Behavioral Channels Linking Firm Characteristics and Environmental, Social, and Governance Performance: Evidence from Chinese Listed Firms

Abstract

This study examines the effects of firm characteristics on environmental, social, and governance (ESG) performance among A-share firms listed in Shanghai and Shenzhen. Drawing on the resource-based view, legitimacy theory, and agency theory, this study examines both direct and indirect mechanisms connecting corporate profitability, firm size, and ownership concentration to enhance ESG performance. This research employs secondary panel data from the CSMAR, DIB, and WIND databases (13,911 observations) and estimates two-way fixed effects models with firm-clustered standard errors. The outcomes show that profitability, firm size, and ESG performance have positive relationships, but ownership concentration is a negative factor in ESG performance. Market share and managers’ risk preferences mediate the relationship between firm characteristics and ESG performance; however, these effects are interpreted as exploratory due to panel data constraints. Internal control enhances these relationships, which emphasize the importance of the process of sustainability itself. The study provides practical insights that managers, policymakers, or investment firms can apply to improve ESG integration accountability in the emerging markets context.

1. Introduction

Environmental, social, and governance (ESG) performance has become a central benchmark for measuring firm sustainability and long-term value creation [1]. Over the past decade, profit-only evaluations by global investors and regulators have shifted towards more comprehensive approaches that incorporate social responsibility and sustainable practices into ESG evaluations of companies [2,3]. However, despite this shift, research on the role of firm-specific factors influencing ESG performance remains fragmented and has received limited empirical evidence, especially in emerging markets with varying enforcement and reporting practices distinct from those of Western nations [4,5]. China, being the world’s second-largest economy and a major player in global environmental pollution [6] and the carbon neutrality initiative, is an optimal setting, as its ESG disclosure reporting system is growing rapidly and shows divergence among firms listed on Chinese stock exchanges [7].

Although existing ESG research has provided some insights, it has some flaws and is not comprehensive enough [4,5]. Firstly, the available literature on ESG is mostly Western- or globally centered, with few studies on the determinants of ESG in the specific Chinese context [8,9]. Secondly, the available literature mainly focuses on the direct impact aspects. It neglects the importance of behavior- and strategy-related mechanisms, such as market structuring or the manager’s risk aversion within firms, in linking firm attributes to ESG outcomes [10,11]. Lovatte et al. [12] highlight the mediating role of corporate social responsibility in the relationship between CO₂ emissions and the Environmental Kuznets Curve in Brazil and call for further empirical research. Thirdly, the effect of ownership concentration in the literature on ESG research is not clear from the perspective of theoretical aspects [13,14].

Based on stakeholder theory [15], previous research has underscored that larger and transparent companies have been more susceptible to ESG pressures, as small firms lack corresponding motivations and resources [16,17]. Alternatively, some studies based on the resource-based view theory proposed that financially strong organizations have used their slack resources for sustainable projects and, as such, met two goals at once—profit maximization and establishment of legitimacy [18,19]. However, these arguments ignore the dynamic nature of firm characteristics’ effects on ESG performances [20]. There is empirical evidence suggesting lagged effects of corporate governance and financial performance on improved ESG ratings [21,22], but these have rarely been examined in China’s A-share market [23].

Moreover, there has been a lack of exploration on the behavioral and structural links between corporate features and ESG outcomes [24,25]. Financial metrics, such as return on equity (ROE), measure corporate profit but not necessarily risk preference or market intention as managerial behavior underlying ESG engagement [26]. Additionally, ownership concentration can be supportive or limiting of sustainability efforts based on whether large shareholders aim for short-term or long-term legitimacy gains [27]. To acknowledge such complex links, it is necessary to integrate agency theory [28], which emphasizes the tensions between large and small shareholders, and legitimacy theory [29], which treats ESG engagement as a behavior through which companies demonstrate congruence between corporate actions and societal expectations [30,31].

The proposed work differs by integrating the resource-based view [32], the theory of legitimacy [29], and agency theory [28] in a structured manner, explicitly connecting each corporate attribute with a theory. Company size and profitability capture resource-based competitiveness, driving ESG investment, market share captures the pressures of legitimacy related to visibility, and ownership concentration and internal control capture governance structures related to agencies influencing managerial discretion. In this manner, our work enables the investigation of direct relationships and, importantly, the mediating and moderating roles of factors related to behavior (risk preference) and governance architecture (internal control) in influencing ESG in the Chinese setting, thereby integrating theories missing in earlier research.

Based on these research gaps, this research aims to answer the following research questions (RQs):

RQ1: Do company-based characteristics, such as ROE, size of firm, and its ownership concentration, impact ESG performance in the Chinese Listed companies?

RQ2: By what means or through what behavioral and strategic channels, specifically risk preference and market share, do these links operate?

RQ3: Do internal control and effective internal control systems act as a moderator of these effects?

By addressing the above RQs, this research adds to the literature in three ways. First, it enhances theoretical explanations on ESG performance issues because it employs all three theories of resource-based view [33], agency [28], and legitimacy theory [29] together in explaining ESG performances, not only as they are directly influenced but through other variables like managerial actions and strategic orientation [34,35]. Second, it contributes methodologically by using panel data on Chinese A-share companies, testing contemporaneous and lagged hypotheses, and emphasizing the diffusion of ESG effects over time [36,37]. Lastly, this work also has practical contributions for policymakers and managers who would like to create sustainable value and link profit and governance structures through emerging markets.

The following research is organized as follows: Section 2 covers the theoretical foundation and develops hypotheses, and Section 3 describes the empirical methods, including the variables. The section covers results and tests of robustness, while covering a discussion of findings, implications, and future directions.

2. Theoretical Support and Hypotheses Formulation

2.1. Theoretical Support

Sustainability at a corporate or organizational level has moved beyond compliance and has become a strategic concern requiring organizations to balance governance, profitability, and stakeholder legitimacy [6,38]. There are three complementary theories (the resource-based view, legitimacy, and agency) that provide the analytical foundation for this study.

The resource-based view traces its origins to Penrose’s conceptualization of firms as bundles of heterogeneous resources [39], formalized by Wernerfelt [33] and extended by Barney [32], identifying four criteria for strategic resources (valuable, rare, inimitable, non-substitutable). Peteraf [40] further extended the RBV by emphasizing the role of organizational and human capital resources. The RBV theory is especially relevant for ESG studies, given that organizations, possessing superior financial and managerial resources, can allocate slack resources toward sustainability efforts [41]. Considering this, the company size and profitability (ROE) provide the capacity to facilitate investment in ESG practices, thereby improving reputation and performance [16].

According to legitimacy theory [29], organizations aim to ensure that their actions are viewed as consistent and compatible with societal values and expectations regarding the environment [42]. Through ESG practices, firms attain legitimacy, which enhances survival prospects and resource accessibility, as ESG reporting helps firms uphold a “social contract.” [43]. In China’s institutional environment, characterized by heightened public scrutiny and regulatory attention, firms with greater visibility face stronger pressures. In this context, market share acts as an essential mechanism through which organizations respond to these pressures, as highly visible organizations face greater stakeholder expectations.

The agency theory [28] focuses on conflicts of interest between managers, controllers, and minority shareholders. Ownership structures and internal control mechanisms shape whether managerial actions are consistent with shareholder preferences. Ownership concentration can either support or constrain ESG investment, depending on whether dominant shareholders prioritize long-term legitimacy or short-term private gains [27]. Internal control affects these dynamics by protecting against agency problems, improving transparency, and ensuring oversight during decision-making for ESG investments [44]. As a result, well-designed internal controls play a critical role in governance, particularly in contexts where corporations’ ESG disclosures and practices undergo intensive scrutiny.

These three theories together offer complementary insights into the role of a firm’s characteristics in determining ESG outcomes. More significantly, these theories together offer insights into the two behavioral channels identified in the research, providing a framework for analyzing the influence of a firm’s resources and governance mechanisms on ESG performance. A resource-based view of the firm helps establish the extent to which a firm can make ESG investments; legitimacy theory helps establish the pressure from the outside to make these ESG investments, while agency theory helps establish the internal governance mechanisms for ESG spending.

This theory-driven model shows that among Chinese listed companies, ESG performance is influenced not only through governance structure and availability of resource but also via behavior-driven channels such as managerial risk orientation and market visibility. These channels shape how financial capacity of an organization is translated into sustainability-oriented engagement under internal control and legitimacy pressure.

2.2. Hypotheses Formulation

2.2.1. Firm Profitability and ESG Performance

Based on the resource-based view, organizations with superior financial performance have greater slack resources, enabling them to allocate organizational and managerial attention towards sustainability-driven activities [45,46]. Higher profitability increases an organization’s ability to absorb the costs and risks related to ESG efforts while enhancing its reputation and competitiveness [47]. Financially healthy firms can allocate more funds toward green technology R&D, environmental facility upgrades, employee welfare programs, and social contributions, thereby enhancing all dimensions of ESG performance [48]. Research indicates that profitable firms engage in proactive environmental practices and provide clear social disclosures [49]. Indrawan et al. [50] observed that ROE strengthens the effects on ESG performance. Most recently, D’Amato et al. [49] and Aydoğmuş et al. [51] also reported the significant link between these constructs. Therefore, it is hypothesized that

H1: 

Corporate profitability (e.g., ROE) positively influences ESG performance.

