Impact of Tax Administration on ESG Performance—A Quasi-Natural Experiment Based on China’s Golden Tax Project III

Abstract

With the growing importance of sustainable development, ESG is receiving attention from governments, firms, and investors. This study uses the reform of China’s tax inspection system—Golden Tax Project III—as a quasi-natural experiment to examine the impact of the enhanced tax administration on firms’ ESG performance with a sample of Chinese A-share listed companies from 2008 to 2020. The findings, which were identified via a time-varying difference-in-differences (time-varying DID) approach, indicate that Golden Tax Project III significantly reduces firms’ ESG performance and has the most significant impact on the environmental dimension and the minor impact on the social dimension. The negative impact of Golden Tax Project III on ESG performance is more substantial among non-state firms and firms in regions with higher levels of marketization than other firms. The mechanism test shows that Golden Tax Project III negatively affects ESG performance by increasing the tax burden on firms. The moderating effect test shows that tax incentives can effectively mitigate the dampening effect of strengthened tax administration on ESG performance. This study enriches the literature on the factors influencing firms’ ESG performance to a certain extent; it helps the government use taxation policies to inspire firms to improve ESG performance, contributing to sustainable development.

1. Introduction

ESG is a comprehensive indicator for evaluating companies in terms of environment (E), social responsibility (S), and corporate governance (G) and was first introduced in 2004 in the United Nations study Who Cares Wins, and is an extension of corporate social responsibility. With the introduction of sustainable development, ESG has received increasing attention from stakeholders such as companies, investors, and governments. Moreover, ESG performance has become an essential indicator for evaluating the non-financial performance of a firm and its sustainability [1]. However, ESG practices need to input specific costs, so there is a conflict between improving ESG performance and maximizing firms’ profits [2]. Additionally, improving ESG performance is a process that requires long-term investment, which leads to a need for more internal motivation for firms to engage in ESG activities and improve their ESG performance. Taxes are an essential expense for enterprises, and the greater the tax bill, the greater the loss of corporate value. Hence, firms usually resort to various means of tax avoidance. As shown by the 2012–2017 China Business Operators Questionnaire Tracking Survey Report, over half of entrepreneurs have long considered the tax burden the most significant difficulty in running a business today. The extent of firms’ tax avoidance depends on the strength and efficiency of tax collection and administration, which in turn depends on improving the infrastructure of the tax collection system. Rahman found that effective use of computer software can enhance management efficiency with a reduced workforce [3]; in recent years, Golden Tax Project III, implemented by the Chinese government, has improved tax administration system infrastructure by applying information technology such as big data and cloud computing. Golden Tax Project III enhanced tax administration and reduced opportunities for tax avoidance, thus increasing the tax and cost burden on firms [4,5,6]. In addition, studies have pointed out that corporate tax payments and corporate social responsibility are substitutes for each other and that degree of corporate tax avoidance is significantly and positively related with ESG performance [7,8]. This suggests that taxation may increase cost pressure on firms and reduce their willingness to engage in ESG activities. ESG construction is still at an early stage of development in China, and most firms have not made ESG practices a priority for development. According to the ESG Research Report for Chinese Listed Companies (2021), only 8.07% of Chinese listed companies have an ESG rating of A. Therefore, how to use tax policies to stimulate corporate willingness to practice ESG and promote their ESG performance is an urgent issue in achieving sustainable development, which is of great significance to the government, regulators, and firms.

Therefore, against the backdrop of deepening tax reforms and increasing stakeholder attention to firms’ ESG performance, what is the impact of enhanced tax administration on corporate ESG performance? How do tax incentives impact the relationship between enhanced tax administration and firms’ ESG performance? These are the two main questions of this paper, which are important for the relevant government departments in order to clarify the relationship between tax administration and firms’ ESG performance and then formulate appropriate tax incentives or other tax policies to promote firms’ ESG performance.

In the existing literature on the impact of tax administration on firms, scholars have mainly studied the economic consequences of tax administration for different countries and regional samples. For example, MAXIM found that stricter tax administration improved business performance, with a one standard deviation increase in collection intensity corresponding to a 2.6% increase in annual company revenue growth [9]. Omrane (2007) found that more robust tax administration made bond issuance cheaper for private companies [10]. Ali et al., Lorenzo et al., and Fan et al. argue that the development of information technology has significantly increased tax authorities’ ability to collect taxes, inhibiting corporate tax avoidance and increasing their tax burden [11,12,13]. In the existing literature on ESG performance, scholars have mainly focused on the impact of ESG performance on financial performance [14], corporate value [15], financing constraints, and other aspects of economic consequences [16] and less on the factors influencing ESG performance.

This study uses China’s Golden Tax Project III as an exogenous policy experiment to investigate whether a more robust tax administration reduces firms’ ESG performance. This paper aims to examine the relationship between tax administration and firms’ ESG performance, provide some empirical support to government departments in formulating tax policies, and provide methods to improve firms’ ESG performance. Golden Tax Project III enables comparative analysis and cross-checking of tax payable by providing a nationwide network of tax departments, which has tightened the process of tax payment by enterprises. In this way, Golden Tax Project III could reduce the possibility of tax evasion and increase tax payable by firms [4,17,18]. The exogenous nature of the policy is evident and provides a quasi-natural experiment to accurately examine the impact of enhanced tax administration on firms’ ESG performance, alleviating the endogeneity problem to some extent.

This study contributes to the literature in several main aspects: First, up to now, most literature has focused on the economic consequences of ESG performance, and few studies have investigated the influencing factors of ESG performance. This study explores the impact of Golden Tax Project III on firms’ ESG performance, which enriches and expands the research on the external influencing factors of ESG performance and the policy economic consequences of Golden Tax Project III. Moreover, using Golden Tax Project III as a policy shock, the exogenous nature of the shock mitigates the possible endogeneity in assessing the impact of tax administrations on firms’ ESG performance. Secondly, this paper confirms that firms’ ESG performance is more sensitive to tax burden from the perspective of increased tax administration and that once tax administration is strengthened and tax burden increases, firms’ ESG performance decreases significantly. Therefore, in the context of downward pressure on the Chinese economy and the need to balance economic concerns with sustainable development, the findings of this paper help government departments to implement tax reduction policies and use taxation tools to actively foster the motivation of market players and increase the willingness of enterprises to participate in ESG practices. Thirdly, based on the heterogeneity of firm characteristics, this paper reveals the heterogeneous impact and transmission mechanism of tax administration on ESG performance and improves the understanding of the mechanism of the tax burden. It helps the government formulate targeted tax incentives for non-state enterprises that do not have credit advantages and are more affected by tax regulation to alleviate their financial pressure and thus motivate them to engage in ESG practices and meet stakeholders’ requirements actively.

The remainder of this study is presented below. Section 2 reviews the relevant literature and presents the research hypotheses based on the existing literature. To test the research hypotheses presented in Section 2, Section 3 presents the data sources, variable definitions, and the empirical model of this paper. Section 4 reports the empirical results. Section 5 performs robustness tests on the empirical results and reports the results. Section 6 discusses the features, contributions, and limitations of the paper. Section 7 gives conclusions and recommendations.

