR might reduce information transparency, thereby amplifying rating disagreements. To further validate this mechanism, we follow [38,44] to measure information opacity (opaque) using the absolute value of discretionary accruals calculated via the modified Jones model. Higher opaque values indicate greater information asymmetry.

Table 9. MOR, information transparency, and ESG rating divergence.

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Esgdiv	Opaque	Esgdiv	Esgdiv	Opaque	Esgdiv
Opaque			0.191 *			0.183 *
			(1.81)			(1.74)
Mix_dum	0.057 ***	0.005 ***	0.056 ***
	(3.11)	(2.98)	(3.06)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$				0.002 ***	0.000 ***	0.002 ***
				(3.52)	(3.53)	(3.45)
Constant	−4.026 ***	0.097 ***	−4.044 ***	−4.113 ***	0.089 ***	−4.130 ***
	(−15.58)	(3.83)	(−15.62)	(−15.97)	(3.51)	(−16.00)
Observations	9019	9019	9019	9019	9019	9019
R-squared	0.513	0.177	0.513	0.513	0.178	0.514
year/industry	YES	YES	YES	YES	YES	YES
Sobel	−0.005 *** (Z = −3.39)			−0.0002 *** (Z = −4.744)
Goodman-1	−0.005 *** (Z = −3.365)			−0.0002 *** (Z = −4.719)
Goodman-2	−0.005 *** (Z = −3.415)			−0.0002 *** (Z= −4.77)

The t-statistics are reported in parentheses on robust standard errors clustered at the firm level. *, ** and *** designate statistical significance at the 10%, 5%, and 1% level, respectively.

presents the mediation effect analysis. The coefficients of Mix_dum and Mix_ratio are both positive and significant, confirming that MOR reduces information transparency, which in turn intensifies ESG rating divergence. The Sobel test and Goodman test further validate the significance of the mediation effect, demonstrating the robustness of our conclusions. These results suggest that information opacity serves as a critical transmission channel between MOR and ESG rating discrepancies.

The mixed-ownership reform of state-owned enterprises can be seen as an opportunity to optimize the internal employee structure of companies. The introduction of socialized capital will enhance the incentives for SOEs to reduce costs and increase efficiency, prompting them to streamline or eliminate departments and employees with lower efficiency [45]. Thus, it reduces the level of corporate social responsibility [31]. However, subjective interpretations of this information could amplify ESG rating divergence. In addition, although the participation of non-state shareholders can reduce overstaffing and increase the scale of high-level talents [46], such high-level talents mostly focus on production efficiency and innovation, and may not possess ESG expertise. In traditional state-owned enterprises, the implicit management experience related to the fulfillment of social responsibility is often attached to the original administrative system. However, the change in management concepts brought about by MOR may disrupt this balance. This leads enterprises to be unable to systematically integrate key ESG data—such as employee rights protection and environmental compliance—into the disclosure system due to a lack of ESG expertise. As a result, rating agencies can only make subjective judgments based on fragmented information, and the differences in evaluation logic among different agencies are amplified. Drawing on [47], we measure the investment in human capital by taking the logarithm of the average employee compensation plus one.

Table 10. MOR, human capital investment, and ESG rating divergence.

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Esgdiv	ln_Human	Esgdiv	Esgdiv	ln_Human	Esgdiv
ln_Human			0.035 *			0.039 *
			(1.65)			(1.83)
Mix_dum	0.046 ***	−0.065 ***	0.048 ***
	(2.70)	(−3.95)	(2.82)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$				0.001 ***	−0.002 ***	0.001 ***
				(3.11)	(−5.17)	(3.28)
Constant	−4.360 ***	8.769 ***	−4.667 ***	−4.428 ***	8.883 ***	−4.775 ***
	(−17.06)	(32.29)	(−14.66)	(−17.38)	(32.65)	(−14.95)
Observations	10,342	10,342	10,342	10,342	10,342	10,342
R-squared	0.546	0.422	0.546	0.546	0.427	0.546
year/industry	YES	YES	YES	YES	YES	YES
Sobel	−0.034 *** (Z = −8.98)			−0.001 *** (Z = −12.58)
Goodman-1	−0.034 *** (Z = −8.98)			−0.001 *** (Z = −12.57)
Goodman-2	−0.034 *** (Z = −8.99)			−0.001 *** (Z = −12.59)

The t-statistics are reported in parentheses on robust standard errors clustered at the firm level. *, ** and *** designate statistical significance at the 10%, 5%, and 1% level, respectively.

shows that coefficients for Mix_dum and Mix_ratio are significantly negative in columns (2) and (5), indicating that MOR reduces human capital investment (ln_Human). The coefficients of the remaining Mix_dum and Mix_ratio are all positive and significant, indicating that the mixed reform of state-owned enterprises increases the ESG rating divergence by reducing human capital input. The Sobel test and Goodman test confirm the robustness of this pathway.

