Can China’s leading officials’ accountability audit of natural resources policy promote corporate emissions reduction?

Abstract

The Chinese government proposed the policy of the leading officials’ accountability audit of natural resources (AANR) in 2013 and launched pilot audits in some regions in 2014. In order to explore whether AANR reduces corporate environmental pollution, this paper validates the impact of AANR on corporate emission reduction using a multi-period difference-in-difference (DID) method based on a sample of Chinese A-share heavy-polluting corporations from 2010 to 2017. The results indicated that AANR can significantly reduce corporate pollutant emissions, with legitimacy pressures and resource dependence playing a mediating role. Further research found that the AANR promoted emission reduction more strongly for large-sized corporations and political connection corporations, and more significantly for corporations with higher internal and external governance levels. Policy effects were more significant in regions with favorable economic growth and legalization. The above findings verify the effect of government audits on micro corporation pollution management, enrich the theoretical system of AANR research, and provide meaningful empirical evidence for AANR's further improvement.

Keywords:

• Leading official’s accountability audit of natural resources
• micro corporations
• emissions reduction
• government audits

Introduction

With the continuous development of ecological civilization, ecological problems such as environmental pollution and resource depletion brought about by economic development have received increasing attention [1], becoming a significant challenge in China’s new development stage. Since the 18th National Congress of the Communist Party of China(CPC), the coordinated development of the economy and environmental protection has risen to the level of national strategy. As the main body of micro-economic activities, corporations’ production and operation activities are important sources of environmental pollution [2]. The macro-level green development strategy ultimately needs to be implemented at the level of micro corporations. Due to the externality of environmental pollution, it is difficult for corporations to reduce emissions without adequate external supervision. As a fundamental part of China’s supervision and governance system, how government audit plays an immune function at the micro level and promotes corporate pollution governance is essential for audit development in the new era.

Under the three-tier pressure transmission structure of central-local-corporation, the local government’s attention to environmental governance can effectively motivate corporations to treat pollution [3]. However, the traditional assessment mechanism of local officials in China is precisely a triggering factor for resource and environmental problems. For a long time, government assessment followed the "promotion tournament" model with GDP as the core indicator [4]. Distorted incentives induce short-sighted behavior of local governments, which often leads to "bottom-up competition" in implementing relevant environmental regulations [5]. Some local governments even tolerate excessive corporate emissions to generate tax revenue, leading to failure in environmental governance [6]. The central government has gradually implemented a series of environmental assessment policies since 2005, driving environmental performance to become an assessment indicator for local officials’ promotion [7]. However, these policies failed to change the preference of local officials for economic development [8] and even led some local officials to whitewash their performance by modifying environmental quality data [9]. To this point, the local officials’ assessment system still has severe institutional flaws.

In 2013, the Third Plenary Session of the 18th CPC Central Committee proposed to "explore the compilation of natural resources assets and liabilities balance sheet, the implementation of off-the-job audits of natural resources assets for leading cadres, and the establishment of a lifelong accountability system for ecological and environmental damage." The Third Plenary Session of the 18th Central Committee of the CPC adopted "The Decision of the Central Committee of the Communist Party of China on Several Major Issues Concerning Comprehensively Deepening Reforms," which made a precise deployment of the leading officials’ accountability audit of natural resources (AANR). Following the requirements of the above policy document, pilot audits have gradually begun to be conducted by cities in various regions of the country from 2014 to 2017. Until 2018, the pilot work was comprehensively implemented and became a regular work for auditing institutions. The AANR has filled the supervision gaps of ecological civilization construction. It is a crucial action to refine officials’ evaluation system and environmental accountability system, aiming to achieve coordinated development of the economy and the environment.

Since its implementation, the AANR has received wide attention from academics. The existing literature has demonstrated that AANR can significantly improve environmental performance in pilot regions, including upgrading environmental quality [10], reducing pollutant emission intensity [11], and treating air and water pollution [12–14]. Recently, research has gradually begun to focus on the impact of AANR on corporate environmental performance. The research shows that the implementation of the AANR has increased corporate environmental investment [15,16] and quality of environmental disclosure [17], reduced corporate environmental violation [18], and promoted corporate green innovation [19–21]. The above studies verified that AANR has induced positive corporate responses in the environmental field, providing valuable references for AANR policy evaluation and laying an essential theoretical foundation for this research. However, existing research focuses on changes in corporate behavior, and there is a lack of validation on whether these changes reduce the negative environmental externalities generated by corporate production and operations. Few studies have explored the impact of AANR on corporate pollutant emissions, although research points to pollutant emissions as an ideal indicator for evaluating the impact of AANR on corporations [17]. At the same time, existing research also lacks a systematic analysis and validation on the path of AANR's impact on corporate environmental performance. Therefore, this paper relies on a quasi-natural experimental scenario formed by the gradual piloting of AANR from 2010-2017 to conduct the study. A multi-period DID model is used to explore the impact of AANR on corporate pollution emission and verify the impact path based on legitimacy theory and resource dependence theory.

Compared with the existing literature, the marginal contributions of this paper are as follows. First, AANR is an audit model innovation based on the unique political background in China, and there is an urgent need for a systematic evaluation of its policy effects. This paper verifies the impact of AANR on micro corporations from the perspective of emission reduction, which provides valuable empirical evidence for the micro governance role of national auditing. Second, based on existing research, this paper systematically analyzes and validates the path of AANR affecting corporate environmental performance, enriching the theoretical system in this field. Third, this paper explores the individual heterogeneity and environmental heterogeneity of the impact that AANR has on corporate emission reduction behavior from four aspects: corporate characteristics, internal governance mechanism, external environment, and external governance mechanism, which provides a meaningful theoretical basis for further improvement and differentiated implementation of AANR. Finally, this paper relies on the quasi-natural experimental scenario of AANR and uses a multi-period DID model for testing. This approach avoids the endogeneity problem of using cross-sectional data for empirical research to some degree, thus increasing the reliability of the findings.

The remaining part of the paper proceeds as follows: "Literature review" reviews the relevant literature of previous studies; "Theoretical analysis and hypotheses" discusses the theoretical mechanisms and proposes research hypotheses; "Research design" outlines the data collection, variable selection, and empirical model; "Empirical analysis and testing" analyzes the empirical results and tests the robustness; "Further analysis" provides heterogeneity analysis; "Conclusion and implications" summarizes the research findings and proposes relevant recommendations.

Policy background and literature review

Policy background

In 2013, "The Decision of the Central Committee of the Communist Party of China on Several Major Issues Concerning Comprehensively Deepening Reform," which was considered and adopted at the Third Plenary Session of the 18th Central Committee of the CPC, explicitly proposes to conduct the AANR. In 2015, under the guidance and concern of the Chinese president, "The Overall Plan for the Reform of the Ecological Civilization System" issued by the Central Committee of the CPC and the State Council included AANR in the ecological civilization performance evaluation and accountability system. AANR is an audit conducted by the auditing organization on the fulfillment of the responsibility for the natural resource asset management and eco-environmental protection of key leading cadres during their tenure of office. Responsibility for natural resource asset management and eco-environment protection consists of the following items: (1) Management, development, and utilization of natural resources such as land, water, forests, grasslands, minerals, and oceans. (2) Environmental protection and improvement of the atmosphere, water, soil, etc. (3) Conservation and restoration of ecosystems such as forests, grasslands, deserts, rivers, lakes, wetlands, oceans, etc. (4) Other items related to natural resource asset management and eco-environmental protection.

The distribution of the AANR pilot regions each year is shown in Figure 1. The annual AANR pilot information for each region was collected manually, mainly through the China Audit Yearbook, which was checked and corrected by comparing the announcements on the websites of the regional audit offices (bureaus) and news reported by the media. As shown in Figure 1, 11 cities from eight provinces and all cities in Guangxi Province conducted AANR pilots in 2014, most of which were concentrated in the southern region. In 2015, the audit pilot began to spread to the eastern and northern regions. The number of pilot cities surged in 2016-2017, with pilots appearing in the west and northeast, covering essentially every province across China.