2. Literature Review and Research Hypothesis

2.1. Literature Review

2.1.1. Tax Administration

Most research on tax administration focuses on identifying and evaluating the economic consequences, dividing the impact of tax administration on enterprise into tax effect and governance effect. The tax effect refers to the fact that increased taxation regulation by the government makes it more difficult for firms to seek tax avoidance. On the one hand, the opportunity cost of tax avoidance rises significantly, the space for tax avoidance is compressed, and the tax burden on firms increases [1,19]. On the other hand, tax avoidance reduces the quality of accounting information [20] and provides room for opportunistic behavior by managers, exacerbating information asymmetry and agency problems [21]. The governance effect refers to the ability of the tax administration to govern the corporate administration, including reducing the opportunistic behavior of management and increasing firm value.

In response to the tax effect, Jin and Huang found that enhanced tax administration limits firms’ tax avoidance and breaks the implicit contract between companies and the government, thus reducing the fulfillment of corporate social responsibility [18]. Ji and Zhen found that firms’ R&D investment and R&D output dropped significantly after the implementation of Golden Tax Project III, suggesting that strengthened tax administration and enforcement would inhibit enterprise innovation [4]. Li and Wang concluded that enhanced tax administration significantly reduced the total productivity factor of firms by weakening the efficiency of capital allocation and inhibiting technological innovation, and that this effect was more pronounced for firms managed by local tax bureaus, firms with high financing constraints, and innovative firms [5]. Li and Yao found that an increase in the intensity of tax administration reduces the scope for tax avoidance and increases the financial burden of firms, leading them to prefer debt financing with a “tax shield effect” [22]. Pham et al. found that mandatory tax administration can lead to accounting treatment (e.g., depreciation, income recognition, etc.) under tax management requirements, which can harm the quality of financial reporting [23]. Lin et al. argued that political ties between board members and government officials could weaken the tax effect of tax administration, i.e., affect the disincentive effect of tax enforcement on corporate tax avoidance [24]. Lim found that stronger tax administration can increase firms’ financing costs [25]. Moynihan et al. showed that tax collection and administration activities increase the administrative burden on firms [26].

From the perspective of the governance effect, the separation of powers in the modern firm system provides managers with the possibility of using their authority to commit self-interested acts, which impairs shareholders’ interests. Tax administration can be used as an external governance mechanism to mitigate the principal-agent problem [27,28]. Guedhami and Pittman found that tax administration and enforcement can induce firms to improve the quality of accounting information, increase information transparency, and reduce information asymmetry, thereby alleviating corporate financing constraints [10]. Gallemore and Jacob also argued that tax administration could improve the information environment of banks and SMEs, enabling banks to make more rational lending decisions and thus increase the probability of receiving credit support for SMEs [29]. Ye et al. found that tax administration via big data can significantly improve the robustness of corporate accounting by increasing the transparency of corporate information, and this effect is more pronounced in firms with weaker external monitoring and tax collection [17]. In addition, Rego et al. showed that once tax authorities become suspicious of a company’s tax behavior, the risk of tax audits and the cost of non-compliance for firms increases significantly [30]. This means that improving the tax authorities’ tax-related information-monitoring techniques can help curb management’s incentives to hide bad news, commit opportunistic behaviors such as surplus management, and discourage corporate financialization from achieving governance effects on the firm [31,32]. Zhang et al. found that stronger tax administration can improve corporate investment efficiency by reducing excessive expenditures [33]. Chen et al. found that the higher the intensity of tax administration, the lower the risk of the corporate share price collapse, suggesting that tax administration can play some external governance role [34]. Wang et al. found that tax administration can significantly contribute to CSR performance [35]. Mironov concluded that improved tax collection and administration capacity enhances the monitoring of tax sources and limits the diversion of income by managers, which in turn contributes to improved corporate performance [9].

2.1.2. ESG Performance

Regarding ESG performance, scholars have focused on its economic consequences and influencing factors. Concerning the influencing factors of ESG performance, Sun et al. found that the green finance pilot zone policy had a significant and positive impact on firms’ ESG performance [36]. Wang D et al. argued that the development of financial technology could alleviate corporate financing constraints by reducing transaction costs and information access costs in the financing process, thereby enhancing firms’ ESG performance [37]. Using a sample of Egyptian firms from 2007–2016, Abdelfattah and Aboud found that corporate tax avoidance is positively associated with ESG performance, i.e., the lower the intensity of tax enforcement, the higher firms’ ESG performance [38]. Using a sample from the Italian banking sector, Menicucci and Paolucci concluded that banks with large board size, high levels of board independence, and CSR development committees have better ESG performance [39]. Dyck et al. found that institutional investors can influence corporate social responsibility policies by putting pressure on firms to improve their ESG performance [40]. Chen and Liu also suggested that investor attention can positively affect firms’ ESG performance by alleviating corporate financing constraints and improving information transparency [41]. Barros et al. found that merger and acquisition activities can improve firms’ ESG performance, but the impact lags by one year [42]. Welch and Yoon showed that improving the competence of managers, who are the primary decision-makers in a company’s business activities, can effectively improve ESG performance [43]. Birindelli et al. and Meng and Zhu both found a non-linear inverted U-shaped relationship between the proportion of female board members and firms’ ESG performance when examining the impact of board composition [44,45]. Hamdi et al. found that the higher the firm’s financial performance and cash holdings, the better its ESG performance [46]. Drempetic et al. found that the larger the company, the more resources it can call upon, and therefore it has a greater ability and willingness to engage in ESG practices, which leads to better ESG performance [47]. Based on legitimacy theory, Abdul Rahman et al. found that companies with better economic sustainability performance and higher financial performance and leverage had higher-quality ESG disclosures [48].

In response to the economic consequences of ESG performance, considerable research has found that improving ESG performance can contribute to financial performance growth and enhance firm value [14,15,49,50,51,52]. Profitability is one of the most critical indicators of a company and is related to shareholders’ equity, long-term liabilities, provisions, deferred income within one year, total liabilities, working capital, and current assets [53]. Baran et al. examined the relationship between ESG performance and financial performance in a sample of eight companies that are leaders in the energy sector in Poland and found no statistically significant relationship between ESG performance and return on equity (ROE), return on assets (ROA), return on sales (ROS), and current ratio (CR). However, there is a significant positive correlation between ESG performance and total asset turnover (TAT) and equity multiplier (A/E) [54]. Ma and Sun found that firms with high ESG performance can increase total factor productivity by alleviating financing constraints and enhancing innovation levels [16]. Velte P found that ESG performance is negatively related to accrual-based surplus management but does not affect actual surplus management in a sample of German firms. Furthermore, when the three dimensions of ESG performance are examined, governance performance has the most substantial negative impact on accrual surplus management compared to environmental and social performance [55]. Chang et al. and Zhang et al. found that ESG performance can improve the efficiency of finance and alleviate financial constraints [56,57]. Chen et al. concluded that ESG performance enhances customer relationship stability by reducing information asymmetry, enhancing product differentiation, and improving firm performance [58]. Xie and Li found that improving ESG performance reduces financial risk, which is not significant for firms with very high levels of financial risk [59]. Kim and Park showed that ESG performance reduces information asymmetry; moreover, assurance services can enhance the negative relationship between ESG performance and information asymmetry, acting as a positive moderator [60]. Lian et al. showed that high ESG performance could reduce bond credit spreads and corporate financial risk by improving corporate transparency and reducing debt agency costs [61]. He et al. found that firms with high ESG performance have better risk-taking capacity, and the contribution of ESG performance to corporate risk-taking capacity is more pronounced in firms with lower information transparency and weaker corporate governance [62]. Lee et al. concluded that firms with improved ESG performance could significantly reduce the risk of share price crashes [63]. Sciarelli et al. argue that integrating ESG criteria, gradually incorporating ESG criteria into corporate financial communications, and improving the accuracy and transparency of disclosure on relevant issues can help make investors’ investments more efficient and promote sustainable financial growth [64]. Weaver et al. point out that the publication of ESG tax transparency reports by companies can improve the information for investors and stakeholders, helping them to assess the value impact of an organization’s tax strategy from an economic, environmental, and sustainability perspective [65].