2.2.2. Firm Size and ESG Performance

Firm size reflects market position and capacity for resource integration. Larger organizations have more resources, diversified stakeholders, and greater visibility to the public and regular attention [52]. Legitimacy theory argues that greater visibility increases the need to adopt sustainability practices to gain social acceptance [53]. Furthermore, larger organizations enjoy economies of scale advantages in developing their governance structures and sustainability reporting practices [54]. While the same cannot be said about small businesses, which may neither have the capabilities nor the motives to embrace ESG strategies holistically [55]. Large corporations receive more scrutiny from stakeholders due to their greater visibility, thereby increasing legitimacy pressures to implement ESG measures [56]. At the same time, the resource-based view argues that larger corporations have more abundant resources (tangible and intangible) to facilitate the implementation of ESG measures [41]. Consistent with these arguments, most recently, Martínez et al. [57] and Shen [58] observed a significant link between these factors. We therefore hypothesize:

H2: 

Firm size (lnSize) positively influences ESG performance.

2.2.3. Ownership Concentration and ESG Performance

The ownership structure identifies the degree to which controlling stakeholders influence strategic decisions [59]. Agency theory suggests that different ownership structures influence the behavior of management and the extent to which the decisions made balance the needs of shareholders and stakeholders [29]. Ownership concentration can lead controlling shareholders to prioritize short-term gains over long-term sustainability, potentially impeding ESG investment [60,61]. Prior studies emphasize that the concentration of ownership limits a firm’s attention to ESG concerns. For instance, Srairi [62] observed a negative correlation between ownership concentration and ESG performance in the banking sector. Likewise, Yavuz et al. [63] reported the negative association in the six financial and 16 non-financial Istanbul companies. We therefore hypothesize:

H3: 

Ownership Concentration Negatively Influence ESG Performance.

2.2.4. The Mediating Role of Market Share (Industry Competitive Position)

Market share refers to the competitive position or visibility of the enterprise in the industry [64]. Legitimacy theory emphasizes that organizations with high visibility face greater expectations and scrutiny regarding environmental and social responsibility [65,66]. Even though small organizations sometimes rely on ESG to some extent for distinction, the nature of the Chinese A-share market tends to involve highly dominant scales and significant public focus on larger corporations. In this context, the net impact of market share is expected to follow a linear pattern: as market share rises, legitimacy pressure intensifies, leading to greater ESG engagement. Firm financial characteristics (ROE and firm size) and governance factors (ownership concentration) shape market position [10,67], which in turn form ESG engagement. Recently, Nimusima et al. [68] used competitive advantage as a mediator between interactive practices and market performance, while Muis et al. [69] used it as a mediator between market orientation, transformational leadership, partnership strategy, and firm performance. However, the exact role of the industry’s competitive position (market share) as a mediator is missing in the existing literature. We therefore hypothesize:

H4: 

Market share (industry competitive position) mediates the relationship between firm characteristics (ROE, lnSize, and ownership concentration) and ESG performance.

2.2.5. The Mediating Role of Managerial Risk Preferences

Managerial risk preference reflects the willingness of decision-makers to participate in long-run, uncertain investments like ESG efforts [70]. Organizations with higher risk tolerance are more likely to pursue sustainability-oriented projects with delayed returns [71]. Risk-averse managers, on the other hand, might shun sustainability because of uncertain benefits [72]. These behavioral channels link firm characteristics to ESG outcomes. In this research, managerial risk preference is proxied by the ratio of R&D investment and fixed investment to total assets [73]. Whereas fixed and R&D assets differ in their risk profiles, this ratio fosters a general orientation towards long-run investment, a theme that has often been emphasized in Chinese literature [74].

To reconcile with the agency problem without conflicting concepts within the agency-theory framework, we trace the behavioral mechanisms: risk-focused executives support long-term ESG actions, while internal governance offsets short-term opportunism to improve ESG engagement [75,76]. Company profitability, size, and ownership structure determine risk preferences [77,78], which subsequently impact ESG investment choices [79,80]. Recently, Cui et al. [81] used management risk executive as a mediator between their experience and firm sustainable innovation. Carina et al. [82] and Ahmed et al. [83] used risk perception as a mediator between constructs such as financial literacy, entrepreneurial orientation, behavioral biases, and investor investment. Accordingly, this paper proposes:

H5: 

Risk preferences mediate the relationship between firm characteristics (ROE, lnSize, and ownership concentration) and ESG performance.

2.2.6. The Moderating Role of Internal Control Level

Internal control systems play an essential role in managing agency problems by promoting compliance, transparency, and alignment between management decisions and the organization’s overall goals [84]. It is argued that improved internal control practices complement the positive impact of profitability and size on ESG disclosures and may reduce asymmetry and align resource allocation with sustainability goals [70,85]. Additionally, proper management of internal controls may offset the negative impact of ownership structure. Further, the importance of effective internal control is that it enhances the procedural integrity of ESG commitments, making ESG practices more credible [85].

As a moderating variable, the level of internal control may function as a “control valve” in the relationship between firm characteristics and ESG performance [86]. High-quality internal controls can strengthen the positive effects of profitability, firm size, and ownership structure on ESG performance [70], while simultaneously mitigating the influence of adverse factors [87]. Recently, Feng and Saleh [70] observed that internal control moderated the effect of managerial ability on ESG risks. Quan and Zhou [88] reported that internal control moderates the relationship between corporate governance and ESG performance in China. In addition, Boulhaga et al. [86] noted that internal controls negatively moderated the relationship between ESG ratings and firm performance among French companies.

This study distinguishes between two indicators: internal control (continuous index), which focuses on improvements in internal governance, and effective internal control (binary indicator), which indicates a threshold level of effectiveness for internal controls. Both indicators describe and represent two forms of governance (gradual and categorical).

Accordingly, this paper proposes:

H6: 

Internal control (EIC and IC) moderates the relationship between firm characteristics and ESG performance such that organizations with stronger internal control reflect more substantial impacts.

Figure 1 illustrates the proposed links among the constructs. Black arrows show direct, green arrows indirect, and red arrows show moderation effects.

3. Materials and Methods

3.1. Sampling and Data Collection

The sample consists of Chinese A-listed firms listed on Shanghai and Shenzhen from 2021 to 2023. Firm-specific information, such as ownership concentration and financial indicators, was collected from the CSMAR database, and ESG information was collected from the Wind ESG rating system. WIND offers ESG_SCORE with a scale from 0 to 100 for each pillar (E = 30%, S = 30%, G = 40%) and also offers a normalized composite index from 0 to 10 [89]. Consistent with previous literature and for easy comparison across regressions, ESG_SCORE is taken as the normalized score from 0 to 10 [90,91]. Every mention of ESG_SCORE is made with reference to this normalized score between 0 and 10.

In building the final balanced panel, the study used the following filters:

(a)

Excluded ST and ST companies (special treatment companies flagged for abnormal financial conditions).

(b)

Excluded companies in the financial industry, because of their different regulatory environment.

(c)

Excluded companies having missing ESG ratings across any years.

(d)

Excluded companies that delisted or listed between 2021 and 2023, to ensure a balanced panel.

(e)

Used listwise deletion for missing control-construct data [92].

This method resulted in 13,911 final observations, reflecting a wide cross-section of A-listed Chinese firms. A concise flow summary of the sample selection is included in Appendix A,

Table A1. Sample selection procedure.

Step	Description	Observation
1	Initial sample of all A-shared listed companies on Shanghai and Shenzhen	15,291
2	Financial companies were excluded due to their different reporting and regulatory systems	−913
3	Companies excluded due to having ST and *ST (special treatment companies flagged for abnormal financial performance)	−412
4	Companies excluded due to missing ESG ratings	−257
5	Companies excluded due to missing financial, governance, and ownership information	−198
6	Companies excluded due to being delisted or listed between 2021 and 2023	−200
Final panel	Companies having complete information across 2021–2023	13,911

.

Other variables like risk preference, market share, and internal control are taken from the China internal control index database, firm disclosures, or inferred from company information on risk preferences [93,94]. The sample was clarified by excluding firms categorized under special treatment, finance sector companies, and those whose ESG_SCORE were missing. Missing data were handled by listwise deletion, just the same as the ESG studies on panels conducted before [92].

In addition, ROE, growth, and risk preference constructs were winsorized at the 1st and 99th percentiles to reduce the influence of extreme values [95]. However, owing to the denominator effect observed for emerging market revenue and investment observations, some outliers still exist after winsorization. A manual investigation was carried out for the top-10 observations for growth and risk preference to confirm that the observations were not entry errors and were actual observations for firms that yielded unusual outcomes for small denominators. Further analysis is included for observers at the 99.5th and 99.9th percentiles in Appendix B,

Table A2. Descriptive statistics prior to and post Windsorisation.

