Does external assurance on CSR reporting contribute to its higher quality? Empirical evidence from China

This paper examines the difference that the assurance brings to the quality of CSR reports in the Chinese institutional setting, in particular, the difference in quality (proxy – RKS ranking) of assured and unassured CSR reports, as well as whether the high ownership concentration and corresponding to it “entrenchment effect” obstruct the positive impact the assurance exerts on the quality of CSR reports. The paper examines CSR reports on 2,292 firm-year observations of large Chinese companies over three years (2015–2018). The hypothesis development process predicates on the signaling and stakeholder theories, whilst this study applies regression analysis to test the hypotheses. Consistent with the predictions of signaling and stakeholder theories, the paper finds that assurance contributes to the higher quality of CSR reports. Moreover, the study finds that assured CSR reports have higher sub-scores in all four aspects of RKS ranking. However, as ownership concentration exceeds 50 per cent and reaches the majority, it thwarts the advancement in the quality of CSR reports through its assurance. The paper provides an initial empirical account of the role of assurance in the emerging CSR reporting practice in China. The paper contributes to the modest body of empirical research on the function of external assurance in the CSR area by explicating the role played both by the accounting (external assurance) and corporate governance (ownership concentration) infrastructure to ensure high quality of CSR reporting. The paper briefs local, international regulatory authorities and the business community about the importance of external assurance for the CSR reporting quality.


INTRODUCTION
Unlike CSRR, CSRA remains mostly voluntary, which means that firms volunteer to verify CSRRs by initiating an external quality assurance process led by independent auditors, experts or consultants who produce their assurance report communicating through it their opinion (Chi et al., 2020;Miralles Quirós et al., 2021;Venter & van Eck, 2021). This external verification process incurs extra costs that companies will only be predisposed to cover if they can discern that this will bring them some benefits (Miralles Quirós et al., 2021).
Independent third-party assurance of the content and structure of CSRR is intended to improve the relevance, reliability and comparability of these reports and, for this reason, increase the over-  Zorio et al., 2013). However, the unregulated and loosely controlled nature of the assurance process, the presence of several varying standards for CSRA (as well as the scope and level of assurance), panoply of assurance providers entails that the assurance can vary significantly (De Beelde & Tuybens, 2015; Perego & Kolk, 2012). Besides the external benefits already mentioned, CSRA can play a crucial role in establishing/enhancing the internal mechanisms and management of firms, as it is believed that companies strive for accurate data confirmed by a third party to build their decision-making processes and business strategy (Zorio et al., 2013).
Despite growing demand for CSRR, few studies so far have investigated whether CSRA contributes to the higher quality of CSRR (Du & Wu, 2019; Maroun, 2019; Moroney et al., 2012). Sensing that dearth of research into this developing field, the paper sets to investigate this link in a Chinese institutional setting. The paper aims to find out whether the CSRA affects the quality of the CSRR itself and what mechanics is here at play (which theory best suitable to explicate this). In addition to this, the paper investigates whether the ownership concentration reaching the level of "entrenchment effect" hinders a positive impact the assurance exerts on the quality of CSRR.
The remainder of this paper is structured as follows. Section 1 provides a brief overview of theories explicating the incentives for assurance on CSRR, debates a literature review, and develops hypotheses. Section 2 discusses the method, including sample selection and research design, and Section 3 presents the findings. Section 4 is a discussion, including limitations and possible directions for future research, and finally, the paper wraps up with conclusions.

THEORETICAL BACKGROUND, LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
Extant research to capture the assurance effects focuses on several theories conditionally divided into two groups: economics-based (e.g., agency theory) and social-political theories (e.g., legitimacy theory, signaling theory, stakeholders' theory) (

Agency theory
Agency theory regards an audit as an independent monitoring function, which is carried out to verify management actions and deeds to meet the needs of shareholders, thus leading to reduced information asymmetry. Previous research shows that external assurance reduces agency costs associated with information asymmetry (Carey et al., 2000) and safeguards a firm's reputation (O'Dwyer et al., 2011). In accordance with the basic principles of agency theory, the independent third party assurance might compensate for the inability of outside shareholders to follow the actions of insiders directly (Forst & Hettler, 2019