2.2. Research Hypothesis

Tax administration affects the cost of tax evasion for firms and has a comprehensive and profound impact on firms’ actual tax burden [66,67]. Golden Tax Project III utilizes big data, cloud computing, and other technical means to enhance the collection and processing of tax-related information. Improved tax administration systems strengthen anti-counterfeiting capabilities, significantly improving the timeliness, accuracy, and extensiveness of tax-related information, while increasing the intensity and efficiency of tax administration [5,12,68]. Therefore, Golden Tax Project III has compressed the space for firms to evade tax administration, inhibited their tax avoidance motives, and increased their tax burden [19,69]. Improving ESG performance requires continuous cost investment. Specifically, environmental (E) performance includes green products, environmental violations, etc. Improving environmental performance requires companies to invest more in R&D and green innovation to improve product technology and reduce environmental pollution. Social (S) performance includes health and safety, social contributions, and quality management, where additional financial investment is required for facilities or insurance purchased by the company to protect the health and safety of its employees, proper handling of after-sales and philanthropic acts such as charitable donations and community activities to enhance social contribution; governance (G) performance includes governance structures, operational risk, external discipline, etc. Actions such as more transparent disclosure, executive incentives, and companies’ assumption of shareholder responsibility to achieve better corporate governance also require some financial support. According to the theory of tax avoidance motivation and the tax effect of tax administration, Golden Tax Project III could raise firms’ cost burden and reduce free cash flow by strengthening tax administration. Therefore, firms’ willingness to invest in ESG activities will decline since they do not have enough financial resources for long-term sustainable development. On the basis of the above analysis, Hypotheses 1 and 2 are therefore proposed in this study.

Hypothesis 1 (H1).

Implementation of Golden Tax Project III will significantly reduce firms’ ESG performance, i.e., the strengthened tax administration will reduce ESG performance.

Hypothesis 2 (H2).

Implementation of Golden Tax Project III will reduce firms’ ESG performance by increasing their tax burden.

Tax incentives refer to the use of tax policies to reduce or exempt particular tax objects from the tax burden in accordance with the provisions of tax laws and administrative regulations. Compared with financial subsidies, tax incentives are more transparent and have a broader scope of application. Therefore, tax incentives can provide financial support for firms to improve their ESG performance, compensate them for the cost of fulfilling their ESG responsibilities, and mitigate the negative effects of enhanced tax administration on ESG performance [70,71]. On the basis of the above analysis, Hypothesis 3 is proposed in this study.

Hypothesis 3 (H3).

Tax incentives can mitigate the negative effect of Golden Tax Project III implementation on firms’ ESG performance.

3. Methodology

The primary research method of this paper is the empirical research method, via the collection, analysis, and testing of a sample of Chinese A-share listed companies, here employed to understand the basic situation of each data of the main variables in the sample and to clarify the relationship between the subjects of the study. Stata statistical software was used to carry out basic processing and analysis of the data, and the time-varying differences-in-differences method was used to empirically test the research hypotheses proposed in this paper and draw relevant research conclusions; finally, methods such as propensity score matching were used to conduct robustness tests.

3.1. Sample Selection and Data Sources

This study uses Chinese A-share listed companies from 2008 to 2020 as the sample. The exact time when each province started to run Golden Tax Project III was manually collected from the websites of provincial tax bureaus and relevant tax administration news. The ESG performance data use the rating results provided by Sino-Securities Index Information Service, which was obtained from the Wind database, and other financial data were obtained from the CSMAR database. To avoid interference from other factors with the available data, the sample was treated as follows: (1) companies with two consecutive years of losses, net assets below stock par value (ST), and three consecutive years of losses with delisting risk indication (*ST) were excluded; (2) companies in the financial and insurance sectors were excluded; (3) companies with too many missing values for the main variables were excluded; (4) all continuous variables were subjected to a 1% before-and-after tailing process. The final unbalanced panel data was obtained for a sample of 29,261 observations.

3.2. Variable Definition

3.2.1. Dependent Variable

In this paper, the Sino-Securities ESG rating was selected to measure the ESG performance of the dependent variable. Compared with other ESG rating systems, such as the SynTao Green Finance ESG rating and the Wind ESG rating, the Sino-Securities ESG rating is relatively closer to the actual situation of the Chinese market. Moreover, the Sino-Securities ESG rating covers all A-share listed companies, covering three primary indicators (environmental, social, and corporate governance), 14 secondary indicators (environmental management systems, green business objectives, green products, external environmental certifications, environmental violations, institutional systems, health and safety, social contributions, quality management, system-building, governance structure, operational activities, operational risks, and external sanctions), 26 tertiary indicators, and over 130 underlying data indicators, resulting in a nine-grade rating from low to high (denoted as C-AAA). The indicators selected by the Sino-Securities ESG rating combine the mainstream international ESG assessment framework, consider Chinese characteristics and specific practical experience, and incorporate the opinions of external market experts, which are reasonable and have been widely used by academics [58,72,73,74,75]. Therefore, this study draws on Fang and Hu and He and Zhuang [76,77], the C-AAA ratings are assigned as 1–9, and ESG performance is measured by taking the mean of each quarterly rating.

3.2.2. Independent Variable

The independent variable in this study is the Golden Tax Project III dummy variable (GTP), which is used as a proxy variable for enhanced tax administration. In 2013, Golden Tax Project III was launched in Chongqing, Shandong, and Shanxi Provinces, and in 2015 in Guangdong, Inner Mongolia, and Henan. After that, all provinces in China started to implement Golden Tax Project III one after another. The project was completed in all provinces in China by the end of 2016. If Golden Tax Project III is launched in the region where i is located, the GTP is assigned a value of 1; otherwise, it is 0.

3.2.3. Mediation Variable

This study refers to the study by Ji and Wang that uses income tax expense-to-gross operating income ratio as a measure of tax burden [4].

3.2.4. Other Variables

Referring to existing studies [36,37,41], this study selects firm size (Size), gearing ratio (Lev), CEO duality (Dual), profitability (Roa), firm growth (Growth), operating cash flow (Cashflow), board size (Board), board independence (Indep), equity concentration (Top1), and the nature of ownership (Soe) as control variables. In addition, this study also controls for the year-fixed effect (Year) and industry-fixed effect (Ind). The industry classification is subdivided using the three-digit codes in the latest 2012 edition of the Classification Guide for Listed Companies in China. This study measures tax incentives (Taxbenefit) as the ratio of various tax refunds to the firm’s operating income in the current year.

Table 1. Main variables definition table.