Variable	Mean (Raw)	Mean (Wins.)	Min (Raw)	Min (Wins.)	Max (Rax)	Max (Wins.)
ROE	−0.135	−0.118	−12.34	−2.11	15.21	3.78
Growth	5.412	2.983	−1242.05	−0.999	1308.33	944.10
Risk-preference	0.083	0.076	−0.112	−0.085	1.431	0.673

and

Table A3. Robustness to outlier exclusion.

Model	Main Findings	99.5th Percentile Excluded	99.9th Percentile Excluded
Baseline fixed effect (H1 to H3)	Outcomes unchanged in sign and level of significance	Consistent	Consistent
Mediation (H4 and H5)	Direction of mediator unchanged (partial mediation)	Consistent	Consistent
Moderations (H6)	Internal control and effective internal control remain positive and significant	Consistent	Consistent

.

3.2. Variable Selection and Definitions

3.2.1. Dependent Variable

The measurement of ESG performance lacks a universal standard, though professional rating systems have gained prominence for their objectivity and cross-industry comparability [96]. This study utilizes Wind ESG ratings (2018-present) to formulate the ESG index (dependent variable) due to their extensive coverage of Chinese firms and detailed scoring methodology [97]. The system consists of three weighted pillars, with environmental E = 30%, social S = 30%, and governance G = 40%, respectively, ranging from 0 to 100 points, with higher scores indicating better ESG practices. WIND provides a normalized 0–10 rating used in this research [89]. The index is often used in current ESG literature, for instance, the study in Rajesh and Rajendran [98].

3.2.2. Independent Variables

This study used firm profitability (ROE), firm size (lnSize), and ownership concentration (TOP10) as independent constructs. Specifically, ROE is calculated as net income divided by shareholders’ equity. ROE measures a company’s financial efficiency and allocates resources for ESG investing [99]. lnSize is calculated by taking the natural logarithm of total assets [100]. The larger the firm, the greater its visibility and resources, leading to greater ESG dedication. TOP10 (ownership concentration) is calculated as the percentage of outstanding shares owned by the ten largest shareholders, obtained from the CSMAR ownership structure module [101,102]. The greater the TOP10, the greater the agency problem.

The Herfindahl-Hirschman index (HHI) is subsequently introduced in the robustness test as an alternative indicator of ownership concentration.

$$\mathit{HHI}={\sum }_{i=1}^N{s}_i^2$$

where s_i is the shareholding proportion of each major stakeholder.

Considering the literature on governance, this research used the TOP10 shareholders in its computations to identify pertinent concentration patterns.

3.2.3. Mediating Variables

This study used two key mediators (market share and managerial risk preference) to capture behavioral channels. Market share, calculated as a firm’s main business revenue divided by total industry revenue [103], reflects a firm’s competitive visibility and legitimacy pressure [104].

Managerial risk preference is proxied by the ratio of R&D and fixed asset investments to total assets [105]. However, this proxy aggregates different investment forms; consistent with earlier studies on the Chinese market, and enables managerial orientation towards long-run risk-oriented investment behavior [106].

3.2.4. Moderating Variable

For moderating effects, we measure internal control quality using two complementary indicators: the normalized internal control index from the Dibo database (divided by 100) and a binary effectiveness measure from CSMAR (1 = effective, 0 = ineffective) [107]. Internal control captures continuous improvement in governance quality, whereas effective internal control provides a cutoff-based certification of internal control effectiveness, enabling us to capture discontinuous and linear moderating impacts.

3.2.5. Control Variables

Drawing on previous research, this study includes firm age (AGE), financial leverage (Lev), growth (Growth), board size (BoardSize), proportion of independent directors (IndepRatio), and cash flow level (CF) as control variables. The names and definitions of all variables used in this study are shown in

Table 1. Variable definitions.

Variable Type	Variable Name	Symbol	Measurement Method
Dependent	ESG performance	ESG_SCORE	Comprehensive ESG_SCORE from WIND (0–10), integrating management practice and controversy event scores
Independent	Return on equity	ROE	Net profit/shareholders’ equity
	Log of total assets	LnSize	Natural logarithm of total assets (A): Ln(A)
	Ownership concentration	TOP10	Sum of shareholding ratios of the top ten shareholders
Mediator	Industry competitiveness	MS	Proportion of a firm’s main business revenue to total industry revenue
	Managerial risk preference	Risk_Preference	(R&D investment + fixed asset investment)/total assets
Moderaor	Internal control level	IC	Dibo internal control index divided by 100
	Effective internal control	ErIC	1 if internal control is effective, 0 otherwise
Control variable	Firm age	AGE	Number of years since establishment (up to research year)
	Financial leverage	Lev	Asset-liability ratio (total liabilities/total assets)
	Growth	Growth	Revenue growth rate (current revenue–previous revenue)/previous revenue)
	Board	BoardSize	Total number of board members
	Ratio of independent directors	IndepRatio	Ratio of independent directors to total board members
	Cash flow level	CF	Net cash flow from operating activities/total assets

Note: ESG_SCORE shows the normalized index developed from WIND ESG ratings; Risk_Preference shows managerial risk preference; and TOP10 reveals ownership concentration.

.

3.3. Assessment of Common Method Variance

Since all the constructs were extracted from the secondary datasets (e.g., CSMAR, DIB, or WIND), the potential of common method variance is already minimal in these constructs [108]. Since the data collection methods are different and variable content, e.g., ESG ratings, financial constructs, or governance, further removes single-source bias. In addition, the variance inflation factor (VIF) for all variables was below 5, reflecting no multicollinearity issues [109].

To mitigate the potential of simultaneity bias, we also used one-period-lagged regressors (see Equation (5)) and replaced ROE with ROA, as a substitution method, and the HHI was used [110]. The findings of robustness analysis are qualitatively in line with the baseline results. Given the complexity of mediation and moderation processes in panels, we interpreted the results as exploratory rather than causal, and we suggest that future studies use structural equation modeling for mediation and bootstrapped panel models.

3.4. Model Design

All analyses were conducted using Stata 17.0, employing a two-way fixed-effects (firm and year) model with standard errors clustered within firms. This model controls for unobserved economy-wide temporal shocks and heterogeneity. Various control variables, such as firm age, financial leverage, growth, board size, and the proportion of independent directors, are incorporated to ensure a comprehensive examination.

First, to examine the effects of firm profitability, firm size, and ownership concentration on ESG performance (i.e., H1, H2, and H3), the following baseline regression model is constructed with reference to relevant literature:

$$ES{G}_{i,t}={\beta }_0+{\beta }_{1}RO{E}_{i,t}+{\beta }_2\ln (\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}{)}_{i,t}+{\beta }_3{\mathrm{T}\mathrm{O}\mathrm{P}10}_{i,t}+{{\beta }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{i,t}+\mu i+\lambda \mathrm{t}+{ϵ}_i$$

(1)

where μi captures unobserved, time-invariant firm-specific heterogeneity, including managerial philosophy, organizational culture, industry positioning, and other characteristics that do not change during the sample period.

${\lambda }_t$ denotes year fixed effects, controlling for macroeconomic conditions, regulatory changes, ESG disclosure reforms, and market-wide shocks that affect all firms in a given year.

${ϵ}_{it}$ is the idiosyncratic error term, capturing random shocks and unobserved factors specific to each firm-year observation.

To further examine how firm characteristics indirectly affect ESG performance through mediating variables (industry competitive position and managerial risk preference), the following mediation effect models are constructed:

$${\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}}_{i,t}={\alpha }_0+{\alpha }_1{\mathrm{R}\mathrm{O}\mathrm{E}}_{i,t}+{\alpha }_2\ln (\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}{)}_{i,t}+{\alpha }_3{\mathrm{T}\mathrm{O}\mathrm{P}10}_{i,t}+{\alpha }_4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{i,t}+\mu i+\lambda \mathrm{t}+{ϵ}_{i,t}$$

(2)

To test the joint impact of core independent variables and mediating variables on ESG performance, the following regression model is constructed:

$${\mathrm{E}\mathrm{S}\mathrm{G}}_{i,t}={\beta }_0+{\beta }_1{\mathrm{R}\mathrm{O}\mathrm{E}}_{i,t}+{\beta }_2\ln (\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}{)}_{i,t}+{\beta }_3{\mathrm{T}\mathrm{O}\mathrm{P}10}_{i,t}+{\beta }_4{\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}}_{i,t}+{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{i,t}+\mu i+\lambda \mathrm{t}+{ϵ}_{i,t}$$

(3)

Finally, to examine the moderating role of internal control level in the relationship between ESG ratings and firm characteristics, the following moderating effect model is constructed:

$${\mathrm{E}\mathrm{S}\mathrm{G}}_{i,t}={\beta }_0+{\beta }_1{X}_{i,t}+{\beta }_2{\mathrm{I}\mathrm{C}/\mathrm{E}\mathrm{I}\mathrm{C}}_{i,t}+{\beta }_3\left(X_{i,t}\times {{\displaystyle \frac{\mathrm{I}\mathrm{C}}{\mathrm{E}\mathrm{I}\mathrm{C}}}}_{i,t}\right)+{{\beta }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{i,t}+\mu i+\lambda \mathrm{t}+{ϵ}_{i,t}$$

(4)

where X_i,t represents each independent variable (ROE, lnSize, TOP10).