Stakeholder-agency theory
Stakeholder-agency theory considers managers as unique stakeholders who are at the center of the nexus of contracts. Given a manager's central position and insider knowledge, managers are "agents", but not shareholders' agents, but stakeholders (Hill & Jones, 1992). Stakeholder-agency theory links management to their stakeholders, who have a claim on a company (Moroney et al., 2012). Stakeholder-agency theory presupposes that managers have contractual relationships with all types of stakeholders and thus work as "agents" for them (Dutta & Dutta, 2020). In situation of "stakeholder diffusion" (when stakeholders are dispersed and no one individually controls serious resources on their own), the need for "monitoring structures" arises to monitor management performance (Hill & Jones, 1992). External assurance on CSRR, giv-en its mostly voluntary status, qualifies to be and may also be regarded as a "monitoring structures".
Counter to agency theory, social-political theories, such as legitimacy or stakeholder theories, bring a more all-encompassing perspective on CSRR as they explicitly acknowledge that "organizations evolve within a society that encompasses many political, social and institutional frameworks" (Cormier et al., 2005, p. 7).

Legitimacy theory
In legitimacy theory, corporations are seen as social creations, and their existence relies on the inclination of society to persist in allowing them to operate (O'Donovan, 2002). Thus, companies will behave in such a way that society will acknowledge them to be socially responsible (

Signaling theory
Signaling theory posits that non-financial reporting over various channels can lessen information asymmetry between managers and stakeholders (Connelly et al., 2011;Karaman et al., 2021).
Companies can determine to 'signal' their behavior to the outward parties because of this information asymmetry by assuring its CSRR (Simaens & Koster, 2013). Signaling theory springs from Spence's paper from year 1973 exploring signaling at the labor market (Spence, 1973). Thereafter, management researchers have more and more often utilized signaling theory 'to help explain the influence of information asymmetry in a wide array of research contexts' (Connelly et al., 2011, p. 40). This theory implies the presence of three principal components in the signaling process such as the signaler, the signal itself, and the receiver of the signal (Connelly et al., 2011).
Overall, signaling is utilized to differentiate and highlight quality, intent and risk issues (Connelly et al., 2011;Li & Zhang, 2010). According to theory of Melumad and Thoman (1990), in audits within the voluntary context, the choice to recruit an external auditor is going to signal a firm's low-risk type, whereas dodging an audit altogether signals to the contrary -a high-risk type (Melumad & Thoman, 1990). In light of this theory, and provided that a firm's CSRR quality is indicative of its risk type, good CSR performers would be more inclined to attain external assurance on their CSRR as a signal of their low-risk type (Li & Zhang, 2010). Similarly, signaling distinguishes between high-quality firms and low-quality firms. The companies may be aware of their own actual quality, while outsiders are not, so information asymmetry exists. In this context, each company might avails itself of the opportunity to signal or not to signal its actual quality to outsiders (Connelly et al., 2011). In our case, the signaling takes place through the assurance provided by the independent third party. Hence, as a rule, low quality of CSRR would not make it to endure the stringency of assurance providers as the company chose a path to assure its CSRR signals to stakeholders the CSRR higher quality. Therefore, signaling theory assumes that higher corporate social responsibility performers are more prone to assure their CSRR externally.

Stakeholder theory
Stakeholder theory postulates that firms are part and parcel of a broader social system in which their businesses influence and are influenced This study is completely on board with Cormier et al. (2005), arguing that CSRR is a compound phenomenon that cannot be interpreted restrictively by one single theory. Nevertheless, it is believed that the assurance on CSRR best suitable to describe is the signaling theory. Despite pressure from stakeholders and companies' efforts to obtain legitimate and compelling arguments, signaling theory is more convincing given the topic of this study. Firms 'signal' their higher quality through assurance on their CSRR, distinqishing themselves from low-quality firms that shun scrutiny from assurance providers. Companies incur additional costs to overcome the asymmetry of information, which is present between them and stakeholders, and add credibility to its CSRR. This study adapts a "signaling model" of Bagnoli and Watts (2017); however, here the approach can be called "stakeholder-signalling", since it is the expectations of stakeholders and their positive assessment of what the companies are striving to attain.