Variable Name	Symbols	Variable Definitions
ESG performance	ESG	Sino-Securities ESG rating results are assigned from low to high (1–9)
Golden Tax Project III	GTP	A dummy variable with a value of 1 in the year of implementation of Golden Tax Project III and subsequent years; in other years, it takes the value of 0
Firm size	Size	Total assets at the end of the period as the natural logarithm
Gearing ratio	Lev	Total liabilities at the end of the period/total assets at the end of the period
Profitability	Roa	Net profit/total assets at the end of the period
Operating cash flow	Cashflow	Net cash flow from operations/total assets at the end of the period
Growth	Growth	(Operating revenue for the period—operating revenue for the previous period)/operating revenue for the previous period
Board size	Board	The number of board members, taken as the natural logarithm
Board independence	Indep	Independent directors divided by the number of directors
CEO duality	Dual	1 if the chairman of the company and the general manager are the same person; otherwise, 0
Equity nature	SOE	1 if the firm is effectively controlled by the state; 0 otherwise
Equity concentration	Top1	Number of shares held by the largest shareholder/total number of shares
Tax burden	Tax	Income tax expense/gross operating income at the end of the period
Tax incentives	Taxbenefit	Various tax refunds/operating income of the firm at the end of the period

shows the definitions of the main variables.

3.3. Model Setting

The timing of the implementation of Golden Tax Project III varies widely across Chinese provinces, from Chongqing, Shandong (except Qingdao), and Shanxi in 2013 to Guangdong (except Shenzhen), Inner Mongolia and Henan in 2014, and so on. Golden Tax Project III provides a natural experimental scenario for this study to use the implementation of Golden Tax Project III as an exogenous shock to examine how enhanced tax administration affects firms’ ESG performance. Drawing on Jin and Huang and Wang et al. [18,35], this study constructs the following the time-varying differences-in-differences (time-varying DID) model (1) to test the effect of Golden Tax Project III on firms’ ESG performance, i.e., Hypothesis 1 (H1).

There are three main reasons for choosing the time-varying differences-in-differences model. Firstly, the differences-in-differences model is a widely used econometric method in policy analysis and engineering assessment, mainly applied to evaluate the degree of impact of an event or policy in mixed cross-sectional datasets. As Golden Tax Project III was implemented at different times in each province in China, it is more appropriate to choose the time-varying differences-in-differences model. Secondly, the model can largely avoid the endogeneity problem; policies are generally exogenous for microeconomic agents and thus do not suffer from reverse causality. Third, compared to the traditional approach of assessing policy effects under which a dummy variable for the occurrence or non-occurrence of a policy is set and then regressed, the time-varying differences-in-differences model is set up more scientifically to control for unobservable individual heterogeneity between samples and the effects of unobservable factors that change over time, and to estimate policy effects more accurately. The specific regression model is as follows.

$${ESG}_{i,t}={\alpha }_0+{\alpha }_1{GTP}_{i,t}+{\alpha }_2{Controls}_{i,t}+{\alpha }_3{Ind}_{i,t}+{\alpha }_4{Year}_{i,t}+{\epsilon }_{i,t}$$

(1)

${ESG}_{i,t}$ is a proxy variable for a firm’s ESG performance. ${GTP}_{i,t}$ is an independent variable that measures the enhanced tax administration and takes a value of 1 in the year when the firm’s location implemented Golden Tax Project III and in subsequent years; otherwise, it takes a value of 0. ${Controls}_{i,t}$ represents the set of all control variables. In addition, this study also controls for industry-fixed effect (Ind) and year-fixed effect (Year).

In order to further validate the path Golden Tax Project III takes in affecting firms’ ESG performance, this study constructs models (2)–(4) to test the mediating role of the tax burden, drawing on Ye et al. and Wen and Ye’s approach [17,78].

$${ESG}_{i,t}={\alpha }_0+{\alpha }_1{GTP}_{i,t}+{\alpha }_2{Controls}_{i,t}+{\alpha }_3{Ind}_{i,t}+{\alpha }_4{Year}_{i,t}+{\epsilon }_{i,t}$$

(2)

$${Tax}_{i,t}={\alpha }_0+{\alpha }_1{GTP}_{i,t}+{\alpha }_2{Controls}_{i,t}+{\alpha }_3{Ind}_{i,t}+{\alpha }_4{Year}_{i,t}+{\epsilon }_{i,t}$$

(3)

$${ESG}_{i,t}={\alpha }_0+{\alpha }_1{GTP}_{i,t}+{\alpha }_2{Tax}_{i,t}+{\alpha }_3{Controls}_{i,t}{+\alpha }_4{Ind}_{i,t}+{\alpha }_5{Year}_{i,t}+{\epsilon }_{i,t}$$

(4)

Drawing on the study by Lin et al. which used the triple difference (DDD) method to analyze the variability of study results under different conditions [79], this study constructs model (5) to test Hypothesis 3 (H3) based on model (1), focusing on the coefficients ${\alpha }_2$.

$${ESG}_{i,t}={\alpha }_0+{\alpha }_1{GTP}_{i,t}+{{\alpha }_2{GTP}_{i,t}\ast Taxbenefit+\alpha }_3{Controls}_{i,t}+{\alpha }_4{Ind}_{i,t}+{\alpha }_5{Year}_{i,t}+{\epsilon }_{i,t}$$

(5)

4. Empirical Analysis

4.1. Descriptive Statistics

Table 2. Results of descriptive statistics of the main variables.

Variable	N	Mean	sd	Min	p50	Max
GTP	29,261	0.473	0.500	0	1	1
ESG	29,261	6.514	1.093	4	6	9
Size	29,261	22.16	1.309	19.69	22.00	26.17
Lev	29,261	0.437	0.210	0.0500	0.434	0.904
Roa	29,261	0.0380	0.0640	−0.250	0.0370	0.211
Cashflow	29,261	0.0440	0.0720	−0.180	0.0440	0.245
Growth	29,261	0.170	0.447	−0.600	0.100	2.910
Board	29,261	2.144	0.200	1.609	2.197	2.708
Indep	29,261	0.374	0.0530	0.333	0.333	0.571
Dual	29,261	0.245	0.430	0	0	1
SOE	29,261	0.398	0.490	0	0	1
Top1	29,261	0.346	0.150	0.0860	0.324	0.744

shows the descriptive statistics of the main variables. The mean value of the independent variable ESG is 6.514, which indicates that the average rating of the sample companies is relatively high; however, the standard deviation is 1.093, and the difference between the maximum and minimum value is 5, which shows that the ESG performance of different sample companies varies greatly. The mean value of the dependent variable GTP is 0.473, indicating that 47.3% of the sample were affected by Golden Tax Project III. This suggests that the difference between the number of samples that received a policy shock and those that did not is insignificant. Hence, the sample size of the treatment and control groups is more balanced and better serves further research. The mean value of firm size (Size) is 22.16, and the median value is 22.00, indicating that the size of firms in the sample is more concentrated and less diverse, which, to a certain extent, reduces the impact of significant differences in firm size on ESG performance. The difference between the mean and median values of 0.437 and 0.434 indicates that the average gearing level of the sample companies is relatively normal; however, the difference between the maximum and minimum values reflects the fact that there are significant differences in debt-servicing levels between different firms. The difference between the maximum and minimum values of Growth shows a significant difference in the level of growth among the sample companies. The mean values of Roa (0.038) and Cashflow (0.044) indicate that the ability of the sample companies to generate earnings from their total assets is not satisfactory, and there is considerable heterogeneity between the firms. The mean value of the proportion of sole directors (Indep) is 0.374, which meets the regulatory policy requirements, but there is a significant disparity between different firms. The values related to the descriptive statistics of each variable are generally consistent with the existing research, so this study has some credibility in its research based on these data.