To examine short-term dynamic effects and deal with potential simultaneity, this study re-estimates the baseline specification using one-period lagged independent variables. The lagged model is specified as

$${\mathrm{E}\mathrm{S}\mathrm{G}}_{i,t}={\beta }_0+{\beta }_1{ROE}_{i,t-1}+{\beta }_2{\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}}_{i,t-1}+{\beta }_3{\mathrm{T}\mathrm{O}\mathrm{P}10}_{i,t-1}+{{\beta }_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{i,t-1}+\mu i+\lambda \mathrm{t}+{ϵ}_{i,t}$$

(5)

Because fixed effect estimation does not facilitate bootstrapped mediation impact for the panel models, this study used a stepwise method.

4. Empirical Analysis

4.1. Descriptive Statistical Analysis

4.1.1. Descriptive Statistics

Table 2. Results of descriptive statistics.

Variable	Mean	SD	Min	Max
ESG_SCORE	6.092	0.813	2.92	9.62
lnSize	22.409	1.521	18.316	31.4309
ROE	0.118	0.973	77.771	1.697
TOP10	0.674	0.202	0.0973	0.999
MS	0.0169	0.062	0	1
Risk_preference	0.076	0.059	0	0.6747
IC	6.090	0.855	3.1	9.2
EIC	0.506	0.499	0	1
AGE	23.045	6.478	5	69
lev	0.416	0.213	0.0177	1.6036
Growth	0.2926	9.945	−0.9991	944.099
Boardsize	8.291	1.670	4	19
IndepRatio	0.379	0.055	0	0.818
CF	0.0469	0.076	−0.9965	1.170

presents the descriptive statistics, revealing several notable patterns. The sample shows substantial variation in firm growth (Growth), ranging from −0.9991 to 944.0996, indicating extreme disparities in revenue development across Chinese listed companies. Such extreme values are common in emerging datasets, where very small denominators in revenue calculations yield a large proportion, even after winsorization. Risk_preference averages 0.0766522, suggesting generally conservative investment strategies among sampled firms. Both firm size (lnSize) and return on equity (ROE) exhibit wide dispersion, with ROE’s negative mean (−0.1184138) reflecting widespread profitability challenges. Governance measures show average internal control levels (IC) of 6.090051 and independent director ratios (IndepRatio) of 0.3792896, though some firms demonstrate weaker governance structures. HHI indicates varying industry concentration levels, while firm age (AGE) spans up to 64 years, capturing diverse lifecycle stages.

4.1.2. Correlation Analysis

Table 3. Correlation analysis results.

Variables	ESG_SCORE	ln	ROE	TOP10	MS	Risk_Preference	AGE	Lev	Growth	Boardsize	IndepRatio	CF
ESG_SCORE	1	0.257 ***	0.127 ***	−0.567 ***	0.113 ***	0.192 ***	−0.052 ***	−0.056 ***	−0.003	0.105 ***	0.028 **	0.092 ***
lnSize	0.258 ***	1	0.076 ***	−0.042 ***	0.312 ***	−0.181 ***	0.214 ***	0.511 ***	−0.007	0.393 ***	0.020 **	0.084 ***
ROE	0.127 ***	0.076 ***	1	−0.014	0.029 **	0.0248 **	−0.031 **	−0.142 ***	−0.014	0.039 ***	−0.022 **	0.173 ***
TOP10	−0.567 ***	−0.042 ***	−0.014	1	0.033 **	−0.132 ***	−0.026 **	0.046 ***	0.001	−0.024 **	−0.018 **	0.006
MS	0.113 ***	0.312 ***	0.029 **	0.033 **	1	−0.059 ***	0.051 ***	0.141 ***	−0.003	0.096 ***	0.033 ***	0.079 ***
Risk_Preference	0.193 ***	−0.181 ***	0.0248 **	−0.132 ***	−0.060 ***	1	−0.242 ***	−0.180 ***	0.047 ***	−0.121 ***	0.010	0.048 ***
AGE	−0.052 ***	0.214 ***	−0.039 **	−0.026 **	0.050 ***	−0.242 ***	1	0.208 ***	−0.019 **	0.145 ***	−0.024 **	0.020 **
Lev	−0.056 ***	0.511 ***	−0.141 ***	0.047 ***	0.141 ***	−0.179 ***	0.208 ***	1	−0.007	0.208 ***	−0.001	−0.144 ***
Growth	−0.003	−0.007	−0.014	0.0013	−0.003	0.047 ***	−0.020 **	−0.006	1	0.016 **	−0.009	−0.038 ***
Boar-sz	0.105 ***	0.3932 ***	0.040 ***	−0.025 **	0.095 ***	−0.121 ***	0.145 ***	0.208 ***	0.017 **	1	−0.480 ***	0.016
IndepRatio	0.028 **	0.020 **	−0.022 **	−0.018 **	0.033 ***	0.010	−0.024 **	−0.001	−0.001	−0.480 ***	1	0.007
CF	0.092 ***	0.084 ***	0.173 ***	0.006	0.079 ***	0.048 ***	0.020 **	−0.144 ***	−0.038 ***	0.016	0.007	1

** p < 0.05, *** p < 0.01.

presents the Pearson correlation analysis results, revealing several significant relationships. The ESG_SCORE shows a positive correlation with firm size (ln) (r = 0.2577, p < 0.01), indicating larger firms tend to have better ESG performance. Conversely, ESG_SCORE demonstrates a negative association with ownership concentration (TOP10) (r = −0.5667, p < 0.01), suggesting that more dispersed ownership correlates with higher ESG_SCORE. The analysis also reveals that risk preference negatively relates to firm age (p < 0.01), implying that younger firms exhibit greater risk appetite. Additionally, financial leverage (lev) positively correlates with market share (MS) (r = 0.1410, p < 0.01), potentially indicating that leveraged firms maintain stronger market positions. These findings provide preliminary support for our hypothesized relationships while revealing additional patterns in the data.

4.1.3. Multicollinearity Test

The VIF test was conducted to assess potential multicollinearity among the explanatory variables. As a rule of thumb, VIF values exceeding 10 indicate severe multicollinearity, where explanatory variables collectively explain over 90% of a variable’s variance, potentially requiring remedial measures like stepwise regression [111]. However, as presented in

Table 4. VIF (variance inflation factor) test results.

Variable	VIF	1/VIF
lnSize	1.73	0.579663
Boardsize	1.66	0.603768
lev	1.5	0.667068
IndepRatio	1.4	0.716582
CF	1.08	0.926916
ROE	1.07	0.932877
AGE	1.07	0.933952
TOP10	1.01	0.989875
Growth	1	0.997504
Mean VIF	1.28

, all VIF values remain well below 5 [109], with the maximum value of 1.73 observed for firm size (ln). These results confirm that multicollinearity poses no substantial threat to the reliability of our regression estimates.

4.2. Regression Analysis

4.2.1. Baseline Regression Analysis

To ensure methodological rigor, we conducted preliminary model specification tests before performing regression analysis. As shown in

Table 5. Model tests.

Test Method	t-Value	p-Value	Conclusion
F-test	5.44	0.0000	The fixed effects model is superior to pooled OLS.
Hausman test	10.87	0.0000	The fixed effects model is superior to the random effects

, both the F-test and Hausman test yielded statistically significant results, justifying our selection of a fixed effects model.

We first employed Model (1) to examine how key firm characteristics-profitability, firm size, and ownership concentration-influence ESG performance, thereby testing hypotheses H1–H3. The fixed effects regression results for these relationships are presented in

Table 6. Baseline regression results.

Variable	ESG_SCORE	ESG_SCORE
lnSize	0.141 ***	0.175 ***
	(0.0290)	(0.00601)
ROE	0.0204 ***	0.0290 ***
	(0.00515)	(0.00472)
TOP10	−1.828 ***	−2.028 ***
	(0.0328)	(0.0262)
AGE	0.0352	−0.0126 ***
	(0.0738)	(0.00116)
lev	−0.0883	−0.575 ***
	(0.0770)	(0.0376)
Growth	−0.000327	−0.000233
	(0.000447)	(0.000422)
Boardsize	0.0211 **	0.00991 **
	(0.00914)	(0.00491)
IndepRatio	−0.174	0.116
	(0.200)	(0.130)
CF	−0.102	0.122 *
	(0.0755)	(0.0654)
Constant	3.353 **	3.947 ***
	(1.651)	(0.128)

* p < 0.1, ** p < 0.05, *** p < 0.01.

.