Figure 1. Distribution of AANR pilot regions in 2014–2017.

AANR and regional environmental governance

As an innovative environmental auditing policy in China, existing research has examined the impact of AANR on regional environmental governance to assess whether its policy effects have reached expectations. Wu et al. [10] found that implementing the AANR pilot has enhanced the government’s environmental regulation and significantly improved regional environmental quality. Cao et al. [11] verified that the pollutant emission intensity in the AANR pilot regions was significantly reduced and also demonstrated that this change was not due to an increase in environmental fiscal expenditures but rather to an improvement in the efficiency of environmental inputs and outputs. Zhang et al. [12] indicated that AANR could improve air quality in the pilot regions over a long period, with government environmental regulation playing a mediating role. Feng et al. [13] obtained similar research findings but found that AANR also reduced air pollution in neighboring non-pilot regions, demonstrating spatial spillover effects. Ma et al. [14] verified that AANR can also effectively reduce water pollution, and this effect is becoming more noticeable over time. Liu et al. [22] used Chinese provincial data to explore the relationship between AANR, agricultural technological progress, and agricultural carbon emission intensity, proving that AANR has decreased agricultural carbon emissions by promoting technological progress, with the role of mechanical progress playing a more significant role than that of proprietary progress.

AANR and corporate environmental performance

Several researchers attempted to evaluate the effects that AANR has caused on micro corporations. Most research agrees that AANR can facilitate the fulfillment of corporate environmental responsibility, including enhancing corporate environmental investment and environmental performance [15]. Before the implementation of AANR, the complicity of local governments with highly polluting corporations for the sake of economic development targets was a significant impediment to the building of an ecological civilization. After implementing AANR, local governments reduced their support for highly polluting corporations by reducing their policy subsidies and tax incentives [23]. Wei and Du [18] explored that the accountability pressure from AANR has compelled officials to strengthen environmental enforcement, effectively deterring corporate environmental violations in their jurisdictions. Wang [16] found that the mobilizing behavior of the government will effectively enhance the role of AANR in promoting corporate environmental investment. Huang [17] demonstrated that AANR induced corporate timely responses. Specifically, the quality of corporate environmental disclosure in the pilot regions has improved significantly compared to non-pilot regions. The contribution of AANR to corporate green innovation has been proved in several studies. Jiang et al. [19] verified that the promotional effect of AANR on corporate green innovation is realized through the enhancement of government environmental governance and government innovation-driven strategy. Liu et al. [20] found that Environmental investments can enhance the relationship between AANR and corporate green innovation. Zeng et al. [21] pointed out that AANR can only stimulate corporate green innovation but cannot influence corporate investment in environmental governance.

Influencing factors of corporate emission reduction

Corporate pollution emission is an important indicator of environmental performance. Many factors influence corporate pollution emissions due to its negative externalities. Up to now, several studies have explored the factors influencing corporate emissions reduction from the perspectives of environmental regulation, society, institutions, and policies. Laplante and Rilstone [24] investigated the paper industry in Quebec and noted that government environmental regulation effectively reduced air and water pollution emissions from paper corporations. Fan et al. [25] demonstrated that green credits could significantly reduce corporate pollution emissions as a typical market-based environmental regulation. Based on data from Chinese-listed chemical and steel corporations, Zhang et al. [26] comprehensively analyzed institutional factors’ impact on corporate pollution emissions. The empirical results indicated that tax policies, government subsidies, credit policies, and media monitoring could significantly promote corporate energy conservation and emission reduction. Cherniwchan [27] directly examined the relationship between NAFTA and corporate pollution emissions. The research showed that trade liberalization decreased respirable particulate matter (PM10) and sulfur dioxide (SO₂) emissions from U.S. manufacturing corporations. He and Wang [28] based on 2001-2007 Chinese manufacturing plant-level data and found that declining tariffs on intermediate goods helped corporations reduce emissions.

Theoretical analysis and hypotheses

AANR and corporate pollutant emissions

AANR provides an overall evaluation of local government officials’ fulfillment of their responsibilities for natural resource asset management and environmental protection during their tenure. Its audit results are the main reference for officials’ promotion assessment. AANR breaks the traditional official assessment system that is centered on GDP growth. It adopts a dual assessment mechanism of economic development and environmental management to motivate local officials to fulfill their environmental management responsibilities, thus prompting local governments to pursue balanced development of the economy and environment [17]. Unlike previous environmental policies that acted directly on emission subjects, the AANR promotes environmental governance by stimulating the initiative of local officials [10]. AANR sends a signal to local officials of a shift in the assessment system. Based on public choice theory, officials will adjust their objective functions to maximize their interests and increase the importance of environmental protection work to gain future promotion opportunities. The AANR has both authoritative and professional characteristics, which can effectively monitor the fulfillment of the government’s public environmental fiduciary responsibility. Audit institutions implementing AANR are headed by the Audit Committee, which represents the authority of the CPC. As an environmental audit program with lifetime accountability, AANR can effectively deter opportunistic behavior by local governments. The use of advanced technologies such as 3S (RS, GIS, GPS) in implementing AANR has alleviated the information asymmetry between upper and lower levels of government [19]. It is no longer possible for local governments to falsify environmental data to meet the environmental protection requirements issued by the central government.

The production and operation of corporations generate negative externalities on the environment. Hefty-polluting corporations cause a considerable amount of pollution [29]. Relying solely on market adjustment cannot realize the optimal allocation, which requires the government to fulfill the fiduciary responsibility for effective management [30]. The gas, wastewater, and sludge emitted by corporations have severe negative impacts on the air, land, forests, and other natural resources, the protection of which is the focus of AANR. Considering the enormous influence that local governments have on corporations [31], under the shock of the institutional environment change, local officials, to gain political promotion or avoid the political risks, will not only increase their environmental investment but also choose to transfer the environmental pressure to the corporations in their jurisdiction. Local governments will reorganize their relationship with heavy polluters and reduce support to force them to transform [23]. At the same time, they will no longer maintain an acquiescent attitude toward corporate environmental pollution behavior but restrict corporate emissions by increasing environmental supervision [32].

Based on the above analysis, this paper proposes research hypothesis H1.

H1: The implementation of the AANR can significantly contribute to the reduction of corporate pollutant emissions.

Mechanism based on legitimacy

AANR audits local officials, but ultimately, emissions need to be controlled by corporations [17]. When local governments increase their environmental concerns due to AANR, legitimacy pressures will be placed on jurisdictional corporations [33]. In a system of norms, values, beliefs, and definitions constructed by society, corporate production and operation should be consistent with the values and beliefs agreed upon by society [34]. A corporation’s legitimacy is threatened when its behavior deviates from the requirements of laws and socially accepted values. Legitimacy pressure is a driving factor for corporate social responsibility [35]. Corporations active in environmental responsibility will receive less censure, thus reducing their legitimacy risk [36]. After implementing the AANR pilot, the legitimacy risk of jurisdictional corporations is increased. In particular, heavily polluting corporations tend to become the focus of government and public attention. At this time, corporations prefer to take measures to reduce the negative impact of their production operations on the environment. This response can reduce the likelihood of penalties for non-compliance with environmental regulations. It also presents a green corporate image to stakeholders such as investors, consumers, and suppliers, demonstrating that the corporation is aligned with the social value system, which increases the corporate competitiveness [37].

Based on the above analysis, this paper proposes research hypothesis H2a.

H2a: AANR promotes corporate emissions reduction through increasing corporate legitimacy pressure.