4.2. Correlation Analysis

In order to test whether there is multicollinearity between the main variables, this research uses Pearson’s correlation coefficient analysis method for verification. As can be seen in

Table 3. Correlation analysis of the main variables.

	ESG	GTP	Size	Lev	Roa	Cashflow	Growth
ESG	1
GTP	−0.043 ***	1
Size	0.362 ***	0.229 ***	1
Lev	0.087 ***	0.013 **	0.458 ***	1
Roa	0.151 ***	−0.110 ***	0.027 ***	−0.360 ***	1
Cashflow	0.073 ***	0.025 ***	0.070 ***	−0.133 ***	0.314 ***	1
Growth	−0.00800	−0.00700	−0.00300	0.00600	0.00100	−0.00100	1
Board	0.140 ***	−0.098 ***	0.238 ***	0.145 ***	0.038 ***	0.058 ***	0.00200
Indep	−0.00500	0.063 ***	0.031 ***	−0.00700	−0.029 ***	−0.027 ***	−0.00500
Dual	−0.103 ***	0.018 ***	−0.152 ***	−0.137 ***	0.025 ***	−0.027 ***	−0.00200
SOE	0.259 ***	−0.077 ***	0.318 ***	0.282 ***	−0.053 ***	0.033 ***	0.00600
Top1	0.137 ***	−0.122 ***	0.215 ***	0.061 ***	0.132 ***	0.076 ***	0.00600
	Board	Indep	Dual	SOE	Top1
Board	1
Indep	−0.501 ***	1
Dual	−0.177 ***	0.113 ***	1
SOE	0.253 ***	−0.049 ***	−0.283 ***	1
Top1	0.039 ***	0.038 ***	−0.061 ***	0.248 ***	1

Note: robust t-statistics are in brackets; ***, **, * represent significance at the levels of 1%, 5%, and 10%, respectively.

, the correlation coefficient values between the variables listed in the table are less than 0.7 in absolute value, indicating that there is no severe multicollinearity between the variables in the research model of this paper.

4.3. Baseline Regression Results

4.3.1. Parallel Trend Test

The differences-in-differences (DID) method is subject to a basic assumption—the parallel trend assumption, which requires that individuals in the treatment group have the same trend of change in the outcome variable as individuals in the control group before they receive an intervention for an event. Drawing on the study by Thorsten Beck et al. [80], model (6) is constructed to test for parallel trends. ${GTP}_{it}^{-n}$ denotes the state n years prior to the implementation of Golden Tax Project III, and ${GTP}_{it}^n$ denotes the state in the nth year after the implementation of Golden Tax Project III. If ${\alpha }_1$ and ${\alpha }_2$ are insignificant, it indicates that the control and treatment groups have approximately the same trend of change in ESG performance before implementing Golden Tax Project III, i.e., the parallel trend hypothesis is passed.

$$\begin{gathered} {ESG}_{it}={\alpha }_0+{\alpha }_1{GTP}_{it}^{-2}+{\alpha }_2{GTP}_{it}^{-1}+{\alpha }_3{GTP}_{it}^0+{\alpha }_4{GTP}_{it}^1+{\alpha }_5{GTP}_{it}^2+{\alpha }_6{GTP}_{it}^3+{\alpha }_7{treat}_i+{\alpha }_8{Controls}_{i,t} \\ +{\alpha }_9{Ind}_{i,t}+{\alpha }_{10}{Year}_{i,t}+{\epsilon }_{i,t} \end{gathered}$$

(6)

As shown in Figure 1, the difference in ESG performance between the sample firms in the treatment and control groups prior to the implementation of Golden Tax Project III fluctuates above and below 0 with confidence intervals of 0%. This indicates that there is no significant difference between ESG performance in the treatment and control groups before the implementation of Golden Tax Project III, i.e., the ex ante parallel trend hypothesis is passed, and the estimates of the DID design are held to identify causality.

4.3.2. Analysis of Basic Results

Table 4. Baseline results.

	(1)	(2)
	ESG	ESG
GTP	−0.0185 **	−0.1407 ***
	(−2.0552)	(−12.5260)
Size		0.2440 ***
		(28.6753)
Lev		−0.5551 ***
		(−12.7216)
ROA		1.1475 ***
		(11.5137)
Cashflow		−0.1731 **
		(−2.2142)
Growth		−0.0625 ***
		(−5.7953)
Board		0.0166
		(0.3581)
Indep		−0.1974
		(−1.3434)
Dual		−0.0254 *
		(−1.6864)
SOE		0.2305 ***
		(9.6980)
Top1		0.4521 ***
		(7.3384)
_cons	6.4931 ***	1.1816 ***
	(10.7513)	(5.6274)
N	29,261	29,261
Ind	Yes	Yes
Year	Yes	Yes
r2_a	0.1951	0.2833

Note: robust t-statistics are in brackets; ***, **, * represent significance at the levels of 1%, 5%, and 10%, respectively.

illustrates the results of the baseline regression based on model (1). Column (1) shows the results of the empirical test without control variables, while column (2) shows the impact of Golden Tax Project III on ESG performance, including control variables, the time-fixed effect, and the industry-fixed effect. As can be seen from (1) of Table 3, the result without control variables has an adjusted R2 of 0.1951, and the whole model is significant at the 5% level, indicating that the underlying model (1) fits well, is feasible to construct, and the test results are somewhat reliable. The main regression coefficient in column (2) is −0.1407, indicating that Golden Tax Project III (GTP) and the ESG performance of enterprises (ESG) are negatively correlated at the 1% level of significance. The results suggest that the enhanced tax administration after Golden Tax Project III inhibits the enhancement of firms’ ESG performance. Based on the above empirical results, H1 is verified.

4.3.3. Analysis of Mediating Effect

Column (1) of

Table 5. Results of mediating and moderating effects tests.