Furthermore, we examine whether an improvement in information disclosure quality helps to alleviate the increase in ESG disagreements resulting from MOR of state-owned enterprises. Following CSMAR’s disclosure ratings (graded A–D, scored 4–1, the higher grade or score stands for higher disclosure quality), we construct a variable DQ to measure information transparency.

Table 11. MOR, information disclosure quality, and ESG rating divergence.

	(1)	(2)
VARIABLES	Esgdiv	Esgdiv
Mix_dum_DQ	−0.061 **
	(−2.47)
Mix_ratio_DQ		−0.001 **
		(−2.14)
Mix_dum	0.246 ***
	(3.07)
$\mathrm{M}\mathrm{i}\mathrm{x}\text{\_}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}$		0.005 ***
		(3.03)
DQ	0.028	0.023
	(1.34)	(1.18)
Constant	−4.159 ***	−4.180 ***
	(−15.67)	(−16.05)
Observations	9019	9019
R-squared	0.514	0.514
year/industry	YES	YES

The t-statistics are reported in parentheses on robust standard errors clustered at the firm level. *, ** and *** designate statistical significance at the 10%, 5%, and 1% level, respectively.

reveals that interaction terms between Mix_dum, Mix_ratio, and DQ are significantly negative, while the coefficients of Mix_dum and Mix_ratio are both positive and significant. These results indicate that while MOR increases ESG divergence, improving disclosure quality effectively alleviates this effect.

6. Conclusions

This study examines the relationship between the mixed-ownership reform (MOR) of Chinese state-owned enterprises (SOEs) and ESG rating divergence using panel data of A-share SOEs from 2015 to 2022. The results reveal that MOR significantly increases ESG rating divergence, especially among firms with lower equity concentration, smaller scale, heavy-pollution industry affiliation, and higher levels of ESG disclosure. These findings remain robust after addressing potential endogeneity concerns through lagged variables, propensity score matching (PSM), and a multi-period difference-in-differences (DID) approach.

This research deepens the understanding of the interplay between corporate governance reforms and sustainability evaluation. Drawing upon information asymmetry and agency theories, we identify two mechanisms—reduced information transparency and decreased human-capital investment—through which MOR amplifies discrepancies in ESG assessments. The study thus extends the theoretical boundary of governance-sustainability research by linking ownership restructuring to information consistency in sustainability ratings. Empirically, this work enriches the literature on the economic consequences of MOR and the determinants of ESG rating divergence. By constructing a multi-agency dataset and employing rigorous robustness tests, we provide the first large-scale evidence that governance reforms within SOEs contribute to inconsistencies in sustainability ratings. The mediation analysis also validates the dual-path mechanism, offering a new empirical framework for evaluating reform outcomes in transitional economies.

For policymakers, the findings underscore the need to strengthen information-disclosure standards and internal governance transparency in restructured SOEs to mitigate ESG rating inconsistency. For rating agencies, understanding ownership heterogeneity can improve evaluation methodologies and enhance cross-agency comparability. For investors, recognizing how MOR affects the reliability of ESG scores assists in making more informed capital-allocation decisions.

Finally, regarding the limitations of the study, this study focuses exclusively on Chinese state-owned enterprises (SOEs) listed in the A-share market. While this provides strong policy relevance and allows for a detailed examination of MOR and ESG rating divergence, the conclusions may not fully generalize to private or non-listed firms, or to SOEs in other institutional environments. The sample period of 2015–2022 is relatively limited, which may constrain the ability to capture longer-term trends in ESG rating divergence. Additionally, although we carefully select ESG rating agencies and control for observable differences in their methodologies, fully accounting for all methodological variations across agencies is challenging, which could influence the measurement of divergence. Future research could extend the analysis to cross-country or multi-ownership samples, consider longer time horizons, and explore additional mechanisms to validate and broaden the applicability of these findings.