Mechanism based on resource dependency

Sometimes, regulatory pressure from the government has a limited impact on corporate environmental performance, and corporations may be proactive in green transformation under some circumstances [38]. According to resource dependence theory, external resources are necessary for corporate survival and development [39]. When environmental performance becomes an important indicator for officials’ assessment due to the implementation of AANR, an active undertaking of corporate environmental responsibility may increase the likelihood of gaining the government’s support. Chinese socialist system determines that the government’s influence on the market is pivotal [40]. The government holds many resources and has the power to allocate them [4], which directly affects the access to resources of corporations in the jurisdiction. Resource leaning or policy support from local government (such as the right to develop or use some natural resources, financing advantages, government subsidy, and project approval) is crucial for corporations to alleviate resource constraints. To gain access to more government resources, corporations in the pilot regions are incentivized to assist local governments in responding to AANR assessment by actively undertaking emission reduction tasks.

Based on the above analysis, this paper proposes research hypothesis H2b.

H2b: Corporations in the AANR pilot areas can alleviate financing constraints by reducing emissions.

The logical framework of the theoretical analysis part of this paper is shown in Figure 2.

Figure 2. Logical framework for AANR promoting corporate emission reduction.

Research design

Sample selection and data sources

This paper selects corporations in the heavy pollution industry listed on Chinese A-shares from 2010-2017 as the research sample. The determination of heavily polluting industry is mainly based on the "List of Listed Companies for Environmental Verification Industry Classification and Management" developed by the Ministry of Ecology and Environment (former Ministry of Environmental Protection) in 2008, which classifies 16 categories of industries, including thermal power, iron, steel, cement, electrolytic aluminum, coal, metallurgy, chemical, petrochemical, building materials, paper, brewing, pharmaceutical, fermentation, textile, tannery, and mining, as heavily polluting industries. The AANR has been conducted in the whole of China since 2018, and the absence of a control group from 2018 and after makes it impossible to use the DID model for testing. Considering this situation, the ending date of the study was chosen as 2017 in this paper. In this paper, the heavily polluting corporations in the 181 pilot regions are selected as the experimental group, and the heavily polluting corporations in other non-pilot regions are chosen as the control group. The pilot data were collected manually from the Audit Yearbook and cross-checked with information from local audit offices’ websites and the news media. The data on corporate emission charges are manually collected from the annual reports of listed corporations. The rest of the data related to corporate characteristics, corporate governance, etc., are obtained from the CSMAR database. To further enhance the data representativeness, this paper treats the samples as follows: (1) exclude sample corporations with a trading status of ST and PT; (2) exclude samples with missing or abnormal variable data; (3) winsorize all continuous variables at the 1% level to eliminate the effect of extreme values. Finally, 4271 valid observations were obtained, including 994 in the experimental group and 3859 in the control group.

Variable definition and measurements

1. Explained variables. The corporate emission fee (CEF) is a fee of a punitive nature charged by the Government’s Environmental Protection Department according to national laws and regulations to internalize corporations’ environmental costs. The emission fee is calculated based on the types and quantities of various solids, liquids, gases, and hazardous wastes discharged by the company and, to some extent, reflects the level of pollutants emissions. This paper uses emissions fees to measure corporate emissions intensity. The lower the emission fee, the fewer pollutants the corporation emits, and vice versa. According to the characteristics of the sample data, the emission fee indicator is measured by taking the natural logarithm of the corporate current year’s emission fee after adding 1.

2. Explanatory variables. This paper constructs policy dummy variables as explanatory variables by the following steps. First, generate the group dummy variable. If a corporation is in a region where the AANR is implemented, the corporation is taken as the experimental group with a value of 1; otherwise, it is taken as the control group with a value of 0. Then, generate the time dummy variable. The value is 1 if the year is in the year of the AANR implementation or later; otherwise, it is 0. Finally, the paper uses the interaction term DID of the group dummy variable and the time dummy variable as the policy dummy variable.

3. Control variables. This paper controls variables at the corporate characteristics and corporate governance levels. At the corporate characteristics level, the control variables selected for this paper are as follows. Corporate size (Size), using the corporation’s total assets at the end of the period and normalized by the natural logarithm. State-owned enterprise (Soe), taking the value of 1 if it is a state-owned enterprise and 0 if it is a non-state-owned enterprise. Leverage (Lev), which is equal to total liabilities/total assets, the higher the indicator, the higher the debt level. Operating capacity (OC), which is equal to Operating Income/Total Assets Ending Balance. Profitability (ROA), which is equal to net profit/total assets balance. Growth capacity (Growth), equal to (Total assets at the end of the period - Total assets at the beginning of the period)/Total assets at the beginning of the period. Risk level (RL), equal to (net income + income tax expense + finance costs + depreciation of fixed assets, depreciation of oil and gas assets, depreciation of productive biological assets + amortization of intangible assets + amortization of long-term pending expenses)/(net income + income tax expense). At the corporate governance level, the control variables selected for this paper are as follows. Shareholding concentration (Concen) equals the percentage of shares held by the first largest shareholder. Board size (BDS) equals the number of board directors. Dual position (DUA), chairman and general manager in one position is 1, otherwise is 0. The Board of Supervisors size (BSS) equals the number of supervisors (including the supervisors’ chairman).

4. Other variables. The Chinese government’s work report reflects the government’s will and its environmental content reflects the government’s concern for environmental issues. This paper uses the percentage of environmental words to total words (ER20) in the prefecture-level city government’s work report to measure the environmental pressure borne by corporations. The environmental word set was constructed with reference to the research of Yu et al. [41]. Considering that financial resources are essential to corporate operations, this paper uses financing constraints (SA) as a proxy for corporate resource constraints. Referring to the treatment of Hadlock and Pierce [42], the absolute value of -0.737 × Size + 0.043 × ${\mathit{\text{Size}}}^2$-0.04 × Age is taken to measure the level of corporate financing constraint. The higher the value of SA, the stronger the corporate financing constraint.

Research model

With reference to the research of Beck et al. [43], this paper constructs a multi-period DID model (1) to test H1. Based on the mediating effects model proposed by Baron and Kenny [44], this paper adds models (2) and (3) to test H2a while adding models (4) and (5) to test H2b. (1) $${\mathit{\text{CEF}}}_{i,t}={\alpha }_0+{\alpha }_1{\mathit{\text{DID}}}_{i,t}+\lambda X_{i,t}+\mathit{\text{Year}}+\mathit{\text{Company}}+{\epsilon }_{i,t}$$(1) (2) $${\mathit{\text{ER}}20}_{i,t}={\beta }_0+{\beta }_1{\mathit{\text{DID}}}_{i,t}+\lambda X_{i,t}+\mathit{\text{Year}}+\mathit{\text{Company}}+{\epsilon }_{i,t}$$(2) (3) $${\mathit{\text{CEF}}}_{i,t}={\gamma }_0+{\gamma }_1{\mathit{\text{DID}}}_{i,t}+{\gamma }_2{\mathit{\text{ER}}20}_{i,t}+\lambda X_{i,t}+\mathit{\text{Year}}+\mathit{\text{Company}}+{\epsilon }_{i,t}$$(3) (4) $${\mathit{\text{SA}}}_{i,t}={\beta }_0+{\beta }_1{\mathit{\text{DID}}}_{i,t}+\lambda X_{i,t}+\mathit{\text{Year}}+\mathit{\text{Company}}+{\epsilon }_{i,t}$$(4) (5) $${\mathit{\text{SA}}}_{i,t}={\gamma }_0+{\gamma }_1{\mathit{\text{DID}}}_{i,t}+{\gamma }_2{\mathit{\text{CEF}}}_{i,t}+\lambda X_{i,t}+\mathit{\text{Year}}+\mathit{\text{Company}}+{\epsilon }_{i,t}$$(5)

In the above models, the subscript i denotes individual variable (i.e. company); the subscript t denotes time series (i.e. year); ${\alpha }_0{,{\beta }_0 \mathit{\text{and}} \gamma }_0$ are the intercept term; $X_{i,t}$ is a series of control variables selected in this paper; Year is a time-fixed effect; Company is a company-fixed effect; ${\epsilon }_{i,t}$ is a random disturbance term.