	(1)	(2)	(3)	(4)
	ESG	Tax	ESG	ESG
GTP	−0.1407 ***	0.0021 *	−0.0886 ***	−0.1226 ***
	(−12.5260)	(0.8666)	(−7.3342)	(−9.8852)
GTP *Taxbenefit				−0.0183
				(−2.0641)
Size	0.2440 ***	0.0013 ***	0.1739 ***	0.2441 ***
	−28.6753	(6.0957)	(16.3199)	(28.6964)
Lev	−0.5551 ***	−0.0116 ***	−0.5917 ***	−0.5495 ***
	(−12.7216)	(−11.5709)	(−12.2270)	(−12.5867)
ROA	1.1475 ***	0.1066 ***	1.0439 ***	1.1389 ***
	−11.5137	(49.3163)	(9.7817)	(11.4257)
Cashflow	−0.1731 **	0.0028 *	−0.2044 **	−0.1711 **
	(−2.2142)	(1.6644)	(−2.5653)	(−2.1891)
Growth	−0.0625 ***	−0.0008 ***	−0.0490 ***	−0.0626 ***
	(−5.7953)	(−3.5106)	(−4.5198)	(−5.8046)
Board	0.0166	0.0001	−0.0719	0.0150
	−0.3581	(0.1366)	(−1.3929)	(0.3244)
Indep	−0.1974	0.0064 *	−0.4341 ***	−0.1946
	(−1.3434)	(1.9527)	(−2.7512)	(−1.3245)
Dual	−0.0254 *	−0.0001	−0.0145	−0.0259 *
	(−1.6864)	(−0.3311)	(−0.9093)	(−1.7199)
SOE	0.2305 ***	−0.0038 ***	−0.0032	0.2289 ***
	−9.698	(−5.5245)	(−0.0935)	(9.6326)
Top1	0.4521 ***	0.0098 ***	0.5150 ***	0.4411 ***
	−7.3384	(6.3168)	(6.7853)	(7.1527)
Tax			−0.9610 ***
			(−3.1960)
_cons	1.1816 ***	−0.0118 **	3.1121 ***	1.1843 ***
	−5.6274	(−2.2377)	(11.6539)	(5.6418)
N	29,261	29,261	29,261	29,261
Ind	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
r2_a	0.2833	0.1487	0.2137	0.2873

Note: robust t-statistics are in brackets; ***, **, * represent significance at the levels of 1%, 5%, and 10%, respectively.

suggests that there is a significant negative correlation between Golden Tax Project III (GTP) and firms’ ESG performance (ESG) at the 1% level. From column (2), it can be seen that there is a positive correlation between Golden Tax Project III (GTP) and the tax burden (Tax) at the 10% level. This shows that the implementation of the Golden Tax Phase III project has enhanced the strength and efficiency of tax collection and administration, curbing the scope for tax avoidance by enterprises, which in turn increases their tax burden. Column (3) shows that the correlation coefficient between corporate tax burden (Tax) and ESG performance (ESG) is −0.9610 and is significant at the 1% level. This is because increased corporate tax burdens lead to reduced free cash flow, which reduces firms’ ESG performance by inhibiting the willingness of companies to make costly investments to improve their ESG performance. Hence, the tax burden (Tax) partially mediates the relationship between Golden Tax Project III (GTP) and firms’ ESG performance (ESG), and H2 is tested.

4.3.4. Analysis of Moderation Effect

As shown in column (4) of Table 5, the correlation coefficient of the impact of Golden Tax Project III (GTP) on corporate ESG performance (ESG) is −0.1226 at the 1% level; after the inclusion of the tax intensives interaction, the correlation coefficient of GTP*Taxbenefit on firms’ ESG performance is −0.0183 and insignificant. This indicates that tax incentives can mitigate the inhibitory effect of Golden Tax Project III on firms’ ESG performance, and H3 is verified. This result may have emerged since tax incentives alleviate cost pressures caused by enhanced tax administration efforts, provide financial resources for corporate ESG practices, and incentivize firms to improve their ESG performance.

4.4. Further Analysis

4.4.1. Impact of Golden Tax Project III on Individual ESG Pillars

As the Sino-Securities ESG rating only has an overall score and lacks scores for each dimension of firms’ ESG performance, this study uses the Bloomberg ESG rating system’s results to examine the impact of Golden Tax Project III on individual ESG pillars, including E (environment), S (society), and G (governance).

Table 6. The effects of Golden Tax Project III on individual ESG pillars.

	(1)	(2)	(3)
	E	S	G
GTP	−2.2627 ***	−2.1267 ***	−1.3629 ***
	(−18.7410)	(−14.2857)	(−18.8268)
Size	2.4732 ***	3.3337 ***	0.9260 ***
	(23.6095)	(25.5963)	(14.5763)
Lev	−2.5575 ***	−3.9806 ***	−2.7788 ***
	(−4.5383)	(−5.7123)	(−8.1922)
ROA	−1.0522	−2.3726	0.0615
	(−0.8686)	(−1.5912)	(0.0849)
Cashflow	2.9968 ***	2.9751 ***	−0.4840
	(3.2799)	(2.6459)	(−0.8858)
Growth	−0.3860 ***	−0.8368 ***	−0.0180
	(−2.9932)	(−5.2778)	(−0.2340)
Board	−0.2333	−1.1474 *	0.3181
	(−0.4616)	(−1.8361)	(1.0457)
Indep	2.2893	1.1044	0.8856
	(1.4528)	(0.5678)	(0.9361)
Dual	−0.2058	−0.4782 **	−0.1433
	(−1.1497)	(−2.1676)	(−1.3359)
SOE	0.8173 ***	0.7170 **	0.5734 ***
	(2.9334)	(2.0584)	(3.3677)
Top1	0.2054	−2.2326 **	2.2652 ***
	(0.2832)	(−2.4757)	(5.1495)
_cons	−48.9899 ***	−50.8762 ***	21.7816 ***
	(−19.0752)	(−15.9170)	(13.9656)
N	29,261	29,261	29,261
Ind	Yes	Yes	Yes
Year	Yes	Yes	Yes
r2_a	0.1790	0.1752	0.1402

Note: robust t-statistics are in brackets; ***, **, * represent significance at the levels of 1%, 5%, and 10%, respectively.

shows the coefficients of Golden Tax Project III (GTP) are all significantly negative, and while the coefficient between GTP and E (environment) is most significant, the coefficient between GTP and G (governance) is minor. This indicates that Golden Tax Project III has the most significant impact on the environmental score and the least significant impact on the governance score of firms’ ESG performance. In terms of this result specifically, environmental performance includes aspects such as green products and environmental pollution. In that case, companies need to invest more in research for green innovation or related upgrading of equipment and improve their production technology to produce green products and reduce pollutant emissions. Social performance includes aspects such as philanthropic activities and charitable giving, which are non-routine operating activities that require additional expenses to be paid by companies. Governance performance includes aspects such as board structure, risk control, etc. Although improving governance performance also requires financial support, corporate governance has weaker externalities than the environmental and social dimensions. There is a contradiction between resource allocation and externalities in both the enhancement of environmental performance and enterprises’ fulfillment of social responsibility. In reality, the externalities of environmental and social responsibility undertaken by enterprises cannot be fully internalized, resulting in a contradiction between the goal of enhancing environmental and social performance and maximizing corporate profits. Thus, when tax administration is more robust, the tax burden on firms is heavier and free cash flow is reduced, there is a greater disincentive to perform in the environmental and social dimensions, which require more financial support, and a relatively smaller disincentive to perform on the governance dimension.

4.4.2. Heterogeneity Test

(1)

Impact of Golden Tax Project III on ESG Performance of Enterprises under Different Ownership Nature

The impact of Golden Tax Project III on firms’ ESG performance when the nature of the property differs is examined via grouping the ownership types of firms into state-owned and non-state-owned. As shown in columns (1) and (2) of

Table 7. Results of the heterogeneity test.

	SOEs	Non-SOEs	Low Marketability Level	High Level of Marketability
	(1)	(2)	(3)	(4)
	ESG	ESG	ESG	ESG
GTP	−0.0255	−0.2709 ***	−0.1329 ***	−0.1624 ***
	(−1.4978)	(−18.3688)	(−7.7395)	(−11.0040)
_cons	−0.1209	1.8271 ***	0.6688 **	1.2062 ***
	(−0.3573)	(6.6517)	(2.1686)	(4.4030)
Controls	Yes	Yes	Yes	Yes
Ind	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
N	9629	19,631	10,391	18,870
r2_a	0.1983	0.182	0.3146	0.2729

Note: robust t-statistics are in brackets; ***, **, * represent significance at the levels of 1%, 5%, and 10%, respectively.