The regression results provide strong empirical support for our hypotheses regarding firm characteristics and ESG performance. For H1, ROE shows significantly positive coefficients of 0.0204 (Model 1) and 0.0290 (Model 2) (p < 0.01), confirming that higher profitability enhances ESG performance. Similarly, firm size (lnSize) demonstrates positive coefficients of 0.141 (Model 1) and 0.175 (Model 2) (p < 0.01), supporting H2’s prediction that larger firms achieve better ESG outcomes. Conversely, ownership concentration (TOP10) exhibits significantly negative coefficients (−1.828 in Model 1 and −2.028 in Model 2, p < 0.01), validating H3’s proposition that concentrated ownership undermines ESG performance.

Control variables reveal additional insights: board size (Boardsize) shows positive effects in some specifications, suggesting diverse governance perspectives favor ESG investments. Financial leverage (lev) negatively impacts ESG in Model 2, indicating that financially constrained firms deprioritize sustainability. Firm age (AGE) presents mixed results, with Model 2’s negative coefficient implying older firms may face ESG implementation barriers due to institutional inertia. These findings collectively demonstrate that financial capacity (profitability, size) and governance structure (ownership concentration) systematically shape corporate ESG performance, while accounting for relevant organizational contingencies.

4.2.2. Regression Results Considering Lagged Effects

The lagged effects analysis builds upon our baseline findings by examining temporal dynamics in the firm characteristics-ESG performance relationship. Using one-period lagged variables (ln, ROE, TOP10),

Table 7. Regression results with explanatory variables lagged by one period.

Construct	(1)	(2)
	ESG_SCORE	ESG_SCORE
lnSize	0.171 ***
	(0.00461)
ROE	0.0504 ***
	(0.00548)
	(1)	(2)
	ESG_SCORE	ESG_SCORE
TOP-10	−2.126 ***
	(0.0259)
MS	0.747 ***	0.637 ***
	(0.0876)	(0.120)
Risk_Preference	1.916 ***	2.502 ***
	(1)	(2)
	ESG_SCORE	ESG_SCORE
	(0.0923)	(0.130)
AGE	−0.00917 ***	−0.00981 ***
	(0.000840)	(0.00118)
LEV	−0.600 ***	−0.608 ***
	(0.0297)	(0.0404)
Growth	−0.000571	−0.000572
	(0.000519)	(0.00128)
BoardSize	0.0139 ***	0.0137 **
	(0.00391)	(0.00545)
IndepRatio	0.333 ***	0.331 **
	(0.111)	(0.153)
CF	0.263 ***	0.108
	(0.0705)	(0.0981)
ln_lag1		0.198 ***
		(0.00627)
ROE_lag1		0.0539 ***
		(0.00684)
TOP10_lag1		−1.284 ***
		(0.0364)
Constant	3.743 ***	2.579 ***
	(0.102)	(0.138)
	0.442	0.283

** p < 0.05, *** p < 0.01.

reveals that both contemporaneous and lagged measures significantly influence ESG_SCORE at the 1% level. Notably, the lagged effects prove stronger than contemporaneous ones-firm size’s impact increases from 0.171 to 0.198, while profitability’s influence grows from 0.0504 to 0.0539. This pattern suggests that ESG performance improvements require time for firms to operationalize resource allocations and strategic changes, with the full effects becoming more pronounced in subsequent periods. The amplified coefficients may reflect cumulative impacts of sustained investments, though diminishing returns could eventually emerge as firms approach performance ceilings. These temporal dynamics underscore the importance of longitudinal perspectives in ESG research, as cross-sectional analyses may underestimate the complete effects of firm characteristics.

4.2.3. Robustness Checks for Simultaneity Bias

As a robustness check for simultaneity bias, we re-estimated the baseline fixed-effects models using one-period-lagged values of ROE, lnSize, TOP10, and all control variables (see

Table 8. Robustness checks for simultaneity.

Construct	Model (1) Lagged FE (ESG_SCORE)	Model (2) ROA FE (ESG_SCORE)
Lagged ROE	0.042 ***
	(3.71)
ROA		0.036 ***
		(3.54)
Lagged lnSize	0.128 ***
	(5.92)
lnSize		0.143 ***
		(6.15)
Lagged TOP10	−0.057 ***
	(−2.28)
TOP10		−0.064 ***
		(−2.46)
LEV	−0.011	−0.009
	(−1.14)	(−1.02)
Growth	0.003 **	0.003 **
	(2.14)	(2.08)
CF	0.019 *	0.21 *
	(1.66)	(1.72)
BoardSize	0.004	0.005
	(0.98)	(1.02)
IndepRatio	0.012	0.014
	(1.48)	(1.53)
Firm FE	Yes	Yes
Year FE	Yes	Yes
Clustered FE	Firm level	Firm level
Adj. R-squared	0.312	0.326
Observations	13,911	13,911

* p < 0.1, ** p < 0.05, *** p < 0.01.

). Model (1) re-estimates the baseline specification using one-period-lagged values of ROE, lnSize, TOP10, and control variables to reduce simultaneity concerns. Model (2) replaces ROE with ROA to verify that profitability-ESG relationships are not driven by leverage effects.

To mitigate potential concerns about information lost with the median split in our examination of heterogeneity, we re-estimated our base model with interactions for ROE, lnSize, TOP10, and a continuous variable for firm age measured as a number of listing years (see Appendix B,

Table A4. Robustness check of the interaction between firm characteristics and listing age.

Model	Two-Way FE Model
ROE	0.032 ***
	(3.45)
lnSize	0.139 ***
	(6.02)
Age (listing years)	−0.061 **
	(−2.41)
ROE × Age	0.0003
	(0.88)
lnSize × Age	0.0011 *
	(1.92)
TOP10 × Age	−0.0004
	(−0.97)
LEV	−0.10
	(−1.12)
Growth	0.003 **
	(2.09)
CF	0.020 *
	(1.71)
BoardSize	0.005
	(1.03)
IndepRatio	0.013
	(1.51)
Firm FE	Yes
Year FE	Yes
Clustered FE	Yes
R	0.138
Observations	13,911

* p < 0.1, ** p < 0.05, *** p < 0.01.

).

4.2.4. Mediation Analysis

Our mediation analysis examines whether industry competitive position (MS) and managerial Risk_preference transmit the effects of firm characteristics to ESG performance (ESG_SCORE). As shown in

Table 9. Results of mediation effect analysis.

Variable	ESG_SCORE	MS	Risk_Preference	ESG_SCORE
lnSize	0.121 *** (0.004)	0.013 *** (0.000)	−0.007 *** (0.000)	0.128 *** (0.004)
ROE	0.085 *** (0.006)	0.000 (0.001)	0.002 *** (0.001)	0.079 *** (0.005)
TOP10	−2.231 *** (0.027)	0.014 *** (0.002)	−0.041 *** (0.002)	−2.150 *** (0.026)
MS	-	-	-	0.825 *** (0.089)
Risk_Preference	-	-	-	2.294 *** (0.092)
Constant	4.891 *** (0.083)	−0.279 *** (0.008)	0.270 *** (0.007)	4.503 *** (0.088)

*** p < 0.01.

, MS, which is industry competitiveness, partially mediates the relationship. Model (2) reveals that firm size (lnSize) and ownership concentration (TOP10) significantly influence MS, while Model (4) demonstrates MS’s significant effect on ESG_SCORE. The attenuated coefficients of ROE and TOP10 in Model (4) versus Model (1) further support this partial mediation. Similarly, Risk_preference serves as another partial mediator, evidenced by: (1) ln, ROE, and TOP10’s significant impacts on Risk_preference in Model (3); (2) Risk_preference’s significant effect on ESG_SCORE in Model (4); and (3) reduced coefficients for ROE and TOP10 in Model (4), supporting the H4 and H5. These findings collectively indicate that both industry position and risk appetite represent important pathways through which profitability and ownership structure ultimately affect ESG performance.

4.2.5. Moderation Analysis

Table 10. Results of moderation effect test.