Empirical analysis and testing

Descriptive statistics

Table 1 shows the descriptive statistics results of the main variables. Regarding the explained variables, the mean value of corporate emission fee (CEF) is 3.112, the maximum value is 17.767, the minimum value is 0, and the standard deviation is 6.129, which indicates that the overall emission fees of listed corporations in China are low, and the level of pollution emission varies widely among different corporations. Regarding core explanatory variables, the number of samples experiencing policy shocks accounts for about 20% of the total sample size. The descriptive statistics results for each control variable are similar to those of related studies, indicating that the sample data are relatively accurate. Except for equity concentration (Concen), the standard deviations of all variables are minor, indicating that the variables are relatively evenly distributed.

Table 1. Descriptive statistics of main variables.

Variable	Obs	Mean	Std. Dev	Min	Max
CEF	4271	3.112	6.129	0	17.767
DID	4271	0.207	0.405	0	1
Size	4271	22.222	1.341	19.999	26.221
Soe	4271	0.446	0.497	0	1
Lev	4271	0.422	0.203	0.049	0.925
OC	4271	0.692	0.41	0.1	2.399
ROA	4271	0.048	0.041	−0.015	0.199
Growth	4271	0.199	0.33	−0.248	1.842
RL	4271	3.551	5.088	0.984	34.917
Concen	4271	37.425	15.489	8.63	79.71
BDS	4271	8.972	1.811	5	15
DUA	4271	0.227	0.419	0	1
BSS	4271	3.807	1.224	3	8

Correlation analysis

In order to get a preliminary understanding of the correlation between the explained variables and the explanatory variables and determine whether there is a multicollinearity problem among the variables, Pearson correlation analysis was performed. As shown in Table 2, the policy dummy variable (DID) is significantly negatively correlated with the corporate emission fee (CEF) at the 1% level. It indicates that implementing AANR reduces corporations’ pollutant emissions, and hypothesis H is initially verified. The vast majority of the control variables showed significant correlations with the CEF, indicating that the control variables in this paper were appropriately selected. The correlation coefficients between the main variables were below 0.50, indicating no severe multicollinearity problem.

Table 2. Correlation analysis.

	CEF	DID	Size	Soe	Lev	OC	ROA	Growth	RL	Concen	BDS	DUA	BSS
CEF	1
DID	−0.058***	1
Size	0.170***	0.099***	1
Soe	0.165***	−0.033**	0.444***	1
Lev	0.181***	−0.053***	0.521***	0.394***	1
OC	−0.006	−0.064***	−0.043***	0.016	0.045***	1
ROA	−0.111***	0.045***	−0.079***	−0.199***	−0.455***	0.119***	1
Growth	−0.086***	0.004	−0.137***	−0.236***	−0.217***	−0.098***	0.276***	1
RL	0.116***	−0.070***	0.185***	0.224***	0.380***	−0.034**	−0.435***	−0.173***	1
Concen	0.055***	−0.013	0.320***	0.221***	0.114***	0.117***	0.074***	−0.028*	0.022	1
BDS	0.069***	−0.055***	0.344***	0.307***	0.236***	0.030**	−0.021	−0.054***	0.103***	0.059***	1
DUA	−0.063***	0.030**	−0.220***	−0.322***	−0.181***	−0.044***	0.064***	0.126***	−0.060***	−0.071***	−0.176***	1
BSS	0.171***	−0.073***	0.361***	0.448***	0.300***	0.051***	−0.107***	−0.146***	0.130***	0.112***	0.402***	−0.204***	1

Note: ***, **, * denote 1%, 5%, 10% significance levels, respectively.

Benchmark regression results

Table 3 shows the benchmark regression results. Among them, column (1) only includes explained and explanatory variables; column (2) adds control variables on the basis of column (1); column (3) adds controls for company-fixed effects and time-fixed effects based on column (2). The results in columns (1) to (3) all indicate that the DID is negatively related to the CEF (with the results in column (3) being ${\alpha }_1$= −0.911 and t = −3.29). It is obvious that implementing AANR can significantly curb corporate pollutant emissions. Hypothesis H1 is verified. Similar to the existing literature [15–21], the benchmark regression results again validate the environmental governance role of AANR on the micro corporations. However, we have new findings that this governance effect is not only reflected in the change of corporate behavior but also in the timely reduction of the corporate negative impact on the environment.

Table 3. Benchmark regression results.

Variable	(1)	(2)	(3)
	CEF	CEF	CEF
DID	−0.915***	−0.879***	−0.911***
	(−4.26)	(−4.11)	(−3.29)
Size		0.398***	0.415***
		(4.05)	(4.14)
Soe		0.423*	0.700***
		(1.71)	(2.76)
Lev		1.432**	2.227***
		(2.27)	(3.49)
OC		−0.240	−0.882***
		(−1.10)	(−3.97)
ROA		−5.410**	−4.971*
		(−2.00)	(−1.85)
Growth		−0.427*	−0.715***
		(−1.73)	(−2.81)
RL		0.049**	0.041*
		(2.03)	(1.74)
Concen		0.003	0.004
		(0.43)	(0.54)
BDS		−0.112*	−0.092
		(−1.88)	(−1.52)
DUA		0.015	0.021
		(0.07)	(0.10)
BSS		0.554***	0.565***
		(5.34)	(5.50)
Constant	3.302***	−7.213***	−7.780***
	(30.68)	(−3.72)	(−3.91)
Company	NO	NO	YES
Year	NO	NO	YES
N	4,271	4,271	4,271
R-squared	0.004	0.060	0.079

Note: The t-values are in parentheses. ***, **, * denote 1%, 5%, and 10% significance levels, respectively. Same as below.

In terms of control variables, many significant relationships can also be found in column (3). The corporate size (Size) positively correlates with CEF, indicating that the larger the corporate size, the more significant the negative impact on the environment is caused by its production and operation activities. State-owned enterprises (Soe) are positively correlated with CEF, indicating that compared to non-state enterprises, state-owned enterprises with a special political background may face looser constraints on emissions when there are no new policy incentives. Leverage (Lev) is positively correlated with CEF, indicating that the higher the debt level, the more difficult it is for corporations to treat emissions due to insufficient funds. Operating capacity (OC) and growth capacity (Growth) are negatively correlated with CEF, indicating that the stronger a corporation is, the better it is likely to do pollution control work. Shareholding concentration (Concen) is positively correlated with CEF, which indicates that the more concentrated the shareholding, the more difficult it is for its internal control mechanisms to function and the more likely it is to engage in short-sighted, environmentally destructive behavior for the sake of profitability. The board of supervisors size (BSS) positively correlates with the CEF. Due to the complementary nature of the board size and the BSS, a large BSS may lead to institutional bloat and inadequate supervision of corporate emissions practices.

Mechanism test results

Table 4 shows the mechanism test results. Column (1) shows a significant negative correlation between DID and ER20 (${\beta }_1$=0.062, t = 3.37), demonstrating that the environmental pressures in the pilot regions are elevated after AANR implementation. A plausible explanation for this result is that AANR puts environmental performance evaluation pressure on local officials, who are forced to shift the pressure by increasing environmental regulation and enforcement [10,18] in their jurisdictions. Column (2) shows a negative correlation between ER20 and CEF (${\gamma }_2$=-0.568, t=-2.54), indicating that environmental pressure can significantly curb corporate pollution emissions. This finding re-validates the government’s significant influence on corporate behavior [31]. Combining columns (1)-(2) of Table 4 and column (3) of Table 3, it is clear that all the coefficients of DID as well as ER20 are significant, and the results of the Sobel test are significant at the 5% level. These results indicate the presence of partial mediation effects, implying that AANR reduces corporate emissions by increasing environmental pressures. Based on the legitimacy theory [34], the mechanism of AANR's effect on corporate emission reduction is verified, and H2a is valid.

Table 4. Mechanism test results.