, the correlation coefficient between Golden Tax Project III (GTP) and firms’ ESG performance (ESG) is not significantly negative at −0.0255 in state-owned firms (SOEs), while in non-state-owned firms (non-SOEs), the GTP correlation coefficient is −0.2709 and significant at the 1% level, which is about ten times higher than that in state-owned firms. This suggests that enhanced tax administration has a more substantial dampening effect on ESG performance in non-state-owned firms (non-SOEs) than in state-owned firms (SOEs). The possible reasons for this are that, in terms of tax compliance, state-owned firms (SOEs) do not have a strong incentive to evade tax due to the specificity of their business objectives; in terms of the implementation of tax incentives, state-owned firms (SOEs) are more aware of tax incentives and, due to their closer relationship with the government, it is also easier for state-owned firms (SOEs) to obtain tax incentives to alleviate the financial pressure caused by tax administration. Hence, enhanced tax administration has less effect on the ESG performance of state-owned firms (SOEs).

(2)

Impact of Golden Tax Project III on Corporate ESG Performance under Different Levels of Marketization

After the China Marketization Index published by Wang et al. was used to measure the levels of local marketization [81], the sample was then divided into a high marketization degree sample group and a low marketization degree sample group for group testing based on the median marketization index. From columns (3) and (4), it can be seen that Golden Tax Project III has a significant inhibitory effect on the ESG performance of enterprises in regions with different marketization levels, with a more substantial inhibitory effect on the ESG performance of enterprises in regions with high marketization levels.

5. Robustness Test

5.1. PSM-DID

The propensity score-matching method (PSM) is used to alleviate the problem of unreliable test results caused by sample selection bias. This method was first proposed in 1983 by Paul R. Rosenbaum and Donald B. Rubin and was used to eliminate as much as possible the influence of confounding factors on experimental results [82]. A single effect of the test variable on the outcome variable is achieved by matching samples in the control group with characteristics similar to the treatment group and controlling for the effects of other factors on the outcome variable. The specific steps include: (1) selecting matching variables to construct a logit model to calculate the propensity score, which is recorded as the PS value (implemented through Stata); (2) selecting matching methods to match the samples; (3) testing the premise hypotheses, i.e., the balanced hypothesis and the common support hypothesis. The conditions that need to be met to test the balanced hypothesis are that the standardized deviation of each matched variable in the control and treatment group samples after matching fluctuates around 0%, i.e., there is no significant difference between the means of the variables in the two groups of samples. Pre-matching and post-matching samples are generally used in empirical tests.

According to the results of the balanced hypothesis test in Figure 2, the matching variables of the treatment and control groups were significantly different before the matching of the samples, and the resulting individual differences may have affected the test results; however, after the propensity score matching, the standard errors of the matching variables fluctuated slightly around 0%, indicating that the sample characteristics before and after the matching were basically the same, i.e., the “balanced hypothesis” test was passed. According to the results of the common support hypothesis test in Figure 3, the kernel density curves of the PS values of the treatment and control groups after matching have a high degree of overlap, indicating that the characteristics of the treatment and control groups are very close to each other and the matching effect is good, which satisfies the common support hypothesis, meets the premise of the propensity score matching method, and can solve the problem of selectivity bias.

Matching was performed using the 1:1 non-relaxed nearest-neighbor matching method, and regression analysis using the matched samples showed consistent results. After controlling for sample selectivity bias, there is still a significant suppressive effect of Golden Tax Project III on firms’ ESG performance, indicating that the test results have some robustness. See the regression results in column (1) of

Table 8. Robustness tests of the impact of the Golden Tax III Project on corporate ESG performance.

	(1)	(2)	(3)	(4)
	ESG_psm	ESG_Bloomberg	ESG_SynTao	Dynamic Effects
GTP	−0.1534 ***	−1.9576 ***	−0.3249 ***
	(−13.1438)	(−20.0959)	(−6.2344)
${GTP}_{it}^{-2}$				−0.0188
				(−0.5965)
${GTP}_{it}^{-1}$				−0.0057
				(−0.2238)
${GTP}_{it}^0$				−0.0342
				(−1.3766)
${GTP}_{it}^1$				−0.1012 ***
				(−3.3698)
${GTP}_{it}^2$				−0.1238 ***
				(−3.4902)
${GTP}_{it}^3$				−0.1650 ***
				(−3.8020)
Size	0.2395 ***	2.3266 ***	0.2392 ***	0.2469 ***
	(27.9721)	(23.8167)	(8.1445)	(27.2681)
Lev	−0.6151 ***	−2.8171 ***	−0.7667 ***	−0.5332 ***
	(−14.0394)	(−5.9924)	(−4.5590)	(−12.2551)
ROA	0.8973 ***	−0.7739	−0.5591 *	1.1433 ***
	(10.5936)	(−0.8210)	(−1.6808)	(11.4506)
Cashflow	−0.1672 **	1.8289 ***	0.5603 **	−0.0788
	(−2.2860)	(2.5799)	(1.9809)	(−1.0090)
Growth	−0.0007 *	−0.3575 ***	−0.1453 ***	−0.0600 ***
	(−1.8983)	(−3.6159)	(−3.3219)	(−5.5465)
Board	0.0413	−0.8026 *	0.0317	0.0100
	(0.8629)	(−1.9051)	(0.2637)	(0.2147)
Indep	−0.1288	0.9271	0.6542 *	−0.1914
	(−0.8893)	(0.7261)	(1.6530)	(−1.3151)
Dual	−0.0239	−0.2224	−0.1714 ***	−0.0322 **
	(−1.5180)	(−1.5648)	(−3.5362)	(−2.1591)
SOE	0.2438 ***	0.1867	0.1302 **	0.2524 ***
	(9.9233)	(0.6306)	(2.2684)	(10.6571)
Top1	0.4432 ***	0.0310	−0.2036	0.3651 ***
	(7.0147)	(0.0459)	(−1.2096)	(5.9005)
_cons	1.2221 ***	−31.5968 ***	−0.8322	0.8782 ***
	(5.7769)	(−12.8776)	(−1.1944)	(4.0629)
N	29,261	29,261	29,261	29,261
Ind	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
r2_a	0.2773	0.1285	0.1402	0.2861

Note: robust t-statistics are in brackets; ***, **, * represent significance at the levels of 1%, 5%, and 10%, respectively.

for details.

5.2. Alternative Proxy for ESG Performance

An analysis of the ESG rating results of different rating agencies exists. To avoid misjudgment of ESG performance due to the quality of ESG disclosure, this paper draws on relevant studies [76] and further re-estimates model (1) using Bloomberg ESG scores and SynTao Green Finance ESG rating data as key explained variables. As shown in columns (2) and (3), after the explained variables were replaced, firms’ ESG performance remained significantly lower after the implementation of Golden Tax Project III, and the findings remain unchanged.