Variable	(IC Moderation)	(EIC Moderation)
IC	0.323 ***
	(0.036)
Ln	−0.101 ***	0.033 ***
	(0.009)	(0.005)
ROE	−0.054 ***	0.032 ***
	(0.018)	(0.006)
TOP-10	−0.782 ***	−1.367 ***
	(0.093)	(0.032)
IC # Ln	0.021 ***
	(0.001)
IC # ROE	0.010 ***
	(0.003)
IC # TOP-10	0.066 ***
	(0.015)
AGE	−0.002 ***	−0.004 ***
	(0.000)	(0.001)
LEV	−0.113 ***	−0.283 ***
	(0.012)	(0.023)
Growth	−0.000	−0.000
	(0.000)	(0.000)
BoardSize	0.000	−0.000
	(0.002)	(0.003)
IndepRatio	−0.080 *	−0.026
	(0.045)	(0.084)
CF	0.045	0.246 ***
	(0.029)	(0.053)
eIC = 0		0.000
eIC = 1		−2.551 ***
		(0.124)
eIC = 0 # lnSize		0.000
eIC = 1 # lnSize		0.116 ***
		(0.005)
eIC = 0 # ROE		0.000
eIC = 1 # ROE		0.001
		(0.008)
	(IC Moderation)	(EIC Moderation)
eIC = 0 # TOP-10		0.000
eIC = 1 # TOP-10		1.469 ***
		(0.050)
Constant	3.952 ***	6.057 ***
	(0.234)	(0.109)
Adjusted R2	0.906	0.679
Samples	13,911	13,911

* p < 0.1, *** p < 0.01, # = interaction

results reveal distinct moderating patterns: the continuous IC measure consistently strengthens all independent variable-ESG_SCORE relationships through significantly positive interaction terms. For the binary EIC measure, effective internal controls (EIC = 1) significantly enhance the positive effects of firm size (lnSize) and weaken the negative impact of ownership concentration (TOP-10) on ESG_SCORE, while ineffective controls (EIC = 0) show no moderating influence. Control variables demonstrate expected patterns, with AGE and LEV showing significant negative effects on ESG_SCORE across specifications, while other controls exhibit model-dependent effects. Model fit statistics indicate strong explanatory power, with Model 1’s adjusted R² of 0.906 outperforming Model 2’s 0.679, suggesting the continuous IC specification better captures internal control’s moderating role in ESG performance. Figure 2 describes how ROE marginally influences ESG_SCORE based on the range of internal control (IC) observed, as indicated by the interaction effect in Table 9. At a low level of IC, profitability has a marginally negative effect on ESG. As IC levels increase, this effect turns positive and continues to escalate until a stage when IC boosts a company’s ability to leverage profitability for ESG engagement.

4.3. Robustness Test

Substitution of Core Variables

• Substituting the dependent variable

To ensure the robustness of our findings, we conducted two key sensitivity tests. First, we replaced the Wind ESG ratings with Huazheng ESG data as an alternative dependent variable, given their distinct methodologies, coverage, and evaluation frameworks. As

Table 11. Regression results with substituted dependent variable.

Variable	(1)	(2)
	ESG_SCORE	HZESG
lnSize	0.174 ***	1.699 ***
	(0.00587)	(0.0816)
ROE	0.0510 ***	0.476 ***
	(0.00283)	(0.00320)
TOP10	−2.194 ***	−21.15 ***
	(0.0254)	(0.415)
AGE	−0.0128 ***	−0.120 ***
	(0.000863)	(0.0137)
lev	−0.636 ***	−6.212 ***
	(0.0351)	(0.522)
Growth	−0.0000495	0.000388
	(0.000556)	(0.00748)
Boardsize	0.00942 *	0.0644
	(0.00438)	(0.0675)
IndepRatio	0.300 *	3.328
	(0.122)	(1.913)
	(1)	(2)
	ESG_SCORE	HZESG
CF	0.373 ***	2.006
	(0.0876)	(1.250)
Costant	4.023 ***	4.68 ***
	(0.124)	(0.1756)

* p < 0.1, *** p < 0.01.

shows, the core relationships between firm characteristics and ESG performance remain consistent despite this substitution, though some control variables exhibit minor variations. This consistency across different ESG measurement systems strengthens the reliability and generalizability of our conclusions.

2.

Using ROA instead of ROE as an explanatory variable

Second, we substituted ROA for ROE as our profitability measure to address potential leverage distortions in ROE. Since ROE can be artificially inflated through debt financing without genuine operational improvement, ROA provides a more direct assessment of asset utilization efficiency by excluding capital structure effects. This alternative specification maintains our core findings while offering a more stable basis for evaluating firms’ long-term operational capabilities, further confirming the robustness of our results against different profitability measures.

Table 12. Regression results using ROA instead of ROE.

Variable	(1)	(2)
	Baseline model	Robustness model
ROE	0.085 ***
	(0.0049)
AGE	−0.008 ***	−0.008 ***
	(0.001)	(0.001)
LEV	−0.215 ***	−0.214 ***
	(0.040)	(0.039)
Growth	−0.000	−0.000
	(0.001)	(0.001)
BoardSize	0.084 ***	0.082 ***
	(0.005)	(0.005)
IndepRatio	1.670 ***	1.673 ***
	(0.149)	(0.150)
CF	0.683 ***	0.685 ***
	(0.136)	(0.133)
ROA		0.083 ***
		(0.0046)
	(1)	(2)
	Baseline model	Robustness model
Constant	5.007 ***	5.020 ***
	(0.088)	(0.089)
Observations	13,911	13,911

*** p < 0.01.

presents the comparative results of both baseline and robustness models, demonstrating consistent effect directions and significance levels across specifications. The control variables show stable patterns, with AGE and LEV exhibiting significant negative correlations (p < 0.01), while BoardSize, IndepRatio, and CF maintain significant positive relationships (p < 0.01) with the dependent variable in both models. Growth remains statistically insignificant, and the constant term shows consistent significance (p < 0.01) across specifications. These parallel results confirm the model’s robustness, as the core relationships persist despite methodological variations, while most control variables demonstrate reliable effects.

3.

Using the logarithm of main business revenue instead of the logarithm of total assets (Ln) as an explanatory variable

This study employs the logarithm of main business revenue rather than total assets as the explanatory variable because it better captures market-driven scale, especially for asset-light industries and operating efficiency studies. Main business revenue directly measures core sales capability and market position, unaffected by non-operating assets in total assets or variations in depreciation methods and industry asset structures.

Table 13. Regression results using the logarithm of the main business revenue.

Variable	(1)	(2)
	Baseline model	Robustness model
lnSize	0.204 ***
	(0.006)
AGE	−0.011 ***	−0.011 ***
	(0.001)	(0.001)
LEV	−0.899 ***	−0.875 ***
	(0.036)	(0.036)
Growth	−0.000	0.000
	(0.000)	(0.001)
BoardSize	0.016 ***	0.044 ***
	(0.005)	(0.005)
	(1)	(2)
	Baseline model	Robustness model
IndepRatio	0.509 ***	0.965 ***
	(0.144)	(0.143)
CF	0.291 ***	−0.030
	(0.092)	(0.096)
income		0.180 ***
		(0.006)
Constant	1.823 ***	2.079 ***
	(0.125)	(0.124)
Observations	13,911	13,911

*** p < 0.01.

confirms the stability of control variables, with AGE, LEV, BoardSize, IndepRatio, and CF showing consistent significant effects (p < 0.01) across both baseline and robustness models.

4.

Using HHI instead of the sum of the shareholding ratios of the top ten shareholders as an explanatory variable

This study employs the HHI to measure ownership concentration, offering advantages over the conventional top ten shareholders’ ratio (TOP-10). The HHI, calculated as the sum of squared ownership proportions, better captures the distributional nuances and power balances among shareholders while minimizing multicollinearity concerns. As

Table 14. Regression results using HHI.

Construct	(1)	(2)
	Baseline model	Robustness model
lnSize	0.174 ***	0.201 ***
	(38.51)	(36.42)
ROE	0.0510 ***	0.0501 ***
	(9.15)	(7.35)
TOP-10	−2.194 ***
	(−84.33)
AGE	−0.0128 ***	−0.0106 ***
	(1)	(2)
	Baseline model	Robustness model
	(−15.35)	(−10.42)
LEV	−0.636 ***	−0.815 ***
	(−21.08)	(−22.15)
Growth	−0.0000495	−0.000183
	(−0.09)	(−0.28)
BoardSize	0.00942 **	0.0180 ***
	(2.37)	(3.72)
IndepRatio	0.300 ***	0.611 ***
	(2.66)	(4.44)
CF	0.373 ***	0.274 ***
	(5.22)	(3.13)
HHI		−0.989 ***
		(−13.92)
Constant	4.023 ***	1.854 ***
	(41.34)	(16.12)
Adj. R-squared	0.422	0.138
Observations	13,911	13,911

** p < 0.05, *** p < 0.01.

shows, both HHI (−0.989 *) and TOP-10 (−2.194 *) demonstrate significant negative correlations with ESG_SCORE at the 1% level, confirming that higher ownership concentration reduces ESG performance regardless of measurement approach. The models maintain stable coefficients for key variables (lnSize, ROE, AGE, LEV, BoardSize, IndepRatio, CF) across specifications, with adjusted R² values of 0.422 and 0.138, respectively. With 13,911 observations, these consistent results across alternative measures strengthen the findings’ robustness and facilitate international comparability given HHI’s widespread theoretical use.

4.4. Further Research

4.4.1. Sub-Dimension Analysis of ESG

To examine the multidimensional impacts of firm characteristics on ESG performance, we conducted separate regressions using Wind’s environmental (E), social (S), and governance (G) sub-scores as dependent variables.

$$E={\beta }_0+{\beta }_{1}ROE+{\beta }_2\ln +{\beta }_{3}TOP10+{ϵ}_E$$

(6)

$$S={\gamma }_0+{\gamma }_{1}ROE+{\gamma }_2\ln +{\gamma }_{3}TOP10+{ϵ}_S$$

(7)

$$G={\delta }_0+{\delta }_{1}ROE+{\delta }_2\ln +{\delta }_{3}TOP10+{ϵ}_G$$

(8)

As

Table 15. Regression results of sub-dimension analysis.