CSRA and CSRR quality: the role of major shareholders
Ownership concentration is an important indicator of a company's equity distribution, which often determines how it fares conflict-wise internally. When equity is dispersed, potential conflicts between principals and agents may exist, while the external shareholders hope that management discloses more information to reduce the degree of information asymmetry. When the shareholding is concentrated, large shareholders may invade the interests of minority shareholders by preventing a company from disclosing information.
Depending on the difference in the shareholding ratio of major shareholders, the gap size between control rights and cash flow rights, and the nature of major shareholders, the conflict between major shareholders and minority shareholders could result in two scenarios: 1) incentive-alignment effect or 2) entrenchment effect ( In his epochal article, Chow (1982) claims that the main reason for recruiting an independent auditor is to reduce the information asymmetry between managers, shareholders and stakeholders (Chow, 1982). Forst and Hettler, on the sample of U.S. dual-class firms, find that insider ownership is positively related with audit charges, the probability of recruitment of a Big Four firm or expert auditor, and auditor independence, thus, testifying to the self-bonding mechanism through enhanced external assurance (Forst & Hettler, 2019).
The purpose of this study is to examine the contrast that assurance makes to the quality of CSRR in a Chinese institutional setting. Adopting signaling and stakeholder theories, the paper suggests that external assurance enhances the CSSR quality analogous to financial reporting case. Thus, this paper conducts a more refined analysis of interdependent and interrelated influence external assurance has on the quality of CSRR when coupled with ownership concentration and its corresponding entrenchment effect.
Thus, in line with the discussion above, the paper's hypotheses can be formally stated as follows: H1: Sustainability-related assurance significantly improves the quality of CSRR.
H2: Significant ownership concentration and its corresponding entrenchment effect are negatively associated with the quality of CSRR, and thus the positive effect of CSRA on quality of CSRR is thwarted by the "entrenchment effect".

Sample selection and data collection
To test the abovementioned hypotheses, data of all listed companies on Shanghai and Shenzhen stock exchanges (China) were collected covering the period from 2015 to 2018. The CSRR score and CSR assurance data in this paper are all from "Rankins CSR Ratings" or Runlin Global's Rankings rating (hereafter referred to as RKS), and the remaining data originate from the China Stock Market & Accounting Research (CSMAR) database. The initial sample of 13,023 was screened (Table 1), and the final sample comprised 2,292 firms. Note: ST means "special treatment" and occurs when a company suffers losses for two consecutive years or its net assets are lower than the par value of the stock. *ST is added to those companies' names whose operations have not improved in the third year after ST, which means delisting risk.
This paper uses statistical software STATA16.0 for data processing and statistical regression analysis.  CSR assurance. CSR assurance is an independent variable. It is used to measure whether a company's annual CSR report has been inspected and as-sured by an external assurer. If a company's CSR is audited, the value is 1, otherwise it is 0.
Ownership concentration. This paper uses the ratio of the shareholding ratio of the largest shareholder to the second largest shareholder to measure ownership concentration. It is suggested that the greater the ratio, the higher the concentration of the major shareholder's equity and the stronger its entrenchment effect.
Control variables. Control variables are included to enhance the internal validity of this study. The paper controls for firm size, firm age, leverage, return on assets, corporate growth, ownership rights type, board size and the dummy of year and industry.
The main variables are defined in Table A2 (see Appendix).

Model construction
This paper constructs the following two models. It first builds a model (1) to verify the first hypothesis of the study. The quality of the dependent variable CSR is measured as RKS rating total score (Score), and a regression model is constructed.
In the robustness test, when the total rating (Credit) is used as a dependent variable, its value is discrete, and a Poisson regression model is constructed; when the four sub-dimension indicators of RKS rating (M, C, T, and I) are used as a dependent variable, the regression is constructed model.
In the empirical regression of model (1), clustered by a company, control annual effect and industry effect, and report Robust-t value adjusted for heteroscedasticity. If the coefficient α_1 is significantly greater than 0, it indicates that the CSR reporting assurance is instrumental to enhance the quality of CSR information disclosure.
In the regression model (2), this paper introduces the proportion of major shareholders (FstMonitor) as the adjustment variable, and Audit×FstMonitor is the product of the proportion of assurance (Audit) and major shareholders (FstMonitor). If α_2 is significantly less than 0, it indicates that the entrenchment effect of major shareholders is enhanced, which affects the quality of CSRR through the CSRA. Control represents the control variables.
In the paper, the variance inflation factor (VIF) function is run to test multicollinearity in the models. The test returned mean VIF 1.28 and 1.29, respectively. The results testify that there is no multicollinearity in both models, since, in general, when the maximum VIF value does not exceed 10, it is considered that there is no multicollinearity in the model.

Descriptive statistics and correlation analysis
The descriptive statistics of the main variables in this study are shown in Table 2.   It can be seen from Table 2 that the average value of the total score of the CSR reporting of Chinese listed companies is only 42.60, and the standard deviation is as high as 11.91. This shows that the overall quality of China's CSR report is not very high yet, and the CSR reporting quality fluctuates from company to company. Moreover, the average value of Audit (CSR assurance) is only 1.96%, indicating that the practice of CSR assurance is in its infancy.
The correlation analysis results presented in Table  3 show that CSR reporting assurance (Audit) and CSR reporting quality (Score) are significantly (0.366) positively correlated at the 1% level. This indicates that the quality of CSR reports of companies that have conducted its assurance is significantly higher than that of companies that have not conducted audits.
For further analysis, statistical inspections are conducted to study various patterns between companies that assured their CSR reporting and those who did not do that, as well state-owned vs. private firms.