Variable	(1)	(2)	(3)	(4)
	ER20	CEF	SA	SA
DID	0.062***	−0.876***	−0.031**	−0.030**
	(3.37)	(−3.07)	(−2.43)	(−2.33)
ER20		−0.568**
		(−2.54)
CEF				0.001**
				(2.21)
Size	0.023***	0.347***	−0.051***	−0.052***
	(3.52)	(3.43)	(−9.67)	(−9.75)
Soe	−0.072***	0.703***	−0.026***	−0.027***
	(−4.49)	(2.71)	(−2.61)	(−2.71)
Lev	−0.054	2.725***	0.091***	0.087***
	(−1.28)	(4.24)	(3.34)	(3.23)
OC	0.006	−1.000***	0.010	0.011
	(0.36)	(−4.47)	(1.04)	(1.17)
ROA	−0.048	−3.709	0.040	0.047
	(-0.26)	(−1.36)	(0.34)	(0.40)
Growth	0.013	−0.630**	0.019*	0.020*
	(0.58)	(-2.52)	(1.74)	(1.84)
RL	−0.002*	0.032	0.000	0.000
	(−1.77)	(1.34)	(0.38)	(0.31)
Concen	0.001	0.005	−0.001***	−0.001***
	(1.61)	(0.83)	(−4.97)	(−4.98)
BDS	0.013***	−0.106*	−0.013***	−0.013***
	(3.59)	(−1.75)	(−5.17)	(−5.13)
DUA	−0.009	−0.006	−0.017*	−0.017*
	(−0.58)	(−0.03)	(−1.96)	(−1.96)
BSS	−0.009	0.622***	−0.007*	−0.008*
	(−1.41)	(5.95)	(−1.67)	(−1.85)
Constant	1.804***	−5.425***	4.936***	4.947***
	(13.61)	(−2.63)	(46.30)	(46.27)
Company	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	4193	4193	4,271	4,271
R-squared	0.124	0.079	0.237	0.238
Sobel tests	−0.035**		−0.001**
	(−1.96)		(−1.98)

Column (3) in Table 4 shows that the regression coefficient of the DID is significantly negative at the 1% level (${\beta }_1$=-0.031, t=-2.43). This indicates that AANR significantly reduces the corporate financing constraint. Most existing research concludes that AANR raises corporate compliance costs and constrains corporate production and operations [45]. However, we have a different finding that the AANR has improved corporate financial resources. Column (4) shows that the regression coefficient of CEF is significantly positive at the 5% level (${\gamma }_2$=0.001, t = 2.21). It indicates that emission discharge enhances corporate financial constraints. Combining columns (3)-(4) of Table 4 and column (3) of Table 3, it is shown that all the coefficients of DID as well as CEF are significant, and the results of the Sobel test are significant at 5% level, which proves the existence of partial mediation effects. It means that AANR can alleviate the financing constraints of corporations that reduce emissions, and H2b is verified. According to resource dependence theory, corporations are incentivized to cooperate with the government through pollution control to obtain more resources.

Robustness tests

Parallel trend test

The parallel trend assumption is a crucial prerequisite assumption for the DID method. Its purpose is to verify that the experimental and control groups are not significantly different and have a common trend before the policy is implemented. This approach was used to exclude that the policy effect was due to differences between the experimental and control groups. Therefore, this paper conducts parallel trend hypothesis testing and constructs the following model (6). (6) $${\mathit{\text{CEF}}}_{i,t}={\alpha }_0+{\alpha }_1\mathit{\text{pre}}6+{\alpha }_2\mathit{\text{pre}}5+{\alpha }_3\mathit{\text{pre}}4+{\alpha }_4\mathit{\text{pre}}3+{\alpha }_5\mathit{\text{pre}}2+{\alpha }_6\mathit{\text{pre}}1+{\alpha }_{7}c\mathit{\text{urrent}}+{\alpha }_{8}p\mathit{\text{ost}}1+{\alpha }_{9}p\mathit{\text{ost}}2+{\alpha }_{10}p\mathit{\text{ost}}3+\lambda X_{i,t}+\mathit{\text{Year}}+\mathit{\text{Company}}+{\epsilon }_{i,t}$$(6)

The model’s pre(n), current, and post(n) are time dummy variables. Pre(n) indicates that the experimental group takes a value of 1 in the n years before the policy implementation and 0 otherwise; current indicates that the experimental group takes a value of 1 in the year of policy implementation and 0 otherwise; post(n) indicates that the experimental group takes a value of 1 in the n years after the policy implementation and 0 otherwise, respectively. This paper selects the first year of the time series as the base year, so pre7 is not included in the model (6).

According to the model (6), a parallel trend test is conducted. The estimated coefficients ${\alpha }_1-{\alpha }_{10}$ with 95% confidence intervals are shown in Figure 3. The test results showed that none of the estimated coefficients ${\alpha }_1-{\alpha }_6$ significantly differed from 0 before the implementation of the AANR pilot policy. This indicates that the experimental and control groups had a common change trend before the implementation of AANR, which is consistent with the premise hypothesis of DID. The estimated coefficients ${\alpha }_7-{\alpha }_8$ are not significantly different from 0, and the estimated coefficients ${\alpha }_9-{\alpha }_{10}$ are significantly different from 0, which indicates a time lag in the promotion of AANR on corporate emission reduction. The above results prove that the experimental and control groups are comparable.

Figure 3. Parallel trend test.

Placebo test

This paper draws on the treatment of Cai et al. [46] to conduct a placebo test by randomly assigning pilot cities to determine whether random factors drive the facilitative effect of AANR on corporate emissions reductions. Specifically, this paper fictitious the same number of pilot cities as the experimental group through random sampling, assuming they implemented AANR and the remaining cities are selected as the control group. The model (1) is re-estimated using the new sample based on the above design, and one placebo test is completed. The above process is repeated 500 times to obtain a P-value plot of all estimated coefficients, shown in Figure 4.

Figure 4. Placebo test.

According to Figure 4, most DID-estimated coefficients have P-values greater than 0.1. Meanwhile, the actual estimate of this paper (${\alpha }_1$=-0.911) is a significant outlier in the placebo test. These results indicate that the explanatory variable DID, constructed by random sampling, does not affect corporate emission reductions, ruling out that the policy results are due to other unobservable factors or omitted variables. Therefore, the benchmark regression results are robust.

PSM- DID test

This paper uses Propensity Score Matching (PSM) to match the experimental group with the corresponding control group to exclude the bias problem due to possible significant differences between the samples of the experimental group and those of the control group. Specifically, this paper first selects all control variables as covariates and estimates propensity scores using Logit models. Next, matching is performed using the caliper k nearest neighbor matching method with a matching ratio of 1:1 and a matching caliper of 0.01. Then, the common support assumption is used to test the balance of the matched samples and remove the observations that do not satisfy the common support. Finally, regression is performed again based on model (1) to observe the regression coefficients of DID. Column (1) of Table 5 reports the regression results of PSM - DID. As shown in column (1), the DID regression coefficient is significantly negative at the 1% level (${\alpha }_1$=-0.953, t=-3.37), which does not change substantially compared to the benchmark regression, indicating that the benchmark regression results of this paper are robust.

Table 5. Robustness test results.