Variable	(1)	(2)	(3)
	E	S	G
lnSize	0.516 ***	0.496 ***	0.503 ***
	(0.017)	(0.016)	(0.017)
TOP-10	−9.105 ***	−8.771 ***	−8.911 ***
	(0.086)	(0.082)	(0.083)
ROE	0.128 *	0.127 *	0.127 *
	(0.073)	(0.070)	(0.070)
AGE	−0.047 ***	−0.045 ***	−0.047 ***
	(0.003)	(0.003)	(0.003)
LEV	−2.266 ***	−2.115 ***	−2.149 ***
	(0.109)	(0.104)	(0.105)
Growth	0.000	0.000	−0.000
	(0.002)	(0.002)	(0.002)
BoardSize	0.040 ***	0.036 ***	0.039 ***
	(0.014)	(0.013)	(0.014)
IndepRatio	1.101 ***	1.096 ***	0.975 **
	(0.397)	(0.373)	(0.383)
CF	1.073 ***	1.073 ***	1.022 ***
	(0.266)	(0.254)	(0.256)
Constant	0.835 **	0.780 **	0.819 **
	(0.370)	(0.354)	(0.360)
Observations	13,911	13,912	13,912
	(1)	(2)	(3)
	E	S	G
R2	0.500	0.506	0.504
Adjusted R2	0.500	0.506	0.504

* p < 0.1, ** p < 0.05, *** p < 0.01.

shows, firm size (lnSize) consistently demonstrates strong positive effects across all three dimensions (coefficients: 0.516 for E, 0.496 for S, 0.503 for G; p < 0.01), while ownership concentration (TOP-10) shows uniformly negative impacts (coefficients ranging from −9.105 to −8.771). Profitability (ROE) exhibits modest but significant positive effects (β ≈ 0.127, p < 0.10), and control variables maintain their directional consistency-AGE and LEV show negative associations. In contrast, BoardSize, IndepRatio, and CF show positive relationships. These parallel results across E, S, and G components demonstrate that firm characteristics influence all ESG dimensions similarly, suggesting their effects operate through fundamental organizational mechanisms rather than dimension-specific pathways.

4.4.2. The Impact of Listing Duration on Regression Results

This study examines how listing duration moderates the ESG-performance relationship through grouped regression analysis (median split: Group-1 = long-listed, Group-0 = short-listed). The results (

Table 16. Regression results grouped by listing duration.

Variable	(1)	(2)	(3)	(4)	(5)
		Model1_Group1	Model1_Group0	Model2_Group1	Model2_Group0
	ESG_SCORE	ESG_SCORE	ESG_SCORE	ESG_SCORE	ESG_SCORE
lnSize		0.184 ***	0.153 ***	0.184 ***	0.153 ***
		(0.00630)	(0.00661)	(0.00630)	(0.00661)
ROE	0.0270 ***	0.0286 ***	0.207 ***	0.0286 ***	0.207 ***
	(0.0103)	(0.00607)	(0.0155)	(0.00607)	(0.0155)
TOP10		−2.241 ***	−2.162 ***	−2.241 ***	−2.162 ***
		(0.0375)	(0.0360)	(0.0375)	(0.0360)
AGE		−0.00135	−0.0174 ***	−0.00135	−0.0174 ***
		(0.00178)	(0.00188)	(0.00178)	(0.00188)
lev		−0.626 ***	−0.548 ***	−0.626 ***	−0.548 ***
	(1)	(2)	(3)	(4)	(5)
		Model1_Group1	Model1_Group0	Model2_Group1	Model2_Group0
		(0.0439)	(0.0418)	(0.0439)	(0.0418)
Growth		−0.00630	0.000117	−0.00630	0.000117
		(0.00729)	(0.000516)	(0.00729)	(0.000516)
Boardsize		0.0129 **	0.00595	0.0129 **	0.00595
		(0.00558)	(0.00562)	(0.00558)	(0.00562)
IndepRatio		0.294 *	0.355 **	0.294 *	0.355 **
		(0.161)	(0.157)	(0.161)	(0.157)
CF		0.478 ***	0.0714	0.478 ***	0.0714
		(0.109)	(0.0969)	(0.109)	(0.0969)
Constant	5.827 ***	3.463 ***	4.577 ***	3.463 ***	4.577 ***
	(0.0122)	(0.144)	(0.146)	(0.144)	(0.146)
Observations	14,542	6442	7469	6442	7469
R-squared	0.000	0.456	0.399	0.456	0.399

* p < 0.1, ** p < 0.05, *** p < 0.01

) reveal systematic differences (χ² = 63.60, p = 0.000): long-listed firms show stronger size effects (β = 0.184 * vs. 0.153 *) and ownership concentration impacts (β = −2.241 * vs. −2.162 *), with cash flow being uniquely significant (β = 0.478 *). Short-listed firms exhibit greater sensitivity to profitability (ROE β jumps from 0.0286 to 0.207 * across models) and independent directors (β = 0.355 *), while suffering more from age-related ESG declines (β = −0.0174 *). These patterns reflect distinct strategic priorities-mature firms leverage scale advantages while managing ownership conflicts, whereas younger firms prioritize short-term financial performance and governance structures. The findings underscore how the corporate lifecycle stage fundamentally shapes ESG implementation pathways and challenges.

4.4.3. Analysis of Industry Heterogeneity

The grouped regression analysis (

Table 17. Regression results for industry characteristics.

Variable	(1)	(2)
	Heavily polluting Group	Non-heavily polluting Group
lnSize	0.188 ***	0.169 ***
	(0.010)	(0.005)
ROE	0.201 ***	0.045 ***
	(0.026)	(0.006)
TOP-10	−1.978 ***	−2.259 ***
	(0.055)	(0.030)
AGE	−0.015 ***	−0.012 ***
	(0.002)	(0.001)
LEV	−0.548 ***	−0.644 ***
	(0.066)	(0.034)
Growth	−0.000	0.002
	(0.001)	(0.007)
BoardSize	0.010	0.009 *
	(0.008)	(0.005)
IndepRatio	0.783 ***	0.165
	(0.232)	(0.129)
CF	−0.383 **	0.517 ***
	(0.157)	(0.082)
Observations	3123	10789
R2	0.390	0.432
Adjusted R2	0.389	0.432

* p < 0.1, ** p < 0.05, *** p < 0.01.

) reveals systematic differences: while firm size (lnSize) and profitability (ROE) positively influence ESG_SCORE in both sectors, their effects are stronger for heavy polluters (e.g., ROE β = 0.35 vs. 0.22), likely reflecting greater environmental compliance pressures and image-conscious investments. Conversely, ownership concentration (TOP-10) shows a more detrimental impact in non-polluting firms (β = −2.45 vs. −1.89), possibly due to weaker institutional oversight in these sectors. Control variables exhibit varying significance across groups, with the Chow test confirming statistically significant coefficient differences. These findings suggest that regulatory scrutiny in polluting industries amplifies the ESG benefits of firm resources while mitigating governance-related drawbacks, whereas non-polluting firms face different optimization pathways due to their distinct stakeholder pressures and operational contexts.

5. Discussion

5.1. Result Discussions

In this research, the impact of corporate profitability, size, and ownership structure on ESG performance was tested for listed companies in China. Based on the RBV, legitimacy, and agency theories, the hypotheses posit that market share and risk preference serve as mediating mechanisms and that the impact of firms’ characteristics on ESG is tempered by corporate internal control. Our hypotheses were supported with important differences for the specific context of the Chinese capital market.

Regarding the H1, the study observed a positive impact of ROE on ESG performance, consistent with the RBV assumption that financially stronger companies have the slack resources needed to invest in the sustainability efforts [1,112]. This is consistent with the notion advanced by other researchers, who argue that financial success enables firms to exhibit green innovation, thereby achieving legitimacy [113,114]. This finding validates the legitimacy theory assumption, given that profitable companies are often involved in ESG activities to maintain their societal acceptance and improve firm responsibility images.

Concerning H2, the size of the firm is observed to improve ESG outcomes because large firms have greater visibility in the public eye and are therefore more likely to engage in sustainability reporting practices, supporting the legitimacy theory [57,67]. This indicates that firms engage in ESG practices for self-interest rather than for virtuous motives to be morally good or good humanitarians [115]. Hence, the result is in line with legitimacy theory by showing how reputable companies act to maintain their social image and stakeholder support.

Regarding H3, the study observed a significantly negative impact of ownership concentration on ESG performance, consistent with the agency theory assumption [28] that large stockholders often place greater emphasis on financially oriented short-term objectives than on long-run stakeholder interests [34,60]. The negative impact is stable under various specifications, including alternative ownership measures (HHI), suggesting that governance frictions continue to shape sustainability behavior in the Chinese context.