The difference between companies that provide CSR reporting assurance and those that do not
The statistical test of the difference in the quality of CSR of companies that have conducted CSR report audits and those that have not is shown in panel I of Table 4. The results show that the total CSRR score (Score) of companies that conduct CSRA is significantly higher than that of companies that do not, and the significance level reaches 1%.
Moreover, the inspection found out that CSR reports of companies that conduct their assurance have higher sub-scores in all four aspects of RKS rating, namely macrocosm (M), content (C), technicality (T), and industry (I) than those that do not at a significance level of 1%. The two subcategories with the largest difference that CSR assurance brings to are content (13.678) and macrocosm (8.724), which are fully consistent with the hypothesis proposed in this study, since the assurers inspect foremost content of the reports.

The difference between state-owned vs. private companies
The statistical test results of the difference in the quality of CSR reports between state-owned vs. private companies are shown in Table 4, panel II. It can be seen that the total scores and subitems of CSR reports of state-owned enterprises are higher than non-state-owned enterprises, and are significant at the 1% level. Again the RKS indicator where the biggest divergence is noted is content (C) (2.008). This indicates that the quality  Note: ***, **, and * represent statistical significance at the 1%, 5% and 10% levels, respectively. of CSR information disclosure of state-owned enterprises is higher than that of private enterprises, and the overall performance of social responsibility is better.

CSR assurance and CSR disclosure quality
Regression is performed on sample data according to model (1), and the results are listed in Table A3 (see Appendix).
The paper uses the RKS's CSR report total score (Score) as a dependent variable for regression testing. The results in column I show that the CSR assurance (Audit) significantly positively (31.416) affects the quality of CSR information disclosure, and the significance level is 1%. This shows that a company's CSR assurance has indeed improved the quality of CSR information disclosure through supervision, restriction and proper risk assessment. Thus, this test result supports the first hypothesis of this study.
Among the control variables, the size of a company (3.574, p < 0.01), the board size (5.107, p < 0.01), and ownership rights type (1.628, p < 0.01) are significantly positively correlated with the quality of CSR reports, whereas the company age (-1.682, p < 0.01) and financial leverage (-4.726, p < 0.01) are significantly negatively correlated with CSR total scope as assessed by RKS. This shows that, compared to small-scale companies, largescale companies disclose better social responsibility information, the big board is more conducive to promoting corporate social responsibility, and that state-owned companies are more inclined (in effect mandated) to disclose social responsibility information than their non-state owned counterparts. On the contrary, long-time listed companies and companies experiencing deterioration in their financial status, endure non-beneficial influence on the quality of CSR reporting as a result.

CSR report audit and CSR information disclosure quality
The paper uses model (2) to regress the sample.
The results in column II of Table A3 (Appendix) show that the CSR report assurance (Audit) coefficient is significant (24.867) at the 1% level, which still supports the first hypothesis. However, when the major shareholder's shareholding ratio exceeds 50% and reaches absolute holding, a positive correlation between the variable FstMonitor and CSR is no longer significant. Although the coefficient of the variable FstMonitor is still positive, it is no longer significant. The coefficient of the variable Audit×FstMonitor is negative (-0.100, p < 0.1), which confirms that the CSR assurance is negatively correlated with the ownership concentration, that is, the degree to which the largest shareholder effectively controls the company, and is significant at the 10% level. This means that when the majority shareholder's shareholding ratio is increased to be able to effectively control the company, it will hinder and put a cap on the advancement in the quality of CSRR through its assurance. Thus, hypothesis H2 is confirmed.