Variable	(1)	(2)	(3)	(4)
	CEF	CEF	CEF	CEF
DID	−0.953***	−0.751***	−0.914***	−0.015***
	(−3.37)	(−2.67)	(−3.29)	(−4.20)
Talk		0.764***
		(2.91)
Inspect			0.170
			(0.36)
Size	0.469***	0.421***	0.414***	−0.002*
	(4.43)	(4.21)	(4.14)	(−1.81)
Soe	0.715***	0.649**	0.700***	0.010**
	(2.75)	(2.56)	(2.76)	(2.47)
Lev	2.066***	2.148***	2.226***	0.023***
	(3.19)	(3.38)	(3.49)	(2.70)
OC	−0.930***	−0.870***	−0.881***	−0.003
	(−4.08)	(−3.94)	(−3.97)	(−0.95)
ROA	−4.234	−4.964*	−4.955*	0.086
	(−1.54)	(−1.85)	(−1.84)	(1.50)
Growth	−0.727***	−0.691***	−0.715***	−0.011***
	(−2.75)	(−2.72)	(−2.81)	(−3.64)
RL	0.049*	0.041*	0.041*	−0.000
	(1.80)	(1.72)	(1.75)	(−0.61)
Concen	0.005	0.004	0.003	0.000*
	(0.69)	(0.60)	(0.54)	(1.87)
BDS	−0.111*	−0.088	−0.092	−0.002***
	(−1.77)	(−1.46)	(−1.52)	(−2.69)
DUA	0.005	0.029	0.020	0.005*
	(0.02)	(0.14)	(0.10)	(1.82)
BSS	0.575***	0.560***	0.566***	0.003***
	(5.33)	(5.45)	(5.50)	(3.01)
Constant	−8.804***	−8.126***	−7.817***	0.053**
	(−4.23)	(−4.09)	(−3.92)	(2.34)
Company	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	4,126	4,271	4,271	4,271
R-squared	0.081	0.080	0.079	0.019

Exclude other policy factors interference

Since the 18th National Congress of the Communist Party of China (CPC) wrote "building ecological civilization" into the party constitution, the central government has introduced several measures to promote the coordinated development of economic and environmental protection. Therefore, other environmental policies introduced in the same period as AANR also have the possibility to influence the corporate pollutant emission intensity, which in turn leads to biased estimation results in this paper. Existing studies show that environmental protection talk [47] and environmental protection inspector [48] significantly impact the fulfillment of corporate environmental responsibility. Meanwhile, the beginning dates of these two policies are 2014 and 2016, respectively, which roughly coincide with the beginning date of AANR. Based on this, in order to exclude the interference of other policy factors and further verify the robustness of the benchmark regression results, this paper generates environmental protection talk (Talk) and environmental protection inspectors (Inspect) policy dummy variables added to the model (1), respectively. Columns (2) and (3) of Table 5 report the regression results excluding other policy interferences. After controlling for other policy factors, the regression coefficients of DID are all significantly negative (${\alpha }_1$=-0.751 and −0.914, respectively), which is consistent with the benchmark regression results, indicating that the benchmark regression results are robust.

Substitution of explanatory variables

In order to exclude the influence of corporate size on emissions, this paper conducts robustness tests by changing the standardized way of explained variables. We use the value of (Corporate emission fee/Total assets at the end of the period) ×1000 to measure CEF and re-run the regression based on model (1). Columns (4) of Table 5 shows that the regression coefficient of DID remains negative (${\alpha }_1$=-0.015, t=-4.20). The regression results after replacing the explanatory variables still support the original hypothesis, indicating that the benchmark regression results are robust.

Further analysis

Heterogeneity analysis

The above results suggest that the implementation of AANR does help promote micro-corporation emission reduction. In order to deepen the study of AANR pilot effects, this paper explores the individual heterogeneity and environmental heterogeneity of the impact that AANR causes on corporate emission reduction behavior from four aspects: corporate characteristics, internal governance mechanisms, external environment, and external governance mechanisms.

Heterogeneity of corporate characteristics

In the face of local government environmental governance pressure, corporations with different corporate characteristics may show some variability in their environmental behavior. The corporation’s size is an essential factor in controlling emissions, as the pollution treatment sometimes requires enormous expenditures [49]. Small companies are at a competitive disadvantage, and resource constraints make it challenging to adopt new eco-friendly technologies or equipment [50]. Under the policy shock, large companies may have greater capacity to manage pollution than small companies. Compared with non-state-owned enterprises (non-SOEs), state-owned enterprises (SOEs) have greater bargaining power in the face of environmental regulation. Some studies have shown that the political status of some SOE managers is even higher than that of local environmental regulators, which reduces the risk of SOEs being penalized for emission problems [51]. The special status of SOEs may render AANR ineffective against them. However, SOE executives are directly appointed by government departments and have the status of "quasi-officials." Compared with professional managers, they tend to shoulder more social responsibility in exchange for political reputation [52]. SOE executives will be more proactive in combating pollution to meet the environmental goals for promotion purposes [53]. Political connection means that the corporate managers maintain a tight relationship with the government, which benefits corporate access to resources and information. Managers of political connection corporations are sensitive to governmental directives due to their particular political status or political experience [54]. When local officials are under pressure from AANR assessments, political connection corporations must be more active in controlling emissions than ordinary corporations to meet the government’s expectations [55].

This paper uses the annual median of corporate size (Size) as the basis for grouping and divides the sample into large-sized corporations and small-sized corporations. The samples are also grouped into state-owned enterprises (Soe = 1) and non-state-owned enterprises (Soe = 0), as well as politically connected corporations (PC = 1) and non-politically connected corporations (PC = 0) based on the nature of property rights (Soe) and political affiliation (PC), respectively. The regression results in Table 6 show that the coefficients of DID are all significantly negative. However, the regression coefficient of DID for the large-sized corporation group (${\alpha }_1$=-0.987) is much larger than that for the small-sized corporation group (${\alpha }_1$=-0.612) while passing the Chow test. The regression coefficient of DID for the politically connected corporation group (${\alpha }_1$=-1.708) is much larger than that for the non-politically connected corporation group (${\alpha }_1$=-0.569) while passing the Chow test. It means that the promotion effect of AANR on corporate emission reduction is stronger in large-sized corporations and politically connected corporations. The DID regression coefficient for the state-owned corporation group (${\alpha }_1$=-1.154) is much larger than that for the non-state-owned corporation group (${\alpha }_1$=-0.653), but the Chow test does not pass. Although it is impossible to conclude that AANR has a greater impact on discharges in state-owned corporations than in non-state-owned corporations, it is certain that state-owned corporations do not escape from regulation because of their special status.

Table 6. Grouped regression results for different corporate characteristics.

	(1)	(2)	(3)	(4)	(5)	(6)
	CEF	CEF	CEF	CEF	CEF	CEF
Variable	Large size	Small size	SOE	non-SOE	PC	non-PC
DID	−0.987**	−0.612**	−1.154**	−0.653**	−1.708***	−0.569*
	(−2.17)	(−2.01)	(−2.28)	(−2.08)	(−3.42)	(−1.71)
Size	−0.224	0.624***	0.284**	0.732***	0.542***	0.350***
	(−1.26)	(2.99)	(2.00)	(5.20)	(3.10)	(2.88)
Soe	0.896**	0.579*			0.612	1.132***
	(2.29)	(1.81)			(1.30)	(3.69)
Lev	2.100*	0.866	4.175***	0.294	2.165**	2.343***
	(1.80)	(1.15)	(3.99)	(0.36)	(1.97)	(3.04)
OC	−1.036***	−0.578**	−0.243	−1.591***	−1.636***	−0.519*
	(−2.69)	(−2.53)	(−0.66)	(−7.15)	(−4.52)	(−1.85)
ROA	−10.714**	−1.537	−12.144***	4.950	−11.003***	−0.533
	(−2.50)	(−0.48)	(−2.90)	(1.48)	(−2.58)	(−0.15)
Growth	−1.742***	−0.302	−1.196*	−0.807***	−1.053**	−0.597*
	(−3.20)	(−1.11)	(−1.75)	(−2.99)	(−2.46)	(−1.95)
RL	0.028	0.047	−0.011	0.124***	0.062	0.033
	(0.90)	(1.41)	(−0.40)	(2.63)	(1.49)	(1.19)
Concen	0.010	0.007	−0.002	0.020**	0.024**	−0.008
	(0.97)	(0.84)	(−0.20)	(2.49)	(2.31)	(−0.94)
BDS	−0.080	−0.069	−0.050	−0.076	−0.228**	−0.040
	(−0.97)	(−0.86)	(−0.57)	(−1.00)	(−2.05)	(−0.55)
DUA	0.102	0.045	0.023	0.208	−0.248	0.303
	(0.23)	(0.20)	(0.04)	(0.92)	(−0.72)	(1.14)
BSS	0.665***	0.194	0.261**	1.271***	0.310*	0.716***
	(5.13)	(1.31)	(2.14)	(6.92)	(1.83)	(5.56)
Constant	6.989*	−11.539***	−3.962	−17.162***	−7.733**	−8.012***
	(1.88)	(−2.62)	(−1.43)	(−5.92)	(−2.25)	(−3.28)
Company	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES
N	2,127	2,144	1,905	2,366	1,585	2,686
R-squared	0.065	0.036	0.064	0.098	0.108	0.087
Chow tests	15.97***		0.71		7.82***