Furthermore, the mediation results supported H4 and H5, indicating that market share and managerial risk preferences play the role of behavioral channels between firm attributes and ESG performance outcomes. More profitable and larger firms tend to enjoy stronger market power, which may make them more vulnerable to stakeholder monitoring and pressures for legitimacy [64]. There is clear support for the notion of “competitive legitimacy,” according to which firms maintain their leading positions in the marketplace by behaving in socially virtuous ways [30]. While smaller firms may find ways to distinguish themselves on ESG performance at specific instances, the overall environment for the Chinese A-share market is one that focuses on larger and more visible firms. Therefore, empirical findings of a positive connection between market share and ESG performance align with institutional expectations. Most importantly, in line with the study methodology, these mediation effects are recommended to be interpreted in a suggestive, rather than a definitive, causal manner. Further, consistent with the well-known limitations of mediation tests in fixed-effects panel analysis, the finding above is considered tentative evidence.

Where managerial risk preference is concerned, the mediation effect between firm characteristics and ESG performance is affirmatively supported. The proxy for this variable is intensity for fixed assets and R&D investment, which captures a strategic orientation to a longer-run investment compared to behavioral risk preferences. The study result that risk-oriented organizations invest more in ESG efforts is consistent with growing literature connecting long-run strategic horizons with firm sustainability-driven engagement [36,116]. Notably, this outcome does not contradict agency theory. However, it suggests that risk-preference leaders could support long-run sustainability efforts, whereas strong governance within a firm protects against opportunistic short-termism. These distinguish behavioral paths that converge to support the ESG outcome.

Regarding H6, the moderating outcomes indicate that internal control plays a critical role in shaping the relationships among profitability, size, and ESG performance [87]. For companies with stronger internal controls, the positive impact of profitability ratios and size on ESG disclosures is more substantial. This result is more consistent with agency theory prescriptions for governance that focus on monitoring and supporting an environment- and stakeholder-focused management decision-making process [107,117]. The result supports the role of good governance structures as institutions that turn ESG commitment into performance [118]. Furthermore, good governance also provides evidence and assurance related to the ESG efforts, ensuring transparency and legitimacy [119]. Additionally, an effective control system offers insight into the critical threshold levels for governance credibility, with distinct between-company variations from the more general internal control-index trends. However, the strength and direction of such interaction terms could be driven by overall institutional reporting trends; therefore, careful attention is recommended when interpreting.

The study reveals consistent patterns across ESG sub-dimensions (environmental, social, governance). Analysis shows that the effect of firm characteristics is similar across environment, social, and governance factors. This is consistent with the integrated structure of WIND’s ESG rating system and emphasizes that resource capacity and pressures for legitimacy and ownership structures play relatively equal roles across the separate ESG factors. Further research suggested examining ESG factors across specific sectors to determine whether they tend to be more sensitive to resource factors and governance structures.

Finally, lagged models place strong emphasis on the significance of time. Although profits and size lead ESG performance in the following time period, the short time horizon (2021 to 2023) may represent a combination of actual adjustment behaviors and ESG reporting delays. ESG rating reporting is carried out on an annual basis, with known time gaps for observations on ESG ratings within the context of the Chinese economy; therefore, part of the time effect is driven by actual measurements and behaviors. This does not undermine the directionality of the results but introduces caution in interpreting dynamic patterns.

5.2. Implications

5.2.1. Theoretical Implications

First, while the literature has extensively examined the direct impact of corporate features such as size, ownership structure, and profitability on ESG performance [50,57,62], relatively few works explain the underlying channels through which these corporate features affect ESG performance [83]. This research extends and complements the ESG literature, which broadly focuses on ownership types and board traits [80,102], by introducing managerial risk preferences and market share as behavioral channels and by emphasizing the mediating mechanisms linking organizational resources, managerial decision-making, and stakeholder visibility to ESG performance.

Second, this study also extends the literature by developing a framework that integrates the RBV, agency theory, and legitimacy theory. Contrary to much prior research that uses multiple theories simultaneously and separately [31,40,120]. This framework spells out how internal resources (RBV) [33], control dynamics and ownership (agency theory) [28], and external legitimacy (legitimacy theory) [67] combine to form organization ESG performance. This theoretical integration of the study clarifies the theoretical pathways underlying observed ESG heterogeneity in China and shows how distinct theoretical logics converge through behavioral channels.

Third, this study is also significant to the growing literature on the behavioral underpinnings of firm sustainability investing, as it shows that managerial risk-taking behavior partially mediates the relationship between corporate characteristics and ESG performance. The research suggests that risk-taking behavior can serve as a potential underpinning for investing in firm sustainability, and that this can be supported by strong corporate internal governance. Further, this extends the current research by reflecting how managerial behavioral patterns and governance constraints interact, offering a more nuanced understanding of sustainability decision-making in emerging markets [34,121].

Third, the study draws attention to the role of the internal control system itself, emphasizing its significance in the sustainability context, where it plays a crucial role in enabling the sustainability process [86,118]. The study clearly indicates that internal control serves as a moderator between the firm’s attributes and ESG outcomes [87,118]. The concept of distinguishable governance quality (internal control) and categorical governance certification (effective internal control) offers ESG studies a more complex framework for understanding governance processes.

5.2.2. Practical Implications

The study results have important implications for investors, managers, and policymakers. First, the study’s implications for corporate managers highlight the importance of profitability and size, used not only for financial gain but also for sustainability advantages. Managers are advised to adopt a risk-seeking stance, focusing on financial resources directed to innovation linked to ESG rather than other financial investments.

Second, the implication for board members or controlling shareholders is the importance of maintaining control while balancing accountability, as overly concentrated ownership hinders ESG efforts. Diversifying ownership structures is one way to align stakeholders’ interests with sustainability considerations.

Third, regarding policy regulators, improvements in the standard of disclosure of internal controls, coupled with tighter ESG reporting guidelines, are beneficial for firms in terms of transparency and performance. Regulators promote the reporting of “effective internal control” to encourage firms to integrate ESG considerations into their operations.

Lastly, from the investor perspective, the implication is that ownership concentration remains an essential risk indicator in assessing ESG performance. Organizations with dispersed ownership and a strong, effective governance structure are more likely to deliver credible, stable ESG outcomes. In addition, managerial investment and market share behavior also serve as essential indicators of organizations’ sustainability trajectories, providing additional dimensions for screening ESG and portfolio formation.

5.3. Limitations and Future Directions

Although the work has many strengths, there are also some limitations for future studies to explore. Firstly, although the use of secondary data from panels contributes to greater objectivity, there may be unseen managerial or cultural variables shaping ESG practices. In addition, the sample period (2021–2023) is relatively short, limiting the extent to which true adjustment can be measured. A future study is suggested to use archival research methods in partnership with survey or interview methods to more thoroughly tap into managerial cognition. Additionally, a longer time window assists in distinguishing reporting-related lags from strategic ESG transitions.

Secondly, since this work is based on Chinese A-share firms, it makes it difficult to generalize the results to other institutions, possibly creating room for future studies in other Asian or Belt and Road countries. Thirdly, although the model used the variables’ lagged values and instruments, the problem of endogeneity cannot be completely discounted. The study is suggested to be expanded in the future by employing dynamic panels, specifically the Generalized Method of Moments.

Fourthly, the mediation model uses step-wise regression, an approach that is less stringent than the bootstrapped panel model or SEM. Since the bootstrapped panel model and the SEM could not be conducted in this study, the indirect effects and confidence intervals could not be estimated. The results obtained from the model regarding the mediation effect may therefore be considered preliminary.

Fifth, although this study follows a traditional methodological approach to managerial risk preference by quantifying risk preference as the proportion of investments in R&D and fixed assets relative to overall assets, this proxy aggregates investment categories that differ in risk properties. Future studies are suggested to consider the use of surveys and other indicators, such as earnings volatility and innovation intensity. Not only is this true when outliers are dealt with via winsorization and manual checks, but there is also volatility inherent in emerging market data owing to the denominator effect. Future research is suggested to explore the application of strong regression methods and Bayesian estimators.

Finally, this research focuses on the characteristics and channels at the corporate and behavioral levels. Future research is suggested to add board network properties and outside pressures, as well as executive characteristics, to obtain a more complete picture on ESG diversity for emerging markets such as China. Also suggested examining industry-specific ESG factors, as the environmental and social forces acting on high-emission industries differ significantly from those on low-emission sectors, including finance and technology.

6. Conclusions

The current study examined the relationships between the firms’ characteristics (e.g., profitability, size, and ownership concentration) and their ESG performance. Additionally, specific emphasis is placed on the mediating impacts of market share and risk preference, as well as the moderative roles played by the internal control system, utilizing the three theories, namely the agency, legitimacy, and resource-based. The result indicates that profitable firms, large firms, and those with strong risk preferences perform well on ESG, but ownership concentration hinders sustainability engagement.

Market share and managerial risk preference are the behavioral channels transmitting firm attributes to ESG performance, and strong internal controls are also beneficial. The findings support the proposition that financial resources, strategic behaviors, and ESG performance are all crucial in deciding firm sustainability. This work contributes to integrated theoretical and practical insights, prompting managers to connect the profitability of the firm with long-term ESG strategies, while also encouraging regulators to improve the framework of internal control/ESG reporting.