5.3. Dynamic Effects Test

It has been argued that time-varying DID may have some bias if it relies on a single estimator alone. Huang et al. argue that dynamic effects can improve the test results’ robustness; the authors of this paper have therefore drawn on their study to construct the following model (7) for dynamic effects testing [83]. The year of implementation of Golden Tax Project III is used as the base period while the treatment variables two years before and three years after the purchase are used as explanatory variables; subsequently, the main focus is on ${\alpha }_3$~${\alpha }_6$. The test results are shown in column (4) of Table 8. As can be seen from the results in the table, there is no significant effect in the year of implementation of Golden Tax Project III and a significant dampening effect from the first year onwards, which is consistent with the results of the single estimator, indicating that the empirical results of this paper are somewhat robust.

$$\begin{gathered} {ESG}_{it}={\alpha }_0+{\alpha }_1{GTP}_{it}^{-2}+{\alpha }_2{GTP}_{it}^{-1}+{\alpha }_3{GTP}_{it}^0+{\alpha }_4{GTP}_{it}^1+{\alpha }_5{GTP}_{it}^2+{\alpha }_6{GTP}_{it}^3+{\alpha }_7{Controls}_{i,t}+{\alpha }_8{Ind}_{i,t} \\ +{\alpha }_9{Year}_{i,t}+{\epsilon }_{i,t} \end{gathered}$$

(7)

5.4. Placebo Test

To test whether some chance factors drive the baseline empirical results, this paper referred to a study by Ferrara et al. and conducted a placebo test [72]. The treatment and control groups were randomly simulated, Golden Tax Project III was randomly assigned to the sample companies and regressed based on model (1), and the process was repeated 500 times. Figure 4 shows that the proportion of results with significantly negative coefficients is relatively small, indicating that the dummy treatment effect constructed in this paper does not exist and that the change in ESG performance is indeed caused by the implementation of Golden Tax Project III and not by other factors.

6. Discussion

Tax administration, as an external constraint mechanism, often impacts enterprises’ business activities. At the same time, with the emergence of issues such as climate and labor rights, there is a global consensus for sustainable corporate development. ESG performance encompasses corporate efforts in the three areas of environmental, social responsibility, and corporate governance, in line with the concept and requirements of sustainable development. A firm with good ESG performance means that it places more emphasis on sustainable development, has a more scientific governance structure, and is more aware of its social responsibility, and firms’ ESG performance is increasingly attracting the attention of the capital market. Therefore, it is essential to study the impact of tax administration on firms’ ESG performance to build a fiscal and tax policy system and a favorable external environment conducive to enhancing firms’ ESG performance.

The existing literature on ESG performance focuses on the economic consequences of ESG performance. Studies have found that improving ESG performance can enhance financial performance and firms’ value [49,50,51,52], reduce accrued surplus management [55], alleviate financing constraints [56,57], improve customer stability [58], reduce corporate risk [62], etc. Fewer studies have focused on the factors influencing ESG performance. Scholars have now found that financial technology [37], board structure [39], investor focus [41], and firm size can have an impact on corporate ESG performance [47]. However, little literature has examined the relationship between taxation and ESG performance. Abdelfattah and Aboud found a positive relationship between corporate tax avoidance and corporate ESG performance [38]. Jin and Huang pointed out that the increased intensity of tax administration limits corporate tax avoidance and breaks the implicit contract between corporations and governments, thus reducing the fulfillment of corporate social responsibility [18]. However, there is no literature examining the impact of tax administration on firms’ ESG performance at the micro level.

Our research differs from their analysis in several respects. There may be a reciprocal causal relationship between corporate tax avoidance and ESG performance. Firstly, while the existing literature only examines the impact of tax administration on a single aspect of environmental or social responsibility, this paper examines the impact of stronger tax administration on ESG performance, a comprehensive sustainability indicator, and examines the differences caused by tax administration on the three dimensions of E, S, and G. Secondly, this paper uses the implementation of Golden Tax Project III as a quasi-natural experiment to avoid this endogeneity problem and enable a more accurate assessment of the impact of tax administration on firms’ ESG performance to be carried out. In addition, this study also explores the mechanisms by which tax administration affects ESG performance and the moderating effect of tax incentives on the relationship between tax administration and ESG performance.

At the theoretical level, this research extends the study of the factors influencing ESG performance and the “taxation effect” of tax administration. At the practical level, the conclusions of this paper, on the one hand, help government departments to gain a deeper understanding of the relationship between tax policy and business operations and provide some reference for how to optimize tax collection and administration. On the other hand, the findings of this paper can help government departments to formulate targeted tax incentives to alleviate the financial difficulties of enterprises in order to motivate them to actively engage in ESG practices, improve their ESG performance, and contribute to sustainable development.

Regarding the limitations and future directions of this research, firstly, due to the limitations of the data, this paper only selected a sample of Chinese A-share listed companies and did not examine a sample of non-listed companies; future research could expand the sample to include all companies as well as overseas companies. Secondly, this paper uses tax burden as a mediating variable to study the mechanism of the effect of tax administration on ESG performance. However, other variables may exist, and future research can further refine and complement the path of the effect of tax administration on ESG performance. Thirdly, due to the availability of data, this paper has only tested the heterogeneity of firms with different natures of ownership and at different levels of marketization, and future research can be extended to different industries, etc. Fourthly, the findings of this paper are based on the context of China’s economic and social system and may change depending on the country or fiscal system. Future research could compare the variability of the impact of tax administration on ESG performance in different countries or fiscal systems to expand the conclusions.

7. Conclusions

Based on the data of A-share non-financial listed companies in China from 2008 to 2020, this study examines the impact of Golden Tax Project III on firms’ ESG performance and analyzes the mechanism of its effect through theoretical analysis and empirical design; focusing on the context of China’s transition economy, it analyzes the variability of this impact in different environments; finally, it uses propensity score matching, dynamic effects, placebo test and replacement of explained variables to conduct robustness tests and obtains the following conclusions: Firstly, the firms’ ESG performance was significantly reduced after the implementation of Golden Tax Project III, i.e., the strengthening of tax administration suppresses firms’ ESG performance; the mediating effect study suggests that Golden Tax Project III reduces firms’ ESG performance by increasing their tax burden; the moderating effect study finds that tax incentives can significantly mitigate the negative impact of strengthening tax administration on the firms’ ESG performance. Secondly, further research found that Golden Tax Project III had the most significant impact on the environment (E) score and the least significant impact on the social (S) score in firms’ ESG performance. Thirdly, the heterogeneity test found that the dampening effect of Golden Tax Project III on firms’ ESG performance was relatively more significant among state-owned enterprises and enterprises in regions with a higher level of marketization.

The findings of this paper confirm that firms’ ESG performance is sensitive to their tax burden as a result of increased tax administration. After the implementation of Golden Tax Project III and tax collection and administration increased, enterprises’ tax burden and their willingness to engage in ESG practices decreased, and ESG performance has declined. Therefore, in the context of downward pressure on China’s economy and sustainable development, the government should implement tax reduction policies and give appropriate tax concessions to enterprises with good ESG performance; relevant government departments can promote and motivate enterprises with good ESG performance through the news media to improve their social recognition and reputation so that they can gain more benefits from ESG performance; the government should provide targeted tax incentives for non-state enterprises that do not have credit advantages and are more affected by the enhanced tax administration, so as to enhance the motivational effects of tax incentives on the ESG performance of these enterprises.