Heterogeneity of internal governance mechanisms

Whether corporations will actively respond to national policies is closely related to their internal governance mechanisms. According to internal control theory, the corporate internal governance mechanism is often responsible for ensuring the organization’s healthy development and promoting the realization of corporate strategies. Meanwhile, ensuring corporate legal compliance is one of the critical objectives of the internal governance mechanism [56]. After the Auditor General’s Office conducts the AANR, corporations are often under regulation pressure from local government. Corporations with stronger internal governance mechanisms can better restrain corporate management’s short-sightedness and moral hazard [57]. Thus, they are more likely to act actively to fulfill their environmental responsibilities to ensure corporate legitimacy. Based on this, this paper uses the annual median of internal control index (IC), board size (BDS), and percentage of independent directors (Indd) as the basis for grouping the samples into the above-median group (high internal control index, many directors, and high percentage of independent directors) and below-median group (low internal control index, few directors, and low percentage of independent directors), respectively. The regression results in Table 7 show that the DID regression coefficients for the above-median group are all significantly negative at the 1% level (${\alpha }_1$=-1.477, −1.197, and −0.916); the DID regression coefficients for the below-median group are all negative (${\alpha }_1$=-0.485, −0.406, and −1.108), but none are significant. This indicates that the AANR contributes more to emissions reduction in corporations with better internal governance mechanisms.

Table 7. Group regression results for different internal governance mechanisms.

	(1)	(2)	(3)	(4)	(5)	(6)
	CEF	CEF	CEF	CEF	CEF	CEF
Variable	High internal control	Low internal control	Many directors	Few directors	High independent directors percentage	Low independent directors percentage
DID	−1.477***	−0.485	−1.197***	−0.406	−0.916***	−1.108
	(−3.63)	(−1.26)	(−3.49)	(−0.87)	(−3.30)	(−0.23)
Size	0.473***	0.395***	0.492***	0.247	0.430***	−1.435
	(3.42)	(2.65)	(4.07)	(1.29)	(4.29)	(−1.33)
Soe	0.507	0.839**	0.622**	1.209**	0.702***	6.302**
	(1.41)	(2.28)	(2.15)	(2.33)	(2.74)	(2.21)
Lev	2.340**	2.240**	2.157***	2.694**	2.071***	21.739**
	(2.57)	(2.39)	(2.79)	(2.44)	(3.24)	(2.18)
OC	−0.971***	−0.826***	−0.831***	−1.019***	−0.967***	1.715
	(−2.88)	(−2.60)	(−2.91)	(−3.35)	(−4.38)	(0.59)
ROA	−4.885	−6.394	−5.731*	−3.118	−5.020*	16.836
	(−1.33)	(−1.55)	(−1.76)	(−0.66)	(−1.84)	(0.94)
Growth	−0.485	−0.920***	−0.410	−1.376***	−0.746***	6.247*
	(−1.29)	(−2.59)	(−1.28)	(−3.34)	(−2.93)	(1.91)
RL	0.040	0.032	0.056**	−0.027	0.040*	−0.208
	(1.21)	(0.98)	(2.07)	(−0.68)	(1.69)	(−0.80)
Concen	−0.002	0.010	0.001	0.001	0.004	0.035
	(−0.26)	(1.06)	(0.13)	(0.10)	(0.61)	(0.25)
BDS	−0.030	−0.173**	−0.160*	−0.102	−0.080	−0.178
	(−0.34)	(−2.03)	(−1.68)	(−0.55)	(−1.30)	(−0.34)
DUA	0.120	−0.126	0.008	0.026	−0.001	2.059
	(0.39)	(−0.42)	(0.03)	(0.07)	(−0.00)	(0.70)
BSS	0.465***	0.653***	0.678***	0.010	0.534***	1.503**
	(3.09)	(4.57)	(6.03)	(0.04)	(5.13)	(2.31)
Constant	−8.891***	−7.163**	−9.163***	−2.240	−7.995***	11.621
	(−3.24)	(−2.43)	(−3.95)	(−0.53)	(−4.00)	(0.61)
Company	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES
N	2,110	2,112	3,066	1,205	4,228	41
R-squared	0.077	0.088	0.094	0.072	0.078	0.702

Heterogeneity of external environment

The external environment is a crucial factor influencing the government audit to perform its governance function. Whether the AANR pilot can produce significant policy effects is related to the level of economic development and legalization in the pilot regions. Local governments are often trapped in public value conflicts when making administrative decisions [58]. When pursuing environmental performance, it will limit the realization of economic performance. Economic growth is the most widely recognized public value goal, often seen as happiness or a substitute for happiness [59]. Regions with poor economic growth may continue to tolerate high-polluting corporations to achieve economic growth goals, undermining the AANR's effectiveness. An excellent legal environment is a fundamental guarantee for implementing policies and regulations. In regions with a high level of legalization, anyone can effectively prosecute polluters [50]. This advantage will help AANR fulfill its supervisory function. In general, the eastern regions of China have the highest level of legalization, followed by the central and western regions. The paper uses the annual median economic growth rate as the basis of grouping and divides the sample into the high economic growth level group and the low economic growth level group. Meanwhile, the paper uses the city location to measure the degree of regional legalization and divides the sample into the eastern region group and the central and western region group. Columns (1) and (2) of Table 8 show that the DID regression coefficients are significantly negative at the 1% level for the eastern region (${\alpha }_1$=-1.408, t=-4.04) and insignificant for the central and western regions (${\alpha }_1$=0.111, t = 0.23). Columns (3) and (4) show that the DID regression coefficients are significantly negative at the 1% level for the high economic growth level group (${\alpha }_1$=-1.133, t=-3.48) and insignificant for the low economic growth level group (${\alpha }_1$=-0.850, t=-1.64). This indicates that the higher the level of regional legalization and economic growth, the stronger the contribution of AANR to corporate emissions reduction.

Table 8. Grouped regression results for different external environments.

	(1)	(2)	(3)	(4)
	CEF	CEF	CEF	CEF
Variable	Eastern regions	Central and western regions	High economic growth	Low economic growth
DID	−1.408***	0.111	−1.133***	−0.850
	(−4.04)	(0.23)	(−3.48)	(−1.64)
Size	0.557***	0.396**	0.339***	0.551***
	(3.99)	(2.53)	(2.65)	(3.33)
Soe	1.081***	0.125	0.793**	0.705*
	(3.11)	(0.31)	(2.39)	(1.75)
Lev	0.507	3.981***	2.980***	0.608
	(0.64)	(3.78)	(3.66)	(0.59)
OC	−1.694***	0.277	−0.443	−1.262***
	(−7.34)	(0.71)	(−1.46)	(−3.86)
ROA	−3.165	−4.922	−6.453*	−3.026
	(−0.95)	(−1.12)	(−1.93)	(−0.65)
Growth	−0.388	−1.183***	−0.518	−0.852**
	(−1.24)	(−2.72)	(−1.51)	(−2.26)
RL	0.065*	0.020	0.018	0.072*
	(1.74)	(0.65)	(0.63)	(1.77)
Concen	−0.013	0.015	0.018**	−0.019*
	(−1.62)	(1.37)	(2.15)	(−1.82)
BDS	−0.188**	−0.112	−0.048	−0.183*
	(−2.33)	(−1.25)	(−0.60)	(−1.91)
DUA	−0.183	0.728*	0.366	−0.495
	(−0.77)	(1.73)	(1.31)	(−1.54)
BSS	0.339**	0.760***	0.643***	0.456***
	(2.49)	(4.88)	(4.75)	(2.89)
Constant	−7.779***	−9.481***	−7.887***	−7.864**
	(−2.92)	(−3.02)	(−3.01)	(−2.48)
Company	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	2,467	1,804	2,488	1,751
R-squared	0.071	0.111	0.104	0.068