DISCUSSION
This paper contributes to CSRR and CSRA research by analyzing these practices in China. The large sample consists of 2,292 Chinese listed firms covering the period from 2015 to 2018, which assists in generalizing the findings. The data from this study comes from among the largest listed companies on the Shanghai and Shenzhen stock exchanges through information extricated from RKS rankings and CSMAR database.
Adopting signaling and stakeholder theories, the paper anticipates and proves that external assurance enhances the CSSR quality akin to financial reporting case. The study finds that the quality of CSRR of companies that conduct its assurance have higher sub-scores in all four aspects of RKS rating, namely macrocosm (M), content (C), technicality (T), and industry (I), compared to those that chose not to audit it. However, the paper also documents that when the ownership concentration reaches majority, it thwarts the advancement in the quality of CSRR through its assurance. The emergence of the "entrenchment effect" of major shareholders prevents companies from conducting external assurance, and thus puts small and medium shareholders into situation of being deprived of reliable information.
Overall, the results verify the signaling theory as the result is interpreted in a way congruent with signaling theory: more qualified CSR reporters merely take advantage of external assurance as a quality signal, whilst poor quality reporting is not assured (Maroun, 2019). Those findings from Chinese institutional setting are on par with conclusions reached by Maroun (2019) and Moroney et al. (2012) in South Africa and Australia, respectively. These findings (assured CSRR are of higher quality) are to be interpreted in line with signaling theory that higher CSR performers who relate themselves to either low-risk type firms (Melumad & Thoman, 1990) or high-quality firms (Connelly et al., 2011) signal their superiority over high-risk and low-quality firms. Firms, following the term of Bagnoli and Watts (2017), select a "separating equilibrium" path: incurring extra cost but displaying their superiority. Thus, the paper testifies that in the voluntary context, higher corporate social responsibility performers are more likely to assure their corporate social responsibility reports externally.
The paper contributes to the modest body of empirical research on the function of external assurance in the CSR area by explicating the role that the accounting (external assurance) and corporate governance (ownership concentration) infrastructure plays in order to ensure high quality of CSR reporting.
Care should be taken as some caveats are to be considered when interpreting those findings. First, the sample is limited to the available data, as only the data of listed companies included in the rating system by RKS in 2015-2018 are used for testing. As a result, listed companies and non-listed companies not included in the RKS ranking system need to be further studied. Secondly, caution is advised about any interpretation of those findings as suggesting mandatory CSRA, since mere mandating without corresponding to its legal condition is futile. The beneficial impacts of CSR reporting are believed to be crucially dependent on the alignment of legal and economic incentives, which would jointly spur firms in terms of high quality of CSRR. Lastly, this study scrutinizes only Chinese companies, and thus, those findings cannot be generalized and applied to companies in other jurisdictions.
Although this paper demonstrates that external assurance on CSRR contributes to its higher quality, the exact mechanism of how and why is not ascertained yet. This provides the scope for more investigative studies, which could answer the following questions: What is the role of the external assurer in identifying weaknesses in CSRR and areas that could be addressed to improve it? Does auditing make preparers more aware of the need for complete, accurate, and balanced information? Do the CSRA and the interaction with the assurers alter a company's attitude and approaches to the preparation, evaluation and submission of CSRR? Moreover, given the panoply of assurance providers' types, it is advisable to compare the assurance effects of different audit subjects (something we was unable to undertake due to unavailability of corresponding data). Future research could exploit these routes for exploration to extend this lineage of the studies. Furthermore, it is believed that assured CSRR would be of higher quality in mandatory context as well, which would mean that other theories are also instrumental in explicating CSRA phenomenon. Once the mandatory context is out there, the paper urges researchers to explore the underlying relationships and pledge to follow that route as well.

CONCLUSION
This paper aims to test the effect of the external assurance on the CSRR quality in China. From the regression analysis of the data, followed by robustness tests, the following conclusions could be drawn. First, the results indicate that CSRA contributes to the higher quality of CSRR. Moreover, the study finds out that CSRR of companies that conduct its assurance have higher sub-scores in all four aspects of RKS rating, namely macrocosm (M), content (C), technicality (T), and industry (I), than those of companies that do not. However, when the major shareholder's holding ratio exceeds 50% and reaches majority, it will hinder and put a cap on the advancement in the quality of CSRR through its assurance.
Thus, in the Chinese institutional setting, external assurance improves the quality of CSR reporting. Nevertheless, external assurance combined with excessive concentration of ownership accompanied by an 'entrenchment effect' ceases to deliver a positive effect on the quality of CSRR. Thus, local and international regulatory authorities and the business community should emphasize both the significance of external assurance for CSR reporting quality and reasonable concentration of ownership, since the paper's evidence show that the 'entrenchment effect' blocks positive impulses produced by external assurance. The provision of an assurance statement with a sustainability report generates greater credibility in the report from users perspective Level of Assurance No significant association Type of assurance practitioner

AUTHOR CONTRIBUTIONS
Public accounting firm having (as opposed by a specialist consultant) a more positive impact on report users' confidence in the sustainability report Note: * compiled by the authors.  Note: Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.1.