Heterogeneity of external governance mechanisms

As an essential external governance mechanism, the media and securities analysts improve corporate governance mainly through reputation and information dissemination mechanisms [60,61]. Based on legitimacy theory, the attention of the media and analysts is not only a way for corporations to gain legitimacy but also a source of legitimacy-generating crises. Media and analyst attention will make corporations pay more attention to environmental responsibility fulfillment to avoid legitimacy risks, send positive signals to the public, investors, and creditors that corporations are actively fulfilling their social responsibilities, and promote corporate reputation and value [62]. Therefore, in the background of AANR implementation, corporations with higher media and analyst attention are more pressured and more willing to reduce emissions. This paper uses the annual median of analyst and media attention as the grouping basis, respectively. The samples are divided into the above-median group (high media attention and high analyst attention) and the below-median group (low media attention and low analyst attention). The regression results in Table 9 show that the DID regression coefficients are significantly negative at the 1% level for the high media attention group (${\alpha }_1$=-0.980, t=-2.52) and insignificant for the low media attention group (${\alpha }_1$=-0.583, t=-1.39). It can also be seen that the DID regression coefficients are significantly negative at the 1% level for the high analyst attention group (${\alpha }_1$=-1.104, t=-2.62) and insignificant for the low analyst attention group (${\alpha }_1$=-0.618, t=-1.53). These results indicate that the contribution effect of AANR on corporate emissions reduction is stronger in corporations with better external governance mechanisms.

Table 9. Group regression results for different external governance mechanisms.

	(1)	(2)	(3)	(4)
	CEF	CEF	CEF	CEF
Variable	High media attention	Low media attention	High analyst attention	Low analyst attention
DID	−0.980**	−0.583	−1.104***	−0.618
	(−2.52)	(−1.39)	(−2.62)	(−1.53)
Size	0.501***	0.307**	0.545***	0.655***
	(3.33)	(2.11)	(3.73)	(4.07)
Soe	0.968***	0.163	0.614*	0.703*
	(2.58)	(0.46)	(1.71)	(1.77)
Lev	2.353**	2.436***	2.430***	1.336
	(2.50)	(2.60)	(2.79)	(1.33)
OC	−1.172***	−0.812**	−1.206***	−0.793**
	(−3.83)	(−2.44)	(−3.71)	(−2.54)
ROA	2.864	−10.159***	−7.455**	−5.021
	(0.64)	(−2.94)	(−2.00)	(−1.22)
Growth	−1.223***	−0.377	−0.904***	−1.021**
	(−3.22)	(−1.04)	(−2.68)	(−2.56)
RL	0.023	0.062*	0.069*	0.028
	(0.72)	(1.73)	(1.85)	(0.87)
Concen	0.004	0.007	0.009	0.008
	(0.36)	(0.75)	(0.98)	(0.82)
BDS	−0.113	−0.111	0.062	−0.244***
	(−1.30)	(−1.23)	(0.68)	(−2.70)
DUA	0.080	−0.143	−0.113	−0.016
	(0.25)	(−0.47)	(−0.39)	(−0.05)
BSS	0.475***	0.699***	0.655***	0.413**
	(3.13)	(4.87)	(4.64)	(2.52)
Constant	−9.403***	−5.577*	−12.276***	−10.845***
	(−3.18)	(−1.94)	(−4.15)	(−3.43)
Company	YES	YES	YES	YES
Year	YES	YES	YES	YES
N	2,026	2,035	2,092	1,879
R-squared	0.090	0.080	0.116	0.081

Conclusion and implications

Research conclusions

In this paper, we use the multi-period DID method to test the effect of AANR on corporate pollution emissions with a sample of Chinese A-share heavy-polluting corporations from 2010-2017. Then, the impact mechanisms are analyzed and validated based on existing theories. Finally, the influence of different corporate characteristics, internal governance, external environment, and external governance on the policy effects are also explored. The main conclusions of the research are as follows.

1. The AANR can significantly reduce the pollutant emissions of heavy-polluting corporations in the pilot area, and the robustness test results support this conclusion. It suggests that AANR can place environmental performance evaluation pressure on local officials and ultimately transfer that pressure to corporations in their jurisdictions to achieve emission reductions. This finding verifies the micro governance role of the AANR.

2. The impact mechanisms of macro policies on micro corporations are complicated, but this paper explores two possible pathways through which AANR influences corporate emissions reductions. Based on legitimacy theory, we find that AANR forces corporations to reduce emissions by increasing environmental pressure. Based on resource dependence theory, we find that AANR alleviates the financial constraints of corporations that cooperate in reducing emissions, and this resource incentive can promote corporate emissions reductions. These meaningful discoveries enrich the theoretical system of national environmental auditing.

3. The heterogeneity test results indicate that corporate and environmental characteristics significantly impact the implementation effectiveness of AANR. Specifically, the AANR promotes emission reduction more strongly for large-sized corporations and political connection corporations, and more significantly for corporations with higher internal and external governance levels. Audit pilot’s effects on the treatment of corporate pollution are more significant in regions with high economic growth and legalization. These conclusions provide a meaningful theoretical basis for further improvement and differentiated implementation of AANR.

Implications

The conclusions of this paper provide valuable empirical evidence for the policy effects assessment of AANR and validate the micro-environmental governance function of national auditing. The main insights that can be drawn from this research are as follows.

First, the central government should adhere to the implementation of AANR and utilize the "immune function" of national audits for environmental management. Audit institutions need to utilize the professionalism and authority advantage of AANR to solidify local officials’ environmental responsibility further. Governments at all levels should further calculate the "ecological accounts" and "economic accounts" and continuously improve the evaluation system for officials to raise their attention to environmental issues. Relying on the change of evaluation pressure to open up the top-down pressure transmission path and promote energy saving and emission reduction of micro entities. Utilizing the micro-governance function of national audits to help the economy and society achieve the green transformation.

Secondly, in the process of implementing AANR, local governments can take various measures to make corporations in their jurisdictions work together on environmental management. Increasing environmental regulation is usually the preferred option for local governments with enforcement powers. This method can effectively raise the legitimacy pressure on corporations and internalize the cost of corporate pollution, thus leading corporations to reduce their environmental pollution. Overly stringent environmental regulations can affect corporate production and operations, and local governments may also consider rewarding corporations that cooperate in environmental management. Policy or resource support for well-performing corporations will inspire more corporations to take the initiative to reduce pollution.

Furthermore, differentiated policies are needed for corporations and regions with different characteristics in implementing AANR. For small corporations, emphasis should be placed on providing financial or policy support so that they can afford to adopt cleaner technologies and equipment. For corporations with a weak connection with the government, the focus is on strengthening the policy so that corporations understand the benefits of participation in environmental governance and the risks of non-participation. For regions with poor economic development and legalization, emphasis should be placed on the regulation of local protectionist and opportunistic behaviors, including punishing those who sacrifice ecology for economic development and ensuring the effective enforcement of policies.

Finally, to fulfill the governance function of AANR for micro corporations, it is not enough to rely on the power of the government. News media and securities analysts need to pay more attention to the impact of corporations on the environment and disclose corporate positive and negative environmental performance. At the same time, it is necessary for corporations to continuously improve their internal control mechanisms to prevent environmental risks and build an excellent corporate image through active environmental management. In addition, corporations should be sensitive to the policy signals released by AANR, seize first-mover opportunities as much as possible, reduce resource consumption and pollutant emissions through green innovation and industrial technology innovation, and accelerate corporate transformation and upgrading.

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Kewei Hu and Yugui Hao. The first draft of the manuscript was written by Dan Yu and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Data availability statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Zhejiang Provincial Social Science Planning Project "Zhejiang Province Carbon Audit System Design and Digital Collaborative Operation Mechanism Research" (Grant numbers 23NDJC199YB).